2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/jhash.h>
48 #include "current_stateid.h"
53 #define NFSDDBG_FACILITY NFSDDBG_PROC
55 #define all_ones {{~0,~0},~0}
56 static const stateid_t one_stateid
= {
58 .si_opaque
= all_ones
,
60 static const stateid_t zero_stateid
= {
63 static const stateid_t currentstateid
= {
67 static u64 current_sessionid
= 1;
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
71 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
73 /* forward declarations */
74 static bool check_for_locks(struct nfs4_file
*fp
, struct nfs4_lockowner
*lowner
);
75 static void nfs4_free_ol_stateid(struct nfs4_stid
*stid
);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(state_lock
);
87 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
88 * the refcount on the open stateid to drop.
90 static DECLARE_WAIT_QUEUE_HEAD(close_wq
);
92 static struct kmem_cache
*openowner_slab
;
93 static struct kmem_cache
*lockowner_slab
;
94 static struct kmem_cache
*file_slab
;
95 static struct kmem_cache
*stateid_slab
;
96 static struct kmem_cache
*deleg_slab
;
97 static struct kmem_cache
*odstate_slab
;
99 static void free_session(struct nfsd4_session
*);
101 static struct nfsd4_callback_ops nfsd4_cb_recall_ops
;
103 static bool is_session_dead(struct nfsd4_session
*ses
)
105 return ses
->se_flags
& NFS4_SESSION_DEAD
;
108 static __be32
mark_session_dead_locked(struct nfsd4_session
*ses
, int ref_held_by_me
)
110 if (atomic_read(&ses
->se_ref
) > ref_held_by_me
)
111 return nfserr_jukebox
;
112 ses
->se_flags
|= NFS4_SESSION_DEAD
;
116 static bool is_client_expired(struct nfs4_client
*clp
)
118 return clp
->cl_time
== 0;
121 static __be32
get_client_locked(struct nfs4_client
*clp
)
123 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
125 lockdep_assert_held(&nn
->client_lock
);
127 if (is_client_expired(clp
))
128 return nfserr_expired
;
129 atomic_inc(&clp
->cl_refcount
);
133 /* must be called under the client_lock */
135 renew_client_locked(struct nfs4_client
*clp
)
137 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
139 if (is_client_expired(clp
)) {
141 printk("%s: client (clientid %08x/%08x) already expired\n",
143 clp
->cl_clientid
.cl_boot
,
144 clp
->cl_clientid
.cl_id
);
148 dprintk("renewing client (clientid %08x/%08x)\n",
149 clp
->cl_clientid
.cl_boot
,
150 clp
->cl_clientid
.cl_id
);
151 list_move_tail(&clp
->cl_lru
, &nn
->client_lru
);
152 clp
->cl_time
= get_seconds();
155 static void put_client_renew_locked(struct nfs4_client
*clp
)
157 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
159 lockdep_assert_held(&nn
->client_lock
);
161 if (!atomic_dec_and_test(&clp
->cl_refcount
))
163 if (!is_client_expired(clp
))
164 renew_client_locked(clp
);
167 static void put_client_renew(struct nfs4_client
*clp
)
169 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
171 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &nn
->client_lock
))
173 if (!is_client_expired(clp
))
174 renew_client_locked(clp
);
175 spin_unlock(&nn
->client_lock
);
178 static __be32
nfsd4_get_session_locked(struct nfsd4_session
*ses
)
182 if (is_session_dead(ses
))
183 return nfserr_badsession
;
184 status
= get_client_locked(ses
->se_client
);
187 atomic_inc(&ses
->se_ref
);
191 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
193 struct nfs4_client
*clp
= ses
->se_client
;
194 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
196 lockdep_assert_held(&nn
->client_lock
);
198 if (atomic_dec_and_test(&ses
->se_ref
) && is_session_dead(ses
))
200 put_client_renew_locked(clp
);
203 static void nfsd4_put_session(struct nfsd4_session
*ses
)
205 struct nfs4_client
*clp
= ses
->se_client
;
206 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
208 spin_lock(&nn
->client_lock
);
209 nfsd4_put_session_locked(ses
);
210 spin_unlock(&nn
->client_lock
);
213 static inline struct nfs4_stateowner
*
214 nfs4_get_stateowner(struct nfs4_stateowner
*sop
)
216 atomic_inc(&sop
->so_count
);
221 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
)
223 return (sop
->so_owner
.len
== owner
->len
) &&
224 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
);
227 static struct nfs4_openowner
*
228 find_openstateowner_str_locked(unsigned int hashval
, struct nfsd4_open
*open
,
229 struct nfs4_client
*clp
)
231 struct nfs4_stateowner
*so
;
233 lockdep_assert_held(&clp
->cl_lock
);
235 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[hashval
],
237 if (!so
->so_is_open_owner
)
239 if (same_owner_str(so
, &open
->op_owner
))
240 return openowner(nfs4_get_stateowner(so
));
245 static struct nfs4_openowner
*
246 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
,
247 struct nfs4_client
*clp
)
249 struct nfs4_openowner
*oo
;
251 spin_lock(&clp
->cl_lock
);
252 oo
= find_openstateowner_str_locked(hashval
, open
, clp
);
253 spin_unlock(&clp
->cl_lock
);
258 opaque_hashval(const void *ptr
, int nbytes
)
260 unsigned char *cptr
= (unsigned char *) ptr
;
270 static void nfsd4_free_file_rcu(struct rcu_head
*rcu
)
272 struct nfs4_file
*fp
= container_of(rcu
, struct nfs4_file
, fi_rcu
);
274 kmem_cache_free(file_slab
, fp
);
278 put_nfs4_file(struct nfs4_file
*fi
)
280 might_lock(&state_lock
);
282 if (atomic_dec_and_lock(&fi
->fi_ref
, &state_lock
)) {
283 hlist_del_rcu(&fi
->fi_hash
);
284 spin_unlock(&state_lock
);
285 WARN_ON_ONCE(!list_empty(&fi
->fi_clnt_odstate
));
286 WARN_ON_ONCE(!list_empty(&fi
->fi_delegations
));
287 call_rcu(&fi
->fi_rcu
, nfsd4_free_file_rcu
);
292 __nfs4_get_fd(struct nfs4_file
*f
, int oflag
)
294 if (f
->fi_fds
[oflag
])
295 return get_file(f
->fi_fds
[oflag
]);
300 find_writeable_file_locked(struct nfs4_file
*f
)
304 lockdep_assert_held(&f
->fi_lock
);
306 ret
= __nfs4_get_fd(f
, O_WRONLY
);
308 ret
= __nfs4_get_fd(f
, O_RDWR
);
313 find_writeable_file(struct nfs4_file
*f
)
317 spin_lock(&f
->fi_lock
);
318 ret
= find_writeable_file_locked(f
);
319 spin_unlock(&f
->fi_lock
);
324 static struct file
*find_readable_file_locked(struct nfs4_file
*f
)
328 lockdep_assert_held(&f
->fi_lock
);
330 ret
= __nfs4_get_fd(f
, O_RDONLY
);
332 ret
= __nfs4_get_fd(f
, O_RDWR
);
337 find_readable_file(struct nfs4_file
*f
)
341 spin_lock(&f
->fi_lock
);
342 ret
= find_readable_file_locked(f
);
343 spin_unlock(&f
->fi_lock
);
349 find_any_file(struct nfs4_file
*f
)
353 spin_lock(&f
->fi_lock
);
354 ret
= __nfs4_get_fd(f
, O_RDWR
);
356 ret
= __nfs4_get_fd(f
, O_WRONLY
);
358 ret
= __nfs4_get_fd(f
, O_RDONLY
);
360 spin_unlock(&f
->fi_lock
);
364 static atomic_long_t num_delegations
;
365 unsigned long max_delegations
;
368 * Open owner state (share locks)
371 /* hash tables for lock and open owners */
372 #define OWNER_HASH_BITS 8
373 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
374 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
376 static unsigned int ownerstr_hashval(struct xdr_netobj
*ownername
)
380 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
381 return ret
& OWNER_HASH_MASK
;
384 /* hash table for nfs4_file */
385 #define FILE_HASH_BITS 8
386 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
388 static unsigned int nfsd_fh_hashval(struct knfsd_fh
*fh
)
390 return jhash2(fh
->fh_base
.fh_pad
, XDR_QUADLEN(fh
->fh_size
), 0);
393 static unsigned int file_hashval(struct knfsd_fh
*fh
)
395 return nfsd_fh_hashval(fh
) & (FILE_HASH_SIZE
- 1);
398 static struct hlist_head file_hashtbl
[FILE_HASH_SIZE
];
401 __nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
403 lockdep_assert_held(&fp
->fi_lock
);
405 if (access
& NFS4_SHARE_ACCESS_WRITE
)
406 atomic_inc(&fp
->fi_access
[O_WRONLY
]);
407 if (access
& NFS4_SHARE_ACCESS_READ
)
408 atomic_inc(&fp
->fi_access
[O_RDONLY
]);
412 nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
414 lockdep_assert_held(&fp
->fi_lock
);
416 /* Does this access mode make sense? */
417 if (access
& ~NFS4_SHARE_ACCESS_BOTH
)
420 /* Does it conflict with a deny mode already set? */
421 if ((access
& fp
->fi_share_deny
) != 0)
422 return nfserr_share_denied
;
424 __nfs4_file_get_access(fp
, access
);
428 static __be32
nfs4_file_check_deny(struct nfs4_file
*fp
, u32 deny
)
430 /* Common case is that there is no deny mode. */
432 /* Does this deny mode make sense? */
433 if (deny
& ~NFS4_SHARE_DENY_BOTH
)
436 if ((deny
& NFS4_SHARE_DENY_READ
) &&
437 atomic_read(&fp
->fi_access
[O_RDONLY
]))
438 return nfserr_share_denied
;
440 if ((deny
& NFS4_SHARE_DENY_WRITE
) &&
441 atomic_read(&fp
->fi_access
[O_WRONLY
]))
442 return nfserr_share_denied
;
447 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
449 might_lock(&fp
->fi_lock
);
451 if (atomic_dec_and_lock(&fp
->fi_access
[oflag
], &fp
->fi_lock
)) {
452 struct file
*f1
= NULL
;
453 struct file
*f2
= NULL
;
455 swap(f1
, fp
->fi_fds
[oflag
]);
456 if (atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
457 swap(f2
, fp
->fi_fds
[O_RDWR
]);
458 spin_unlock(&fp
->fi_lock
);
466 static void nfs4_file_put_access(struct nfs4_file
*fp
, u32 access
)
468 WARN_ON_ONCE(access
& ~NFS4_SHARE_ACCESS_BOTH
);
470 if (access
& NFS4_SHARE_ACCESS_WRITE
)
471 __nfs4_file_put_access(fp
, O_WRONLY
);
472 if (access
& NFS4_SHARE_ACCESS_READ
)
473 __nfs4_file_put_access(fp
, O_RDONLY
);
477 * Allocate a new open/delegation state counter. This is needed for
478 * pNFS for proper return on close semantics.
480 * Note that we only allocate it for pNFS-enabled exports, otherwise
481 * all pointers to struct nfs4_clnt_odstate are always NULL.
483 static struct nfs4_clnt_odstate
*
484 alloc_clnt_odstate(struct nfs4_client
*clp
)
486 struct nfs4_clnt_odstate
*co
;
488 co
= kmem_cache_zalloc(odstate_slab
, GFP_KERNEL
);
491 atomic_set(&co
->co_odcount
, 1);
497 hash_clnt_odstate_locked(struct nfs4_clnt_odstate
*co
)
499 struct nfs4_file
*fp
= co
->co_file
;
501 lockdep_assert_held(&fp
->fi_lock
);
502 list_add(&co
->co_perfile
, &fp
->fi_clnt_odstate
);
506 get_clnt_odstate(struct nfs4_clnt_odstate
*co
)
509 atomic_inc(&co
->co_odcount
);
513 put_clnt_odstate(struct nfs4_clnt_odstate
*co
)
515 struct nfs4_file
*fp
;
521 if (atomic_dec_and_lock(&co
->co_odcount
, &fp
->fi_lock
)) {
522 list_del(&co
->co_perfile
);
523 spin_unlock(&fp
->fi_lock
);
525 nfsd4_return_all_file_layouts(co
->co_client
, fp
);
526 kmem_cache_free(odstate_slab
, co
);
530 static struct nfs4_clnt_odstate
*
531 find_or_hash_clnt_odstate(struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*new)
533 struct nfs4_clnt_odstate
*co
;
534 struct nfs4_client
*cl
;
541 spin_lock(&fp
->fi_lock
);
542 list_for_each_entry(co
, &fp
->fi_clnt_odstate
, co_perfile
) {
543 if (co
->co_client
== cl
) {
544 get_clnt_odstate(co
);
550 hash_clnt_odstate_locked(new);
552 spin_unlock(&fp
->fi_lock
);
556 struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
,
557 struct kmem_cache
*slab
)
559 struct nfs4_stid
*stid
;
562 stid
= kmem_cache_zalloc(slab
, GFP_KERNEL
);
566 idr_preload(GFP_KERNEL
);
567 spin_lock(&cl
->cl_lock
);
568 new_id
= idr_alloc_cyclic(&cl
->cl_stateids
, stid
, 0, 0, GFP_NOWAIT
);
569 spin_unlock(&cl
->cl_lock
);
573 stid
->sc_client
= cl
;
574 stid
->sc_stateid
.si_opaque
.so_id
= new_id
;
575 stid
->sc_stateid
.si_opaque
.so_clid
= cl
->cl_clientid
;
576 /* Will be incremented before return to client: */
577 atomic_set(&stid
->sc_count
, 1);
580 * It shouldn't be a problem to reuse an opaque stateid value.
581 * I don't think it is for 4.1. But with 4.0 I worry that, for
582 * example, a stray write retransmission could be accepted by
583 * the server when it should have been rejected. Therefore,
584 * adopt a trick from the sctp code to attempt to maximize the
585 * amount of time until an id is reused, by ensuring they always
586 * "increase" (mod INT_MAX):
590 kmem_cache_free(slab
, stid
);
594 static struct nfs4_ol_stateid
* nfs4_alloc_open_stateid(struct nfs4_client
*clp
)
596 struct nfs4_stid
*stid
;
597 struct nfs4_ol_stateid
*stp
;
599 stid
= nfs4_alloc_stid(clp
, stateid_slab
);
603 stp
= openlockstateid(stid
);
604 stp
->st_stid
.sc_free
= nfs4_free_ol_stateid
;
608 static void nfs4_free_deleg(struct nfs4_stid
*stid
)
610 kmem_cache_free(deleg_slab
, stid
);
611 atomic_long_dec(&num_delegations
);
615 * When we recall a delegation, we should be careful not to hand it
616 * out again straight away.
617 * To ensure this we keep a pair of bloom filters ('new' and 'old')
618 * in which the filehandles of recalled delegations are "stored".
619 * If a filehandle appear in either filter, a delegation is blocked.
620 * When a delegation is recalled, the filehandle is stored in the "new"
622 * Every 30 seconds we swap the filters and clear the "new" one,
623 * unless both are empty of course.
625 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
626 * low 3 bytes as hash-table indices.
628 * 'blocked_delegations_lock', which is always taken in block_delegations(),
629 * is used to manage concurrent access. Testing does not need the lock
630 * except when swapping the two filters.
632 static DEFINE_SPINLOCK(blocked_delegations_lock
);
633 static struct bloom_pair
{
634 int entries
, old_entries
;
636 int new; /* index into 'set' */
637 DECLARE_BITMAP(set
[2], 256);
638 } blocked_delegations
;
640 static int delegation_blocked(struct knfsd_fh
*fh
)
643 struct bloom_pair
*bd
= &blocked_delegations
;
645 if (bd
->entries
== 0)
647 if (seconds_since_boot() - bd
->swap_time
> 30) {
648 spin_lock(&blocked_delegations_lock
);
649 if (seconds_since_boot() - bd
->swap_time
> 30) {
650 bd
->entries
-= bd
->old_entries
;
651 bd
->old_entries
= bd
->entries
;
652 memset(bd
->set
[bd
->new], 0,
655 bd
->swap_time
= seconds_since_boot();
657 spin_unlock(&blocked_delegations_lock
);
659 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
660 if (test_bit(hash
&255, bd
->set
[0]) &&
661 test_bit((hash
>>8)&255, bd
->set
[0]) &&
662 test_bit((hash
>>16)&255, bd
->set
[0]))
665 if (test_bit(hash
&255, bd
->set
[1]) &&
666 test_bit((hash
>>8)&255, bd
->set
[1]) &&
667 test_bit((hash
>>16)&255, bd
->set
[1]))
673 static void block_delegations(struct knfsd_fh
*fh
)
676 struct bloom_pair
*bd
= &blocked_delegations
;
678 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
680 spin_lock(&blocked_delegations_lock
);
681 __set_bit(hash
&255, bd
->set
[bd
->new]);
682 __set_bit((hash
>>8)&255, bd
->set
[bd
->new]);
683 __set_bit((hash
>>16)&255, bd
->set
[bd
->new]);
684 if (bd
->entries
== 0)
685 bd
->swap_time
= seconds_since_boot();
687 spin_unlock(&blocked_delegations_lock
);
690 static struct nfs4_delegation
*
691 alloc_init_deleg(struct nfs4_client
*clp
, struct svc_fh
*current_fh
,
692 struct nfs4_clnt_odstate
*odstate
)
694 struct nfs4_delegation
*dp
;
697 dprintk("NFSD alloc_init_deleg\n");
698 n
= atomic_long_inc_return(&num_delegations
);
699 if (n
< 0 || n
> max_delegations
)
701 if (delegation_blocked(¤t_fh
->fh_handle
))
703 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
707 dp
->dl_stid
.sc_free
= nfs4_free_deleg
;
709 * delegation seqid's are never incremented. The 4.1 special
710 * meaning of seqid 0 isn't meaningful, really, but let's avoid
711 * 0 anyway just for consistency and use 1:
713 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
714 INIT_LIST_HEAD(&dp
->dl_perfile
);
715 INIT_LIST_HEAD(&dp
->dl_perclnt
);
716 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
717 dp
->dl_clnt_odstate
= odstate
;
718 get_clnt_odstate(odstate
);
719 dp
->dl_type
= NFS4_OPEN_DELEGATE_READ
;
721 nfsd4_init_cb(&dp
->dl_recall
, dp
->dl_stid
.sc_client
,
722 &nfsd4_cb_recall_ops
, NFSPROC4_CLNT_CB_RECALL
);
725 atomic_long_dec(&num_delegations
);
730 nfs4_put_stid(struct nfs4_stid
*s
)
732 struct nfs4_file
*fp
= s
->sc_file
;
733 struct nfs4_client
*clp
= s
->sc_client
;
735 might_lock(&clp
->cl_lock
);
737 if (!atomic_dec_and_lock(&s
->sc_count
, &clp
->cl_lock
)) {
738 wake_up_all(&close_wq
);
741 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
742 spin_unlock(&clp
->cl_lock
);
748 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
750 struct file
*filp
= NULL
;
752 spin_lock(&fp
->fi_lock
);
753 if (fp
->fi_deleg_file
&& --fp
->fi_delegees
== 0)
754 swap(filp
, fp
->fi_deleg_file
);
755 spin_unlock(&fp
->fi_lock
);
758 vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&fp
);
763 void nfs4_unhash_stid(struct nfs4_stid
*s
)
769 hash_delegation_locked(struct nfs4_delegation
*dp
, struct nfs4_file
*fp
)
771 lockdep_assert_held(&state_lock
);
772 lockdep_assert_held(&fp
->fi_lock
);
774 atomic_inc(&dp
->dl_stid
.sc_count
);
775 dp
->dl_stid
.sc_type
= NFS4_DELEG_STID
;
776 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
777 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
781 unhash_delegation_locked(struct nfs4_delegation
*dp
)
783 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
785 lockdep_assert_held(&state_lock
);
787 dp
->dl_stid
.sc_type
= NFS4_CLOSED_DELEG_STID
;
788 /* Ensure that deleg break won't try to requeue it */
790 spin_lock(&fp
->fi_lock
);
791 list_del_init(&dp
->dl_perclnt
);
792 list_del_init(&dp
->dl_recall_lru
);
793 list_del_init(&dp
->dl_perfile
);
794 spin_unlock(&fp
->fi_lock
);
797 static void destroy_delegation(struct nfs4_delegation
*dp
)
799 spin_lock(&state_lock
);
800 unhash_delegation_locked(dp
);
801 spin_unlock(&state_lock
);
802 put_clnt_odstate(dp
->dl_clnt_odstate
);
803 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
804 nfs4_put_stid(&dp
->dl_stid
);
807 static void revoke_delegation(struct nfs4_delegation
*dp
)
809 struct nfs4_client
*clp
= dp
->dl_stid
.sc_client
;
811 WARN_ON(!list_empty(&dp
->dl_recall_lru
));
813 put_clnt_odstate(dp
->dl_clnt_odstate
);
814 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
816 if (clp
->cl_minorversion
== 0)
817 nfs4_put_stid(&dp
->dl_stid
);
819 dp
->dl_stid
.sc_type
= NFS4_REVOKED_DELEG_STID
;
820 spin_lock(&clp
->cl_lock
);
821 list_add(&dp
->dl_recall_lru
, &clp
->cl_revoked
);
822 spin_unlock(&clp
->cl_lock
);
830 static unsigned int clientid_hashval(u32 id
)
832 return id
& CLIENT_HASH_MASK
;
835 static unsigned int clientstr_hashval(const char *name
)
837 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
841 * We store the NONE, READ, WRITE, and BOTH bits separately in the
842 * st_{access,deny}_bmap field of the stateid, in order to track not
843 * only what share bits are currently in force, but also what
844 * combinations of share bits previous opens have used. This allows us
845 * to enforce the recommendation of rfc 3530 14.2.19 that the server
846 * return an error if the client attempt to downgrade to a combination
847 * of share bits not explicable by closing some of its previous opens.
849 * XXX: This enforcement is actually incomplete, since we don't keep
850 * track of access/deny bit combinations; so, e.g., we allow:
852 * OPEN allow read, deny write
853 * OPEN allow both, deny none
854 * DOWNGRADE allow read, deny none
856 * which we should reject.
859 bmap_to_share_mode(unsigned long bmap
) {
861 unsigned int access
= 0;
863 for (i
= 1; i
< 4; i
++) {
864 if (test_bit(i
, &bmap
))
870 /* set share access for a given stateid */
872 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
874 unsigned char mask
= 1 << access
;
876 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
877 stp
->st_access_bmap
|= mask
;
880 /* clear share access for a given stateid */
882 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
884 unsigned char mask
= 1 << access
;
886 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
887 stp
->st_access_bmap
&= ~mask
;
890 /* test whether a given stateid has access */
892 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
894 unsigned char mask
= 1 << access
;
896 return (bool)(stp
->st_access_bmap
& mask
);
899 /* set share deny for a given stateid */
901 set_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
903 unsigned char mask
= 1 << deny
;
905 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
906 stp
->st_deny_bmap
|= mask
;
909 /* clear share deny for a given stateid */
911 clear_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
913 unsigned char mask
= 1 << deny
;
915 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
916 stp
->st_deny_bmap
&= ~mask
;
919 /* test whether a given stateid is denying specific access */
921 test_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
923 unsigned char mask
= 1 << deny
;
925 return (bool)(stp
->st_deny_bmap
& mask
);
928 static int nfs4_access_to_omode(u32 access
)
930 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
931 case NFS4_SHARE_ACCESS_READ
:
933 case NFS4_SHARE_ACCESS_WRITE
:
935 case NFS4_SHARE_ACCESS_BOTH
:
943 * A stateid that had a deny mode associated with it is being released
944 * or downgraded. Recalculate the deny mode on the file.
947 recalculate_deny_mode(struct nfs4_file
*fp
)
949 struct nfs4_ol_stateid
*stp
;
951 spin_lock(&fp
->fi_lock
);
952 fp
->fi_share_deny
= 0;
953 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
)
954 fp
->fi_share_deny
|= bmap_to_share_mode(stp
->st_deny_bmap
);
955 spin_unlock(&fp
->fi_lock
);
959 reset_union_bmap_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
964 for (i
= 1; i
< 4; i
++) {
965 if ((i
& deny
) != i
) {
971 /* Recalculate per-file deny mode if there was a change */
973 recalculate_deny_mode(stp
->st_stid
.sc_file
);
976 /* release all access and file references for a given stateid */
978 release_all_access(struct nfs4_ol_stateid
*stp
)
981 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
983 if (fp
&& stp
->st_deny_bmap
!= 0)
984 recalculate_deny_mode(fp
);
986 for (i
= 1; i
< 4; i
++) {
987 if (test_access(i
, stp
))
988 nfs4_file_put_access(stp
->st_stid
.sc_file
, i
);
989 clear_access(i
, stp
);
993 static void nfs4_put_stateowner(struct nfs4_stateowner
*sop
)
995 struct nfs4_client
*clp
= sop
->so_client
;
997 might_lock(&clp
->cl_lock
);
999 if (!atomic_dec_and_lock(&sop
->so_count
, &clp
->cl_lock
))
1001 sop
->so_ops
->so_unhash(sop
);
1002 spin_unlock(&clp
->cl_lock
);
1003 kfree(sop
->so_owner
.data
);
1004 sop
->so_ops
->so_free(sop
);
1007 static void unhash_ol_stateid(struct nfs4_ol_stateid
*stp
)
1009 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
1011 lockdep_assert_held(&stp
->st_stateowner
->so_client
->cl_lock
);
1013 spin_lock(&fp
->fi_lock
);
1014 list_del(&stp
->st_perfile
);
1015 spin_unlock(&fp
->fi_lock
);
1016 list_del(&stp
->st_perstateowner
);
1019 static void nfs4_free_ol_stateid(struct nfs4_stid
*stid
)
1021 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1023 put_clnt_odstate(stp
->st_clnt_odstate
);
1024 release_all_access(stp
);
1025 if (stp
->st_stateowner
)
1026 nfs4_put_stateowner(stp
->st_stateowner
);
1027 kmem_cache_free(stateid_slab
, stid
);
1030 static void nfs4_free_lock_stateid(struct nfs4_stid
*stid
)
1032 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1033 struct nfs4_lockowner
*lo
= lockowner(stp
->st_stateowner
);
1036 file
= find_any_file(stp
->st_stid
.sc_file
);
1038 filp_close(file
, (fl_owner_t
)lo
);
1039 nfs4_free_ol_stateid(stid
);
1043 * Put the persistent reference to an already unhashed generic stateid, while
1044 * holding the cl_lock. If it's the last reference, then put it onto the
1045 * reaplist for later destruction.
1047 static void put_ol_stateid_locked(struct nfs4_ol_stateid
*stp
,
1048 struct list_head
*reaplist
)
1050 struct nfs4_stid
*s
= &stp
->st_stid
;
1051 struct nfs4_client
*clp
= s
->sc_client
;
1053 lockdep_assert_held(&clp
->cl_lock
);
1055 WARN_ON_ONCE(!list_empty(&stp
->st_locks
));
1057 if (!atomic_dec_and_test(&s
->sc_count
)) {
1058 wake_up_all(&close_wq
);
1062 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
1063 list_add(&stp
->st_locks
, reaplist
);
1066 static void unhash_lock_stateid(struct nfs4_ol_stateid
*stp
)
1068 struct nfs4_openowner
*oo
= openowner(stp
->st_openstp
->st_stateowner
);
1070 lockdep_assert_held(&oo
->oo_owner
.so_client
->cl_lock
);
1072 list_del_init(&stp
->st_locks
);
1073 unhash_ol_stateid(stp
);
1074 nfs4_unhash_stid(&stp
->st_stid
);
1077 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
1079 struct nfs4_openowner
*oo
= openowner(stp
->st_openstp
->st_stateowner
);
1081 spin_lock(&oo
->oo_owner
.so_client
->cl_lock
);
1082 unhash_lock_stateid(stp
);
1083 spin_unlock(&oo
->oo_owner
.so_client
->cl_lock
);
1084 nfs4_put_stid(&stp
->st_stid
);
1087 static void unhash_lockowner_locked(struct nfs4_lockowner
*lo
)
1089 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
1091 lockdep_assert_held(&clp
->cl_lock
);
1093 list_del_init(&lo
->lo_owner
.so_strhash
);
1097 * Free a list of generic stateids that were collected earlier after being
1101 free_ol_stateid_reaplist(struct list_head
*reaplist
)
1103 struct nfs4_ol_stateid
*stp
;
1104 struct nfs4_file
*fp
;
1108 while (!list_empty(reaplist
)) {
1109 stp
= list_first_entry(reaplist
, struct nfs4_ol_stateid
,
1111 list_del(&stp
->st_locks
);
1112 fp
= stp
->st_stid
.sc_file
;
1113 stp
->st_stid
.sc_free(&stp
->st_stid
);
1119 static void release_lockowner(struct nfs4_lockowner
*lo
)
1121 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
1122 struct nfs4_ol_stateid
*stp
;
1123 struct list_head reaplist
;
1125 INIT_LIST_HEAD(&reaplist
);
1127 spin_lock(&clp
->cl_lock
);
1128 unhash_lockowner_locked(lo
);
1129 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
1130 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
1131 struct nfs4_ol_stateid
, st_perstateowner
);
1132 unhash_lock_stateid(stp
);
1133 put_ol_stateid_locked(stp
, &reaplist
);
1135 spin_unlock(&clp
->cl_lock
);
1136 free_ol_stateid_reaplist(&reaplist
);
1137 nfs4_put_stateowner(&lo
->lo_owner
);
1140 static void release_open_stateid_locks(struct nfs4_ol_stateid
*open_stp
,
1141 struct list_head
*reaplist
)
1143 struct nfs4_ol_stateid
*stp
;
1145 while (!list_empty(&open_stp
->st_locks
)) {
1146 stp
= list_entry(open_stp
->st_locks
.next
,
1147 struct nfs4_ol_stateid
, st_locks
);
1148 unhash_lock_stateid(stp
);
1149 put_ol_stateid_locked(stp
, reaplist
);
1153 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
,
1154 struct list_head
*reaplist
)
1156 lockdep_assert_held(&stp
->st_stid
.sc_client
->cl_lock
);
1158 unhash_ol_stateid(stp
);
1159 release_open_stateid_locks(stp
, reaplist
);
1162 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
1164 LIST_HEAD(reaplist
);
1166 spin_lock(&stp
->st_stid
.sc_client
->cl_lock
);
1167 unhash_open_stateid(stp
, &reaplist
);
1168 put_ol_stateid_locked(stp
, &reaplist
);
1169 spin_unlock(&stp
->st_stid
.sc_client
->cl_lock
);
1170 free_ol_stateid_reaplist(&reaplist
);
1173 static void unhash_openowner_locked(struct nfs4_openowner
*oo
)
1175 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1177 lockdep_assert_held(&clp
->cl_lock
);
1179 list_del_init(&oo
->oo_owner
.so_strhash
);
1180 list_del_init(&oo
->oo_perclient
);
1183 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
1185 struct nfsd_net
*nn
= net_generic(oo
->oo_owner
.so_client
->net
,
1187 struct nfs4_ol_stateid
*s
;
1189 spin_lock(&nn
->client_lock
);
1190 s
= oo
->oo_last_closed_stid
;
1192 list_del_init(&oo
->oo_close_lru
);
1193 oo
->oo_last_closed_stid
= NULL
;
1195 spin_unlock(&nn
->client_lock
);
1197 nfs4_put_stid(&s
->st_stid
);
1200 static void release_openowner(struct nfs4_openowner
*oo
)
1202 struct nfs4_ol_stateid
*stp
;
1203 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1204 struct list_head reaplist
;
1206 INIT_LIST_HEAD(&reaplist
);
1208 spin_lock(&clp
->cl_lock
);
1209 unhash_openowner_locked(oo
);
1210 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
1211 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
1212 struct nfs4_ol_stateid
, st_perstateowner
);
1213 unhash_open_stateid(stp
, &reaplist
);
1214 put_ol_stateid_locked(stp
, &reaplist
);
1216 spin_unlock(&clp
->cl_lock
);
1217 free_ol_stateid_reaplist(&reaplist
);
1218 release_last_closed_stateid(oo
);
1219 nfs4_put_stateowner(&oo
->oo_owner
);
1223 hash_sessionid(struct nfs4_sessionid
*sessionid
)
1225 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
1227 return sid
->sequence
% SESSION_HASH_SIZE
;
1230 #ifdef CONFIG_SUNRPC_DEBUG
1232 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1234 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
1235 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
1239 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1245 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1246 * won't be used for replay.
1248 void nfsd4_bump_seqid(struct nfsd4_compound_state
*cstate
, __be32 nfserr
)
1250 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
1252 if (nfserr
== nfserr_replay_me
)
1255 if (!seqid_mutating_err(ntohl(nfserr
))) {
1256 nfsd4_cstate_clear_replay(cstate
);
1261 if (so
->so_is_open_owner
)
1262 release_last_closed_stateid(openowner(so
));
1268 gen_sessionid(struct nfsd4_session
*ses
)
1270 struct nfs4_client
*clp
= ses
->se_client
;
1271 struct nfsd4_sessionid
*sid
;
1273 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
1274 sid
->clientid
= clp
->cl_clientid
;
1275 sid
->sequence
= current_sessionid
++;
1280 * The protocol defines ca_maxresponssize_cached to include the size of
1281 * the rpc header, but all we need to cache is the data starting after
1282 * the end of the initial SEQUENCE operation--the rest we regenerate
1283 * each time. Therefore we can advertise a ca_maxresponssize_cached
1284 * value that is the number of bytes in our cache plus a few additional
1285 * bytes. In order to stay on the safe side, and not promise more than
1286 * we can cache, those additional bytes must be the minimum possible: 24
1287 * bytes of rpc header (xid through accept state, with AUTH_NULL
1288 * verifier), 12 for the compound header (with zero-length tag), and 44
1289 * for the SEQUENCE op response:
1291 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1294 free_session_slots(struct nfsd4_session
*ses
)
1298 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
1299 kfree(ses
->se_slots
[i
]);
1303 * We don't actually need to cache the rpc and session headers, so we
1304 * can allocate a little less for each slot:
1306 static inline u32
slot_bytes(struct nfsd4_channel_attrs
*ca
)
1310 if (ca
->maxresp_cached
< NFSD_MIN_HDR_SEQ_SZ
)
1313 size
= ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
1314 return size
+ sizeof(struct nfsd4_slot
);
1318 * XXX: If we run out of reserved DRC memory we could (up to a point)
1319 * re-negotiate active sessions and reduce their slot usage to make
1320 * room for new connections. For now we just fail the create session.
1322 static u32
nfsd4_get_drc_mem(struct nfsd4_channel_attrs
*ca
)
1324 u32 slotsize
= slot_bytes(ca
);
1325 u32 num
= ca
->maxreqs
;
1328 spin_lock(&nfsd_drc_lock
);
1329 avail
= min((unsigned long)NFSD_MAX_MEM_PER_SESSION
,
1330 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
1331 num
= min_t(int, num
, avail
/ slotsize
);
1332 nfsd_drc_mem_used
+= num
* slotsize
;
1333 spin_unlock(&nfsd_drc_lock
);
1338 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs
*ca
)
1340 int slotsize
= slot_bytes(ca
);
1342 spin_lock(&nfsd_drc_lock
);
1343 nfsd_drc_mem_used
-= slotsize
* ca
->maxreqs
;
1344 spin_unlock(&nfsd_drc_lock
);
1347 static struct nfsd4_session
*alloc_session(struct nfsd4_channel_attrs
*fattrs
,
1348 struct nfsd4_channel_attrs
*battrs
)
1350 int numslots
= fattrs
->maxreqs
;
1351 int slotsize
= slot_bytes(fattrs
);
1352 struct nfsd4_session
*new;
1355 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
1356 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
1357 mem
= numslots
* sizeof(struct nfsd4_slot
*);
1359 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
1362 /* allocate each struct nfsd4_slot and data cache in one piece */
1363 for (i
= 0; i
< numslots
; i
++) {
1364 new->se_slots
[i
] = kzalloc(slotsize
, GFP_KERNEL
);
1365 if (!new->se_slots
[i
])
1369 memcpy(&new->se_fchannel
, fattrs
, sizeof(struct nfsd4_channel_attrs
));
1370 memcpy(&new->se_bchannel
, battrs
, sizeof(struct nfsd4_channel_attrs
));
1375 kfree(new->se_slots
[i
]);
1380 static void free_conn(struct nfsd4_conn
*c
)
1382 svc_xprt_put(c
->cn_xprt
);
1386 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
1388 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
1389 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
1391 spin_lock(&clp
->cl_lock
);
1392 if (!list_empty(&c
->cn_persession
)) {
1393 list_del(&c
->cn_persession
);
1396 nfsd4_probe_callback(clp
);
1397 spin_unlock(&clp
->cl_lock
);
1400 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
1402 struct nfsd4_conn
*conn
;
1404 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
1407 svc_xprt_get(rqstp
->rq_xprt
);
1408 conn
->cn_xprt
= rqstp
->rq_xprt
;
1409 conn
->cn_flags
= flags
;
1410 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
1414 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1416 conn
->cn_session
= ses
;
1417 list_add(&conn
->cn_persession
, &ses
->se_conns
);
1420 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1422 struct nfs4_client
*clp
= ses
->se_client
;
1424 spin_lock(&clp
->cl_lock
);
1425 __nfsd4_hash_conn(conn
, ses
);
1426 spin_unlock(&clp
->cl_lock
);
1429 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
1431 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
1432 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
1435 static void nfsd4_init_conn(struct svc_rqst
*rqstp
, struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1439 nfsd4_hash_conn(conn
, ses
);
1440 ret
= nfsd4_register_conn(conn
);
1442 /* oops; xprt is already down: */
1443 nfsd4_conn_lost(&conn
->cn_xpt_user
);
1444 /* We may have gained or lost a callback channel: */
1445 nfsd4_probe_callback_sync(ses
->se_client
);
1448 static struct nfsd4_conn
*alloc_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_create_session
*cses
)
1450 u32 dir
= NFS4_CDFC4_FORE
;
1452 if (cses
->flags
& SESSION4_BACK_CHAN
)
1453 dir
|= NFS4_CDFC4_BACK
;
1454 return alloc_conn(rqstp
, dir
);
1457 /* must be called under client_lock */
1458 static void nfsd4_del_conns(struct nfsd4_session
*s
)
1460 struct nfs4_client
*clp
= s
->se_client
;
1461 struct nfsd4_conn
*c
;
1463 spin_lock(&clp
->cl_lock
);
1464 while (!list_empty(&s
->se_conns
)) {
1465 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
1466 list_del_init(&c
->cn_persession
);
1467 spin_unlock(&clp
->cl_lock
);
1469 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
1472 spin_lock(&clp
->cl_lock
);
1474 spin_unlock(&clp
->cl_lock
);
1477 static void __free_session(struct nfsd4_session
*ses
)
1479 free_session_slots(ses
);
1483 static void free_session(struct nfsd4_session
*ses
)
1485 nfsd4_del_conns(ses
);
1486 nfsd4_put_drc_mem(&ses
->se_fchannel
);
1487 __free_session(ses
);
1490 static void init_session(struct svc_rqst
*rqstp
, struct nfsd4_session
*new, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
1493 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
1495 new->se_client
= clp
;
1498 INIT_LIST_HEAD(&new->se_conns
);
1500 new->se_cb_seq_nr
= 1;
1501 new->se_flags
= cses
->flags
;
1502 new->se_cb_prog
= cses
->callback_prog
;
1503 new->se_cb_sec
= cses
->cb_sec
;
1504 atomic_set(&new->se_ref
, 0);
1505 idx
= hash_sessionid(&new->se_sessionid
);
1506 list_add(&new->se_hash
, &nn
->sessionid_hashtbl
[idx
]);
1507 spin_lock(&clp
->cl_lock
);
1508 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
1509 spin_unlock(&clp
->cl_lock
);
1512 struct sockaddr
*sa
= svc_addr(rqstp
);
1514 * This is a little silly; with sessions there's no real
1515 * use for the callback address. Use the peer address
1516 * as a reasonable default for now, but consider fixing
1517 * the rpc client not to require an address in the
1520 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
1521 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
1525 /* caller must hold client_lock */
1526 static struct nfsd4_session
*
1527 __find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
)
1529 struct nfsd4_session
*elem
;
1531 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1533 lockdep_assert_held(&nn
->client_lock
);
1535 dump_sessionid(__func__
, sessionid
);
1536 idx
= hash_sessionid(sessionid
);
1537 /* Search in the appropriate list */
1538 list_for_each_entry(elem
, &nn
->sessionid_hashtbl
[idx
], se_hash
) {
1539 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1540 NFS4_MAX_SESSIONID_LEN
)) {
1545 dprintk("%s: session not found\n", __func__
);
1549 static struct nfsd4_session
*
1550 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
,
1553 struct nfsd4_session
*session
;
1554 __be32 status
= nfserr_badsession
;
1556 session
= __find_in_sessionid_hashtbl(sessionid
, net
);
1559 status
= nfsd4_get_session_locked(session
);
1567 /* caller must hold client_lock */
1569 unhash_session(struct nfsd4_session
*ses
)
1571 struct nfs4_client
*clp
= ses
->se_client
;
1572 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1574 lockdep_assert_held(&nn
->client_lock
);
1576 list_del(&ses
->se_hash
);
1577 spin_lock(&ses
->se_client
->cl_lock
);
1578 list_del(&ses
->se_perclnt
);
1579 spin_unlock(&ses
->se_client
->cl_lock
);
1582 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1584 STALE_CLIENTID(clientid_t
*clid
, struct nfsd_net
*nn
)
1587 * We're assuming the clid was not given out from a boot
1588 * precisely 2^32 (about 136 years) before this one. That seems
1589 * a safe assumption:
1591 if (clid
->cl_boot
== (u32
)nn
->boot_time
)
1593 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1594 clid
->cl_boot
, clid
->cl_id
, nn
->boot_time
);
1599 * XXX Should we use a slab cache ?
1600 * This type of memory management is somewhat inefficient, but we use it
1601 * anyway since SETCLIENTID is not a common operation.
1603 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1605 struct nfs4_client
*clp
;
1608 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1611 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1612 if (clp
->cl_name
.data
== NULL
)
1614 clp
->cl_ownerstr_hashtbl
= kmalloc(sizeof(struct list_head
) *
1615 OWNER_HASH_SIZE
, GFP_KERNEL
);
1616 if (!clp
->cl_ownerstr_hashtbl
)
1617 goto err_no_hashtbl
;
1618 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++)
1619 INIT_LIST_HEAD(&clp
->cl_ownerstr_hashtbl
[i
]);
1620 clp
->cl_name
.len
= name
.len
;
1621 INIT_LIST_HEAD(&clp
->cl_sessions
);
1622 idr_init(&clp
->cl_stateids
);
1623 atomic_set(&clp
->cl_refcount
, 0);
1624 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1625 INIT_LIST_HEAD(&clp
->cl_idhash
);
1626 INIT_LIST_HEAD(&clp
->cl_openowners
);
1627 INIT_LIST_HEAD(&clp
->cl_delegations
);
1628 INIT_LIST_HEAD(&clp
->cl_lru
);
1629 INIT_LIST_HEAD(&clp
->cl_revoked
);
1630 #ifdef CONFIG_NFSD_PNFS
1631 INIT_LIST_HEAD(&clp
->cl_lo_states
);
1633 spin_lock_init(&clp
->cl_lock
);
1634 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1637 kfree(clp
->cl_name
.data
);
1644 free_client(struct nfs4_client
*clp
)
1646 while (!list_empty(&clp
->cl_sessions
)) {
1647 struct nfsd4_session
*ses
;
1648 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1650 list_del(&ses
->se_perclnt
);
1651 WARN_ON_ONCE(atomic_read(&ses
->se_ref
));
1654 rpc_destroy_wait_queue(&clp
->cl_cb_waitq
);
1655 free_svc_cred(&clp
->cl_cred
);
1656 kfree(clp
->cl_ownerstr_hashtbl
);
1657 kfree(clp
->cl_name
.data
);
1658 idr_destroy(&clp
->cl_stateids
);
1662 /* must be called under the client_lock */
1664 unhash_client_locked(struct nfs4_client
*clp
)
1666 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1667 struct nfsd4_session
*ses
;
1669 lockdep_assert_held(&nn
->client_lock
);
1671 /* Mark the client as expired! */
1673 /* Make it invisible */
1674 if (!list_empty(&clp
->cl_idhash
)) {
1675 list_del_init(&clp
->cl_idhash
);
1676 if (test_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
))
1677 rb_erase(&clp
->cl_namenode
, &nn
->conf_name_tree
);
1679 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
1681 list_del_init(&clp
->cl_lru
);
1682 spin_lock(&clp
->cl_lock
);
1683 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1684 list_del_init(&ses
->se_hash
);
1685 spin_unlock(&clp
->cl_lock
);
1689 unhash_client(struct nfs4_client
*clp
)
1691 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1693 spin_lock(&nn
->client_lock
);
1694 unhash_client_locked(clp
);
1695 spin_unlock(&nn
->client_lock
);
1698 static __be32
mark_client_expired_locked(struct nfs4_client
*clp
)
1700 if (atomic_read(&clp
->cl_refcount
))
1701 return nfserr_jukebox
;
1702 unhash_client_locked(clp
);
1707 __destroy_client(struct nfs4_client
*clp
)
1709 struct nfs4_openowner
*oo
;
1710 struct nfs4_delegation
*dp
;
1711 struct list_head reaplist
;
1713 INIT_LIST_HEAD(&reaplist
);
1714 spin_lock(&state_lock
);
1715 while (!list_empty(&clp
->cl_delegations
)) {
1716 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1717 unhash_delegation_locked(dp
);
1718 list_add(&dp
->dl_recall_lru
, &reaplist
);
1720 spin_unlock(&state_lock
);
1721 while (!list_empty(&reaplist
)) {
1722 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1723 list_del_init(&dp
->dl_recall_lru
);
1724 put_clnt_odstate(dp
->dl_clnt_odstate
);
1725 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
1726 nfs4_put_stid(&dp
->dl_stid
);
1728 while (!list_empty(&clp
->cl_revoked
)) {
1729 dp
= list_entry(clp
->cl_revoked
.next
, struct nfs4_delegation
, dl_recall_lru
);
1730 list_del_init(&dp
->dl_recall_lru
);
1731 nfs4_put_stid(&dp
->dl_stid
);
1733 while (!list_empty(&clp
->cl_openowners
)) {
1734 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1735 nfs4_get_stateowner(&oo
->oo_owner
);
1736 release_openowner(oo
);
1738 nfsd4_return_all_client_layouts(clp
);
1739 nfsd4_shutdown_callback(clp
);
1740 if (clp
->cl_cb_conn
.cb_xprt
)
1741 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1746 destroy_client(struct nfs4_client
*clp
)
1749 __destroy_client(clp
);
1752 static void expire_client(struct nfs4_client
*clp
)
1755 nfsd4_client_record_remove(clp
);
1756 __destroy_client(clp
);
1759 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1761 memcpy(target
->cl_verifier
.data
, source
->data
,
1762 sizeof(target
->cl_verifier
.data
));
1765 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1767 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1768 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1771 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1773 if (source
->cr_principal
) {
1774 target
->cr_principal
=
1775 kstrdup(source
->cr_principal
, GFP_KERNEL
);
1776 if (target
->cr_principal
== NULL
)
1779 target
->cr_principal
= NULL
;
1780 target
->cr_flavor
= source
->cr_flavor
;
1781 target
->cr_uid
= source
->cr_uid
;
1782 target
->cr_gid
= source
->cr_gid
;
1783 target
->cr_group_info
= source
->cr_group_info
;
1784 get_group_info(target
->cr_group_info
);
1785 target
->cr_gss_mech
= source
->cr_gss_mech
;
1786 if (source
->cr_gss_mech
)
1787 gss_mech_get(source
->cr_gss_mech
);
1792 compare_blob(const struct xdr_netobj
*o1
, const struct xdr_netobj
*o2
)
1794 if (o1
->len
< o2
->len
)
1796 if (o1
->len
> o2
->len
)
1798 return memcmp(o1
->data
, o2
->data
, o1
->len
);
1801 static int same_name(const char *n1
, const char *n2
)
1803 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1807 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1809 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1813 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1815 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1818 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
1822 if (g1
->ngroups
!= g2
->ngroups
)
1824 for (i
=0; i
<g1
->ngroups
; i
++)
1825 if (!gid_eq(GROUP_AT(g1
, i
), GROUP_AT(g2
, i
)))
1831 * RFC 3530 language requires clid_inuse be returned when the
1832 * "principal" associated with a requests differs from that previously
1833 * used. We use uid, gid's, and gss principal string as our best
1834 * approximation. We also don't want to allow non-gss use of a client
1835 * established using gss: in theory cr_principal should catch that
1836 * change, but in practice cr_principal can be null even in the gss case
1837 * since gssd doesn't always pass down a principal string.
1839 static bool is_gss_cred(struct svc_cred
*cr
)
1841 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1842 return (cr
->cr_flavor
> RPC_AUTH_MAXFLAVOR
);
1847 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1849 if ((is_gss_cred(cr1
) != is_gss_cred(cr2
))
1850 || (!uid_eq(cr1
->cr_uid
, cr2
->cr_uid
))
1851 || (!gid_eq(cr1
->cr_gid
, cr2
->cr_gid
))
1852 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
1854 if (cr1
->cr_principal
== cr2
->cr_principal
)
1856 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
1858 return 0 == strcmp(cr1
->cr_principal
, cr2
->cr_principal
);
1861 static bool svc_rqst_integrity_protected(struct svc_rqst
*rqstp
)
1863 struct svc_cred
*cr
= &rqstp
->rq_cred
;
1866 if (!cr
->cr_gss_mech
)
1868 service
= gss_pseudoflavor_to_service(cr
->cr_gss_mech
, cr
->cr_flavor
);
1869 return service
== RPC_GSS_SVC_INTEGRITY
||
1870 service
== RPC_GSS_SVC_PRIVACY
;
1873 static bool mach_creds_match(struct nfs4_client
*cl
, struct svc_rqst
*rqstp
)
1875 struct svc_cred
*cr
= &rqstp
->rq_cred
;
1877 if (!cl
->cl_mach_cred
)
1879 if (cl
->cl_cred
.cr_gss_mech
!= cr
->cr_gss_mech
)
1881 if (!svc_rqst_integrity_protected(rqstp
))
1883 if (!cr
->cr_principal
)
1885 return 0 == strcmp(cl
->cl_cred
.cr_principal
, cr
->cr_principal
);
1888 static void gen_confirm(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
1893 * This is opaque to client, so no need to byte-swap. Use
1894 * __force to keep sparse happy
1896 verf
[0] = (__force __be32
)get_seconds();
1897 verf
[1] = (__force __be32
)nn
->clientid_counter
;
1898 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
1901 static void gen_clid(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
1903 clp
->cl_clientid
.cl_boot
= nn
->boot_time
;
1904 clp
->cl_clientid
.cl_id
= nn
->clientid_counter
++;
1905 gen_confirm(clp
, nn
);
1908 static struct nfs4_stid
*
1909 find_stateid_locked(struct nfs4_client
*cl
, stateid_t
*t
)
1911 struct nfs4_stid
*ret
;
1913 ret
= idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1914 if (!ret
|| !ret
->sc_type
)
1919 static struct nfs4_stid
*
1920 find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1922 struct nfs4_stid
*s
;
1924 spin_lock(&cl
->cl_lock
);
1925 s
= find_stateid_locked(cl
, t
);
1927 if (typemask
& s
->sc_type
)
1928 atomic_inc(&s
->sc_count
);
1932 spin_unlock(&cl
->cl_lock
);
1936 static struct nfs4_client
*create_client(struct xdr_netobj name
,
1937 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1939 struct nfs4_client
*clp
;
1940 struct sockaddr
*sa
= svc_addr(rqstp
);
1942 struct net
*net
= SVC_NET(rqstp
);
1944 clp
= alloc_client(name
);
1948 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1953 nfsd4_init_cb(&clp
->cl_cb_null
, clp
, NULL
, NFSPROC4_CLNT_CB_NULL
);
1954 clp
->cl_time
= get_seconds();
1955 clear_bit(0, &clp
->cl_cb_slot_busy
);
1956 copy_verf(clp
, verf
);
1957 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1958 clp
->cl_cb_session
= NULL
;
1964 add_clp_to_name_tree(struct nfs4_client
*new_clp
, struct rb_root
*root
)
1966 struct rb_node
**new = &(root
->rb_node
), *parent
= NULL
;
1967 struct nfs4_client
*clp
;
1970 clp
= rb_entry(*new, struct nfs4_client
, cl_namenode
);
1973 if (compare_blob(&clp
->cl_name
, &new_clp
->cl_name
) > 0)
1974 new = &((*new)->rb_left
);
1976 new = &((*new)->rb_right
);
1979 rb_link_node(&new_clp
->cl_namenode
, parent
, new);
1980 rb_insert_color(&new_clp
->cl_namenode
, root
);
1983 static struct nfs4_client
*
1984 find_clp_in_name_tree(struct xdr_netobj
*name
, struct rb_root
*root
)
1987 struct rb_node
*node
= root
->rb_node
;
1988 struct nfs4_client
*clp
;
1991 clp
= rb_entry(node
, struct nfs4_client
, cl_namenode
);
1992 cmp
= compare_blob(&clp
->cl_name
, name
);
1994 node
= node
->rb_left
;
1996 node
= node
->rb_right
;
2004 add_to_unconfirmed(struct nfs4_client
*clp
)
2006 unsigned int idhashval
;
2007 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2009 lockdep_assert_held(&nn
->client_lock
);
2011 clear_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2012 add_clp_to_name_tree(clp
, &nn
->unconf_name_tree
);
2013 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2014 list_add(&clp
->cl_idhash
, &nn
->unconf_id_hashtbl
[idhashval
]);
2015 renew_client_locked(clp
);
2019 move_to_confirmed(struct nfs4_client
*clp
)
2021 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2022 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2024 lockdep_assert_held(&nn
->client_lock
);
2026 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
2027 list_move(&clp
->cl_idhash
, &nn
->conf_id_hashtbl
[idhashval
]);
2028 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
2029 add_clp_to_name_tree(clp
, &nn
->conf_name_tree
);
2030 set_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2031 renew_client_locked(clp
);
2034 static struct nfs4_client
*
2035 find_client_in_id_table(struct list_head
*tbl
, clientid_t
*clid
, bool sessions
)
2037 struct nfs4_client
*clp
;
2038 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
2040 list_for_each_entry(clp
, &tbl
[idhashval
], cl_idhash
) {
2041 if (same_clid(&clp
->cl_clientid
, clid
)) {
2042 if ((bool)clp
->cl_minorversion
!= sessions
)
2044 renew_client_locked(clp
);
2051 static struct nfs4_client
*
2052 find_confirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2054 struct list_head
*tbl
= nn
->conf_id_hashtbl
;
2056 lockdep_assert_held(&nn
->client_lock
);
2057 return find_client_in_id_table(tbl
, clid
, sessions
);
2060 static struct nfs4_client
*
2061 find_unconfirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2063 struct list_head
*tbl
= nn
->unconf_id_hashtbl
;
2065 lockdep_assert_held(&nn
->client_lock
);
2066 return find_client_in_id_table(tbl
, clid
, sessions
);
2069 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
2071 return clp
->cl_exchange_flags
!= 0;
2074 static struct nfs4_client
*
2075 find_confirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2077 lockdep_assert_held(&nn
->client_lock
);
2078 return find_clp_in_name_tree(name
, &nn
->conf_name_tree
);
2081 static struct nfs4_client
*
2082 find_unconfirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2084 lockdep_assert_held(&nn
->client_lock
);
2085 return find_clp_in_name_tree(name
, &nn
->unconf_name_tree
);
2089 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
2091 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
2092 struct sockaddr
*sa
= svc_addr(rqstp
);
2093 u32 scopeid
= rpc_get_scope_id(sa
);
2094 unsigned short expected_family
;
2096 /* Currently, we only support tcp and tcp6 for the callback channel */
2097 if (se
->se_callback_netid_len
== 3 &&
2098 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
2099 expected_family
= AF_INET
;
2100 else if (se
->se_callback_netid_len
== 4 &&
2101 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
2102 expected_family
= AF_INET6
;
2106 conn
->cb_addrlen
= rpc_uaddr2sockaddr(clp
->net
, se
->se_callback_addr_val
,
2107 se
->se_callback_addr_len
,
2108 (struct sockaddr
*)&conn
->cb_addr
,
2109 sizeof(conn
->cb_addr
));
2111 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
2114 if (conn
->cb_addr
.ss_family
== AF_INET6
)
2115 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
2117 conn
->cb_prog
= se
->se_callback_prog
;
2118 conn
->cb_ident
= se
->se_callback_ident
;
2119 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
2122 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
2123 conn
->cb_addrlen
= 0;
2124 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
2125 "will not receive delegations\n",
2126 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
2132 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2135 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
2137 struct xdr_buf
*buf
= resp
->xdr
.buf
;
2138 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2141 dprintk("--> %s slot %p\n", __func__
, slot
);
2143 slot
->sl_opcnt
= resp
->opcnt
;
2144 slot
->sl_status
= resp
->cstate
.status
;
2146 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
2147 if (nfsd4_not_cached(resp
)) {
2148 slot
->sl_datalen
= 0;
2151 base
= resp
->cstate
.data_offset
;
2152 slot
->sl_datalen
= buf
->len
- base
;
2153 if (read_bytes_from_xdr_buf(buf
, base
, slot
->sl_data
, slot
->sl_datalen
))
2154 WARN("%s: sessions DRC could not cache compound\n", __func__
);
2159 * Encode the replay sequence operation from the slot values.
2160 * If cachethis is FALSE encode the uncached rep error on the next
2161 * operation which sets resp->p and increments resp->opcnt for
2162 * nfs4svc_encode_compoundres.
2166 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
2167 struct nfsd4_compoundres
*resp
)
2169 struct nfsd4_op
*op
;
2170 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2172 /* Encode the replayed sequence operation */
2173 op
= &args
->ops
[resp
->opcnt
- 1];
2174 nfsd4_encode_operation(resp
, op
);
2176 /* Return nfserr_retry_uncached_rep in next operation. */
2177 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
2178 op
= &args
->ops
[resp
->opcnt
++];
2179 op
->status
= nfserr_retry_uncached_rep
;
2180 nfsd4_encode_operation(resp
, op
);
2186 * The sequence operation is not cached because we can use the slot and
2190 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
2191 struct nfsd4_sequence
*seq
)
2193 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2194 struct xdr_stream
*xdr
= &resp
->xdr
;
2198 dprintk("--> %s slot %p\n", __func__
, slot
);
2200 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
2204 p
= xdr_reserve_space(xdr
, slot
->sl_datalen
);
2207 return nfserr_serverfault
;
2209 xdr_encode_opaque_fixed(p
, slot
->sl_data
, slot
->sl_datalen
);
2210 xdr_commit_encode(xdr
);
2212 resp
->opcnt
= slot
->sl_opcnt
;
2213 return slot
->sl_status
;
2217 * Set the exchange_id flags returned by the server.
2220 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
2222 #ifdef CONFIG_NFSD_PNFS
2223 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_PNFS_MDS
;
2225 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
2228 /* Referrals are supported, Migration is not. */
2229 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
2231 /* set the wire flags to return to client. */
2232 clid
->flags
= new->cl_exchange_flags
;
2235 static bool client_has_state(struct nfs4_client
*clp
)
2238 * Note clp->cl_openowners check isn't quite right: there's no
2239 * need to count owners without stateid's.
2241 * Also note we should probably be using this in 4.0 case too.
2243 return !list_empty(&clp
->cl_openowners
)
2244 || !list_empty(&clp
->cl_delegations
)
2245 || !list_empty(&clp
->cl_sessions
);
2249 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
2250 struct nfsd4_compound_state
*cstate
,
2251 struct nfsd4_exchange_id
*exid
)
2253 struct nfs4_client
*conf
, *new;
2254 struct nfs4_client
*unconf
= NULL
;
2256 char addr_str
[INET6_ADDRSTRLEN
];
2257 nfs4_verifier verf
= exid
->verifier
;
2258 struct sockaddr
*sa
= svc_addr(rqstp
);
2259 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
2260 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2262 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
2263 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2264 "ip_addr=%s flags %x, spa_how %d\n",
2265 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
2266 addr_str
, exid
->flags
, exid
->spa_how
);
2268 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
2269 return nfserr_inval
;
2271 switch (exid
->spa_how
) {
2273 if (!svc_rqst_integrity_protected(rqstp
))
2274 return nfserr_inval
;
2277 default: /* checked by xdr code */
2280 return nfserr_encr_alg_unsupp
;
2283 new = create_client(exid
->clname
, rqstp
, &verf
);
2285 return nfserr_jukebox
;
2287 /* Cases below refer to rfc 5661 section 18.35.4: */
2288 spin_lock(&nn
->client_lock
);
2289 conf
= find_confirmed_client_by_name(&exid
->clname
, nn
);
2291 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
2292 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
2295 if (!clp_used_exchangeid(conf
)) { /* buggy client */
2296 status
= nfserr_inval
;
2299 if (!mach_creds_match(conf
, rqstp
)) {
2300 status
= nfserr_wrong_cred
;
2303 if (!creds_match
) { /* case 9 */
2304 status
= nfserr_perm
;
2307 if (!verfs_match
) { /* case 8 */
2308 status
= nfserr_not_same
;
2312 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2315 if (!creds_match
) { /* case 3 */
2316 if (client_has_state(conf
)) {
2317 status
= nfserr_clid_inuse
;
2322 if (verfs_match
) { /* case 2 */
2323 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2326 /* case 5, client reboot */
2331 if (update
) { /* case 7 */
2332 status
= nfserr_noent
;
2336 unconf
= find_unconfirmed_client_by_name(&exid
->clname
, nn
);
2337 if (unconf
) /* case 4, possible retry or client restart */
2338 unhash_client_locked(unconf
);
2340 /* case 1 (normal case) */
2343 status
= mark_client_expired_locked(conf
);
2347 new->cl_minorversion
= cstate
->minorversion
;
2348 new->cl_mach_cred
= (exid
->spa_how
== SP4_MACH_CRED
);
2351 add_to_unconfirmed(new);
2354 exid
->clientid
.cl_boot
= conf
->cl_clientid
.cl_boot
;
2355 exid
->clientid
.cl_id
= conf
->cl_clientid
.cl_id
;
2357 exid
->seqid
= conf
->cl_cs_slot
.sl_seqid
+ 1;
2358 nfsd4_set_ex_flags(conf
, exid
);
2360 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2361 conf
->cl_cs_slot
.sl_seqid
, conf
->cl_exchange_flags
);
2365 spin_unlock(&nn
->client_lock
);
2369 expire_client(unconf
);
2374 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
2376 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
2379 /* The slot is in use, and no response has been sent. */
2381 if (seqid
== slot_seqid
)
2382 return nfserr_jukebox
;
2384 return nfserr_seq_misordered
;
2386 /* Note unsigned 32-bit arithmetic handles wraparound: */
2387 if (likely(seqid
== slot_seqid
+ 1))
2389 if (seqid
== slot_seqid
)
2390 return nfserr_replay_cache
;
2391 return nfserr_seq_misordered
;
2395 * Cache the create session result into the create session single DRC
2396 * slot cache by saving the xdr structure. sl_seqid has been set.
2397 * Do this for solo or embedded create session operations.
2400 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
2401 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
2403 slot
->sl_status
= nfserr
;
2404 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
2408 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
2409 struct nfsd4_clid_slot
*slot
)
2411 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
2412 return slot
->sl_status
;
2415 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2416 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2417 1 + /* MIN tag is length with zero, only length */ \
2418 3 + /* version, opcount, opcode */ \
2419 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2420 /* seqid, slotID, slotID, cache */ \
2421 4 ) * sizeof(__be32))
2423 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2424 2 + /* verifier: AUTH_NULL, length 0 */\
2426 1 + /* MIN tag is length with zero, only length */ \
2427 3 + /* opcount, opcode, opstatus*/ \
2428 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2429 /* seqid, slotID, slotID, slotID, status */ \
2430 5 ) * sizeof(__be32))
2432 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs
*ca
, struct nfsd_net
*nn
)
2434 u32 maxrpc
= nn
->nfsd_serv
->sv_max_mesg
;
2436 if (ca
->maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
)
2437 return nfserr_toosmall
;
2438 if (ca
->maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
)
2439 return nfserr_toosmall
;
2440 ca
->headerpadsz
= 0;
2441 ca
->maxreq_sz
= min_t(u32
, ca
->maxreq_sz
, maxrpc
);
2442 ca
->maxresp_sz
= min_t(u32
, ca
->maxresp_sz
, maxrpc
);
2443 ca
->maxops
= min_t(u32
, ca
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
2444 ca
->maxresp_cached
= min_t(u32
, ca
->maxresp_cached
,
2445 NFSD_SLOT_CACHE_SIZE
+ NFSD_MIN_HDR_SEQ_SZ
);
2446 ca
->maxreqs
= min_t(u32
, ca
->maxreqs
, NFSD_MAX_SLOTS_PER_SESSION
);
2448 * Note decreasing slot size below client's request may make it
2449 * difficult for client to function correctly, whereas
2450 * decreasing the number of slots will (just?) affect
2451 * performance. When short on memory we therefore prefer to
2452 * decrease number of slots instead of their size. Clients that
2453 * request larger slots than they need will get poor results:
2455 ca
->maxreqs
= nfsd4_get_drc_mem(ca
);
2457 return nfserr_jukebox
;
2462 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2463 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2464 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2465 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2467 static __be32
check_backchannel_attrs(struct nfsd4_channel_attrs
*ca
)
2469 ca
->headerpadsz
= 0;
2472 * These RPC_MAX_HEADER macros are overkill, especially since we
2473 * don't even do gss on the backchannel yet. But this is still
2474 * less than 1k. Tighten up this estimate in the unlikely event
2475 * it turns out to be a problem for some client:
2477 if (ca
->maxreq_sz
< NFSD_CB_MAX_REQ_SZ
)
2478 return nfserr_toosmall
;
2479 if (ca
->maxresp_sz
< NFSD_CB_MAX_RESP_SZ
)
2480 return nfserr_toosmall
;
2481 ca
->maxresp_cached
= 0;
2483 return nfserr_toosmall
;
2488 static __be32
nfsd4_check_cb_sec(struct nfsd4_cb_sec
*cbs
)
2490 switch (cbs
->flavor
) {
2496 * GSS case: the spec doesn't allow us to return this
2497 * error. But it also doesn't allow us not to support
2499 * I'd rather this fail hard than return some error the
2500 * client might think it can already handle:
2502 return nfserr_encr_alg_unsupp
;
2507 nfsd4_create_session(struct svc_rqst
*rqstp
,
2508 struct nfsd4_compound_state
*cstate
,
2509 struct nfsd4_create_session
*cr_ses
)
2511 struct sockaddr
*sa
= svc_addr(rqstp
);
2512 struct nfs4_client
*conf
, *unconf
;
2513 struct nfs4_client
*old
= NULL
;
2514 struct nfsd4_session
*new;
2515 struct nfsd4_conn
*conn
;
2516 struct nfsd4_clid_slot
*cs_slot
= NULL
;
2518 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2520 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
2521 return nfserr_inval
;
2522 status
= nfsd4_check_cb_sec(&cr_ses
->cb_sec
);
2525 status
= check_forechannel_attrs(&cr_ses
->fore_channel
, nn
);
2528 status
= check_backchannel_attrs(&cr_ses
->back_channel
);
2530 goto out_release_drc_mem
;
2531 status
= nfserr_jukebox
;
2532 new = alloc_session(&cr_ses
->fore_channel
, &cr_ses
->back_channel
);
2534 goto out_release_drc_mem
;
2535 conn
= alloc_conn_from_crses(rqstp
, cr_ses
);
2537 goto out_free_session
;
2539 spin_lock(&nn
->client_lock
);
2540 unconf
= find_unconfirmed_client(&cr_ses
->clientid
, true, nn
);
2541 conf
= find_confirmed_client(&cr_ses
->clientid
, true, nn
);
2542 WARN_ON_ONCE(conf
&& unconf
);
2545 status
= nfserr_wrong_cred
;
2546 if (!mach_creds_match(conf
, rqstp
))
2548 cs_slot
= &conf
->cl_cs_slot
;
2549 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2550 if (status
== nfserr_replay_cache
) {
2551 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
2553 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
2554 status
= nfserr_seq_misordered
;
2557 } else if (unconf
) {
2558 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
2559 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
2560 status
= nfserr_clid_inuse
;
2563 status
= nfserr_wrong_cred
;
2564 if (!mach_creds_match(unconf
, rqstp
))
2566 cs_slot
= &unconf
->cl_cs_slot
;
2567 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2569 /* an unconfirmed replay returns misordered */
2570 status
= nfserr_seq_misordered
;
2573 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
2575 status
= mark_client_expired_locked(old
);
2581 move_to_confirmed(unconf
);
2584 status
= nfserr_stale_clientid
;
2589 * We do not support RDMA or persistent sessions
2591 cr_ses
->flags
&= ~SESSION4_PERSIST
;
2592 cr_ses
->flags
&= ~SESSION4_RDMA
;
2594 init_session(rqstp
, new, conf
, cr_ses
);
2595 nfsd4_get_session_locked(new);
2597 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
2598 NFS4_MAX_SESSIONID_LEN
);
2599 cs_slot
->sl_seqid
++;
2600 cr_ses
->seqid
= cs_slot
->sl_seqid
;
2602 /* cache solo and embedded create sessions under the client_lock */
2603 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
2604 spin_unlock(&nn
->client_lock
);
2605 /* init connection and backchannel */
2606 nfsd4_init_conn(rqstp
, conn
, new);
2607 nfsd4_put_session(new);
2612 spin_unlock(&nn
->client_lock
);
2617 __free_session(new);
2618 out_release_drc_mem
:
2619 nfsd4_put_drc_mem(&cr_ses
->fore_channel
);
2623 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
2626 case NFS4_CDFC4_FORE
:
2627 case NFS4_CDFC4_BACK
:
2629 case NFS4_CDFC4_FORE_OR_BOTH
:
2630 case NFS4_CDFC4_BACK_OR_BOTH
:
2631 *dir
= NFS4_CDFC4_BOTH
;
2634 return nfserr_inval
;
2637 __be32
nfsd4_backchannel_ctl(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_backchannel_ctl
*bc
)
2639 struct nfsd4_session
*session
= cstate
->session
;
2640 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2643 status
= nfsd4_check_cb_sec(&bc
->bc_cb_sec
);
2646 spin_lock(&nn
->client_lock
);
2647 session
->se_cb_prog
= bc
->bc_cb_program
;
2648 session
->se_cb_sec
= bc
->bc_cb_sec
;
2649 spin_unlock(&nn
->client_lock
);
2651 nfsd4_probe_callback(session
->se_client
);
2656 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
2657 struct nfsd4_compound_state
*cstate
,
2658 struct nfsd4_bind_conn_to_session
*bcts
)
2661 struct nfsd4_conn
*conn
;
2662 struct nfsd4_session
*session
;
2663 struct net
*net
= SVC_NET(rqstp
);
2664 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2666 if (!nfsd4_last_compound_op(rqstp
))
2667 return nfserr_not_only_op
;
2668 spin_lock(&nn
->client_lock
);
2669 session
= find_in_sessionid_hashtbl(&bcts
->sessionid
, net
, &status
);
2670 spin_unlock(&nn
->client_lock
);
2672 goto out_no_session
;
2673 status
= nfserr_wrong_cred
;
2674 if (!mach_creds_match(session
->se_client
, rqstp
))
2676 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
2679 conn
= alloc_conn(rqstp
, bcts
->dir
);
2680 status
= nfserr_jukebox
;
2683 nfsd4_init_conn(rqstp
, conn
, session
);
2686 nfsd4_put_session(session
);
2691 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
2695 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
2699 nfsd4_destroy_session(struct svc_rqst
*r
,
2700 struct nfsd4_compound_state
*cstate
,
2701 struct nfsd4_destroy_session
*sessionid
)
2703 struct nfsd4_session
*ses
;
2705 int ref_held_by_me
= 0;
2706 struct net
*net
= SVC_NET(r
);
2707 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2709 status
= nfserr_not_only_op
;
2710 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
2711 if (!nfsd4_last_compound_op(r
))
2715 dump_sessionid(__func__
, &sessionid
->sessionid
);
2716 spin_lock(&nn
->client_lock
);
2717 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
, net
, &status
);
2719 goto out_client_lock
;
2720 status
= nfserr_wrong_cred
;
2721 if (!mach_creds_match(ses
->se_client
, r
))
2722 goto out_put_session
;
2723 status
= mark_session_dead_locked(ses
, 1 + ref_held_by_me
);
2725 goto out_put_session
;
2726 unhash_session(ses
);
2727 spin_unlock(&nn
->client_lock
);
2729 nfsd4_probe_callback_sync(ses
->se_client
);
2731 spin_lock(&nn
->client_lock
);
2734 nfsd4_put_session_locked(ses
);
2736 spin_unlock(&nn
->client_lock
);
2741 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
2743 struct nfsd4_conn
*c
;
2745 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
2746 if (c
->cn_xprt
== xpt
) {
2753 static __be32
nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
2755 struct nfs4_client
*clp
= ses
->se_client
;
2756 struct nfsd4_conn
*c
;
2757 __be32 status
= nfs_ok
;
2760 spin_lock(&clp
->cl_lock
);
2761 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
2764 status
= nfserr_conn_not_bound_to_session
;
2765 if (clp
->cl_mach_cred
)
2767 __nfsd4_hash_conn(new, ses
);
2768 spin_unlock(&clp
->cl_lock
);
2769 ret
= nfsd4_register_conn(new);
2771 /* oops; xprt is already down: */
2772 nfsd4_conn_lost(&new->cn_xpt_user
);
2775 spin_unlock(&clp
->cl_lock
);
2780 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
2782 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
2784 return args
->opcnt
> session
->se_fchannel
.maxops
;
2787 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
2788 struct nfsd4_session
*session
)
2790 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
2792 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
2796 nfsd4_sequence(struct svc_rqst
*rqstp
,
2797 struct nfsd4_compound_state
*cstate
,
2798 struct nfsd4_sequence
*seq
)
2800 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2801 struct xdr_stream
*xdr
= &resp
->xdr
;
2802 struct nfsd4_session
*session
;
2803 struct nfs4_client
*clp
;
2804 struct nfsd4_slot
*slot
;
2805 struct nfsd4_conn
*conn
;
2808 struct net
*net
= SVC_NET(rqstp
);
2809 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2811 if (resp
->opcnt
!= 1)
2812 return nfserr_sequence_pos
;
2815 * Will be either used or freed by nfsd4_sequence_check_conn
2818 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
2820 return nfserr_jukebox
;
2822 spin_lock(&nn
->client_lock
);
2823 session
= find_in_sessionid_hashtbl(&seq
->sessionid
, net
, &status
);
2825 goto out_no_session
;
2826 clp
= session
->se_client
;
2828 status
= nfserr_too_many_ops
;
2829 if (nfsd4_session_too_many_ops(rqstp
, session
))
2830 goto out_put_session
;
2832 status
= nfserr_req_too_big
;
2833 if (nfsd4_request_too_big(rqstp
, session
))
2834 goto out_put_session
;
2836 status
= nfserr_badslot
;
2837 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
2838 goto out_put_session
;
2840 slot
= session
->se_slots
[seq
->slotid
];
2841 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
2843 /* We do not negotiate the number of slots yet, so set the
2844 * maxslots to the session maxreqs which is used to encode
2845 * sr_highest_slotid and the sr_target_slot id to maxslots */
2846 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
2848 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
2849 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
2850 if (status
== nfserr_replay_cache
) {
2851 status
= nfserr_seq_misordered
;
2852 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
2853 goto out_put_session
;
2854 cstate
->slot
= slot
;
2855 cstate
->session
= session
;
2857 /* Return the cached reply status and set cstate->status
2858 * for nfsd4_proc_compound processing */
2859 status
= nfsd4_replay_cache_entry(resp
, seq
);
2860 cstate
->status
= nfserr_replay_cache
;
2864 goto out_put_session
;
2866 status
= nfsd4_sequence_check_conn(conn
, session
);
2869 goto out_put_session
;
2871 buflen
= (seq
->cachethis
) ?
2872 session
->se_fchannel
.maxresp_cached
:
2873 session
->se_fchannel
.maxresp_sz
;
2874 status
= (seq
->cachethis
) ? nfserr_rep_too_big_to_cache
:
2876 if (xdr_restrict_buflen(xdr
, buflen
- rqstp
->rq_auth_slack
))
2877 goto out_put_session
;
2878 svc_reserve(rqstp
, buflen
);
2881 /* Success! bump slot seqid */
2882 slot
->sl_seqid
= seq
->seqid
;
2883 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
2885 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
2887 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
2889 cstate
->slot
= slot
;
2890 cstate
->session
= session
;
2894 switch (clp
->cl_cb_state
) {
2896 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
2898 case NFSD4_CB_FAULT
:
2899 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
2902 seq
->status_flags
= 0;
2904 if (!list_empty(&clp
->cl_revoked
))
2905 seq
->status_flags
|= SEQ4_STATUS_RECALLABLE_STATE_REVOKED
;
2909 spin_unlock(&nn
->client_lock
);
2912 nfsd4_put_session_locked(session
);
2913 goto out_no_session
;
2917 nfsd4_sequence_done(struct nfsd4_compoundres
*resp
)
2919 struct nfsd4_compound_state
*cs
= &resp
->cstate
;
2921 if (nfsd4_has_session(cs
)) {
2922 if (cs
->status
!= nfserr_replay_cache
) {
2923 nfsd4_store_cache_entry(resp
);
2924 cs
->slot
->sl_flags
&= ~NFSD4_SLOT_INUSE
;
2926 /* Drop session reference that was taken in nfsd4_sequence() */
2927 nfsd4_put_session(cs
->session
);
2929 put_client_renew(cs
->clp
);
2933 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
2935 struct nfs4_client
*conf
, *unconf
;
2936 struct nfs4_client
*clp
= NULL
;
2938 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2940 spin_lock(&nn
->client_lock
);
2941 unconf
= find_unconfirmed_client(&dc
->clientid
, true, nn
);
2942 conf
= find_confirmed_client(&dc
->clientid
, true, nn
);
2943 WARN_ON_ONCE(conf
&& unconf
);
2946 if (client_has_state(conf
)) {
2947 status
= nfserr_clientid_busy
;
2950 status
= mark_client_expired_locked(conf
);
2957 status
= nfserr_stale_clientid
;
2960 if (!mach_creds_match(clp
, rqstp
)) {
2962 status
= nfserr_wrong_cred
;
2965 unhash_client_locked(clp
);
2967 spin_unlock(&nn
->client_lock
);
2974 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2978 if (rc
->rca_one_fs
) {
2979 if (!cstate
->current_fh
.fh_dentry
)
2980 return nfserr_nofilehandle
;
2982 * We don't take advantage of the rca_one_fs case.
2983 * That's OK, it's optional, we can safely ignore it.
2988 status
= nfserr_complete_already
;
2989 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
2990 &cstate
->session
->se_client
->cl_flags
))
2993 status
= nfserr_stale_clientid
;
2994 if (is_client_expired(cstate
->session
->se_client
))
2996 * The following error isn't really legal.
2997 * But we only get here if the client just explicitly
2998 * destroyed the client. Surely it no longer cares what
2999 * error it gets back on an operation for the dead
3005 nfsd4_client_record_create(cstate
->session
->se_client
);
3011 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3012 struct nfsd4_setclientid
*setclid
)
3014 struct xdr_netobj clname
= setclid
->se_name
;
3015 nfs4_verifier clverifier
= setclid
->se_verf
;
3016 struct nfs4_client
*conf
, *new;
3017 struct nfs4_client
*unconf
= NULL
;
3019 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3021 new = create_client(clname
, rqstp
, &clverifier
);
3023 return nfserr_jukebox
;
3024 /* Cases below refer to rfc 3530 section 14.2.33: */
3025 spin_lock(&nn
->client_lock
);
3026 conf
= find_confirmed_client_by_name(&clname
, nn
);
3029 status
= nfserr_clid_inuse
;
3030 if (clp_used_exchangeid(conf
))
3032 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
3033 char addr_str
[INET6_ADDRSTRLEN
];
3034 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
3036 dprintk("NFSD: setclientid: string in use by client "
3037 "at %s\n", addr_str
);
3041 unconf
= find_unconfirmed_client_by_name(&clname
, nn
);
3043 unhash_client_locked(unconf
);
3044 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
))
3045 /* case 1: probable callback update */
3046 copy_clid(new, conf
);
3047 else /* case 4 (new client) or cases 2, 3 (client reboot): */
3049 new->cl_minorversion
= 0;
3050 gen_callback(new, setclid
, rqstp
);
3051 add_to_unconfirmed(new);
3052 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
3053 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
3054 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
3058 spin_unlock(&nn
->client_lock
);
3062 expire_client(unconf
);
3068 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
3069 struct nfsd4_compound_state
*cstate
,
3070 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
3072 struct nfs4_client
*conf
, *unconf
;
3073 struct nfs4_client
*old
= NULL
;
3074 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
3075 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
3077 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3079 if (STALE_CLIENTID(clid
, nn
))
3080 return nfserr_stale_clientid
;
3082 spin_lock(&nn
->client_lock
);
3083 conf
= find_confirmed_client(clid
, false, nn
);
3084 unconf
= find_unconfirmed_client(clid
, false, nn
);
3086 * We try hard to give out unique clientid's, so if we get an
3087 * attempt to confirm the same clientid with a different cred,
3088 * there's a bug somewhere. Let's charitably assume it's our
3091 status
= nfserr_serverfault
;
3092 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
3094 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
3096 /* cases below refer to rfc 3530 section 14.2.34: */
3097 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
3098 if (conf
&& !unconf
) /* case 2: probable retransmit */
3100 else /* case 4: client hasn't noticed we rebooted yet? */
3101 status
= nfserr_stale_clientid
;
3105 if (conf
) { /* case 1: callback update */
3107 unhash_client_locked(old
);
3108 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
3109 } else { /* case 3: normal case; new or rebooted client */
3110 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
3112 status
= mark_client_expired_locked(old
);
3118 move_to_confirmed(unconf
);
3121 get_client_locked(conf
);
3122 spin_unlock(&nn
->client_lock
);
3123 nfsd4_probe_callback(conf
);
3124 spin_lock(&nn
->client_lock
);
3125 put_client_renew_locked(conf
);
3127 spin_unlock(&nn
->client_lock
);
3133 static struct nfs4_file
*nfsd4_alloc_file(void)
3135 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
3138 /* OPEN Share state helper functions */
3139 static void nfsd4_init_file(struct knfsd_fh
*fh
, unsigned int hashval
,
3140 struct nfs4_file
*fp
)
3142 lockdep_assert_held(&state_lock
);
3144 atomic_set(&fp
->fi_ref
, 1);
3145 spin_lock_init(&fp
->fi_lock
);
3146 INIT_LIST_HEAD(&fp
->fi_stateids
);
3147 INIT_LIST_HEAD(&fp
->fi_delegations
);
3148 INIT_LIST_HEAD(&fp
->fi_clnt_odstate
);
3149 fh_copy_shallow(&fp
->fi_fhandle
, fh
);
3150 fp
->fi_deleg_file
= NULL
;
3151 fp
->fi_had_conflict
= false;
3152 fp
->fi_share_deny
= 0;
3153 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
3154 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
3155 #ifdef CONFIG_NFSD_PNFS
3156 INIT_LIST_HEAD(&fp
->fi_lo_states
);
3157 atomic_set(&fp
->fi_lo_recalls
, 0);
3159 hlist_add_head_rcu(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
3163 nfsd4_free_slabs(void)
3165 kmem_cache_destroy(odstate_slab
);
3166 kmem_cache_destroy(openowner_slab
);
3167 kmem_cache_destroy(lockowner_slab
);
3168 kmem_cache_destroy(file_slab
);
3169 kmem_cache_destroy(stateid_slab
);
3170 kmem_cache_destroy(deleg_slab
);
3174 nfsd4_init_slabs(void)
3176 openowner_slab
= kmem_cache_create("nfsd4_openowners",
3177 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
3178 if (openowner_slab
== NULL
)
3180 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
3181 sizeof(struct nfs4_lockowner
), 0, 0, NULL
);
3182 if (lockowner_slab
== NULL
)
3183 goto out_free_openowner_slab
;
3184 file_slab
= kmem_cache_create("nfsd4_files",
3185 sizeof(struct nfs4_file
), 0, 0, NULL
);
3186 if (file_slab
== NULL
)
3187 goto out_free_lockowner_slab
;
3188 stateid_slab
= kmem_cache_create("nfsd4_stateids",
3189 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
3190 if (stateid_slab
== NULL
)
3191 goto out_free_file_slab
;
3192 deleg_slab
= kmem_cache_create("nfsd4_delegations",
3193 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
3194 if (deleg_slab
== NULL
)
3195 goto out_free_stateid_slab
;
3196 odstate_slab
= kmem_cache_create("nfsd4_odstate",
3197 sizeof(struct nfs4_clnt_odstate
), 0, 0, NULL
);
3198 if (odstate_slab
== NULL
)
3199 goto out_free_deleg_slab
;
3202 out_free_deleg_slab
:
3203 kmem_cache_destroy(deleg_slab
);
3204 out_free_stateid_slab
:
3205 kmem_cache_destroy(stateid_slab
);
3207 kmem_cache_destroy(file_slab
);
3208 out_free_lockowner_slab
:
3209 kmem_cache_destroy(lockowner_slab
);
3210 out_free_openowner_slab
:
3211 kmem_cache_destroy(openowner_slab
);
3213 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3217 static void init_nfs4_replay(struct nfs4_replay
*rp
)
3219 rp
->rp_status
= nfserr_serverfault
;
3221 rp
->rp_buf
= rp
->rp_ibuf
;
3222 mutex_init(&rp
->rp_mutex
);
3225 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state
*cstate
,
3226 struct nfs4_stateowner
*so
)
3228 if (!nfsd4_has_session(cstate
)) {
3229 mutex_lock(&so
->so_replay
.rp_mutex
);
3230 cstate
->replay_owner
= nfs4_get_stateowner(so
);
3234 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state
*cstate
)
3236 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
3239 cstate
->replay_owner
= NULL
;
3240 mutex_unlock(&so
->so_replay
.rp_mutex
);
3241 nfs4_put_stateowner(so
);
3245 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
3247 struct nfs4_stateowner
*sop
;
3249 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
3253 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
3254 if (!sop
->so_owner
.data
) {
3255 kmem_cache_free(slab
, sop
);
3258 sop
->so_owner
.len
= owner
->len
;
3260 INIT_LIST_HEAD(&sop
->so_stateids
);
3261 sop
->so_client
= clp
;
3262 init_nfs4_replay(&sop
->so_replay
);
3263 atomic_set(&sop
->so_count
, 1);
3267 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
3269 lockdep_assert_held(&clp
->cl_lock
);
3271 list_add(&oo
->oo_owner
.so_strhash
,
3272 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
3273 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
3276 static void nfs4_unhash_openowner(struct nfs4_stateowner
*so
)
3278 unhash_openowner_locked(openowner(so
));
3281 static void nfs4_free_openowner(struct nfs4_stateowner
*so
)
3283 struct nfs4_openowner
*oo
= openowner(so
);
3285 kmem_cache_free(openowner_slab
, oo
);
3288 static const struct nfs4_stateowner_operations openowner_ops
= {
3289 .so_unhash
= nfs4_unhash_openowner
,
3290 .so_free
= nfs4_free_openowner
,
3293 static struct nfs4_openowner
*
3294 alloc_init_open_stateowner(unsigned int strhashval
, struct nfsd4_open
*open
,
3295 struct nfsd4_compound_state
*cstate
)
3297 struct nfs4_client
*clp
= cstate
->clp
;
3298 struct nfs4_openowner
*oo
, *ret
;
3300 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
3303 oo
->oo_owner
.so_ops
= &openowner_ops
;
3304 oo
->oo_owner
.so_is_open_owner
= 1;
3305 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
3307 if (nfsd4_has_session(cstate
))
3308 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3310 oo
->oo_last_closed_stid
= NULL
;
3311 INIT_LIST_HEAD(&oo
->oo_close_lru
);
3312 spin_lock(&clp
->cl_lock
);
3313 ret
= find_openstateowner_str_locked(strhashval
, open
, clp
);
3315 hash_openowner(oo
, clp
, strhashval
);
3318 nfs4_free_openowner(&oo
->oo_owner
);
3319 spin_unlock(&clp
->cl_lock
);
3323 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
3324 struct nfs4_openowner
*oo
= open
->op_openowner
;
3326 atomic_inc(&stp
->st_stid
.sc_count
);
3327 stp
->st_stid
.sc_type
= NFS4_OPEN_STID
;
3328 INIT_LIST_HEAD(&stp
->st_locks
);
3329 stp
->st_stateowner
= nfs4_get_stateowner(&oo
->oo_owner
);
3331 stp
->st_stid
.sc_file
= fp
;
3332 stp
->st_access_bmap
= 0;
3333 stp
->st_deny_bmap
= 0;
3334 stp
->st_openstp
= NULL
;
3335 spin_lock(&oo
->oo_owner
.so_client
->cl_lock
);
3336 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
3337 spin_lock(&fp
->fi_lock
);
3338 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3339 spin_unlock(&fp
->fi_lock
);
3340 spin_unlock(&oo
->oo_owner
.so_client
->cl_lock
);
3344 * In the 4.0 case we need to keep the owners around a little while to handle
3345 * CLOSE replay. We still do need to release any file access that is held by
3346 * them before returning however.
3349 move_to_close_lru(struct nfs4_ol_stateid
*s
, struct net
*net
)
3351 struct nfs4_ol_stateid
*last
;
3352 struct nfs4_openowner
*oo
= openowner(s
->st_stateowner
);
3353 struct nfsd_net
*nn
= net_generic(s
->st_stid
.sc_client
->net
,
3356 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
3359 * We know that we hold one reference via nfsd4_close, and another
3360 * "persistent" reference for the client. If the refcount is higher
3361 * than 2, then there are still calls in progress that are using this
3362 * stateid. We can't put the sc_file reference until they are finished.
3363 * Wait for the refcount to drop to 2. Since it has been unhashed,
3364 * there should be no danger of the refcount going back up again at
3367 wait_event(close_wq
, atomic_read(&s
->st_stid
.sc_count
) == 2);
3369 release_all_access(s
);
3370 if (s
->st_stid
.sc_file
) {
3371 put_nfs4_file(s
->st_stid
.sc_file
);
3372 s
->st_stid
.sc_file
= NULL
;
3375 spin_lock(&nn
->client_lock
);
3376 last
= oo
->oo_last_closed_stid
;
3377 oo
->oo_last_closed_stid
= s
;
3378 list_move_tail(&oo
->oo_close_lru
, &nn
->close_lru
);
3379 oo
->oo_time
= get_seconds();
3380 spin_unlock(&nn
->client_lock
);
3382 nfs4_put_stid(&last
->st_stid
);
3385 /* search file_hashtbl[] for file */
3386 static struct nfs4_file
*
3387 find_file_locked(struct knfsd_fh
*fh
, unsigned int hashval
)
3389 struct nfs4_file
*fp
;
3391 hlist_for_each_entry_rcu(fp
, &file_hashtbl
[hashval
], fi_hash
) {
3392 if (fh_match(&fp
->fi_fhandle
, fh
)) {
3393 if (atomic_inc_not_zero(&fp
->fi_ref
))
3401 find_file(struct knfsd_fh
*fh
)
3403 struct nfs4_file
*fp
;
3404 unsigned int hashval
= file_hashval(fh
);
3407 fp
= find_file_locked(fh
, hashval
);
3412 static struct nfs4_file
*
3413 find_or_add_file(struct nfs4_file
*new, struct knfsd_fh
*fh
)
3415 struct nfs4_file
*fp
;
3416 unsigned int hashval
= file_hashval(fh
);
3419 fp
= find_file_locked(fh
, hashval
);
3424 spin_lock(&state_lock
);
3425 fp
= find_file_locked(fh
, hashval
);
3426 if (likely(fp
== NULL
)) {
3427 nfsd4_init_file(fh
, hashval
, new);
3430 spin_unlock(&state_lock
);
3436 * Called to check deny when READ with all zero stateid or
3437 * WRITE with all zero or all one stateid
3440 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
3442 struct nfs4_file
*fp
;
3443 __be32 ret
= nfs_ok
;
3445 fp
= find_file(¤t_fh
->fh_handle
);
3448 /* Check for conflicting share reservations */
3449 spin_lock(&fp
->fi_lock
);
3450 if (fp
->fi_share_deny
& deny_type
)
3451 ret
= nfserr_locked
;
3452 spin_unlock(&fp
->fi_lock
);
3457 static void nfsd4_cb_recall_prepare(struct nfsd4_callback
*cb
)
3459 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3460 struct nfsd_net
*nn
= net_generic(dp
->dl_stid
.sc_client
->net
,
3463 block_delegations(&dp
->dl_stid
.sc_file
->fi_fhandle
);
3466 * We can't do this in nfsd_break_deleg_cb because it is
3467 * already holding inode->i_lock.
3469 * If the dl_time != 0, then we know that it has already been
3470 * queued for a lease break. Don't queue it again.
3472 spin_lock(&state_lock
);
3473 if (dp
->dl_time
== 0) {
3474 dp
->dl_time
= get_seconds();
3475 list_add_tail(&dp
->dl_recall_lru
, &nn
->del_recall_lru
);
3477 spin_unlock(&state_lock
);
3480 static int nfsd4_cb_recall_done(struct nfsd4_callback
*cb
,
3481 struct rpc_task
*task
)
3483 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3485 switch (task
->tk_status
) {
3489 case -NFS4ERR_BAD_STATEID
:
3491 * Race: client probably got cb_recall before open reply
3492 * granting delegation.
3494 if (dp
->dl_retries
--) {
3495 rpc_delay(task
, 2 * HZ
);
3504 static void nfsd4_cb_recall_release(struct nfsd4_callback
*cb
)
3506 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3508 nfs4_put_stid(&dp
->dl_stid
);
3511 static struct nfsd4_callback_ops nfsd4_cb_recall_ops
= {
3512 .prepare
= nfsd4_cb_recall_prepare
,
3513 .done
= nfsd4_cb_recall_done
,
3514 .release
= nfsd4_cb_recall_release
,
3517 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
3520 * We're assuming the state code never drops its reference
3521 * without first removing the lease. Since we're in this lease
3522 * callback (and since the lease code is serialized by the kernel
3523 * lock) we know the server hasn't removed the lease yet, we know
3524 * it's safe to take a reference.
3526 atomic_inc(&dp
->dl_stid
.sc_count
);
3527 nfsd4_run_cb(&dp
->dl_recall
);
3530 /* Called from break_lease() with i_lock held. */
3532 nfsd_break_deleg_cb(struct file_lock
*fl
)
3535 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
3536 struct nfs4_delegation
*dp
;
3539 WARN(1, "(%p)->fl_owner NULL\n", fl
);
3542 if (fp
->fi_had_conflict
) {
3543 WARN(1, "duplicate break on %p\n", fp
);
3547 * We don't want the locks code to timeout the lease for us;
3548 * we'll remove it ourself if a delegation isn't returned
3551 fl
->fl_break_time
= 0;
3553 spin_lock(&fp
->fi_lock
);
3554 fp
->fi_had_conflict
= true;
3556 * If there are no delegations on the list, then return true
3557 * so that the lease code will go ahead and delete it.
3559 if (list_empty(&fp
->fi_delegations
))
3562 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
3563 nfsd_break_one_deleg(dp
);
3564 spin_unlock(&fp
->fi_lock
);
3569 nfsd_change_deleg_cb(struct file_lock
*onlist
, int arg
,
3570 struct list_head
*dispose
)
3573 return lease_modify(onlist
, arg
, dispose
);
3578 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
3579 .lm_break
= nfsd_break_deleg_cb
,
3580 .lm_change
= nfsd_change_deleg_cb
,
3583 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
3585 if (nfsd4_has_session(cstate
))
3587 if (seqid
== so
->so_seqid
- 1)
3588 return nfserr_replay_me
;
3589 if (seqid
== so
->so_seqid
)
3591 return nfserr_bad_seqid
;
3594 static __be32
lookup_clientid(clientid_t
*clid
,
3595 struct nfsd4_compound_state
*cstate
,
3596 struct nfsd_net
*nn
)
3598 struct nfs4_client
*found
;
3601 found
= cstate
->clp
;
3602 if (!same_clid(&found
->cl_clientid
, clid
))
3603 return nfserr_stale_clientid
;
3607 if (STALE_CLIENTID(clid
, nn
))
3608 return nfserr_stale_clientid
;
3611 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3612 * cached already then we know this is for is for v4.0 and "sessions"
3615 WARN_ON_ONCE(cstate
->session
);
3616 spin_lock(&nn
->client_lock
);
3617 found
= find_confirmed_client(clid
, false, nn
);
3619 spin_unlock(&nn
->client_lock
);
3620 return nfserr_expired
;
3622 atomic_inc(&found
->cl_refcount
);
3623 spin_unlock(&nn
->client_lock
);
3625 /* Cache the nfs4_client in cstate! */
3626 cstate
->clp
= found
;
3631 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
3632 struct nfsd4_open
*open
, struct nfsd_net
*nn
)
3634 clientid_t
*clientid
= &open
->op_clientid
;
3635 struct nfs4_client
*clp
= NULL
;
3636 unsigned int strhashval
;
3637 struct nfs4_openowner
*oo
= NULL
;
3640 if (STALE_CLIENTID(&open
->op_clientid
, nn
))
3641 return nfserr_stale_clientid
;
3643 * In case we need it later, after we've already created the
3644 * file and don't want to risk a further failure:
3646 open
->op_file
= nfsd4_alloc_file();
3647 if (open
->op_file
== NULL
)
3648 return nfserr_jukebox
;
3650 status
= lookup_clientid(clientid
, cstate
, nn
);
3655 strhashval
= ownerstr_hashval(&open
->op_owner
);
3656 oo
= find_openstateowner_str(strhashval
, open
, clp
);
3657 open
->op_openowner
= oo
;
3661 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
3662 /* Replace unconfirmed owners without checking for replay. */
3663 release_openowner(oo
);
3664 open
->op_openowner
= NULL
;
3667 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
3672 oo
= alloc_init_open_stateowner(strhashval
, open
, cstate
);
3674 return nfserr_jukebox
;
3675 open
->op_openowner
= oo
;
3677 open
->op_stp
= nfs4_alloc_open_stateid(clp
);
3679 return nfserr_jukebox
;
3681 if (nfsd4_has_session(cstate
) &&
3682 (cstate
->current_fh
.fh_export
->ex_flags
& NFSEXP_PNFS
)) {
3683 open
->op_odstate
= alloc_clnt_odstate(clp
);
3684 if (!open
->op_odstate
)
3685 return nfserr_jukebox
;
3691 static inline __be32
3692 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
3694 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
3695 return nfserr_openmode
;
3700 static int share_access_to_flags(u32 share_access
)
3702 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
3705 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
3707 struct nfs4_stid
*ret
;
3709 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
3712 return delegstateid(ret
);
3715 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
3717 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
3718 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
3722 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfsd4_open
*open
,
3723 struct nfs4_delegation
**dp
)
3726 __be32 status
= nfserr_bad_stateid
;
3727 struct nfs4_delegation
*deleg
;
3729 deleg
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
3732 flags
= share_access_to_flags(open
->op_share_access
);
3733 status
= nfs4_check_delegmode(deleg
, flags
);
3735 nfs4_put_stid(&deleg
->dl_stid
);
3740 if (!nfsd4_is_deleg_cur(open
))
3744 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3748 static struct nfs4_ol_stateid
*
3749 nfsd4_find_existing_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3751 struct nfs4_ol_stateid
*local
, *ret
= NULL
;
3752 struct nfs4_openowner
*oo
= open
->op_openowner
;
3754 spin_lock(&fp
->fi_lock
);
3755 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
3756 /* ignore lock owners */
3757 if (local
->st_stateowner
->so_is_open_owner
== 0)
3759 if (local
->st_stateowner
== &oo
->oo_owner
) {
3761 atomic_inc(&ret
->st_stid
.sc_count
);
3765 spin_unlock(&fp
->fi_lock
);
3769 static inline int nfs4_access_to_access(u32 nfs4_access
)
3773 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
3774 flags
|= NFSD_MAY_READ
;
3775 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
3776 flags
|= NFSD_MAY_WRITE
;
3780 static inline __be32
3781 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
3782 struct nfsd4_open
*open
)
3784 struct iattr iattr
= {
3785 .ia_valid
= ATTR_SIZE
,
3788 if (!open
->op_truncate
)
3790 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
3791 return nfserr_inval
;
3792 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
3795 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
3796 struct svc_fh
*cur_fh
, struct nfs4_ol_stateid
*stp
,
3797 struct nfsd4_open
*open
)
3799 struct file
*filp
= NULL
;
3801 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
3802 int access
= nfs4_access_to_access(open
->op_share_access
);
3803 unsigned char old_access_bmap
, old_deny_bmap
;
3805 spin_lock(&fp
->fi_lock
);
3808 * Are we trying to set a deny mode that would conflict with
3811 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
3812 if (status
!= nfs_ok
) {
3813 spin_unlock(&fp
->fi_lock
);
3817 /* set access to the file */
3818 status
= nfs4_file_get_access(fp
, open
->op_share_access
);
3819 if (status
!= nfs_ok
) {
3820 spin_unlock(&fp
->fi_lock
);
3824 /* Set access bits in stateid */
3825 old_access_bmap
= stp
->st_access_bmap
;
3826 set_access(open
->op_share_access
, stp
);
3828 /* Set new deny mask */
3829 old_deny_bmap
= stp
->st_deny_bmap
;
3830 set_deny(open
->op_share_deny
, stp
);
3831 fp
->fi_share_deny
|= (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
3833 if (!fp
->fi_fds
[oflag
]) {
3834 spin_unlock(&fp
->fi_lock
);
3835 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
, &filp
);
3837 goto out_put_access
;
3838 spin_lock(&fp
->fi_lock
);
3839 if (!fp
->fi_fds
[oflag
]) {
3840 fp
->fi_fds
[oflag
] = filp
;
3844 spin_unlock(&fp
->fi_lock
);
3848 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
3850 goto out_put_access
;
3854 stp
->st_access_bmap
= old_access_bmap
;
3855 nfs4_file_put_access(fp
, open
->op_share_access
);
3856 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap
), stp
);
3861 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
)
3864 unsigned char old_deny_bmap
= stp
->st_deny_bmap
;
3866 if (!test_access(open
->op_share_access
, stp
))
3867 return nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, stp
, open
);
3869 /* test and set deny mode */
3870 spin_lock(&fp
->fi_lock
);
3871 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
3872 if (status
== nfs_ok
) {
3873 set_deny(open
->op_share_deny
, stp
);
3874 fp
->fi_share_deny
|=
3875 (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
3877 spin_unlock(&fp
->fi_lock
);
3879 if (status
!= nfs_ok
)
3882 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
3883 if (status
!= nfs_ok
)
3884 reset_union_bmap_deny(old_deny_bmap
, stp
);
3889 nfs4_set_claim_prev(struct nfsd4_open
*open
, bool has_session
)
3891 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3894 /* Should we give out recallable state?: */
3895 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
3897 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
3900 * In the sessions case, since we don't have to establish a
3901 * separate connection for callbacks, we assume it's OK
3902 * until we hear otherwise:
3904 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
3907 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_file
*fp
, int flag
)
3909 struct file_lock
*fl
;
3911 fl
= locks_alloc_lock();
3914 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
3915 fl
->fl_flags
= FL_DELEG
;
3916 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
3917 fl
->fl_end
= OFFSET_MAX
;
3918 fl
->fl_owner
= (fl_owner_t
)fp
;
3919 fl
->fl_pid
= current
->tgid
;
3923 static int nfs4_setlease(struct nfs4_delegation
*dp
)
3925 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
3926 struct file_lock
*fl
, *ret
;
3930 fl
= nfs4_alloc_init_lease(fp
, NFS4_OPEN_DELEGATE_READ
);
3933 filp
= find_readable_file(fp
);
3935 /* We should always have a readable file here */
3941 status
= vfs_setlease(filp
, fl
->fl_type
, &fl
, NULL
);
3943 locks_free_lock(fl
);
3946 spin_lock(&state_lock
);
3947 spin_lock(&fp
->fi_lock
);
3948 /* Did the lease get broken before we took the lock? */
3950 if (fp
->fi_had_conflict
)
3953 if (fp
->fi_deleg_file
) {
3956 hash_delegation_locked(dp
, fp
);
3959 fp
->fi_deleg_file
= filp
;
3960 fp
->fi_delegees
= 1;
3961 hash_delegation_locked(dp
, fp
);
3962 spin_unlock(&fp
->fi_lock
);
3963 spin_unlock(&state_lock
);
3966 spin_unlock(&fp
->fi_lock
);
3967 spin_unlock(&state_lock
);
3973 static struct nfs4_delegation
*
3974 nfs4_set_delegation(struct nfs4_client
*clp
, struct svc_fh
*fh
,
3975 struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*odstate
)
3978 struct nfs4_delegation
*dp
;
3980 if (fp
->fi_had_conflict
)
3981 return ERR_PTR(-EAGAIN
);
3983 dp
= alloc_init_deleg(clp
, fh
, odstate
);
3985 return ERR_PTR(-ENOMEM
);
3988 spin_lock(&state_lock
);
3989 spin_lock(&fp
->fi_lock
);
3990 dp
->dl_stid
.sc_file
= fp
;
3991 if (!fp
->fi_deleg_file
) {
3992 spin_unlock(&fp
->fi_lock
);
3993 spin_unlock(&state_lock
);
3994 status
= nfs4_setlease(dp
);
3997 if (fp
->fi_had_conflict
) {
4002 hash_delegation_locked(dp
, fp
);
4005 spin_unlock(&fp
->fi_lock
);
4006 spin_unlock(&state_lock
);
4009 put_clnt_odstate(dp
->dl_clnt_odstate
);
4010 nfs4_put_stid(&dp
->dl_stid
);
4011 return ERR_PTR(status
);
4016 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
4018 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4019 if (status
== -EAGAIN
)
4020 open
->op_why_no_deleg
= WND4_CONTENTION
;
4022 open
->op_why_no_deleg
= WND4_RESOURCE
;
4023 switch (open
->op_deleg_want
) {
4024 case NFS4_SHARE_WANT_READ_DELEG
:
4025 case NFS4_SHARE_WANT_WRITE_DELEG
:
4026 case NFS4_SHARE_WANT_ANY_DELEG
:
4028 case NFS4_SHARE_WANT_CANCEL
:
4029 open
->op_why_no_deleg
= WND4_CANCELLED
;
4031 case NFS4_SHARE_WANT_NO_DELEG
:
4038 * Attempt to hand out a delegation.
4040 * Note we don't support write delegations, and won't until the vfs has
4041 * proper support for them.
4044 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
,
4045 struct nfs4_ol_stateid
*stp
)
4047 struct nfs4_delegation
*dp
;
4048 struct nfs4_openowner
*oo
= openowner(stp
->st_stateowner
);
4049 struct nfs4_client
*clp
= stp
->st_stid
.sc_client
;
4053 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
4054 open
->op_recall
= 0;
4055 switch (open
->op_claim_type
) {
4056 case NFS4_OPEN_CLAIM_PREVIOUS
:
4058 open
->op_recall
= 1;
4059 if (open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_READ
)
4062 case NFS4_OPEN_CLAIM_NULL
:
4063 case NFS4_OPEN_CLAIM_FH
:
4065 * Let's not give out any delegations till everyone's
4066 * had the chance to reclaim theirs....
4068 if (locks_in_grace(clp
->net
))
4070 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
4073 * Also, if the file was opened for write or
4074 * create, there's a good chance the client's
4075 * about to write to it, resulting in an
4076 * immediate recall (since we don't support
4077 * write delegations):
4079 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
4081 if (open
->op_create
== NFS4_OPEN_CREATE
)
4087 dp
= nfs4_set_delegation(clp
, fh
, stp
->st_stid
.sc_file
, stp
->st_clnt_odstate
);
4091 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
4093 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
4094 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
4095 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_READ
;
4096 nfs4_put_stid(&dp
->dl_stid
);
4099 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE
;
4100 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
4101 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
) {
4102 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4103 open
->op_recall
= 1;
4106 /* 4.1 client asking for a delegation? */
4107 if (open
->op_deleg_want
)
4108 nfsd4_open_deleg_none_ext(open
, status
);
4112 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
4113 struct nfs4_delegation
*dp
)
4115 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
4116 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4117 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4118 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
4119 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
4120 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4121 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4122 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
4124 /* Otherwise the client must be confused wanting a delegation
4125 * it already has, therefore we don't return
4126 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4131 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
4133 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
4134 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
4135 struct nfs4_file
*fp
= NULL
;
4136 struct nfs4_ol_stateid
*stp
= NULL
;
4137 struct nfs4_delegation
*dp
= NULL
;
4141 * Lookup file; if found, lookup stateid and check open request,
4142 * and check for delegations in the process of being recalled.
4143 * If not found, create the nfs4_file struct
4145 fp
= find_or_add_file(open
->op_file
, ¤t_fh
->fh_handle
);
4146 if (fp
!= open
->op_file
) {
4147 status
= nfs4_check_deleg(cl
, open
, &dp
);
4150 stp
= nfsd4_find_existing_open(fp
, open
);
4152 open
->op_file
= NULL
;
4153 status
= nfserr_bad_stateid
;
4154 if (nfsd4_is_deleg_cur(open
))
4159 * OPEN the file, or upgrade an existing OPEN.
4160 * If truncate fails, the OPEN fails.
4163 /* Stateid was found, this is an OPEN upgrade */
4164 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
4169 open
->op_stp
= NULL
;
4170 init_open_stateid(stp
, fp
, open
);
4171 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, stp
, open
);
4173 release_open_stateid(stp
);
4177 stp
->st_clnt_odstate
= find_or_hash_clnt_odstate(fp
,
4179 if (stp
->st_clnt_odstate
== open
->op_odstate
)
4180 open
->op_odstate
= NULL
;
4182 update_stateid(&stp
->st_stid
.sc_stateid
);
4183 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4185 if (nfsd4_has_session(&resp
->cstate
)) {
4186 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
4187 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4188 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
4194 * Attempt to hand out a delegation. No error return, because the
4195 * OPEN succeeds even if we fail.
4197 nfs4_open_delegation(current_fh
, open
, stp
);
4201 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
4202 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
4204 /* 4.1 client trying to upgrade/downgrade delegation? */
4205 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
4206 open
->op_deleg_want
)
4207 nfsd4_deleg_xgrade_none_ext(open
, dp
);
4211 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
4212 nfs4_set_claim_prev(open
, nfsd4_has_session(&resp
->cstate
));
4214 * To finish the open response, we just need to set the rflags.
4216 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
4217 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
4218 !nfsd4_has_session(&resp
->cstate
))
4219 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
4221 nfs4_put_stid(&dp
->dl_stid
);
4223 nfs4_put_stid(&stp
->st_stid
);
4228 void nfsd4_cleanup_open_state(struct nfsd4_compound_state
*cstate
,
4229 struct nfsd4_open
*open
)
4231 if (open
->op_openowner
) {
4232 struct nfs4_stateowner
*so
= &open
->op_openowner
->oo_owner
;
4234 nfsd4_cstate_assign_replay(cstate
, so
);
4235 nfs4_put_stateowner(so
);
4238 kmem_cache_free(file_slab
, open
->op_file
);
4240 nfs4_put_stid(&open
->op_stp
->st_stid
);
4241 if (open
->op_odstate
)
4242 kmem_cache_free(odstate_slab
, open
->op_odstate
);
4246 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4249 struct nfs4_client
*clp
;
4251 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
4253 dprintk("process_renew(%08x/%08x): starting\n",
4254 clid
->cl_boot
, clid
->cl_id
);
4255 status
= lookup_clientid(clid
, cstate
, nn
);
4259 status
= nfserr_cb_path_down
;
4260 if (!list_empty(&clp
->cl_delegations
)
4261 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
4269 nfsd4_end_grace(struct nfsd_net
*nn
)
4271 /* do nothing if grace period already ended */
4272 if (nn
->grace_ended
)
4275 dprintk("NFSD: end of grace period\n");
4276 nn
->grace_ended
= true;
4278 * If the server goes down again right now, an NFSv4
4279 * client will still be allowed to reclaim after it comes back up,
4280 * even if it hasn't yet had a chance to reclaim state this time.
4283 nfsd4_record_grace_done(nn
);
4285 * At this point, NFSv4 clients can still reclaim. But if the
4286 * server crashes, any that have not yet reclaimed will be out
4287 * of luck on the next boot.
4289 * (NFSv4.1+ clients are considered to have reclaimed once they
4290 * call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
4291 * have reclaimed after their first OPEN.)
4293 locks_end_grace(&nn
->nfsd4_manager
);
4295 * At this point, and once lockd and/or any other containers
4296 * exit their grace period, further reclaims will fail and
4297 * regular locking can resume.
4302 nfs4_laundromat(struct nfsd_net
*nn
)
4304 struct nfs4_client
*clp
;
4305 struct nfs4_openowner
*oo
;
4306 struct nfs4_delegation
*dp
;
4307 struct nfs4_ol_stateid
*stp
;
4308 struct list_head
*pos
, *next
, reaplist
;
4309 time_t cutoff
= get_seconds() - nn
->nfsd4_lease
;
4310 time_t t
, new_timeo
= nn
->nfsd4_lease
;
4312 dprintk("NFSD: laundromat service - starting\n");
4313 nfsd4_end_grace(nn
);
4314 INIT_LIST_HEAD(&reaplist
);
4315 spin_lock(&nn
->client_lock
);
4316 list_for_each_safe(pos
, next
, &nn
->client_lru
) {
4317 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4318 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
4319 t
= clp
->cl_time
- cutoff
;
4320 new_timeo
= min(new_timeo
, t
);
4323 if (mark_client_expired_locked(clp
)) {
4324 dprintk("NFSD: client in use (clientid %08x)\n",
4325 clp
->cl_clientid
.cl_id
);
4328 list_add(&clp
->cl_lru
, &reaplist
);
4330 spin_unlock(&nn
->client_lock
);
4331 list_for_each_safe(pos
, next
, &reaplist
) {
4332 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4333 dprintk("NFSD: purging unused client (clientid %08x)\n",
4334 clp
->cl_clientid
.cl_id
);
4335 list_del_init(&clp
->cl_lru
);
4338 spin_lock(&state_lock
);
4339 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
4340 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4341 if (net_generic(dp
->dl_stid
.sc_client
->net
, nfsd_net_id
) != nn
)
4343 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
4344 t
= dp
->dl_time
- cutoff
;
4345 new_timeo
= min(new_timeo
, t
);
4348 unhash_delegation_locked(dp
);
4349 list_add(&dp
->dl_recall_lru
, &reaplist
);
4351 spin_unlock(&state_lock
);
4352 while (!list_empty(&reaplist
)) {
4353 dp
= list_first_entry(&reaplist
, struct nfs4_delegation
,
4355 list_del_init(&dp
->dl_recall_lru
);
4356 revoke_delegation(dp
);
4359 spin_lock(&nn
->client_lock
);
4360 while (!list_empty(&nn
->close_lru
)) {
4361 oo
= list_first_entry(&nn
->close_lru
, struct nfs4_openowner
,
4363 if (time_after((unsigned long)oo
->oo_time
,
4364 (unsigned long)cutoff
)) {
4365 t
= oo
->oo_time
- cutoff
;
4366 new_timeo
= min(new_timeo
, t
);
4369 list_del_init(&oo
->oo_close_lru
);
4370 stp
= oo
->oo_last_closed_stid
;
4371 oo
->oo_last_closed_stid
= NULL
;
4372 spin_unlock(&nn
->client_lock
);
4373 nfs4_put_stid(&stp
->st_stid
);
4374 spin_lock(&nn
->client_lock
);
4376 spin_unlock(&nn
->client_lock
);
4378 new_timeo
= max_t(time_t, new_timeo
, NFSD_LAUNDROMAT_MINTIMEOUT
);
4382 static struct workqueue_struct
*laundry_wq
;
4383 static void laundromat_main(struct work_struct
*);
4386 laundromat_main(struct work_struct
*laundry
)
4389 struct delayed_work
*dwork
= container_of(laundry
, struct delayed_work
,
4391 struct nfsd_net
*nn
= container_of(dwork
, struct nfsd_net
,
4394 t
= nfs4_laundromat(nn
);
4395 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
4396 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, t
*HZ
);
4399 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stid
*stp
)
4401 if (!fh_match(&fhp
->fh_handle
, &stp
->sc_file
->fi_fhandle
))
4402 return nfserr_bad_stateid
;
4407 access_permit_read(struct nfs4_ol_stateid
*stp
)
4409 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
4410 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
4411 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
4415 access_permit_write(struct nfs4_ol_stateid
*stp
)
4417 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
4418 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
4422 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
4424 __be32 status
= nfserr_openmode
;
4426 /* For lock stateid's, we test the parent open, not the lock: */
4427 if (stp
->st_openstp
)
4428 stp
= stp
->st_openstp
;
4429 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
4431 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
4438 static inline __be32
4439 check_special_stateids(struct net
*net
, svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
4441 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
4443 else if (locks_in_grace(net
)) {
4444 /* Answer in remaining cases depends on existence of
4445 * conflicting state; so we must wait out the grace period. */
4446 return nfserr_grace
;
4447 } else if (flags
& WR_STATE
)
4448 return nfs4_share_conflict(current_fh
,
4449 NFS4_SHARE_DENY_WRITE
);
4450 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4451 return nfs4_share_conflict(current_fh
,
4452 NFS4_SHARE_DENY_READ
);
4456 * Allow READ/WRITE during grace period on recovered state only for files
4457 * that are not able to provide mandatory locking.
4460 grace_disallows_io(struct net
*net
, struct inode
*inode
)
4462 return locks_in_grace(net
) && mandatory_lock(inode
);
4465 /* Returns true iff a is later than b: */
4466 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
4468 return (s32
)(a
->si_generation
- b
->si_generation
) > 0;
4471 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
4474 * When sessions are used the stateid generation number is ignored
4477 if (has_session
&& in
->si_generation
== 0)
4480 if (in
->si_generation
== ref
->si_generation
)
4483 /* If the client sends us a stateid from the future, it's buggy: */
4484 if (stateid_generation_after(in
, ref
))
4485 return nfserr_bad_stateid
;
4487 * However, we could see a stateid from the past, even from a
4488 * non-buggy client. For example, if the client sends a lock
4489 * while some IO is outstanding, the lock may bump si_generation
4490 * while the IO is still in flight. The client could avoid that
4491 * situation by waiting for responses on all the IO requests,
4492 * but better performance may result in retrying IO that
4493 * receives an old_stateid error if requests are rarely
4494 * reordered in flight:
4496 return nfserr_old_stateid
;
4499 static __be32
nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid
*ols
)
4501 if (ols
->st_stateowner
->so_is_open_owner
&&
4502 !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
4503 return nfserr_bad_stateid
;
4507 static __be32
nfsd4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
4509 struct nfs4_stid
*s
;
4510 __be32 status
= nfserr_bad_stateid
;
4512 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
4514 /* Client debugging aid. */
4515 if (!same_clid(&stateid
->si_opaque
.so_clid
, &cl
->cl_clientid
)) {
4516 char addr_str
[INET6_ADDRSTRLEN
];
4517 rpc_ntop((struct sockaddr
*)&cl
->cl_addr
, addr_str
,
4519 pr_warn_ratelimited("NFSD: client %s testing state ID "
4520 "with incorrect client ID\n", addr_str
);
4523 spin_lock(&cl
->cl_lock
);
4524 s
= find_stateid_locked(cl
, stateid
);
4527 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
4530 switch (s
->sc_type
) {
4531 case NFS4_DELEG_STID
:
4534 case NFS4_REVOKED_DELEG_STID
:
4535 status
= nfserr_deleg_revoked
;
4537 case NFS4_OPEN_STID
:
4538 case NFS4_LOCK_STID
:
4539 status
= nfsd4_check_openowner_confirmed(openlockstateid(s
));
4542 printk("unknown stateid type %x\n", s
->sc_type
);
4544 case NFS4_CLOSED_STID
:
4545 case NFS4_CLOSED_DELEG_STID
:
4546 status
= nfserr_bad_stateid
;
4549 spin_unlock(&cl
->cl_lock
);
4554 nfsd4_lookup_stateid(struct nfsd4_compound_state
*cstate
,
4555 stateid_t
*stateid
, unsigned char typemask
,
4556 struct nfs4_stid
**s
, struct nfsd_net
*nn
)
4560 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
4561 return nfserr_bad_stateid
;
4562 status
= lookup_clientid(&stateid
->si_opaque
.so_clid
, cstate
, nn
);
4563 if (status
== nfserr_stale_clientid
) {
4564 if (cstate
->session
)
4565 return nfserr_bad_stateid
;
4566 return nfserr_stale_stateid
;
4570 *s
= find_stateid_by_type(cstate
->clp
, stateid
, typemask
);
4572 return nfserr_bad_stateid
;
4576 static struct file
*
4577 nfs4_find_file(struct nfs4_stid
*s
, int flags
)
4582 switch (s
->sc_type
) {
4583 case NFS4_DELEG_STID
:
4584 if (WARN_ON_ONCE(!s
->sc_file
->fi_deleg_file
))
4586 return get_file(s
->sc_file
->fi_deleg_file
);
4587 case NFS4_OPEN_STID
:
4588 case NFS4_LOCK_STID
:
4589 if (flags
& RD_STATE
)
4590 return find_readable_file(s
->sc_file
);
4592 return find_writeable_file(s
->sc_file
);
4600 nfs4_check_olstateid(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*ols
, int flags
)
4604 status
= nfsd4_check_openowner_confirmed(ols
);
4607 return nfs4_check_openmode(ols
, flags
);
4611 nfs4_check_file(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct nfs4_stid
*s
,
4612 struct file
**filpp
, bool *tmp_file
, int flags
)
4614 int acc
= (flags
& RD_STATE
) ? NFSD_MAY_READ
: NFSD_MAY_WRITE
;
4618 file
= nfs4_find_file(s
, flags
);
4620 status
= nfsd_permission(rqstp
, fhp
->fh_export
, fhp
->fh_dentry
,
4621 acc
| NFSD_MAY_OWNER_OVERRIDE
);
4629 status
= nfsd_open(rqstp
, fhp
, S_IFREG
, acc
, filpp
);
4641 * Checks for stateid operations
4644 nfs4_preprocess_stateid_op(struct svc_rqst
*rqstp
,
4645 struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
,
4646 int flags
, struct file
**filpp
, bool *tmp_file
)
4648 struct svc_fh
*fhp
= &cstate
->current_fh
;
4649 struct inode
*ino
= d_inode(fhp
->fh_dentry
);
4650 struct net
*net
= SVC_NET(rqstp
);
4651 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4652 struct nfs4_stid
*s
= NULL
;
4660 if (grace_disallows_io(net
, ino
))
4661 return nfserr_grace
;
4663 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
4664 status
= check_special_stateids(net
, fhp
, stateid
, flags
);
4668 status
= nfsd4_lookup_stateid(cstate
, stateid
,
4669 NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
,
4673 status
= check_stateid_generation(stateid
, &s
->sc_stateid
,
4674 nfsd4_has_session(cstate
));
4678 switch (s
->sc_type
) {
4679 case NFS4_DELEG_STID
:
4680 status
= nfs4_check_delegmode(delegstateid(s
), flags
);
4682 case NFS4_OPEN_STID
:
4683 case NFS4_LOCK_STID
:
4684 status
= nfs4_check_olstateid(fhp
, openlockstateid(s
), flags
);
4687 status
= nfserr_bad_stateid
;
4692 status
= nfs4_check_fh(fhp
, s
);
4695 if (!status
&& filpp
)
4696 status
= nfs4_check_file(rqstp
, fhp
, s
, filpp
, tmp_file
, flags
);
4704 * Test if the stateid is valid
4707 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4708 struct nfsd4_test_stateid
*test_stateid
)
4710 struct nfsd4_test_stateid_id
*stateid
;
4711 struct nfs4_client
*cl
= cstate
->session
->se_client
;
4713 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
4714 stateid
->ts_id_status
=
4715 nfsd4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
4721 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4722 struct nfsd4_free_stateid
*free_stateid
)
4724 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
4725 struct nfs4_stid
*s
;
4726 struct nfs4_delegation
*dp
;
4727 struct nfs4_ol_stateid
*stp
;
4728 struct nfs4_client
*cl
= cstate
->session
->se_client
;
4729 __be32 ret
= nfserr_bad_stateid
;
4731 spin_lock(&cl
->cl_lock
);
4732 s
= find_stateid_locked(cl
, stateid
);
4735 switch (s
->sc_type
) {
4736 case NFS4_DELEG_STID
:
4737 ret
= nfserr_locks_held
;
4739 case NFS4_OPEN_STID
:
4740 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
4743 ret
= nfserr_locks_held
;
4745 case NFS4_LOCK_STID
:
4746 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
4749 stp
= openlockstateid(s
);
4750 ret
= nfserr_locks_held
;
4751 if (check_for_locks(stp
->st_stid
.sc_file
,
4752 lockowner(stp
->st_stateowner
)))
4754 unhash_lock_stateid(stp
);
4755 spin_unlock(&cl
->cl_lock
);
4759 case NFS4_REVOKED_DELEG_STID
:
4760 dp
= delegstateid(s
);
4761 list_del_init(&dp
->dl_recall_lru
);
4762 spin_unlock(&cl
->cl_lock
);
4766 /* Default falls through and returns nfserr_bad_stateid */
4769 spin_unlock(&cl
->cl_lock
);
4777 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
4778 RD_STATE
: WR_STATE
;
4781 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
4783 struct svc_fh
*current_fh
= &cstate
->current_fh
;
4784 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
4787 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
4790 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
4791 || stp
->st_stid
.sc_type
== NFS4_REVOKED_DELEG_STID
)
4793 * "Closed" stateid's exist *only* to return
4794 * nfserr_replay_me from the previous step, and
4795 * revoked delegations are kept only for free_stateid.
4797 return nfserr_bad_stateid
;
4798 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
4801 return nfs4_check_fh(current_fh
, &stp
->st_stid
);
4805 * Checks for sequence id mutating operations.
4808 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
4809 stateid_t
*stateid
, char typemask
,
4810 struct nfs4_ol_stateid
**stpp
,
4811 struct nfsd_net
*nn
)
4814 struct nfs4_stid
*s
;
4815 struct nfs4_ol_stateid
*stp
= NULL
;
4817 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
4818 seqid
, STATEID_VAL(stateid
));
4821 status
= nfsd4_lookup_stateid(cstate
, stateid
, typemask
, &s
, nn
);
4824 stp
= openlockstateid(s
);
4825 nfsd4_cstate_assign_replay(cstate
, stp
->st_stateowner
);
4827 status
= nfs4_seqid_op_checks(cstate
, stateid
, seqid
, stp
);
4831 nfs4_put_stid(&stp
->st_stid
);
4835 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
4836 stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
, struct nfsd_net
*nn
)
4839 struct nfs4_openowner
*oo
;
4840 struct nfs4_ol_stateid
*stp
;
4842 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
4843 NFS4_OPEN_STID
, &stp
, nn
);
4846 oo
= openowner(stp
->st_stateowner
);
4847 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
4848 nfs4_put_stid(&stp
->st_stid
);
4849 return nfserr_bad_stateid
;
4856 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4857 struct nfsd4_open_confirm
*oc
)
4860 struct nfs4_openowner
*oo
;
4861 struct nfs4_ol_stateid
*stp
;
4862 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
4864 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4865 cstate
->current_fh
.fh_dentry
);
4867 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
4871 status
= nfs4_preprocess_seqid_op(cstate
,
4872 oc
->oc_seqid
, &oc
->oc_req_stateid
,
4873 NFS4_OPEN_STID
, &stp
, nn
);
4876 oo
= openowner(stp
->st_stateowner
);
4877 status
= nfserr_bad_stateid
;
4878 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
)
4880 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
4881 update_stateid(&stp
->st_stid
.sc_stateid
);
4882 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4883 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
4884 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
4886 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
4889 nfs4_put_stid(&stp
->st_stid
);
4891 nfsd4_bump_seqid(cstate
, status
);
4895 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
4897 if (!test_access(access
, stp
))
4899 nfs4_file_put_access(stp
->st_stid
.sc_file
, access
);
4900 clear_access(access
, stp
);
4903 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
4905 switch (to_access
) {
4906 case NFS4_SHARE_ACCESS_READ
:
4907 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
4908 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
4910 case NFS4_SHARE_ACCESS_WRITE
:
4911 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
4912 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
4914 case NFS4_SHARE_ACCESS_BOTH
:
4922 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
4923 struct nfsd4_compound_state
*cstate
,
4924 struct nfsd4_open_downgrade
*od
)
4927 struct nfs4_ol_stateid
*stp
;
4928 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
4930 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4931 cstate
->current_fh
.fh_dentry
);
4933 /* We don't yet support WANT bits: */
4934 if (od
->od_deleg_want
)
4935 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
4938 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
4939 &od
->od_stateid
, &stp
, nn
);
4942 status
= nfserr_inval
;
4943 if (!test_access(od
->od_share_access
, stp
)) {
4944 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
4945 stp
->st_access_bmap
, od
->od_share_access
);
4948 if (!test_deny(od
->od_share_deny
, stp
)) {
4949 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
4950 stp
->st_deny_bmap
, od
->od_share_deny
);
4953 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
4955 reset_union_bmap_deny(od
->od_share_deny
, stp
);
4957 update_stateid(&stp
->st_stid
.sc_stateid
);
4958 memcpy(&od
->od_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4961 nfs4_put_stid(&stp
->st_stid
);
4963 nfsd4_bump_seqid(cstate
, status
);
4967 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
4969 struct nfs4_client
*clp
= s
->st_stid
.sc_client
;
4970 LIST_HEAD(reaplist
);
4972 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
4973 spin_lock(&clp
->cl_lock
);
4974 unhash_open_stateid(s
, &reaplist
);
4976 if (clp
->cl_minorversion
) {
4977 put_ol_stateid_locked(s
, &reaplist
);
4978 spin_unlock(&clp
->cl_lock
);
4979 free_ol_stateid_reaplist(&reaplist
);
4981 spin_unlock(&clp
->cl_lock
);
4982 free_ol_stateid_reaplist(&reaplist
);
4983 move_to_close_lru(s
, clp
->net
);
4988 * nfs4_unlock_state() called after encode
4991 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4992 struct nfsd4_close
*close
)
4995 struct nfs4_ol_stateid
*stp
;
4996 struct net
*net
= SVC_NET(rqstp
);
4997 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4999 dprintk("NFSD: nfsd4_close on file %pd\n",
5000 cstate
->current_fh
.fh_dentry
);
5002 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
5004 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
5006 nfsd4_bump_seqid(cstate
, status
);
5009 update_stateid(&stp
->st_stid
.sc_stateid
);
5010 memcpy(&close
->cl_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
5012 nfsd4_close_open_stateid(stp
);
5014 /* put reference from nfs4_preprocess_seqid_op */
5015 nfs4_put_stid(&stp
->st_stid
);
5021 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5022 struct nfsd4_delegreturn
*dr
)
5024 struct nfs4_delegation
*dp
;
5025 stateid_t
*stateid
= &dr
->dr_stateid
;
5026 struct nfs4_stid
*s
;
5028 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5030 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
5033 status
= nfsd4_lookup_stateid(cstate
, stateid
, NFS4_DELEG_STID
, &s
, nn
);
5036 dp
= delegstateid(s
);
5037 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
5041 destroy_delegation(dp
);
5043 nfs4_put_stid(&dp
->dl_stid
);
5049 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
5052 end_offset(u64 start
, u64 len
)
5057 return end
>= start
? end
: NFS4_MAX_UINT64
;
5060 /* last octet in a range */
5062 last_byte_offset(u64 start
, u64 len
)
5068 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
5072 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5073 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5074 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
5075 * locking, this prevents us from being completely protocol-compliant. The
5076 * real solution to this problem is to start using unsigned file offsets in
5077 * the VFS, but this is a very deep change!
5080 nfs4_transform_lock_offset(struct file_lock
*lock
)
5082 if (lock
->fl_start
< 0)
5083 lock
->fl_start
= OFFSET_MAX
;
5084 if (lock
->fl_end
< 0)
5085 lock
->fl_end
= OFFSET_MAX
;
5089 nfsd4_fl_get_owner(fl_owner_t owner
)
5091 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5093 nfs4_get_stateowner(&lo
->lo_owner
);
5098 nfsd4_fl_put_owner(fl_owner_t owner
)
5100 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5103 nfs4_put_stateowner(&lo
->lo_owner
);
5106 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
5107 .lm_get_owner
= nfsd4_fl_get_owner
,
5108 .lm_put_owner
= nfsd4_fl_put_owner
,
5112 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
5114 struct nfs4_lockowner
*lo
;
5116 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
5117 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
5118 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
5119 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
5120 if (!deny
->ld_owner
.data
)
5121 /* We just don't care that much */
5123 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
5124 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
5127 deny
->ld_owner
.len
= 0;
5128 deny
->ld_owner
.data
= NULL
;
5129 deny
->ld_clientid
.cl_boot
= 0;
5130 deny
->ld_clientid
.cl_id
= 0;
5132 deny
->ld_start
= fl
->fl_start
;
5133 deny
->ld_length
= NFS4_MAX_UINT64
;
5134 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
5135 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
5136 deny
->ld_type
= NFS4_READ_LT
;
5137 if (fl
->fl_type
!= F_RDLCK
)
5138 deny
->ld_type
= NFS4_WRITE_LT
;
5141 static struct nfs4_lockowner
*
5142 find_lockowner_str_locked(clientid_t
*clid
, struct xdr_netobj
*owner
,
5143 struct nfs4_client
*clp
)
5145 unsigned int strhashval
= ownerstr_hashval(owner
);
5146 struct nfs4_stateowner
*so
;
5148 lockdep_assert_held(&clp
->cl_lock
);
5150 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[strhashval
],
5152 if (so
->so_is_open_owner
)
5154 if (same_owner_str(so
, owner
))
5155 return lockowner(nfs4_get_stateowner(so
));
5160 static struct nfs4_lockowner
*
5161 find_lockowner_str(clientid_t
*clid
, struct xdr_netobj
*owner
,
5162 struct nfs4_client
*clp
)
5164 struct nfs4_lockowner
*lo
;
5166 spin_lock(&clp
->cl_lock
);
5167 lo
= find_lockowner_str_locked(clid
, owner
, clp
);
5168 spin_unlock(&clp
->cl_lock
);
5172 static void nfs4_unhash_lockowner(struct nfs4_stateowner
*sop
)
5174 unhash_lockowner_locked(lockowner(sop
));
5177 static void nfs4_free_lockowner(struct nfs4_stateowner
*sop
)
5179 struct nfs4_lockowner
*lo
= lockowner(sop
);
5181 kmem_cache_free(lockowner_slab
, lo
);
5184 static const struct nfs4_stateowner_operations lockowner_ops
= {
5185 .so_unhash
= nfs4_unhash_lockowner
,
5186 .so_free
= nfs4_free_lockowner
,
5190 * Alloc a lock owner structure.
5191 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
5194 * strhashval = ownerstr_hashval
5196 static struct nfs4_lockowner
*
5197 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
,
5198 struct nfs4_ol_stateid
*open_stp
,
5199 struct nfsd4_lock
*lock
)
5201 struct nfs4_lockowner
*lo
, *ret
;
5203 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
5206 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
5207 lo
->lo_owner
.so_is_open_owner
= 0;
5208 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
;
5209 lo
->lo_owner
.so_ops
= &lockowner_ops
;
5210 spin_lock(&clp
->cl_lock
);
5211 ret
= find_lockowner_str_locked(&clp
->cl_clientid
,
5212 &lock
->lk_new_owner
, clp
);
5214 list_add(&lo
->lo_owner
.so_strhash
,
5215 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
5218 nfs4_free_lockowner(&lo
->lo_owner
);
5219 spin_unlock(&clp
->cl_lock
);
5224 init_lock_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_lockowner
*lo
,
5225 struct nfs4_file
*fp
, struct inode
*inode
,
5226 struct nfs4_ol_stateid
*open_stp
)
5228 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5230 lockdep_assert_held(&clp
->cl_lock
);
5232 atomic_inc(&stp
->st_stid
.sc_count
);
5233 stp
->st_stid
.sc_type
= NFS4_LOCK_STID
;
5234 stp
->st_stateowner
= nfs4_get_stateowner(&lo
->lo_owner
);
5236 stp
->st_stid
.sc_file
= fp
;
5237 stp
->st_stid
.sc_free
= nfs4_free_lock_stateid
;
5238 stp
->st_access_bmap
= 0;
5239 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
5240 stp
->st_openstp
= open_stp
;
5241 list_add(&stp
->st_locks
, &open_stp
->st_locks
);
5242 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
5243 spin_lock(&fp
->fi_lock
);
5244 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
5245 spin_unlock(&fp
->fi_lock
);
5248 static struct nfs4_ol_stateid
*
5249 find_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
)
5251 struct nfs4_ol_stateid
*lst
;
5252 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5254 lockdep_assert_held(&clp
->cl_lock
);
5256 list_for_each_entry(lst
, &lo
->lo_owner
.so_stateids
, st_perstateowner
) {
5257 if (lst
->st_stid
.sc_file
== fp
) {
5258 atomic_inc(&lst
->st_stid
.sc_count
);
5265 static struct nfs4_ol_stateid
*
5266 find_or_create_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fi
,
5267 struct inode
*inode
, struct nfs4_ol_stateid
*ost
,
5270 struct nfs4_stid
*ns
= NULL
;
5271 struct nfs4_ol_stateid
*lst
;
5272 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5273 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
5275 spin_lock(&clp
->cl_lock
);
5276 lst
= find_lock_stateid(lo
, fi
);
5278 spin_unlock(&clp
->cl_lock
);
5279 ns
= nfs4_alloc_stid(clp
, stateid_slab
);
5283 spin_lock(&clp
->cl_lock
);
5284 lst
= find_lock_stateid(lo
, fi
);
5286 lst
= openlockstateid(ns
);
5287 init_lock_stateid(lst
, lo
, fi
, inode
, ost
);
5292 spin_unlock(&clp
->cl_lock
);
5299 check_lock_length(u64 offset
, u64 length
)
5301 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
5302 LOFF_OVERFLOW(offset
, length
)));
5305 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
5307 struct nfs4_file
*fp
= lock_stp
->st_stid
.sc_file
;
5309 lockdep_assert_held(&fp
->fi_lock
);
5311 if (test_access(access
, lock_stp
))
5313 __nfs4_file_get_access(fp
, access
);
5314 set_access(access
, lock_stp
);
5318 lookup_or_create_lock_state(struct nfsd4_compound_state
*cstate
,
5319 struct nfs4_ol_stateid
*ost
,
5320 struct nfsd4_lock
*lock
,
5321 struct nfs4_ol_stateid
**lst
, bool *new)
5324 struct nfs4_file
*fi
= ost
->st_stid
.sc_file
;
5325 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5326 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
5327 struct inode
*inode
= d_inode(cstate
->current_fh
.fh_dentry
);
5328 struct nfs4_lockowner
*lo
;
5329 unsigned int strhashval
;
5331 lo
= find_lockowner_str(&cl
->cl_clientid
, &lock
->v
.new.owner
, cl
);
5333 strhashval
= ownerstr_hashval(&lock
->v
.new.owner
);
5334 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
5336 return nfserr_jukebox
;
5338 /* with an existing lockowner, seqids must be the same */
5339 status
= nfserr_bad_seqid
;
5340 if (!cstate
->minorversion
&&
5341 lock
->lk_new_lock_seqid
!= lo
->lo_owner
.so_seqid
)
5345 *lst
= find_or_create_lock_stateid(lo
, fi
, inode
, ost
, new);
5347 status
= nfserr_jukebox
;
5352 nfs4_put_stateowner(&lo
->lo_owner
);
5360 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5361 struct nfsd4_lock
*lock
)
5363 struct nfs4_openowner
*open_sop
= NULL
;
5364 struct nfs4_lockowner
*lock_sop
= NULL
;
5365 struct nfs4_ol_stateid
*lock_stp
= NULL
;
5366 struct nfs4_ol_stateid
*open_stp
= NULL
;
5367 struct nfs4_file
*fp
;
5368 struct file
*filp
= NULL
;
5369 struct file_lock
*file_lock
= NULL
;
5370 struct file_lock
*conflock
= NULL
;
5375 struct net
*net
= SVC_NET(rqstp
);
5376 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5378 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5379 (long long) lock
->lk_offset
,
5380 (long long) lock
->lk_length
);
5382 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
5383 return nfserr_inval
;
5385 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
5386 S_IFREG
, NFSD_MAY_LOCK
))) {
5387 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5391 if (lock
->lk_is_new
) {
5392 if (nfsd4_has_session(cstate
))
5393 /* See rfc 5661 18.10.3: given clientid is ignored: */
5394 memcpy(&lock
->v
.new.clientid
,
5395 &cstate
->session
->se_client
->cl_clientid
,
5396 sizeof(clientid_t
));
5398 status
= nfserr_stale_clientid
;
5399 if (STALE_CLIENTID(&lock
->lk_new_clientid
, nn
))
5402 /* validate and update open stateid and open seqid */
5403 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
5404 lock
->lk_new_open_seqid
,
5405 &lock
->lk_new_open_stateid
,
5409 open_sop
= openowner(open_stp
->st_stateowner
);
5410 status
= nfserr_bad_stateid
;
5411 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
5412 &lock
->v
.new.clientid
))
5414 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
5417 status
= nfs4_preprocess_seqid_op(cstate
,
5418 lock
->lk_old_lock_seqid
,
5419 &lock
->lk_old_lock_stateid
,
5420 NFS4_LOCK_STID
, &lock_stp
, nn
);
5424 lock_sop
= lockowner(lock_stp
->st_stateowner
);
5426 lkflg
= setlkflg(lock
->lk_type
);
5427 status
= nfs4_check_openmode(lock_stp
, lkflg
);
5431 status
= nfserr_grace
;
5432 if (locks_in_grace(net
) && !lock
->lk_reclaim
)
5434 status
= nfserr_no_grace
;
5435 if (!locks_in_grace(net
) && lock
->lk_reclaim
)
5438 file_lock
= locks_alloc_lock();
5440 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5441 status
= nfserr_jukebox
;
5445 fp
= lock_stp
->st_stid
.sc_file
;
5446 switch (lock
->lk_type
) {
5449 spin_lock(&fp
->fi_lock
);
5450 filp
= find_readable_file_locked(fp
);
5452 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
5453 spin_unlock(&fp
->fi_lock
);
5454 file_lock
->fl_type
= F_RDLCK
;
5457 case NFS4_WRITEW_LT
:
5458 spin_lock(&fp
->fi_lock
);
5459 filp
= find_writeable_file_locked(fp
);
5461 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
5462 spin_unlock(&fp
->fi_lock
);
5463 file_lock
->fl_type
= F_WRLCK
;
5466 status
= nfserr_inval
;
5470 status
= nfserr_openmode
;
5474 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(&lock_sop
->lo_owner
));
5475 file_lock
->fl_pid
= current
->tgid
;
5476 file_lock
->fl_file
= filp
;
5477 file_lock
->fl_flags
= FL_POSIX
;
5478 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
5479 file_lock
->fl_start
= lock
->lk_offset
;
5480 file_lock
->fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
5481 nfs4_transform_lock_offset(file_lock
);
5483 conflock
= locks_alloc_lock();
5485 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5486 status
= nfserr_jukebox
;
5490 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, conflock
);
5492 case 0: /* success! */
5493 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
5494 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
5498 case (EAGAIN
): /* conflock holds conflicting lock */
5499 status
= nfserr_denied
;
5500 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5501 nfs4_set_lock_denied(conflock
, &lock
->lk_denied
);
5504 status
= nfserr_deadlock
;
5507 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
5508 status
= nfserrno(err
);
5515 /* Bump seqid manually if the 4.0 replay owner is openowner */
5516 if (cstate
->replay_owner
&&
5517 cstate
->replay_owner
!= &lock_sop
->lo_owner
&&
5518 seqid_mutating_err(ntohl(status
)))
5519 lock_sop
->lo_owner
.so_seqid
++;
5522 * If this is a new, never-before-used stateid, and we are
5523 * returning an error, then just go ahead and release it.
5526 release_lock_stateid(lock_stp
);
5528 nfs4_put_stid(&lock_stp
->st_stid
);
5531 nfs4_put_stid(&open_stp
->st_stid
);
5532 nfsd4_bump_seqid(cstate
, status
);
5534 locks_free_lock(file_lock
);
5536 locks_free_lock(conflock
);
5541 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5542 * so we do a temporary open here just to get an open file to pass to
5543 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5546 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
5549 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
5551 err
= nfserrno(vfs_test_lock(file
, lock
));
5561 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5562 struct nfsd4_lockt
*lockt
)
5564 struct file_lock
*file_lock
= NULL
;
5565 struct nfs4_lockowner
*lo
= NULL
;
5567 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5569 if (locks_in_grace(SVC_NET(rqstp
)))
5570 return nfserr_grace
;
5572 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
5573 return nfserr_inval
;
5575 if (!nfsd4_has_session(cstate
)) {
5576 status
= lookup_clientid(&lockt
->lt_clientid
, cstate
, nn
);
5581 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
5584 file_lock
= locks_alloc_lock();
5586 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5587 status
= nfserr_jukebox
;
5591 switch (lockt
->lt_type
) {
5594 file_lock
->fl_type
= F_RDLCK
;
5597 case NFS4_WRITEW_LT
:
5598 file_lock
->fl_type
= F_WRLCK
;
5601 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5602 status
= nfserr_inval
;
5606 lo
= find_lockowner_str(&lockt
->lt_clientid
, &lockt
->lt_owner
,
5609 file_lock
->fl_owner
= (fl_owner_t
)lo
;
5610 file_lock
->fl_pid
= current
->tgid
;
5611 file_lock
->fl_flags
= FL_POSIX
;
5613 file_lock
->fl_start
= lockt
->lt_offset
;
5614 file_lock
->fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
5616 nfs4_transform_lock_offset(file_lock
);
5618 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, file_lock
);
5622 if (file_lock
->fl_type
!= F_UNLCK
) {
5623 status
= nfserr_denied
;
5624 nfs4_set_lock_denied(file_lock
, &lockt
->lt_denied
);
5628 nfs4_put_stateowner(&lo
->lo_owner
);
5630 locks_free_lock(file_lock
);
5635 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5636 struct nfsd4_locku
*locku
)
5638 struct nfs4_ol_stateid
*stp
;
5639 struct file
*filp
= NULL
;
5640 struct file_lock
*file_lock
= NULL
;
5643 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5645 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5646 (long long) locku
->lu_offset
,
5647 (long long) locku
->lu_length
);
5649 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
5650 return nfserr_inval
;
5652 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
5653 &locku
->lu_stateid
, NFS4_LOCK_STID
,
5657 filp
= find_any_file(stp
->st_stid
.sc_file
);
5659 status
= nfserr_lock_range
;
5662 file_lock
= locks_alloc_lock();
5664 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5665 status
= nfserr_jukebox
;
5669 file_lock
->fl_type
= F_UNLCK
;
5670 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(stp
->st_stateowner
));
5671 file_lock
->fl_pid
= current
->tgid
;
5672 file_lock
->fl_file
= filp
;
5673 file_lock
->fl_flags
= FL_POSIX
;
5674 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
5675 file_lock
->fl_start
= locku
->lu_offset
;
5677 file_lock
->fl_end
= last_byte_offset(locku
->lu_offset
,
5679 nfs4_transform_lock_offset(file_lock
);
5681 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, NULL
);
5683 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5686 update_stateid(&stp
->st_stid
.sc_stateid
);
5687 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
5691 nfs4_put_stid(&stp
->st_stid
);
5693 nfsd4_bump_seqid(cstate
, status
);
5695 locks_free_lock(file_lock
);
5699 status
= nfserrno(err
);
5705 * true: locks held by lockowner
5706 * false: no locks held by lockowner
5709 check_for_locks(struct nfs4_file
*fp
, struct nfs4_lockowner
*lowner
)
5711 struct file_lock
*fl
;
5713 struct file
*filp
= find_any_file(fp
);
5714 struct inode
*inode
;
5715 struct file_lock_context
*flctx
;
5718 /* Any valid lock stateid should have some sort of access */
5723 inode
= file_inode(filp
);
5724 flctx
= inode
->i_flctx
;
5726 if (flctx
&& !list_empty_careful(&flctx
->flc_posix
)) {
5727 spin_lock(&flctx
->flc_lock
);
5728 list_for_each_entry(fl
, &flctx
->flc_posix
, fl_list
) {
5729 if (fl
->fl_owner
== (fl_owner_t
)lowner
) {
5734 spin_unlock(&flctx
->flc_lock
);
5741 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
5742 struct nfsd4_compound_state
*cstate
,
5743 struct nfsd4_release_lockowner
*rlockowner
)
5745 clientid_t
*clid
= &rlockowner
->rl_clientid
;
5746 struct nfs4_stateowner
*sop
;
5747 struct nfs4_lockowner
*lo
= NULL
;
5748 struct nfs4_ol_stateid
*stp
;
5749 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
5750 unsigned int hashval
= ownerstr_hashval(owner
);
5752 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5753 struct nfs4_client
*clp
;
5755 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5756 clid
->cl_boot
, clid
->cl_id
);
5758 status
= lookup_clientid(clid
, cstate
, nn
);
5763 /* Find the matching lock stateowner */
5764 spin_lock(&clp
->cl_lock
);
5765 list_for_each_entry(sop
, &clp
->cl_ownerstr_hashtbl
[hashval
],
5768 if (sop
->so_is_open_owner
|| !same_owner_str(sop
, owner
))
5771 /* see if there are still any locks associated with it */
5772 lo
= lockowner(sop
);
5773 list_for_each_entry(stp
, &sop
->so_stateids
, st_perstateowner
) {
5774 if (check_for_locks(stp
->st_stid
.sc_file
, lo
)) {
5775 status
= nfserr_locks_held
;
5776 spin_unlock(&clp
->cl_lock
);
5781 nfs4_get_stateowner(sop
);
5784 spin_unlock(&clp
->cl_lock
);
5786 release_lockowner(lo
);
5790 static inline struct nfs4_client_reclaim
*
5793 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
5797 nfs4_has_reclaimed_state(const char *name
, struct nfsd_net
*nn
)
5799 struct nfs4_client_reclaim
*crp
;
5801 crp
= nfsd4_find_reclaim_client(name
, nn
);
5802 return (crp
&& crp
->cr_clp
);
5806 * failure => all reset bets are off, nfserr_no_grace...
5808 struct nfs4_client_reclaim
*
5809 nfs4_client_to_reclaim(const char *name
, struct nfsd_net
*nn
)
5811 unsigned int strhashval
;
5812 struct nfs4_client_reclaim
*crp
;
5814 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
5815 crp
= alloc_reclaim();
5817 strhashval
= clientstr_hashval(name
);
5818 INIT_LIST_HEAD(&crp
->cr_strhash
);
5819 list_add(&crp
->cr_strhash
, &nn
->reclaim_str_hashtbl
[strhashval
]);
5820 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
5822 nn
->reclaim_str_hashtbl_size
++;
5828 nfs4_remove_reclaim_record(struct nfs4_client_reclaim
*crp
, struct nfsd_net
*nn
)
5830 list_del(&crp
->cr_strhash
);
5832 nn
->reclaim_str_hashtbl_size
--;
5836 nfs4_release_reclaim(struct nfsd_net
*nn
)
5838 struct nfs4_client_reclaim
*crp
= NULL
;
5841 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
5842 while (!list_empty(&nn
->reclaim_str_hashtbl
[i
])) {
5843 crp
= list_entry(nn
->reclaim_str_hashtbl
[i
].next
,
5844 struct nfs4_client_reclaim
, cr_strhash
);
5845 nfs4_remove_reclaim_record(crp
, nn
);
5848 WARN_ON_ONCE(nn
->reclaim_str_hashtbl_size
);
5852 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5853 struct nfs4_client_reclaim
*
5854 nfsd4_find_reclaim_client(const char *recdir
, struct nfsd_net
*nn
)
5856 unsigned int strhashval
;
5857 struct nfs4_client_reclaim
*crp
= NULL
;
5859 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir
);
5861 strhashval
= clientstr_hashval(recdir
);
5862 list_for_each_entry(crp
, &nn
->reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
5863 if (same_name(crp
->cr_recdir
, recdir
)) {
5871 * Called from OPEN. Look for clientid in reclaim list.
5874 nfs4_check_open_reclaim(clientid_t
*clid
,
5875 struct nfsd4_compound_state
*cstate
,
5876 struct nfsd_net
*nn
)
5880 /* find clientid in conf_id_hashtbl */
5881 status
= lookup_clientid(clid
, cstate
, nn
);
5883 return nfserr_reclaim_bad
;
5885 if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
, &cstate
->clp
->cl_flags
))
5886 return nfserr_no_grace
;
5888 if (nfsd4_client_record_check(cstate
->clp
))
5889 return nfserr_reclaim_bad
;
5894 #ifdef CONFIG_NFSD_FAULT_INJECTION
5896 put_client(struct nfs4_client
*clp
)
5898 atomic_dec(&clp
->cl_refcount
);
5901 static struct nfs4_client
*
5902 nfsd_find_client(struct sockaddr_storage
*addr
, size_t addr_size
)
5904 struct nfs4_client
*clp
;
5905 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
5908 if (!nfsd_netns_ready(nn
))
5911 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
5912 if (memcmp(&clp
->cl_addr
, addr
, addr_size
) == 0)
5919 nfsd_inject_print_clients(void)
5921 struct nfs4_client
*clp
;
5923 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
5925 char buf
[INET6_ADDRSTRLEN
];
5927 if (!nfsd_netns_ready(nn
))
5930 spin_lock(&nn
->client_lock
);
5931 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
5932 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
5933 pr_info("NFS Client: %s\n", buf
);
5936 spin_unlock(&nn
->client_lock
);
5942 nfsd_inject_forget_client(struct sockaddr_storage
*addr
, size_t addr_size
)
5945 struct nfs4_client
*clp
;
5946 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
5949 if (!nfsd_netns_ready(nn
))
5952 spin_lock(&nn
->client_lock
);
5953 clp
= nfsd_find_client(addr
, addr_size
);
5955 if (mark_client_expired_locked(clp
) == nfs_ok
)
5960 spin_unlock(&nn
->client_lock
);
5969 nfsd_inject_forget_clients(u64 max
)
5972 struct nfs4_client
*clp
, *next
;
5973 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
5975 LIST_HEAD(reaplist
);
5977 if (!nfsd_netns_ready(nn
))
5980 spin_lock(&nn
->client_lock
);
5981 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
5982 if (mark_client_expired_locked(clp
) == nfs_ok
) {
5983 list_add(&clp
->cl_lru
, &reaplist
);
5984 if (max
!= 0 && ++count
>= max
)
5988 spin_unlock(&nn
->client_lock
);
5990 list_for_each_entry_safe(clp
, next
, &reaplist
, cl_lru
)
5996 static void nfsd_print_count(struct nfs4_client
*clp
, unsigned int count
,
5999 char buf
[INET6_ADDRSTRLEN
];
6000 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
6001 printk(KERN_INFO
"NFS Client: %s has %u %s\n", buf
, count
, type
);
6005 nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid
*lst
,
6006 struct list_head
*collect
)
6008 struct nfs4_client
*clp
= lst
->st_stid
.sc_client
;
6009 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6015 lockdep_assert_held(&nn
->client_lock
);
6016 atomic_inc(&clp
->cl_refcount
);
6017 list_add(&lst
->st_locks
, collect
);
6020 static u64
nfsd_foreach_client_lock(struct nfs4_client
*clp
, u64 max
,
6021 struct list_head
*collect
,
6022 void (*func
)(struct nfs4_ol_stateid
*))
6024 struct nfs4_openowner
*oop
;
6025 struct nfs4_ol_stateid
*stp
, *st_next
;
6026 struct nfs4_ol_stateid
*lst
, *lst_next
;
6029 spin_lock(&clp
->cl_lock
);
6030 list_for_each_entry(oop
, &clp
->cl_openowners
, oo_perclient
) {
6031 list_for_each_entry_safe(stp
, st_next
,
6032 &oop
->oo_owner
.so_stateids
, st_perstateowner
) {
6033 list_for_each_entry_safe(lst
, lst_next
,
6034 &stp
->st_locks
, st_locks
) {
6037 nfsd_inject_add_lock_to_list(lst
,
6042 * Despite the fact that these functions deal
6043 * with 64-bit integers for "count", we must
6044 * ensure that it doesn't blow up the
6045 * clp->cl_refcount. Throw a warning if we
6046 * start to approach INT_MAX here.
6048 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6055 spin_unlock(&clp
->cl_lock
);
6061 nfsd_collect_client_locks(struct nfs4_client
*clp
, struct list_head
*collect
,
6064 return nfsd_foreach_client_lock(clp
, max
, collect
, unhash_lock_stateid
);
6068 nfsd_print_client_locks(struct nfs4_client
*clp
)
6070 u64 count
= nfsd_foreach_client_lock(clp
, 0, NULL
, NULL
);
6071 nfsd_print_count(clp
, count
, "locked files");
6076 nfsd_inject_print_locks(void)
6078 struct nfs4_client
*clp
;
6080 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6083 if (!nfsd_netns_ready(nn
))
6086 spin_lock(&nn
->client_lock
);
6087 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6088 count
+= nfsd_print_client_locks(clp
);
6089 spin_unlock(&nn
->client_lock
);
6095 nfsd_reap_locks(struct list_head
*reaplist
)
6097 struct nfs4_client
*clp
;
6098 struct nfs4_ol_stateid
*stp
, *next
;
6100 list_for_each_entry_safe(stp
, next
, reaplist
, st_locks
) {
6101 list_del_init(&stp
->st_locks
);
6102 clp
= stp
->st_stid
.sc_client
;
6103 nfs4_put_stid(&stp
->st_stid
);
6109 nfsd_inject_forget_client_locks(struct sockaddr_storage
*addr
, size_t addr_size
)
6111 unsigned int count
= 0;
6112 struct nfs4_client
*clp
;
6113 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6115 LIST_HEAD(reaplist
);
6117 if (!nfsd_netns_ready(nn
))
6120 spin_lock(&nn
->client_lock
);
6121 clp
= nfsd_find_client(addr
, addr_size
);
6123 count
= nfsd_collect_client_locks(clp
, &reaplist
, 0);
6124 spin_unlock(&nn
->client_lock
);
6125 nfsd_reap_locks(&reaplist
);
6130 nfsd_inject_forget_locks(u64 max
)
6133 struct nfs4_client
*clp
;
6134 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6136 LIST_HEAD(reaplist
);
6138 if (!nfsd_netns_ready(nn
))
6141 spin_lock(&nn
->client_lock
);
6142 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6143 count
+= nfsd_collect_client_locks(clp
, &reaplist
, max
- count
);
6144 if (max
!= 0 && count
>= max
)
6147 spin_unlock(&nn
->client_lock
);
6148 nfsd_reap_locks(&reaplist
);
6153 nfsd_foreach_client_openowner(struct nfs4_client
*clp
, u64 max
,
6154 struct list_head
*collect
,
6155 void (*func
)(struct nfs4_openowner
*))
6157 struct nfs4_openowner
*oop
, *next
;
6158 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6162 lockdep_assert_held(&nn
->client_lock
);
6164 spin_lock(&clp
->cl_lock
);
6165 list_for_each_entry_safe(oop
, next
, &clp
->cl_openowners
, oo_perclient
) {
6169 atomic_inc(&clp
->cl_refcount
);
6170 list_add(&oop
->oo_perclient
, collect
);
6175 * Despite the fact that these functions deal with
6176 * 64-bit integers for "count", we must ensure that
6177 * it doesn't blow up the clp->cl_refcount. Throw a
6178 * warning if we start to approach INT_MAX here.
6180 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6184 spin_unlock(&clp
->cl_lock
);
6190 nfsd_print_client_openowners(struct nfs4_client
*clp
)
6192 u64 count
= nfsd_foreach_client_openowner(clp
, 0, NULL
, NULL
);
6194 nfsd_print_count(clp
, count
, "openowners");
6199 nfsd_collect_client_openowners(struct nfs4_client
*clp
,
6200 struct list_head
*collect
, u64 max
)
6202 return nfsd_foreach_client_openowner(clp
, max
, collect
,
6203 unhash_openowner_locked
);
6207 nfsd_inject_print_openowners(void)
6209 struct nfs4_client
*clp
;
6211 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6214 if (!nfsd_netns_ready(nn
))
6217 spin_lock(&nn
->client_lock
);
6218 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6219 count
+= nfsd_print_client_openowners(clp
);
6220 spin_unlock(&nn
->client_lock
);
6226 nfsd_reap_openowners(struct list_head
*reaplist
)
6228 struct nfs4_client
*clp
;
6229 struct nfs4_openowner
*oop
, *next
;
6231 list_for_each_entry_safe(oop
, next
, reaplist
, oo_perclient
) {
6232 list_del_init(&oop
->oo_perclient
);
6233 clp
= oop
->oo_owner
.so_client
;
6234 release_openowner(oop
);
6240 nfsd_inject_forget_client_openowners(struct sockaddr_storage
*addr
,
6243 unsigned int count
= 0;
6244 struct nfs4_client
*clp
;
6245 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6247 LIST_HEAD(reaplist
);
6249 if (!nfsd_netns_ready(nn
))
6252 spin_lock(&nn
->client_lock
);
6253 clp
= nfsd_find_client(addr
, addr_size
);
6255 count
= nfsd_collect_client_openowners(clp
, &reaplist
, 0);
6256 spin_unlock(&nn
->client_lock
);
6257 nfsd_reap_openowners(&reaplist
);
6262 nfsd_inject_forget_openowners(u64 max
)
6265 struct nfs4_client
*clp
;
6266 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6268 LIST_HEAD(reaplist
);
6270 if (!nfsd_netns_ready(nn
))
6273 spin_lock(&nn
->client_lock
);
6274 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6275 count
+= nfsd_collect_client_openowners(clp
, &reaplist
,
6277 if (max
!= 0 && count
>= max
)
6280 spin_unlock(&nn
->client_lock
);
6281 nfsd_reap_openowners(&reaplist
);
6285 static u64
nfsd_find_all_delegations(struct nfs4_client
*clp
, u64 max
,
6286 struct list_head
*victims
)
6288 struct nfs4_delegation
*dp
, *next
;
6289 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6293 lockdep_assert_held(&nn
->client_lock
);
6295 spin_lock(&state_lock
);
6296 list_for_each_entry_safe(dp
, next
, &clp
->cl_delegations
, dl_perclnt
) {
6299 * It's not safe to mess with delegations that have a
6300 * non-zero dl_time. They might have already been broken
6301 * and could be processed by the laundromat outside of
6302 * the state_lock. Just leave them be.
6304 if (dp
->dl_time
!= 0)
6307 atomic_inc(&clp
->cl_refcount
);
6308 unhash_delegation_locked(dp
);
6309 list_add(&dp
->dl_recall_lru
, victims
);
6313 * Despite the fact that these functions deal with
6314 * 64-bit integers for "count", we must ensure that
6315 * it doesn't blow up the clp->cl_refcount. Throw a
6316 * warning if we start to approach INT_MAX here.
6318 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6322 spin_unlock(&state_lock
);
6327 nfsd_print_client_delegations(struct nfs4_client
*clp
)
6329 u64 count
= nfsd_find_all_delegations(clp
, 0, NULL
);
6331 nfsd_print_count(clp
, count
, "delegations");
6336 nfsd_inject_print_delegations(void)
6338 struct nfs4_client
*clp
;
6340 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6343 if (!nfsd_netns_ready(nn
))
6346 spin_lock(&nn
->client_lock
);
6347 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6348 count
+= nfsd_print_client_delegations(clp
);
6349 spin_unlock(&nn
->client_lock
);
6355 nfsd_forget_delegations(struct list_head
*reaplist
)
6357 struct nfs4_client
*clp
;
6358 struct nfs4_delegation
*dp
, *next
;
6360 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
6361 list_del_init(&dp
->dl_recall_lru
);
6362 clp
= dp
->dl_stid
.sc_client
;
6363 revoke_delegation(dp
);
6369 nfsd_inject_forget_client_delegations(struct sockaddr_storage
*addr
,
6373 struct nfs4_client
*clp
;
6374 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6376 LIST_HEAD(reaplist
);
6378 if (!nfsd_netns_ready(nn
))
6381 spin_lock(&nn
->client_lock
);
6382 clp
= nfsd_find_client(addr
, addr_size
);
6384 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
6385 spin_unlock(&nn
->client_lock
);
6387 nfsd_forget_delegations(&reaplist
);
6392 nfsd_inject_forget_delegations(u64 max
)
6395 struct nfs4_client
*clp
;
6396 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6398 LIST_HEAD(reaplist
);
6400 if (!nfsd_netns_ready(nn
))
6403 spin_lock(&nn
->client_lock
);
6404 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6405 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
6406 if (max
!= 0 && count
>= max
)
6409 spin_unlock(&nn
->client_lock
);
6410 nfsd_forget_delegations(&reaplist
);
6415 nfsd_recall_delegations(struct list_head
*reaplist
)
6417 struct nfs4_client
*clp
;
6418 struct nfs4_delegation
*dp
, *next
;
6420 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
6421 list_del_init(&dp
->dl_recall_lru
);
6422 clp
= dp
->dl_stid
.sc_client
;
6424 * We skipped all entries that had a zero dl_time before,
6425 * so we can now reset the dl_time back to 0. If a delegation
6426 * break comes in now, then it won't make any difference since
6427 * we're recalling it either way.
6429 spin_lock(&state_lock
);
6431 spin_unlock(&state_lock
);
6432 nfsd_break_one_deleg(dp
);
6438 nfsd_inject_recall_client_delegations(struct sockaddr_storage
*addr
,
6442 struct nfs4_client
*clp
;
6443 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6445 LIST_HEAD(reaplist
);
6447 if (!nfsd_netns_ready(nn
))
6450 spin_lock(&nn
->client_lock
);
6451 clp
= nfsd_find_client(addr
, addr_size
);
6453 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
6454 spin_unlock(&nn
->client_lock
);
6456 nfsd_recall_delegations(&reaplist
);
6461 nfsd_inject_recall_delegations(u64 max
)
6464 struct nfs4_client
*clp
, *next
;
6465 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6467 LIST_HEAD(reaplist
);
6469 if (!nfsd_netns_ready(nn
))
6472 spin_lock(&nn
->client_lock
);
6473 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
6474 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
6475 if (max
!= 0 && ++count
>= max
)
6478 spin_unlock(&nn
->client_lock
);
6479 nfsd_recall_delegations(&reaplist
);
6482 #endif /* CONFIG_NFSD_FAULT_INJECTION */
6485 * Since the lifetime of a delegation isn't limited to that of an open, a
6486 * client may quite reasonably hang on to a delegation as long as it has
6487 * the inode cached. This becomes an obvious problem the first time a
6488 * client's inode cache approaches the size of the server's total memory.
6490 * For now we avoid this problem by imposing a hard limit on the number
6491 * of delegations, which varies according to the server's memory size.
6494 set_max_delegations(void)
6497 * Allow at most 4 delegations per megabyte of RAM. Quick
6498 * estimates suggest that in the worst case (where every delegation
6499 * is for a different inode), a delegation could take about 1.5K,
6500 * giving a worst case usage of about 6% of memory.
6502 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
6505 static int nfs4_state_create_net(struct net
*net
)
6507 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6510 nn
->conf_id_hashtbl
= kmalloc(sizeof(struct list_head
) *
6511 CLIENT_HASH_SIZE
, GFP_KERNEL
);
6512 if (!nn
->conf_id_hashtbl
)
6514 nn
->unconf_id_hashtbl
= kmalloc(sizeof(struct list_head
) *
6515 CLIENT_HASH_SIZE
, GFP_KERNEL
);
6516 if (!nn
->unconf_id_hashtbl
)
6518 nn
->sessionid_hashtbl
= kmalloc(sizeof(struct list_head
) *
6519 SESSION_HASH_SIZE
, GFP_KERNEL
);
6520 if (!nn
->sessionid_hashtbl
)
6523 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6524 INIT_LIST_HEAD(&nn
->conf_id_hashtbl
[i
]);
6525 INIT_LIST_HEAD(&nn
->unconf_id_hashtbl
[i
]);
6527 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
6528 INIT_LIST_HEAD(&nn
->sessionid_hashtbl
[i
]);
6529 nn
->conf_name_tree
= RB_ROOT
;
6530 nn
->unconf_name_tree
= RB_ROOT
;
6531 INIT_LIST_HEAD(&nn
->client_lru
);
6532 INIT_LIST_HEAD(&nn
->close_lru
);
6533 INIT_LIST_HEAD(&nn
->del_recall_lru
);
6534 spin_lock_init(&nn
->client_lock
);
6536 INIT_DELAYED_WORK(&nn
->laundromat_work
, laundromat_main
);
6542 kfree(nn
->unconf_id_hashtbl
);
6544 kfree(nn
->conf_id_hashtbl
);
6550 nfs4_state_destroy_net(struct net
*net
)
6553 struct nfs4_client
*clp
= NULL
;
6554 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6556 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6557 while (!list_empty(&nn
->conf_id_hashtbl
[i
])) {
6558 clp
= list_entry(nn
->conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
6559 destroy_client(clp
);
6563 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6564 while (!list_empty(&nn
->unconf_id_hashtbl
[i
])) {
6565 clp
= list_entry(nn
->unconf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
6566 destroy_client(clp
);
6570 kfree(nn
->sessionid_hashtbl
);
6571 kfree(nn
->unconf_id_hashtbl
);
6572 kfree(nn
->conf_id_hashtbl
);
6577 nfs4_state_start_net(struct net
*net
)
6579 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6582 ret
= nfs4_state_create_net(net
);
6585 nn
->boot_time
= get_seconds();
6586 nn
->grace_ended
= false;
6587 locks_start_grace(net
, &nn
->nfsd4_manager
);
6588 nfsd4_client_tracking_init(net
);
6589 printk(KERN_INFO
"NFSD: starting %ld-second grace period (net %p)\n",
6590 nn
->nfsd4_grace
, net
);
6591 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, nn
->nfsd4_grace
* HZ
);
6595 /* initialization to perform when the nfsd service is started: */
6598 nfs4_state_start(void)
6602 ret
= set_callback_cred();
6605 laundry_wq
= create_singlethread_workqueue("nfsd4");
6606 if (laundry_wq
== NULL
) {
6610 ret
= nfsd4_create_callback_queue();
6612 goto out_free_laundry
;
6614 set_max_delegations();
6619 destroy_workqueue(laundry_wq
);
6625 nfs4_state_shutdown_net(struct net
*net
)
6627 struct nfs4_delegation
*dp
= NULL
;
6628 struct list_head
*pos
, *next
, reaplist
;
6629 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6631 cancel_delayed_work_sync(&nn
->laundromat_work
);
6632 locks_end_grace(&nn
->nfsd4_manager
);
6634 INIT_LIST_HEAD(&reaplist
);
6635 spin_lock(&state_lock
);
6636 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
6637 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
6638 unhash_delegation_locked(dp
);
6639 list_add(&dp
->dl_recall_lru
, &reaplist
);
6641 spin_unlock(&state_lock
);
6642 list_for_each_safe(pos
, next
, &reaplist
) {
6643 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
6644 list_del_init(&dp
->dl_recall_lru
);
6645 put_clnt_odstate(dp
->dl_clnt_odstate
);
6646 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
6647 nfs4_put_stid(&dp
->dl_stid
);
6650 nfsd4_client_tracking_exit(net
);
6651 nfs4_state_destroy_net(net
);
6655 nfs4_state_shutdown(void)
6657 destroy_workqueue(laundry_wq
);
6658 nfsd4_destroy_callback_queue();
6662 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
6664 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
6665 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
6669 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
6671 if (cstate
->minorversion
) {
6672 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
6673 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
6678 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
6680 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
6684 * functions to set current state id
6687 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
6689 put_stateid(cstate
, &odp
->od_stateid
);
6693 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
6695 put_stateid(cstate
, &open
->op_stateid
);
6699 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
6701 put_stateid(cstate
, &close
->cl_stateid
);
6705 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
6707 put_stateid(cstate
, &lock
->lk_resp_stateid
);
6711 * functions to consume current state id
6715 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
6717 get_stateid(cstate
, &odp
->od_stateid
);
6721 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
6723 get_stateid(cstate
, &drp
->dr_stateid
);
6727 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
6729 get_stateid(cstate
, &fsp
->fr_stateid
);
6733 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
6735 get_stateid(cstate
, &setattr
->sa_stateid
);
6739 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
6741 get_stateid(cstate
, &close
->cl_stateid
);
6745 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
6747 get_stateid(cstate
, &locku
->lu_stateid
);
6751 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
6753 get_stateid(cstate
, &read
->rd_stateid
);
6757 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
6759 get_stateid(cstate
, &write
->wr_stateid
);