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 const struct nfsd4_callback_ops nfsd4_cb_recall_ops
;
102 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops
;
104 static bool is_session_dead(struct nfsd4_session
*ses
)
106 return ses
->se_flags
& NFS4_SESSION_DEAD
;
109 static __be32
mark_session_dead_locked(struct nfsd4_session
*ses
, int ref_held_by_me
)
111 if (atomic_read(&ses
->se_ref
) > ref_held_by_me
)
112 return nfserr_jukebox
;
113 ses
->se_flags
|= NFS4_SESSION_DEAD
;
117 static bool is_client_expired(struct nfs4_client
*clp
)
119 return clp
->cl_time
== 0;
122 static __be32
get_client_locked(struct nfs4_client
*clp
)
124 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
126 lockdep_assert_held(&nn
->client_lock
);
128 if (is_client_expired(clp
))
129 return nfserr_expired
;
130 atomic_inc(&clp
->cl_refcount
);
134 /* must be called under the client_lock */
136 renew_client_locked(struct nfs4_client
*clp
)
138 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
140 if (is_client_expired(clp
)) {
142 printk("%s: client (clientid %08x/%08x) already expired\n",
144 clp
->cl_clientid
.cl_boot
,
145 clp
->cl_clientid
.cl_id
);
149 dprintk("renewing client (clientid %08x/%08x)\n",
150 clp
->cl_clientid
.cl_boot
,
151 clp
->cl_clientid
.cl_id
);
152 list_move_tail(&clp
->cl_lru
, &nn
->client_lru
);
153 clp
->cl_time
= get_seconds();
156 static void put_client_renew_locked(struct nfs4_client
*clp
)
158 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
160 lockdep_assert_held(&nn
->client_lock
);
162 if (!atomic_dec_and_test(&clp
->cl_refcount
))
164 if (!is_client_expired(clp
))
165 renew_client_locked(clp
);
168 static void put_client_renew(struct nfs4_client
*clp
)
170 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
172 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &nn
->client_lock
))
174 if (!is_client_expired(clp
))
175 renew_client_locked(clp
);
176 spin_unlock(&nn
->client_lock
);
179 static __be32
nfsd4_get_session_locked(struct nfsd4_session
*ses
)
183 if (is_session_dead(ses
))
184 return nfserr_badsession
;
185 status
= get_client_locked(ses
->se_client
);
188 atomic_inc(&ses
->se_ref
);
192 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
194 struct nfs4_client
*clp
= ses
->se_client
;
195 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
197 lockdep_assert_held(&nn
->client_lock
);
199 if (atomic_dec_and_test(&ses
->se_ref
) && is_session_dead(ses
))
201 put_client_renew_locked(clp
);
204 static void nfsd4_put_session(struct nfsd4_session
*ses
)
206 struct nfs4_client
*clp
= ses
->se_client
;
207 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
209 spin_lock(&nn
->client_lock
);
210 nfsd4_put_session_locked(ses
);
211 spin_unlock(&nn
->client_lock
);
214 static struct nfsd4_blocked_lock
*
215 find_blocked_lock(struct nfs4_lockowner
*lo
, struct knfsd_fh
*fh
,
218 struct nfsd4_blocked_lock
*cur
, *found
= NULL
;
220 spin_lock(&nn
->blocked_locks_lock
);
221 list_for_each_entry(cur
, &lo
->lo_blocked
, nbl_list
) {
222 if (fh_match(fh
, &cur
->nbl_fh
)) {
223 list_del_init(&cur
->nbl_list
);
224 list_del_init(&cur
->nbl_lru
);
229 spin_unlock(&nn
->blocked_locks_lock
);
231 posix_unblock_lock(&found
->nbl_lock
);
235 static struct nfsd4_blocked_lock
*
236 find_or_allocate_block(struct nfs4_lockowner
*lo
, struct knfsd_fh
*fh
,
239 struct nfsd4_blocked_lock
*nbl
;
241 nbl
= find_blocked_lock(lo
, fh
, nn
);
243 nbl
= kmalloc(sizeof(*nbl
), GFP_KERNEL
);
245 fh_copy_shallow(&nbl
->nbl_fh
, fh
);
246 locks_init_lock(&nbl
->nbl_lock
);
247 nfsd4_init_cb(&nbl
->nbl_cb
, lo
->lo_owner
.so_client
,
248 &nfsd4_cb_notify_lock_ops
,
249 NFSPROC4_CLNT_CB_NOTIFY_LOCK
);
256 free_blocked_lock(struct nfsd4_blocked_lock
*nbl
)
258 locks_release_private(&nbl
->nbl_lock
);
263 nfsd4_cb_notify_lock_done(struct nfsd4_callback
*cb
, struct rpc_task
*task
)
266 * Since this is just an optimization, we don't try very hard if it
267 * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
268 * just quit trying on anything else.
270 switch (task
->tk_status
) {
272 rpc_delay(task
, 1 * HZ
);
280 nfsd4_cb_notify_lock_release(struct nfsd4_callback
*cb
)
282 struct nfsd4_blocked_lock
*nbl
= container_of(cb
,
283 struct nfsd4_blocked_lock
, nbl_cb
);
285 free_blocked_lock(nbl
);
288 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops
= {
289 .done
= nfsd4_cb_notify_lock_done
,
290 .release
= nfsd4_cb_notify_lock_release
,
293 static inline struct nfs4_stateowner
*
294 nfs4_get_stateowner(struct nfs4_stateowner
*sop
)
296 atomic_inc(&sop
->so_count
);
301 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
)
303 return (sop
->so_owner
.len
== owner
->len
) &&
304 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
);
307 static struct nfs4_openowner
*
308 find_openstateowner_str_locked(unsigned int hashval
, struct nfsd4_open
*open
,
309 struct nfs4_client
*clp
)
311 struct nfs4_stateowner
*so
;
313 lockdep_assert_held(&clp
->cl_lock
);
315 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[hashval
],
317 if (!so
->so_is_open_owner
)
319 if (same_owner_str(so
, &open
->op_owner
))
320 return openowner(nfs4_get_stateowner(so
));
325 static struct nfs4_openowner
*
326 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
,
327 struct nfs4_client
*clp
)
329 struct nfs4_openowner
*oo
;
331 spin_lock(&clp
->cl_lock
);
332 oo
= find_openstateowner_str_locked(hashval
, open
, clp
);
333 spin_unlock(&clp
->cl_lock
);
338 opaque_hashval(const void *ptr
, int nbytes
)
340 unsigned char *cptr
= (unsigned char *) ptr
;
350 static void nfsd4_free_file_rcu(struct rcu_head
*rcu
)
352 struct nfs4_file
*fp
= container_of(rcu
, struct nfs4_file
, fi_rcu
);
354 kmem_cache_free(file_slab
, fp
);
358 put_nfs4_file(struct nfs4_file
*fi
)
360 might_lock(&state_lock
);
362 if (atomic_dec_and_lock(&fi
->fi_ref
, &state_lock
)) {
363 hlist_del_rcu(&fi
->fi_hash
);
364 spin_unlock(&state_lock
);
365 WARN_ON_ONCE(!list_empty(&fi
->fi_clnt_odstate
));
366 WARN_ON_ONCE(!list_empty(&fi
->fi_delegations
));
367 call_rcu(&fi
->fi_rcu
, nfsd4_free_file_rcu
);
372 __nfs4_get_fd(struct nfs4_file
*f
, int oflag
)
374 if (f
->fi_fds
[oflag
])
375 return get_file(f
->fi_fds
[oflag
]);
380 find_writeable_file_locked(struct nfs4_file
*f
)
384 lockdep_assert_held(&f
->fi_lock
);
386 ret
= __nfs4_get_fd(f
, O_WRONLY
);
388 ret
= __nfs4_get_fd(f
, O_RDWR
);
393 find_writeable_file(struct nfs4_file
*f
)
397 spin_lock(&f
->fi_lock
);
398 ret
= find_writeable_file_locked(f
);
399 spin_unlock(&f
->fi_lock
);
404 static struct file
*find_readable_file_locked(struct nfs4_file
*f
)
408 lockdep_assert_held(&f
->fi_lock
);
410 ret
= __nfs4_get_fd(f
, O_RDONLY
);
412 ret
= __nfs4_get_fd(f
, O_RDWR
);
417 find_readable_file(struct nfs4_file
*f
)
421 spin_lock(&f
->fi_lock
);
422 ret
= find_readable_file_locked(f
);
423 spin_unlock(&f
->fi_lock
);
429 find_any_file(struct nfs4_file
*f
)
433 spin_lock(&f
->fi_lock
);
434 ret
= __nfs4_get_fd(f
, O_RDWR
);
436 ret
= __nfs4_get_fd(f
, O_WRONLY
);
438 ret
= __nfs4_get_fd(f
, O_RDONLY
);
440 spin_unlock(&f
->fi_lock
);
444 static atomic_long_t num_delegations
;
445 unsigned long max_delegations
;
448 * Open owner state (share locks)
451 /* hash tables for lock and open owners */
452 #define OWNER_HASH_BITS 8
453 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
454 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
456 static unsigned int ownerstr_hashval(struct xdr_netobj
*ownername
)
460 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
461 return ret
& OWNER_HASH_MASK
;
464 /* hash table for nfs4_file */
465 #define FILE_HASH_BITS 8
466 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
468 static unsigned int nfsd_fh_hashval(struct knfsd_fh
*fh
)
470 return jhash2(fh
->fh_base
.fh_pad
, XDR_QUADLEN(fh
->fh_size
), 0);
473 static unsigned int file_hashval(struct knfsd_fh
*fh
)
475 return nfsd_fh_hashval(fh
) & (FILE_HASH_SIZE
- 1);
478 static struct hlist_head file_hashtbl
[FILE_HASH_SIZE
];
481 __nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
483 lockdep_assert_held(&fp
->fi_lock
);
485 if (access
& NFS4_SHARE_ACCESS_WRITE
)
486 atomic_inc(&fp
->fi_access
[O_WRONLY
]);
487 if (access
& NFS4_SHARE_ACCESS_READ
)
488 atomic_inc(&fp
->fi_access
[O_RDONLY
]);
492 nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
494 lockdep_assert_held(&fp
->fi_lock
);
496 /* Does this access mode make sense? */
497 if (access
& ~NFS4_SHARE_ACCESS_BOTH
)
500 /* Does it conflict with a deny mode already set? */
501 if ((access
& fp
->fi_share_deny
) != 0)
502 return nfserr_share_denied
;
504 __nfs4_file_get_access(fp
, access
);
508 static __be32
nfs4_file_check_deny(struct nfs4_file
*fp
, u32 deny
)
510 /* Common case is that there is no deny mode. */
512 /* Does this deny mode make sense? */
513 if (deny
& ~NFS4_SHARE_DENY_BOTH
)
516 if ((deny
& NFS4_SHARE_DENY_READ
) &&
517 atomic_read(&fp
->fi_access
[O_RDONLY
]))
518 return nfserr_share_denied
;
520 if ((deny
& NFS4_SHARE_DENY_WRITE
) &&
521 atomic_read(&fp
->fi_access
[O_WRONLY
]))
522 return nfserr_share_denied
;
527 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
529 might_lock(&fp
->fi_lock
);
531 if (atomic_dec_and_lock(&fp
->fi_access
[oflag
], &fp
->fi_lock
)) {
532 struct file
*f1
= NULL
;
533 struct file
*f2
= NULL
;
535 swap(f1
, fp
->fi_fds
[oflag
]);
536 if (atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
537 swap(f2
, fp
->fi_fds
[O_RDWR
]);
538 spin_unlock(&fp
->fi_lock
);
546 static void nfs4_file_put_access(struct nfs4_file
*fp
, u32 access
)
548 WARN_ON_ONCE(access
& ~NFS4_SHARE_ACCESS_BOTH
);
550 if (access
& NFS4_SHARE_ACCESS_WRITE
)
551 __nfs4_file_put_access(fp
, O_WRONLY
);
552 if (access
& NFS4_SHARE_ACCESS_READ
)
553 __nfs4_file_put_access(fp
, O_RDONLY
);
557 * Allocate a new open/delegation state counter. This is needed for
558 * pNFS for proper return on close semantics.
560 * Note that we only allocate it for pNFS-enabled exports, otherwise
561 * all pointers to struct nfs4_clnt_odstate are always NULL.
563 static struct nfs4_clnt_odstate
*
564 alloc_clnt_odstate(struct nfs4_client
*clp
)
566 struct nfs4_clnt_odstate
*co
;
568 co
= kmem_cache_zalloc(odstate_slab
, GFP_KERNEL
);
571 atomic_set(&co
->co_odcount
, 1);
577 hash_clnt_odstate_locked(struct nfs4_clnt_odstate
*co
)
579 struct nfs4_file
*fp
= co
->co_file
;
581 lockdep_assert_held(&fp
->fi_lock
);
582 list_add(&co
->co_perfile
, &fp
->fi_clnt_odstate
);
586 get_clnt_odstate(struct nfs4_clnt_odstate
*co
)
589 atomic_inc(&co
->co_odcount
);
593 put_clnt_odstate(struct nfs4_clnt_odstate
*co
)
595 struct nfs4_file
*fp
;
601 if (atomic_dec_and_lock(&co
->co_odcount
, &fp
->fi_lock
)) {
602 list_del(&co
->co_perfile
);
603 spin_unlock(&fp
->fi_lock
);
605 nfsd4_return_all_file_layouts(co
->co_client
, fp
);
606 kmem_cache_free(odstate_slab
, co
);
610 static struct nfs4_clnt_odstate
*
611 find_or_hash_clnt_odstate(struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*new)
613 struct nfs4_clnt_odstate
*co
;
614 struct nfs4_client
*cl
;
621 spin_lock(&fp
->fi_lock
);
622 list_for_each_entry(co
, &fp
->fi_clnt_odstate
, co_perfile
) {
623 if (co
->co_client
== cl
) {
624 get_clnt_odstate(co
);
630 hash_clnt_odstate_locked(new);
632 spin_unlock(&fp
->fi_lock
);
636 struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
,
637 void (*sc_free
)(struct nfs4_stid
*))
639 struct nfs4_stid
*stid
;
642 stid
= kmem_cache_zalloc(slab
, GFP_KERNEL
);
646 idr_preload(GFP_KERNEL
);
647 spin_lock(&cl
->cl_lock
);
648 new_id
= idr_alloc_cyclic(&cl
->cl_stateids
, stid
, 0, 0, GFP_NOWAIT
);
649 spin_unlock(&cl
->cl_lock
);
654 stid
->sc_free
= sc_free
;
655 stid
->sc_client
= cl
;
656 stid
->sc_stateid
.si_opaque
.so_id
= new_id
;
657 stid
->sc_stateid
.si_opaque
.so_clid
= cl
->cl_clientid
;
658 /* Will be incremented before return to client: */
659 atomic_set(&stid
->sc_count
, 1);
660 spin_lock_init(&stid
->sc_lock
);
663 * It shouldn't be a problem to reuse an opaque stateid value.
664 * I don't think it is for 4.1. But with 4.0 I worry that, for
665 * example, a stray write retransmission could be accepted by
666 * the server when it should have been rejected. Therefore,
667 * adopt a trick from the sctp code to attempt to maximize the
668 * amount of time until an id is reused, by ensuring they always
669 * "increase" (mod INT_MAX):
673 kmem_cache_free(slab
, stid
);
677 static struct nfs4_ol_stateid
* nfs4_alloc_open_stateid(struct nfs4_client
*clp
)
679 struct nfs4_stid
*stid
;
681 stid
= nfs4_alloc_stid(clp
, stateid_slab
, nfs4_free_ol_stateid
);
685 return openlockstateid(stid
);
688 static void nfs4_free_deleg(struct nfs4_stid
*stid
)
690 kmem_cache_free(deleg_slab
, stid
);
691 atomic_long_dec(&num_delegations
);
695 * When we recall a delegation, we should be careful not to hand it
696 * out again straight away.
697 * To ensure this we keep a pair of bloom filters ('new' and 'old')
698 * in which the filehandles of recalled delegations are "stored".
699 * If a filehandle appear in either filter, a delegation is blocked.
700 * When a delegation is recalled, the filehandle is stored in the "new"
702 * Every 30 seconds we swap the filters and clear the "new" one,
703 * unless both are empty of course.
705 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
706 * low 3 bytes as hash-table indices.
708 * 'blocked_delegations_lock', which is always taken in block_delegations(),
709 * is used to manage concurrent access. Testing does not need the lock
710 * except when swapping the two filters.
712 static DEFINE_SPINLOCK(blocked_delegations_lock
);
713 static struct bloom_pair
{
714 int entries
, old_entries
;
716 int new; /* index into 'set' */
717 DECLARE_BITMAP(set
[2], 256);
718 } blocked_delegations
;
720 static int delegation_blocked(struct knfsd_fh
*fh
)
723 struct bloom_pair
*bd
= &blocked_delegations
;
725 if (bd
->entries
== 0)
727 if (seconds_since_boot() - bd
->swap_time
> 30) {
728 spin_lock(&blocked_delegations_lock
);
729 if (seconds_since_boot() - bd
->swap_time
> 30) {
730 bd
->entries
-= bd
->old_entries
;
731 bd
->old_entries
= bd
->entries
;
732 memset(bd
->set
[bd
->new], 0,
735 bd
->swap_time
= seconds_since_boot();
737 spin_unlock(&blocked_delegations_lock
);
739 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
740 if (test_bit(hash
&255, bd
->set
[0]) &&
741 test_bit((hash
>>8)&255, bd
->set
[0]) &&
742 test_bit((hash
>>16)&255, bd
->set
[0]))
745 if (test_bit(hash
&255, bd
->set
[1]) &&
746 test_bit((hash
>>8)&255, bd
->set
[1]) &&
747 test_bit((hash
>>16)&255, bd
->set
[1]))
753 static void block_delegations(struct knfsd_fh
*fh
)
756 struct bloom_pair
*bd
= &blocked_delegations
;
758 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
760 spin_lock(&blocked_delegations_lock
);
761 __set_bit(hash
&255, bd
->set
[bd
->new]);
762 __set_bit((hash
>>8)&255, bd
->set
[bd
->new]);
763 __set_bit((hash
>>16)&255, bd
->set
[bd
->new]);
764 if (bd
->entries
== 0)
765 bd
->swap_time
= seconds_since_boot();
767 spin_unlock(&blocked_delegations_lock
);
770 static struct nfs4_delegation
*
771 alloc_init_deleg(struct nfs4_client
*clp
, struct svc_fh
*current_fh
,
772 struct nfs4_clnt_odstate
*odstate
)
774 struct nfs4_delegation
*dp
;
777 dprintk("NFSD alloc_init_deleg\n");
778 n
= atomic_long_inc_return(&num_delegations
);
779 if (n
< 0 || n
> max_delegations
)
781 if (delegation_blocked(¤t_fh
->fh_handle
))
783 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
, nfs4_free_deleg
));
788 * delegation seqid's are never incremented. The 4.1 special
789 * meaning of seqid 0 isn't meaningful, really, but let's avoid
790 * 0 anyway just for consistency and use 1:
792 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
793 INIT_LIST_HEAD(&dp
->dl_perfile
);
794 INIT_LIST_HEAD(&dp
->dl_perclnt
);
795 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
796 dp
->dl_clnt_odstate
= odstate
;
797 get_clnt_odstate(odstate
);
798 dp
->dl_type
= NFS4_OPEN_DELEGATE_READ
;
800 nfsd4_init_cb(&dp
->dl_recall
, dp
->dl_stid
.sc_client
,
801 &nfsd4_cb_recall_ops
, NFSPROC4_CLNT_CB_RECALL
);
804 atomic_long_dec(&num_delegations
);
809 nfs4_put_stid(struct nfs4_stid
*s
)
811 struct nfs4_file
*fp
= s
->sc_file
;
812 struct nfs4_client
*clp
= s
->sc_client
;
814 might_lock(&clp
->cl_lock
);
816 if (!atomic_dec_and_lock(&s
->sc_count
, &clp
->cl_lock
)) {
817 wake_up_all(&close_wq
);
820 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
821 spin_unlock(&clp
->cl_lock
);
828 nfs4_inc_and_copy_stateid(stateid_t
*dst
, struct nfs4_stid
*stid
)
830 stateid_t
*src
= &stid
->sc_stateid
;
832 spin_lock(&stid
->sc_lock
);
833 if (unlikely(++src
->si_generation
== 0))
834 src
->si_generation
= 1;
835 memcpy(dst
, src
, sizeof(*dst
));
836 spin_unlock(&stid
->sc_lock
);
839 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
841 struct file
*filp
= NULL
;
843 spin_lock(&fp
->fi_lock
);
844 if (fp
->fi_deleg_file
&& --fp
->fi_delegees
== 0)
845 swap(filp
, fp
->fi_deleg_file
);
846 spin_unlock(&fp
->fi_lock
);
849 vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&fp
);
854 void nfs4_unhash_stid(struct nfs4_stid
*s
)
860 * nfs4_get_existing_delegation - Discover if this delegation already exists
861 * @clp: a pointer to the nfs4_client we're granting a delegation to
862 * @fp: a pointer to the nfs4_file we're granting a delegation on
865 * On success: NULL if an existing delegation was not found.
867 * On error: -EAGAIN if one was previously granted to this nfs4_client
868 * for this nfs4_file.
873 nfs4_get_existing_delegation(struct nfs4_client
*clp
, struct nfs4_file
*fp
)
875 struct nfs4_delegation
*searchdp
= NULL
;
876 struct nfs4_client
*searchclp
= NULL
;
878 lockdep_assert_held(&state_lock
);
879 lockdep_assert_held(&fp
->fi_lock
);
881 list_for_each_entry(searchdp
, &fp
->fi_delegations
, dl_perfile
) {
882 searchclp
= searchdp
->dl_stid
.sc_client
;
883 if (clp
== searchclp
) {
891 * hash_delegation_locked - Add a delegation to the appropriate lists
892 * @dp: a pointer to the nfs4_delegation we are adding.
893 * @fp: a pointer to the nfs4_file we're granting a delegation on
896 * On success: NULL if the delegation was successfully hashed.
898 * On error: -EAGAIN if one was previously granted to this
899 * nfs4_client for this nfs4_file. Delegation is not hashed.
904 hash_delegation_locked(struct nfs4_delegation
*dp
, struct nfs4_file
*fp
)
907 struct nfs4_client
*clp
= dp
->dl_stid
.sc_client
;
909 lockdep_assert_held(&state_lock
);
910 lockdep_assert_held(&fp
->fi_lock
);
912 status
= nfs4_get_existing_delegation(clp
, fp
);
916 atomic_inc(&dp
->dl_stid
.sc_count
);
917 dp
->dl_stid
.sc_type
= NFS4_DELEG_STID
;
918 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
919 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
924 unhash_delegation_locked(struct nfs4_delegation
*dp
)
926 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
928 lockdep_assert_held(&state_lock
);
930 if (list_empty(&dp
->dl_perfile
))
933 dp
->dl_stid
.sc_type
= NFS4_CLOSED_DELEG_STID
;
934 /* Ensure that deleg break won't try to requeue it */
936 spin_lock(&fp
->fi_lock
);
937 list_del_init(&dp
->dl_perclnt
);
938 list_del_init(&dp
->dl_recall_lru
);
939 list_del_init(&dp
->dl_perfile
);
940 spin_unlock(&fp
->fi_lock
);
944 static void destroy_delegation(struct nfs4_delegation
*dp
)
948 spin_lock(&state_lock
);
949 unhashed
= unhash_delegation_locked(dp
);
950 spin_unlock(&state_lock
);
952 put_clnt_odstate(dp
->dl_clnt_odstate
);
953 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
954 nfs4_put_stid(&dp
->dl_stid
);
958 static void revoke_delegation(struct nfs4_delegation
*dp
)
960 struct nfs4_client
*clp
= dp
->dl_stid
.sc_client
;
962 WARN_ON(!list_empty(&dp
->dl_recall_lru
));
964 put_clnt_odstate(dp
->dl_clnt_odstate
);
965 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
967 if (clp
->cl_minorversion
== 0)
968 nfs4_put_stid(&dp
->dl_stid
);
970 dp
->dl_stid
.sc_type
= NFS4_REVOKED_DELEG_STID
;
971 spin_lock(&clp
->cl_lock
);
972 list_add(&dp
->dl_recall_lru
, &clp
->cl_revoked
);
973 spin_unlock(&clp
->cl_lock
);
981 static unsigned int clientid_hashval(u32 id
)
983 return id
& CLIENT_HASH_MASK
;
986 static unsigned int clientstr_hashval(const char *name
)
988 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
992 * We store the NONE, READ, WRITE, and BOTH bits separately in the
993 * st_{access,deny}_bmap field of the stateid, in order to track not
994 * only what share bits are currently in force, but also what
995 * combinations of share bits previous opens have used. This allows us
996 * to enforce the recommendation of rfc 3530 14.2.19 that the server
997 * return an error if the client attempt to downgrade to a combination
998 * of share bits not explicable by closing some of its previous opens.
1000 * XXX: This enforcement is actually incomplete, since we don't keep
1001 * track of access/deny bit combinations; so, e.g., we allow:
1003 * OPEN allow read, deny write
1004 * OPEN allow both, deny none
1005 * DOWNGRADE allow read, deny none
1007 * which we should reject.
1010 bmap_to_share_mode(unsigned long bmap
) {
1012 unsigned int access
= 0;
1014 for (i
= 1; i
< 4; i
++) {
1015 if (test_bit(i
, &bmap
))
1021 /* set share access for a given stateid */
1023 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1025 unsigned char mask
= 1 << access
;
1027 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
1028 stp
->st_access_bmap
|= mask
;
1031 /* clear share access for a given stateid */
1033 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1035 unsigned char mask
= 1 << access
;
1037 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
1038 stp
->st_access_bmap
&= ~mask
;
1041 /* test whether a given stateid has access */
1043 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1045 unsigned char mask
= 1 << access
;
1047 return (bool)(stp
->st_access_bmap
& mask
);
1050 /* set share deny for a given stateid */
1052 set_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1054 unsigned char mask
= 1 << deny
;
1056 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
1057 stp
->st_deny_bmap
|= mask
;
1060 /* clear share deny for a given stateid */
1062 clear_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1064 unsigned char mask
= 1 << deny
;
1066 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
1067 stp
->st_deny_bmap
&= ~mask
;
1070 /* test whether a given stateid is denying specific access */
1072 test_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1074 unsigned char mask
= 1 << deny
;
1076 return (bool)(stp
->st_deny_bmap
& mask
);
1079 static int nfs4_access_to_omode(u32 access
)
1081 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
1082 case NFS4_SHARE_ACCESS_READ
:
1084 case NFS4_SHARE_ACCESS_WRITE
:
1086 case NFS4_SHARE_ACCESS_BOTH
:
1094 * A stateid that had a deny mode associated with it is being released
1095 * or downgraded. Recalculate the deny mode on the file.
1098 recalculate_deny_mode(struct nfs4_file
*fp
)
1100 struct nfs4_ol_stateid
*stp
;
1102 spin_lock(&fp
->fi_lock
);
1103 fp
->fi_share_deny
= 0;
1104 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
)
1105 fp
->fi_share_deny
|= bmap_to_share_mode(stp
->st_deny_bmap
);
1106 spin_unlock(&fp
->fi_lock
);
1110 reset_union_bmap_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1113 bool change
= false;
1115 for (i
= 1; i
< 4; i
++) {
1116 if ((i
& deny
) != i
) {
1122 /* Recalculate per-file deny mode if there was a change */
1124 recalculate_deny_mode(stp
->st_stid
.sc_file
);
1127 /* release all access and file references for a given stateid */
1129 release_all_access(struct nfs4_ol_stateid
*stp
)
1132 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
1134 if (fp
&& stp
->st_deny_bmap
!= 0)
1135 recalculate_deny_mode(fp
);
1137 for (i
= 1; i
< 4; i
++) {
1138 if (test_access(i
, stp
))
1139 nfs4_file_put_access(stp
->st_stid
.sc_file
, i
);
1140 clear_access(i
, stp
);
1144 static inline void nfs4_free_stateowner(struct nfs4_stateowner
*sop
)
1146 kfree(sop
->so_owner
.data
);
1147 sop
->so_ops
->so_free(sop
);
1150 static void nfs4_put_stateowner(struct nfs4_stateowner
*sop
)
1152 struct nfs4_client
*clp
= sop
->so_client
;
1154 might_lock(&clp
->cl_lock
);
1156 if (!atomic_dec_and_lock(&sop
->so_count
, &clp
->cl_lock
))
1158 sop
->so_ops
->so_unhash(sop
);
1159 spin_unlock(&clp
->cl_lock
);
1160 nfs4_free_stateowner(sop
);
1163 static bool unhash_ol_stateid(struct nfs4_ol_stateid
*stp
)
1165 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
1167 lockdep_assert_held(&stp
->st_stateowner
->so_client
->cl_lock
);
1169 if (list_empty(&stp
->st_perfile
))
1172 spin_lock(&fp
->fi_lock
);
1173 list_del_init(&stp
->st_perfile
);
1174 spin_unlock(&fp
->fi_lock
);
1175 list_del(&stp
->st_perstateowner
);
1179 static void nfs4_free_ol_stateid(struct nfs4_stid
*stid
)
1181 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1183 put_clnt_odstate(stp
->st_clnt_odstate
);
1184 release_all_access(stp
);
1185 if (stp
->st_stateowner
)
1186 nfs4_put_stateowner(stp
->st_stateowner
);
1187 kmem_cache_free(stateid_slab
, stid
);
1190 static void nfs4_free_lock_stateid(struct nfs4_stid
*stid
)
1192 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1193 struct nfs4_lockowner
*lo
= lockowner(stp
->st_stateowner
);
1196 file
= find_any_file(stp
->st_stid
.sc_file
);
1198 filp_close(file
, (fl_owner_t
)lo
);
1199 nfs4_free_ol_stateid(stid
);
1203 * Put the persistent reference to an already unhashed generic stateid, while
1204 * holding the cl_lock. If it's the last reference, then put it onto the
1205 * reaplist for later destruction.
1207 static void put_ol_stateid_locked(struct nfs4_ol_stateid
*stp
,
1208 struct list_head
*reaplist
)
1210 struct nfs4_stid
*s
= &stp
->st_stid
;
1211 struct nfs4_client
*clp
= s
->sc_client
;
1213 lockdep_assert_held(&clp
->cl_lock
);
1215 WARN_ON_ONCE(!list_empty(&stp
->st_locks
));
1217 if (!atomic_dec_and_test(&s
->sc_count
)) {
1218 wake_up_all(&close_wq
);
1222 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
1223 list_add(&stp
->st_locks
, reaplist
);
1226 static bool unhash_lock_stateid(struct nfs4_ol_stateid
*stp
)
1228 lockdep_assert_held(&stp
->st_stid
.sc_client
->cl_lock
);
1230 list_del_init(&stp
->st_locks
);
1231 nfs4_unhash_stid(&stp
->st_stid
);
1232 return unhash_ol_stateid(stp
);
1235 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
1237 struct nfs4_client
*clp
= stp
->st_stid
.sc_client
;
1240 spin_lock(&clp
->cl_lock
);
1241 unhashed
= unhash_lock_stateid(stp
);
1242 spin_unlock(&clp
->cl_lock
);
1244 nfs4_put_stid(&stp
->st_stid
);
1247 static void unhash_lockowner_locked(struct nfs4_lockowner
*lo
)
1249 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
1251 lockdep_assert_held(&clp
->cl_lock
);
1253 list_del_init(&lo
->lo_owner
.so_strhash
);
1257 * Free a list of generic stateids that were collected earlier after being
1261 free_ol_stateid_reaplist(struct list_head
*reaplist
)
1263 struct nfs4_ol_stateid
*stp
;
1264 struct nfs4_file
*fp
;
1268 while (!list_empty(reaplist
)) {
1269 stp
= list_first_entry(reaplist
, struct nfs4_ol_stateid
,
1271 list_del(&stp
->st_locks
);
1272 fp
= stp
->st_stid
.sc_file
;
1273 stp
->st_stid
.sc_free(&stp
->st_stid
);
1279 static void release_open_stateid_locks(struct nfs4_ol_stateid
*open_stp
,
1280 struct list_head
*reaplist
)
1282 struct nfs4_ol_stateid
*stp
;
1284 lockdep_assert_held(&open_stp
->st_stid
.sc_client
->cl_lock
);
1286 while (!list_empty(&open_stp
->st_locks
)) {
1287 stp
= list_entry(open_stp
->st_locks
.next
,
1288 struct nfs4_ol_stateid
, st_locks
);
1289 WARN_ON(!unhash_lock_stateid(stp
));
1290 put_ol_stateid_locked(stp
, reaplist
);
1294 static bool unhash_open_stateid(struct nfs4_ol_stateid
*stp
,
1295 struct list_head
*reaplist
)
1299 lockdep_assert_held(&stp
->st_stid
.sc_client
->cl_lock
);
1301 unhashed
= unhash_ol_stateid(stp
);
1302 release_open_stateid_locks(stp
, reaplist
);
1306 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
1308 LIST_HEAD(reaplist
);
1310 spin_lock(&stp
->st_stid
.sc_client
->cl_lock
);
1311 if (unhash_open_stateid(stp
, &reaplist
))
1312 put_ol_stateid_locked(stp
, &reaplist
);
1313 spin_unlock(&stp
->st_stid
.sc_client
->cl_lock
);
1314 free_ol_stateid_reaplist(&reaplist
);
1317 static void unhash_openowner_locked(struct nfs4_openowner
*oo
)
1319 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1321 lockdep_assert_held(&clp
->cl_lock
);
1323 list_del_init(&oo
->oo_owner
.so_strhash
);
1324 list_del_init(&oo
->oo_perclient
);
1327 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
1329 struct nfsd_net
*nn
= net_generic(oo
->oo_owner
.so_client
->net
,
1331 struct nfs4_ol_stateid
*s
;
1333 spin_lock(&nn
->client_lock
);
1334 s
= oo
->oo_last_closed_stid
;
1336 list_del_init(&oo
->oo_close_lru
);
1337 oo
->oo_last_closed_stid
= NULL
;
1339 spin_unlock(&nn
->client_lock
);
1341 nfs4_put_stid(&s
->st_stid
);
1344 static void release_openowner(struct nfs4_openowner
*oo
)
1346 struct nfs4_ol_stateid
*stp
;
1347 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1348 struct list_head reaplist
;
1350 INIT_LIST_HEAD(&reaplist
);
1352 spin_lock(&clp
->cl_lock
);
1353 unhash_openowner_locked(oo
);
1354 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
1355 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
1356 struct nfs4_ol_stateid
, st_perstateowner
);
1357 if (unhash_open_stateid(stp
, &reaplist
))
1358 put_ol_stateid_locked(stp
, &reaplist
);
1360 spin_unlock(&clp
->cl_lock
);
1361 free_ol_stateid_reaplist(&reaplist
);
1362 release_last_closed_stateid(oo
);
1363 nfs4_put_stateowner(&oo
->oo_owner
);
1367 hash_sessionid(struct nfs4_sessionid
*sessionid
)
1369 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
1371 return sid
->sequence
% SESSION_HASH_SIZE
;
1374 #ifdef CONFIG_SUNRPC_DEBUG
1376 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1378 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
1379 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
1383 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1389 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1390 * won't be used for replay.
1392 void nfsd4_bump_seqid(struct nfsd4_compound_state
*cstate
, __be32 nfserr
)
1394 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
1396 if (nfserr
== nfserr_replay_me
)
1399 if (!seqid_mutating_err(ntohl(nfserr
))) {
1400 nfsd4_cstate_clear_replay(cstate
);
1405 if (so
->so_is_open_owner
)
1406 release_last_closed_stateid(openowner(so
));
1412 gen_sessionid(struct nfsd4_session
*ses
)
1414 struct nfs4_client
*clp
= ses
->se_client
;
1415 struct nfsd4_sessionid
*sid
;
1417 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
1418 sid
->clientid
= clp
->cl_clientid
;
1419 sid
->sequence
= current_sessionid
++;
1424 * The protocol defines ca_maxresponssize_cached to include the size of
1425 * the rpc header, but all we need to cache is the data starting after
1426 * the end of the initial SEQUENCE operation--the rest we regenerate
1427 * each time. Therefore we can advertise a ca_maxresponssize_cached
1428 * value that is the number of bytes in our cache plus a few additional
1429 * bytes. In order to stay on the safe side, and not promise more than
1430 * we can cache, those additional bytes must be the minimum possible: 24
1431 * bytes of rpc header (xid through accept state, with AUTH_NULL
1432 * verifier), 12 for the compound header (with zero-length tag), and 44
1433 * for the SEQUENCE op response:
1435 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1438 free_session_slots(struct nfsd4_session
*ses
)
1442 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
1443 kfree(ses
->se_slots
[i
]);
1447 * We don't actually need to cache the rpc and session headers, so we
1448 * can allocate a little less for each slot:
1450 static inline u32
slot_bytes(struct nfsd4_channel_attrs
*ca
)
1454 if (ca
->maxresp_cached
< NFSD_MIN_HDR_SEQ_SZ
)
1457 size
= ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
1458 return size
+ sizeof(struct nfsd4_slot
);
1462 * XXX: If we run out of reserved DRC memory we could (up to a point)
1463 * re-negotiate active sessions and reduce their slot usage to make
1464 * room for new connections. For now we just fail the create session.
1466 static u32
nfsd4_get_drc_mem(struct nfsd4_channel_attrs
*ca
)
1468 u32 slotsize
= slot_bytes(ca
);
1469 u32 num
= ca
->maxreqs
;
1472 spin_lock(&nfsd_drc_lock
);
1473 avail
= min((unsigned long)NFSD_MAX_MEM_PER_SESSION
,
1474 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
1475 num
= min_t(int, num
, avail
/ slotsize
);
1476 nfsd_drc_mem_used
+= num
* slotsize
;
1477 spin_unlock(&nfsd_drc_lock
);
1482 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs
*ca
)
1484 int slotsize
= slot_bytes(ca
);
1486 spin_lock(&nfsd_drc_lock
);
1487 nfsd_drc_mem_used
-= slotsize
* ca
->maxreqs
;
1488 spin_unlock(&nfsd_drc_lock
);
1491 static struct nfsd4_session
*alloc_session(struct nfsd4_channel_attrs
*fattrs
,
1492 struct nfsd4_channel_attrs
*battrs
)
1494 int numslots
= fattrs
->maxreqs
;
1495 int slotsize
= slot_bytes(fattrs
);
1496 struct nfsd4_session
*new;
1499 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
1500 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
1501 mem
= numslots
* sizeof(struct nfsd4_slot
*);
1503 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
1506 /* allocate each struct nfsd4_slot and data cache in one piece */
1507 for (i
= 0; i
< numslots
; i
++) {
1508 new->se_slots
[i
] = kzalloc(slotsize
, GFP_KERNEL
);
1509 if (!new->se_slots
[i
])
1513 memcpy(&new->se_fchannel
, fattrs
, sizeof(struct nfsd4_channel_attrs
));
1514 memcpy(&new->se_bchannel
, battrs
, sizeof(struct nfsd4_channel_attrs
));
1519 kfree(new->se_slots
[i
]);
1524 static void free_conn(struct nfsd4_conn
*c
)
1526 svc_xprt_put(c
->cn_xprt
);
1530 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
1532 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
1533 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
1535 spin_lock(&clp
->cl_lock
);
1536 if (!list_empty(&c
->cn_persession
)) {
1537 list_del(&c
->cn_persession
);
1540 nfsd4_probe_callback(clp
);
1541 spin_unlock(&clp
->cl_lock
);
1544 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
1546 struct nfsd4_conn
*conn
;
1548 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
1551 svc_xprt_get(rqstp
->rq_xprt
);
1552 conn
->cn_xprt
= rqstp
->rq_xprt
;
1553 conn
->cn_flags
= flags
;
1554 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
1558 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1560 conn
->cn_session
= ses
;
1561 list_add(&conn
->cn_persession
, &ses
->se_conns
);
1564 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1566 struct nfs4_client
*clp
= ses
->se_client
;
1568 spin_lock(&clp
->cl_lock
);
1569 __nfsd4_hash_conn(conn
, ses
);
1570 spin_unlock(&clp
->cl_lock
);
1573 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
1575 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
1576 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
1579 static void nfsd4_init_conn(struct svc_rqst
*rqstp
, struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1583 nfsd4_hash_conn(conn
, ses
);
1584 ret
= nfsd4_register_conn(conn
);
1586 /* oops; xprt is already down: */
1587 nfsd4_conn_lost(&conn
->cn_xpt_user
);
1588 /* We may have gained or lost a callback channel: */
1589 nfsd4_probe_callback_sync(ses
->se_client
);
1592 static struct nfsd4_conn
*alloc_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_create_session
*cses
)
1594 u32 dir
= NFS4_CDFC4_FORE
;
1596 if (cses
->flags
& SESSION4_BACK_CHAN
)
1597 dir
|= NFS4_CDFC4_BACK
;
1598 return alloc_conn(rqstp
, dir
);
1601 /* must be called under client_lock */
1602 static void nfsd4_del_conns(struct nfsd4_session
*s
)
1604 struct nfs4_client
*clp
= s
->se_client
;
1605 struct nfsd4_conn
*c
;
1607 spin_lock(&clp
->cl_lock
);
1608 while (!list_empty(&s
->se_conns
)) {
1609 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
1610 list_del_init(&c
->cn_persession
);
1611 spin_unlock(&clp
->cl_lock
);
1613 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
1616 spin_lock(&clp
->cl_lock
);
1618 spin_unlock(&clp
->cl_lock
);
1621 static void __free_session(struct nfsd4_session
*ses
)
1623 free_session_slots(ses
);
1627 static void free_session(struct nfsd4_session
*ses
)
1629 nfsd4_del_conns(ses
);
1630 nfsd4_put_drc_mem(&ses
->se_fchannel
);
1631 __free_session(ses
);
1634 static void init_session(struct svc_rqst
*rqstp
, struct nfsd4_session
*new, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
1637 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
1639 new->se_client
= clp
;
1642 INIT_LIST_HEAD(&new->se_conns
);
1644 new->se_cb_seq_nr
= 1;
1645 new->se_flags
= cses
->flags
;
1646 new->se_cb_prog
= cses
->callback_prog
;
1647 new->se_cb_sec
= cses
->cb_sec
;
1648 atomic_set(&new->se_ref
, 0);
1649 idx
= hash_sessionid(&new->se_sessionid
);
1650 list_add(&new->se_hash
, &nn
->sessionid_hashtbl
[idx
]);
1651 spin_lock(&clp
->cl_lock
);
1652 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
1653 spin_unlock(&clp
->cl_lock
);
1656 struct sockaddr
*sa
= svc_addr(rqstp
);
1658 * This is a little silly; with sessions there's no real
1659 * use for the callback address. Use the peer address
1660 * as a reasonable default for now, but consider fixing
1661 * the rpc client not to require an address in the
1664 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
1665 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
1669 /* caller must hold client_lock */
1670 static struct nfsd4_session
*
1671 __find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
)
1673 struct nfsd4_session
*elem
;
1675 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1677 lockdep_assert_held(&nn
->client_lock
);
1679 dump_sessionid(__func__
, sessionid
);
1680 idx
= hash_sessionid(sessionid
);
1681 /* Search in the appropriate list */
1682 list_for_each_entry(elem
, &nn
->sessionid_hashtbl
[idx
], se_hash
) {
1683 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1684 NFS4_MAX_SESSIONID_LEN
)) {
1689 dprintk("%s: session not found\n", __func__
);
1693 static struct nfsd4_session
*
1694 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
,
1697 struct nfsd4_session
*session
;
1698 __be32 status
= nfserr_badsession
;
1700 session
= __find_in_sessionid_hashtbl(sessionid
, net
);
1703 status
= nfsd4_get_session_locked(session
);
1711 /* caller must hold client_lock */
1713 unhash_session(struct nfsd4_session
*ses
)
1715 struct nfs4_client
*clp
= ses
->se_client
;
1716 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1718 lockdep_assert_held(&nn
->client_lock
);
1720 list_del(&ses
->se_hash
);
1721 spin_lock(&ses
->se_client
->cl_lock
);
1722 list_del(&ses
->se_perclnt
);
1723 spin_unlock(&ses
->se_client
->cl_lock
);
1726 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1728 STALE_CLIENTID(clientid_t
*clid
, struct nfsd_net
*nn
)
1731 * We're assuming the clid was not given out from a boot
1732 * precisely 2^32 (about 136 years) before this one. That seems
1733 * a safe assumption:
1735 if (clid
->cl_boot
== (u32
)nn
->boot_time
)
1737 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1738 clid
->cl_boot
, clid
->cl_id
, nn
->boot_time
);
1743 * XXX Should we use a slab cache ?
1744 * This type of memory management is somewhat inefficient, but we use it
1745 * anyway since SETCLIENTID is not a common operation.
1747 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1749 struct nfs4_client
*clp
;
1752 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1755 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1756 if (clp
->cl_name
.data
== NULL
)
1758 clp
->cl_ownerstr_hashtbl
= kmalloc(sizeof(struct list_head
) *
1759 OWNER_HASH_SIZE
, GFP_KERNEL
);
1760 if (!clp
->cl_ownerstr_hashtbl
)
1761 goto err_no_hashtbl
;
1762 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++)
1763 INIT_LIST_HEAD(&clp
->cl_ownerstr_hashtbl
[i
]);
1764 clp
->cl_name
.len
= name
.len
;
1765 INIT_LIST_HEAD(&clp
->cl_sessions
);
1766 idr_init(&clp
->cl_stateids
);
1767 atomic_set(&clp
->cl_refcount
, 0);
1768 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1769 INIT_LIST_HEAD(&clp
->cl_idhash
);
1770 INIT_LIST_HEAD(&clp
->cl_openowners
);
1771 INIT_LIST_HEAD(&clp
->cl_delegations
);
1772 INIT_LIST_HEAD(&clp
->cl_lru
);
1773 INIT_LIST_HEAD(&clp
->cl_revoked
);
1774 #ifdef CONFIG_NFSD_PNFS
1775 INIT_LIST_HEAD(&clp
->cl_lo_states
);
1777 spin_lock_init(&clp
->cl_lock
);
1778 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1781 kfree(clp
->cl_name
.data
);
1788 free_client(struct nfs4_client
*clp
)
1790 while (!list_empty(&clp
->cl_sessions
)) {
1791 struct nfsd4_session
*ses
;
1792 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1794 list_del(&ses
->se_perclnt
);
1795 WARN_ON_ONCE(atomic_read(&ses
->se_ref
));
1798 rpc_destroy_wait_queue(&clp
->cl_cb_waitq
);
1799 free_svc_cred(&clp
->cl_cred
);
1800 kfree(clp
->cl_ownerstr_hashtbl
);
1801 kfree(clp
->cl_name
.data
);
1802 idr_destroy(&clp
->cl_stateids
);
1806 /* must be called under the client_lock */
1808 unhash_client_locked(struct nfs4_client
*clp
)
1810 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1811 struct nfsd4_session
*ses
;
1813 lockdep_assert_held(&nn
->client_lock
);
1815 /* Mark the client as expired! */
1817 /* Make it invisible */
1818 if (!list_empty(&clp
->cl_idhash
)) {
1819 list_del_init(&clp
->cl_idhash
);
1820 if (test_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
))
1821 rb_erase(&clp
->cl_namenode
, &nn
->conf_name_tree
);
1823 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
1825 list_del_init(&clp
->cl_lru
);
1826 spin_lock(&clp
->cl_lock
);
1827 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1828 list_del_init(&ses
->se_hash
);
1829 spin_unlock(&clp
->cl_lock
);
1833 unhash_client(struct nfs4_client
*clp
)
1835 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1837 spin_lock(&nn
->client_lock
);
1838 unhash_client_locked(clp
);
1839 spin_unlock(&nn
->client_lock
);
1842 static __be32
mark_client_expired_locked(struct nfs4_client
*clp
)
1844 if (atomic_read(&clp
->cl_refcount
))
1845 return nfserr_jukebox
;
1846 unhash_client_locked(clp
);
1851 __destroy_client(struct nfs4_client
*clp
)
1853 struct nfs4_openowner
*oo
;
1854 struct nfs4_delegation
*dp
;
1855 struct list_head reaplist
;
1857 INIT_LIST_HEAD(&reaplist
);
1858 spin_lock(&state_lock
);
1859 while (!list_empty(&clp
->cl_delegations
)) {
1860 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1861 WARN_ON(!unhash_delegation_locked(dp
));
1862 list_add(&dp
->dl_recall_lru
, &reaplist
);
1864 spin_unlock(&state_lock
);
1865 while (!list_empty(&reaplist
)) {
1866 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1867 list_del_init(&dp
->dl_recall_lru
);
1868 put_clnt_odstate(dp
->dl_clnt_odstate
);
1869 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
1870 nfs4_put_stid(&dp
->dl_stid
);
1872 while (!list_empty(&clp
->cl_revoked
)) {
1873 dp
= list_entry(clp
->cl_revoked
.next
, struct nfs4_delegation
, dl_recall_lru
);
1874 list_del_init(&dp
->dl_recall_lru
);
1875 nfs4_put_stid(&dp
->dl_stid
);
1877 while (!list_empty(&clp
->cl_openowners
)) {
1878 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1879 nfs4_get_stateowner(&oo
->oo_owner
);
1880 release_openowner(oo
);
1882 nfsd4_return_all_client_layouts(clp
);
1883 nfsd4_shutdown_callback(clp
);
1884 if (clp
->cl_cb_conn
.cb_xprt
)
1885 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1890 destroy_client(struct nfs4_client
*clp
)
1893 __destroy_client(clp
);
1896 static void expire_client(struct nfs4_client
*clp
)
1899 nfsd4_client_record_remove(clp
);
1900 __destroy_client(clp
);
1903 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1905 memcpy(target
->cl_verifier
.data
, source
->data
,
1906 sizeof(target
->cl_verifier
.data
));
1909 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1911 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1912 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1915 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1917 target
->cr_principal
= kstrdup(source
->cr_principal
, GFP_KERNEL
);
1918 target
->cr_raw_principal
= kstrdup(source
->cr_raw_principal
,
1920 if ((source
->cr_principal
&& ! target
->cr_principal
) ||
1921 (source
->cr_raw_principal
&& ! target
->cr_raw_principal
))
1924 target
->cr_flavor
= source
->cr_flavor
;
1925 target
->cr_uid
= source
->cr_uid
;
1926 target
->cr_gid
= source
->cr_gid
;
1927 target
->cr_group_info
= source
->cr_group_info
;
1928 get_group_info(target
->cr_group_info
);
1929 target
->cr_gss_mech
= source
->cr_gss_mech
;
1930 if (source
->cr_gss_mech
)
1931 gss_mech_get(source
->cr_gss_mech
);
1936 compare_blob(const struct xdr_netobj
*o1
, const struct xdr_netobj
*o2
)
1938 if (o1
->len
< o2
->len
)
1940 if (o1
->len
> o2
->len
)
1942 return memcmp(o1
->data
, o2
->data
, o1
->len
);
1945 static int same_name(const char *n1
, const char *n2
)
1947 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1951 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1953 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1957 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1959 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1962 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
1966 if (g1
->ngroups
!= g2
->ngroups
)
1968 for (i
=0; i
<g1
->ngroups
; i
++)
1969 if (!gid_eq(g1
->gid
[i
], g2
->gid
[i
]))
1975 * RFC 3530 language requires clid_inuse be returned when the
1976 * "principal" associated with a requests differs from that previously
1977 * used. We use uid, gid's, and gss principal string as our best
1978 * approximation. We also don't want to allow non-gss use of a client
1979 * established using gss: in theory cr_principal should catch that
1980 * change, but in practice cr_principal can be null even in the gss case
1981 * since gssd doesn't always pass down a principal string.
1983 static bool is_gss_cred(struct svc_cred
*cr
)
1985 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1986 return (cr
->cr_flavor
> RPC_AUTH_MAXFLAVOR
);
1991 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1993 if ((is_gss_cred(cr1
) != is_gss_cred(cr2
))
1994 || (!uid_eq(cr1
->cr_uid
, cr2
->cr_uid
))
1995 || (!gid_eq(cr1
->cr_gid
, cr2
->cr_gid
))
1996 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
1998 if (cr1
->cr_principal
== cr2
->cr_principal
)
2000 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
2002 return 0 == strcmp(cr1
->cr_principal
, cr2
->cr_principal
);
2005 static bool svc_rqst_integrity_protected(struct svc_rqst
*rqstp
)
2007 struct svc_cred
*cr
= &rqstp
->rq_cred
;
2010 if (!cr
->cr_gss_mech
)
2012 service
= gss_pseudoflavor_to_service(cr
->cr_gss_mech
, cr
->cr_flavor
);
2013 return service
== RPC_GSS_SVC_INTEGRITY
||
2014 service
== RPC_GSS_SVC_PRIVACY
;
2017 bool nfsd4_mach_creds_match(struct nfs4_client
*cl
, struct svc_rqst
*rqstp
)
2019 struct svc_cred
*cr
= &rqstp
->rq_cred
;
2021 if (!cl
->cl_mach_cred
)
2023 if (cl
->cl_cred
.cr_gss_mech
!= cr
->cr_gss_mech
)
2025 if (!svc_rqst_integrity_protected(rqstp
))
2027 if (cl
->cl_cred
.cr_raw_principal
)
2028 return 0 == strcmp(cl
->cl_cred
.cr_raw_principal
,
2029 cr
->cr_raw_principal
);
2030 if (!cr
->cr_principal
)
2032 return 0 == strcmp(cl
->cl_cred
.cr_principal
, cr
->cr_principal
);
2035 static void gen_confirm(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
2040 * This is opaque to client, so no need to byte-swap. Use
2041 * __force to keep sparse happy
2043 verf
[0] = (__force __be32
)get_seconds();
2044 verf
[1] = (__force __be32
)nn
->clverifier_counter
++;
2045 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
2048 static void gen_clid(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
2050 clp
->cl_clientid
.cl_boot
= nn
->boot_time
;
2051 clp
->cl_clientid
.cl_id
= nn
->clientid_counter
++;
2052 gen_confirm(clp
, nn
);
2055 static struct nfs4_stid
*
2056 find_stateid_locked(struct nfs4_client
*cl
, stateid_t
*t
)
2058 struct nfs4_stid
*ret
;
2060 ret
= idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
2061 if (!ret
|| !ret
->sc_type
)
2066 static struct nfs4_stid
*
2067 find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
2069 struct nfs4_stid
*s
;
2071 spin_lock(&cl
->cl_lock
);
2072 s
= find_stateid_locked(cl
, t
);
2074 if (typemask
& s
->sc_type
)
2075 atomic_inc(&s
->sc_count
);
2079 spin_unlock(&cl
->cl_lock
);
2083 static struct nfs4_client
*create_client(struct xdr_netobj name
,
2084 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
2086 struct nfs4_client
*clp
;
2087 struct sockaddr
*sa
= svc_addr(rqstp
);
2089 struct net
*net
= SVC_NET(rqstp
);
2091 clp
= alloc_client(name
);
2095 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
2100 nfsd4_init_cb(&clp
->cl_cb_null
, clp
, NULL
, NFSPROC4_CLNT_CB_NULL
);
2101 clp
->cl_time
= get_seconds();
2102 clear_bit(0, &clp
->cl_cb_slot_busy
);
2103 copy_verf(clp
, verf
);
2104 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
2105 clp
->cl_cb_session
= NULL
;
2111 add_clp_to_name_tree(struct nfs4_client
*new_clp
, struct rb_root
*root
)
2113 struct rb_node
**new = &(root
->rb_node
), *parent
= NULL
;
2114 struct nfs4_client
*clp
;
2117 clp
= rb_entry(*new, struct nfs4_client
, cl_namenode
);
2120 if (compare_blob(&clp
->cl_name
, &new_clp
->cl_name
) > 0)
2121 new = &((*new)->rb_left
);
2123 new = &((*new)->rb_right
);
2126 rb_link_node(&new_clp
->cl_namenode
, parent
, new);
2127 rb_insert_color(&new_clp
->cl_namenode
, root
);
2130 static struct nfs4_client
*
2131 find_clp_in_name_tree(struct xdr_netobj
*name
, struct rb_root
*root
)
2134 struct rb_node
*node
= root
->rb_node
;
2135 struct nfs4_client
*clp
;
2138 clp
= rb_entry(node
, struct nfs4_client
, cl_namenode
);
2139 cmp
= compare_blob(&clp
->cl_name
, name
);
2141 node
= node
->rb_left
;
2143 node
= node
->rb_right
;
2151 add_to_unconfirmed(struct nfs4_client
*clp
)
2153 unsigned int idhashval
;
2154 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2156 lockdep_assert_held(&nn
->client_lock
);
2158 clear_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2159 add_clp_to_name_tree(clp
, &nn
->unconf_name_tree
);
2160 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2161 list_add(&clp
->cl_idhash
, &nn
->unconf_id_hashtbl
[idhashval
]);
2162 renew_client_locked(clp
);
2166 move_to_confirmed(struct nfs4_client
*clp
)
2168 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2169 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2171 lockdep_assert_held(&nn
->client_lock
);
2173 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
2174 list_move(&clp
->cl_idhash
, &nn
->conf_id_hashtbl
[idhashval
]);
2175 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
2176 add_clp_to_name_tree(clp
, &nn
->conf_name_tree
);
2177 set_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2178 renew_client_locked(clp
);
2181 static struct nfs4_client
*
2182 find_client_in_id_table(struct list_head
*tbl
, clientid_t
*clid
, bool sessions
)
2184 struct nfs4_client
*clp
;
2185 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
2187 list_for_each_entry(clp
, &tbl
[idhashval
], cl_idhash
) {
2188 if (same_clid(&clp
->cl_clientid
, clid
)) {
2189 if ((bool)clp
->cl_minorversion
!= sessions
)
2191 renew_client_locked(clp
);
2198 static struct nfs4_client
*
2199 find_confirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2201 struct list_head
*tbl
= nn
->conf_id_hashtbl
;
2203 lockdep_assert_held(&nn
->client_lock
);
2204 return find_client_in_id_table(tbl
, clid
, sessions
);
2207 static struct nfs4_client
*
2208 find_unconfirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2210 struct list_head
*tbl
= nn
->unconf_id_hashtbl
;
2212 lockdep_assert_held(&nn
->client_lock
);
2213 return find_client_in_id_table(tbl
, clid
, sessions
);
2216 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
2218 return clp
->cl_exchange_flags
!= 0;
2221 static struct nfs4_client
*
2222 find_confirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2224 lockdep_assert_held(&nn
->client_lock
);
2225 return find_clp_in_name_tree(name
, &nn
->conf_name_tree
);
2228 static struct nfs4_client
*
2229 find_unconfirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2231 lockdep_assert_held(&nn
->client_lock
);
2232 return find_clp_in_name_tree(name
, &nn
->unconf_name_tree
);
2236 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
2238 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
2239 struct sockaddr
*sa
= svc_addr(rqstp
);
2240 u32 scopeid
= rpc_get_scope_id(sa
);
2241 unsigned short expected_family
;
2243 /* Currently, we only support tcp and tcp6 for the callback channel */
2244 if (se
->se_callback_netid_len
== 3 &&
2245 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
2246 expected_family
= AF_INET
;
2247 else if (se
->se_callback_netid_len
== 4 &&
2248 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
2249 expected_family
= AF_INET6
;
2253 conn
->cb_addrlen
= rpc_uaddr2sockaddr(clp
->net
, se
->se_callback_addr_val
,
2254 se
->se_callback_addr_len
,
2255 (struct sockaddr
*)&conn
->cb_addr
,
2256 sizeof(conn
->cb_addr
));
2258 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
2261 if (conn
->cb_addr
.ss_family
== AF_INET6
)
2262 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
2264 conn
->cb_prog
= se
->se_callback_prog
;
2265 conn
->cb_ident
= se
->se_callback_ident
;
2266 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
2269 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
2270 conn
->cb_addrlen
= 0;
2271 dprintk("NFSD: this client (clientid %08x/%08x) "
2272 "will not receive delegations\n",
2273 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
2279 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2282 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
2284 struct xdr_buf
*buf
= resp
->xdr
.buf
;
2285 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2288 dprintk("--> %s slot %p\n", __func__
, slot
);
2290 slot
->sl_opcnt
= resp
->opcnt
;
2291 slot
->sl_status
= resp
->cstate
.status
;
2293 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
2294 if (nfsd4_not_cached(resp
)) {
2295 slot
->sl_datalen
= 0;
2298 base
= resp
->cstate
.data_offset
;
2299 slot
->sl_datalen
= buf
->len
- base
;
2300 if (read_bytes_from_xdr_buf(buf
, base
, slot
->sl_data
, slot
->sl_datalen
))
2301 WARN(1, "%s: sessions DRC could not cache compound\n",
2307 * Encode the replay sequence operation from the slot values.
2308 * If cachethis is FALSE encode the uncached rep error on the next
2309 * operation which sets resp->p and increments resp->opcnt for
2310 * nfs4svc_encode_compoundres.
2314 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
2315 struct nfsd4_compoundres
*resp
)
2317 struct nfsd4_op
*op
;
2318 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2320 /* Encode the replayed sequence operation */
2321 op
= &args
->ops
[resp
->opcnt
- 1];
2322 nfsd4_encode_operation(resp
, op
);
2324 /* Return nfserr_retry_uncached_rep in next operation. */
2325 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
2326 op
= &args
->ops
[resp
->opcnt
++];
2327 op
->status
= nfserr_retry_uncached_rep
;
2328 nfsd4_encode_operation(resp
, op
);
2334 * The sequence operation is not cached because we can use the slot and
2338 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
2339 struct nfsd4_sequence
*seq
)
2341 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2342 struct xdr_stream
*xdr
= &resp
->xdr
;
2346 dprintk("--> %s slot %p\n", __func__
, slot
);
2348 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
2352 p
= xdr_reserve_space(xdr
, slot
->sl_datalen
);
2355 return nfserr_serverfault
;
2357 xdr_encode_opaque_fixed(p
, slot
->sl_data
, slot
->sl_datalen
);
2358 xdr_commit_encode(xdr
);
2360 resp
->opcnt
= slot
->sl_opcnt
;
2361 return slot
->sl_status
;
2365 * Set the exchange_id flags returned by the server.
2368 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
2370 #ifdef CONFIG_NFSD_PNFS
2371 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_PNFS_MDS
;
2373 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
2376 /* Referrals are supported, Migration is not. */
2377 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
2379 /* set the wire flags to return to client. */
2380 clid
->flags
= new->cl_exchange_flags
;
2383 static bool client_has_openowners(struct nfs4_client
*clp
)
2385 struct nfs4_openowner
*oo
;
2387 list_for_each_entry(oo
, &clp
->cl_openowners
, oo_perclient
) {
2388 if (!list_empty(&oo
->oo_owner
.so_stateids
))
2394 static bool client_has_state(struct nfs4_client
*clp
)
2396 return client_has_openowners(clp
)
2397 #ifdef CONFIG_NFSD_PNFS
2398 || !list_empty(&clp
->cl_lo_states
)
2400 || !list_empty(&clp
->cl_delegations
)
2401 || !list_empty(&clp
->cl_sessions
);
2405 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
2406 struct nfsd4_compound_state
*cstate
,
2407 struct nfsd4_exchange_id
*exid
)
2409 struct nfs4_client
*conf
, *new;
2410 struct nfs4_client
*unconf
= NULL
;
2412 char addr_str
[INET6_ADDRSTRLEN
];
2413 nfs4_verifier verf
= exid
->verifier
;
2414 struct sockaddr
*sa
= svc_addr(rqstp
);
2415 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
2416 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2418 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
2419 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2420 "ip_addr=%s flags %x, spa_how %d\n",
2421 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
2422 addr_str
, exid
->flags
, exid
->spa_how
);
2424 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
2425 return nfserr_inval
;
2427 new = create_client(exid
->clname
, rqstp
, &verf
);
2429 return nfserr_jukebox
;
2431 switch (exid
->spa_how
) {
2433 exid
->spo_must_enforce
[0] = 0;
2434 exid
->spo_must_enforce
[1] = (
2435 1 << (OP_BIND_CONN_TO_SESSION
- 32) |
2436 1 << (OP_EXCHANGE_ID
- 32) |
2437 1 << (OP_CREATE_SESSION
- 32) |
2438 1 << (OP_DESTROY_SESSION
- 32) |
2439 1 << (OP_DESTROY_CLIENTID
- 32));
2441 exid
->spo_must_allow
[0] &= (1 << (OP_CLOSE
) |
2442 1 << (OP_OPEN_DOWNGRADE
) |
2444 1 << (OP_DELEGRETURN
));
2446 exid
->spo_must_allow
[1] &= (
2447 1 << (OP_TEST_STATEID
- 32) |
2448 1 << (OP_FREE_STATEID
- 32));
2449 if (!svc_rqst_integrity_protected(rqstp
)) {
2450 status
= nfserr_inval
;
2454 * Sometimes userspace doesn't give us a principal.
2455 * Which is a bug, really. Anyway, we can't enforce
2456 * MACH_CRED in that case, better to give up now:
2458 if (!new->cl_cred
.cr_principal
&&
2459 !new->cl_cred
.cr_raw_principal
) {
2460 status
= nfserr_serverfault
;
2463 new->cl_mach_cred
= true;
2466 default: /* checked by xdr code */
2469 status
= nfserr_encr_alg_unsupp
;
2473 /* Cases below refer to rfc 5661 section 18.35.4: */
2474 spin_lock(&nn
->client_lock
);
2475 conf
= find_confirmed_client_by_name(&exid
->clname
, nn
);
2477 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
2478 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
2481 if (!clp_used_exchangeid(conf
)) { /* buggy client */
2482 status
= nfserr_inval
;
2485 if (!nfsd4_mach_creds_match(conf
, rqstp
)) {
2486 status
= nfserr_wrong_cred
;
2489 if (!creds_match
) { /* case 9 */
2490 status
= nfserr_perm
;
2493 if (!verfs_match
) { /* case 8 */
2494 status
= nfserr_not_same
;
2498 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2501 if (!creds_match
) { /* case 3 */
2502 if (client_has_state(conf
)) {
2503 status
= nfserr_clid_inuse
;
2508 if (verfs_match
) { /* case 2 */
2509 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2512 /* case 5, client reboot */
2517 if (update
) { /* case 7 */
2518 status
= nfserr_noent
;
2522 unconf
= find_unconfirmed_client_by_name(&exid
->clname
, nn
);
2523 if (unconf
) /* case 4, possible retry or client restart */
2524 unhash_client_locked(unconf
);
2526 /* case 1 (normal case) */
2529 status
= mark_client_expired_locked(conf
);
2533 new->cl_minorversion
= cstate
->minorversion
;
2534 new->cl_spo_must_allow
.u
.words
[0] = exid
->spo_must_allow
[0];
2535 new->cl_spo_must_allow
.u
.words
[1] = exid
->spo_must_allow
[1];
2538 add_to_unconfirmed(new);
2541 exid
->clientid
.cl_boot
= conf
->cl_clientid
.cl_boot
;
2542 exid
->clientid
.cl_id
= conf
->cl_clientid
.cl_id
;
2544 exid
->seqid
= conf
->cl_cs_slot
.sl_seqid
+ 1;
2545 nfsd4_set_ex_flags(conf
, exid
);
2547 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2548 conf
->cl_cs_slot
.sl_seqid
, conf
->cl_exchange_flags
);
2552 spin_unlock(&nn
->client_lock
);
2557 expire_client(unconf
);
2562 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
2564 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
2567 /* The slot is in use, and no response has been sent. */
2569 if (seqid
== slot_seqid
)
2570 return nfserr_jukebox
;
2572 return nfserr_seq_misordered
;
2574 /* Note unsigned 32-bit arithmetic handles wraparound: */
2575 if (likely(seqid
== slot_seqid
+ 1))
2577 if (seqid
== slot_seqid
)
2578 return nfserr_replay_cache
;
2579 return nfserr_seq_misordered
;
2583 * Cache the create session result into the create session single DRC
2584 * slot cache by saving the xdr structure. sl_seqid has been set.
2585 * Do this for solo or embedded create session operations.
2588 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
2589 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
2591 slot
->sl_status
= nfserr
;
2592 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
2596 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
2597 struct nfsd4_clid_slot
*slot
)
2599 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
2600 return slot
->sl_status
;
2603 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2604 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2605 1 + /* MIN tag is length with zero, only length */ \
2606 3 + /* version, opcount, opcode */ \
2607 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2608 /* seqid, slotID, slotID, cache */ \
2609 4 ) * sizeof(__be32))
2611 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2612 2 + /* verifier: AUTH_NULL, length 0 */\
2614 1 + /* MIN tag is length with zero, only length */ \
2615 3 + /* opcount, opcode, opstatus*/ \
2616 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2617 /* seqid, slotID, slotID, slotID, status */ \
2618 5 ) * sizeof(__be32))
2620 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs
*ca
, struct nfsd_net
*nn
)
2622 u32 maxrpc
= nn
->nfsd_serv
->sv_max_mesg
;
2624 if (ca
->maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
)
2625 return nfserr_toosmall
;
2626 if (ca
->maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
)
2627 return nfserr_toosmall
;
2628 ca
->headerpadsz
= 0;
2629 ca
->maxreq_sz
= min_t(u32
, ca
->maxreq_sz
, maxrpc
);
2630 ca
->maxresp_sz
= min_t(u32
, ca
->maxresp_sz
, maxrpc
);
2631 ca
->maxops
= min_t(u32
, ca
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
2632 ca
->maxresp_cached
= min_t(u32
, ca
->maxresp_cached
,
2633 NFSD_SLOT_CACHE_SIZE
+ NFSD_MIN_HDR_SEQ_SZ
);
2634 ca
->maxreqs
= min_t(u32
, ca
->maxreqs
, NFSD_MAX_SLOTS_PER_SESSION
);
2636 * Note decreasing slot size below client's request may make it
2637 * difficult for client to function correctly, whereas
2638 * decreasing the number of slots will (just?) affect
2639 * performance. When short on memory we therefore prefer to
2640 * decrease number of slots instead of their size. Clients that
2641 * request larger slots than they need will get poor results:
2643 ca
->maxreqs
= nfsd4_get_drc_mem(ca
);
2645 return nfserr_jukebox
;
2651 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
2652 * These are based on similar macros in linux/sunrpc/msg_prot.h .
2654 #define RPC_MAX_HEADER_WITH_AUTH_SYS \
2655 (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
2657 #define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
2658 (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
2660 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2661 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
2662 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2663 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
2666 static __be32
check_backchannel_attrs(struct nfsd4_channel_attrs
*ca
)
2668 ca
->headerpadsz
= 0;
2670 if (ca
->maxreq_sz
< NFSD_CB_MAX_REQ_SZ
)
2671 return nfserr_toosmall
;
2672 if (ca
->maxresp_sz
< NFSD_CB_MAX_RESP_SZ
)
2673 return nfserr_toosmall
;
2674 ca
->maxresp_cached
= 0;
2676 return nfserr_toosmall
;
2681 static __be32
nfsd4_check_cb_sec(struct nfsd4_cb_sec
*cbs
)
2683 switch (cbs
->flavor
) {
2689 * GSS case: the spec doesn't allow us to return this
2690 * error. But it also doesn't allow us not to support
2692 * I'd rather this fail hard than return some error the
2693 * client might think it can already handle:
2695 return nfserr_encr_alg_unsupp
;
2700 nfsd4_create_session(struct svc_rqst
*rqstp
,
2701 struct nfsd4_compound_state
*cstate
,
2702 struct nfsd4_create_session
*cr_ses
)
2704 struct sockaddr
*sa
= svc_addr(rqstp
);
2705 struct nfs4_client
*conf
, *unconf
;
2706 struct nfs4_client
*old
= NULL
;
2707 struct nfsd4_session
*new;
2708 struct nfsd4_conn
*conn
;
2709 struct nfsd4_clid_slot
*cs_slot
= NULL
;
2711 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2713 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
2714 return nfserr_inval
;
2715 status
= nfsd4_check_cb_sec(&cr_ses
->cb_sec
);
2718 status
= check_forechannel_attrs(&cr_ses
->fore_channel
, nn
);
2721 status
= check_backchannel_attrs(&cr_ses
->back_channel
);
2723 goto out_release_drc_mem
;
2724 status
= nfserr_jukebox
;
2725 new = alloc_session(&cr_ses
->fore_channel
, &cr_ses
->back_channel
);
2727 goto out_release_drc_mem
;
2728 conn
= alloc_conn_from_crses(rqstp
, cr_ses
);
2730 goto out_free_session
;
2732 spin_lock(&nn
->client_lock
);
2733 unconf
= find_unconfirmed_client(&cr_ses
->clientid
, true, nn
);
2734 conf
= find_confirmed_client(&cr_ses
->clientid
, true, nn
);
2735 WARN_ON_ONCE(conf
&& unconf
);
2738 status
= nfserr_wrong_cred
;
2739 if (!nfsd4_mach_creds_match(conf
, rqstp
))
2741 cs_slot
= &conf
->cl_cs_slot
;
2742 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2744 if (status
== nfserr_replay_cache
)
2745 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
2748 } else if (unconf
) {
2749 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
2750 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
2751 status
= nfserr_clid_inuse
;
2754 status
= nfserr_wrong_cred
;
2755 if (!nfsd4_mach_creds_match(unconf
, rqstp
))
2757 cs_slot
= &unconf
->cl_cs_slot
;
2758 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2760 /* an unconfirmed replay returns misordered */
2761 status
= nfserr_seq_misordered
;
2764 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
2766 status
= mark_client_expired_locked(old
);
2772 move_to_confirmed(unconf
);
2775 status
= nfserr_stale_clientid
;
2779 /* Persistent sessions are not supported */
2780 cr_ses
->flags
&= ~SESSION4_PERSIST
;
2781 /* Upshifting from TCP to RDMA is not supported */
2782 cr_ses
->flags
&= ~SESSION4_RDMA
;
2784 init_session(rqstp
, new, conf
, cr_ses
);
2785 nfsd4_get_session_locked(new);
2787 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
2788 NFS4_MAX_SESSIONID_LEN
);
2789 cs_slot
->sl_seqid
++;
2790 cr_ses
->seqid
= cs_slot
->sl_seqid
;
2792 /* cache solo and embedded create sessions under the client_lock */
2793 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
2794 spin_unlock(&nn
->client_lock
);
2795 /* init connection and backchannel */
2796 nfsd4_init_conn(rqstp
, conn
, new);
2797 nfsd4_put_session(new);
2802 spin_unlock(&nn
->client_lock
);
2807 __free_session(new);
2808 out_release_drc_mem
:
2809 nfsd4_put_drc_mem(&cr_ses
->fore_channel
);
2813 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
2816 case NFS4_CDFC4_FORE
:
2817 case NFS4_CDFC4_BACK
:
2819 case NFS4_CDFC4_FORE_OR_BOTH
:
2820 case NFS4_CDFC4_BACK_OR_BOTH
:
2821 *dir
= NFS4_CDFC4_BOTH
;
2824 return nfserr_inval
;
2827 __be32
nfsd4_backchannel_ctl(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_backchannel_ctl
*bc
)
2829 struct nfsd4_session
*session
= cstate
->session
;
2830 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2833 status
= nfsd4_check_cb_sec(&bc
->bc_cb_sec
);
2836 spin_lock(&nn
->client_lock
);
2837 session
->se_cb_prog
= bc
->bc_cb_program
;
2838 session
->se_cb_sec
= bc
->bc_cb_sec
;
2839 spin_unlock(&nn
->client_lock
);
2841 nfsd4_probe_callback(session
->se_client
);
2846 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
2847 struct nfsd4_compound_state
*cstate
,
2848 struct nfsd4_bind_conn_to_session
*bcts
)
2851 struct nfsd4_conn
*conn
;
2852 struct nfsd4_session
*session
;
2853 struct net
*net
= SVC_NET(rqstp
);
2854 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2856 if (!nfsd4_last_compound_op(rqstp
))
2857 return nfserr_not_only_op
;
2858 spin_lock(&nn
->client_lock
);
2859 session
= find_in_sessionid_hashtbl(&bcts
->sessionid
, net
, &status
);
2860 spin_unlock(&nn
->client_lock
);
2862 goto out_no_session
;
2863 status
= nfserr_wrong_cred
;
2864 if (!nfsd4_mach_creds_match(session
->se_client
, rqstp
))
2866 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
2869 conn
= alloc_conn(rqstp
, bcts
->dir
);
2870 status
= nfserr_jukebox
;
2873 nfsd4_init_conn(rqstp
, conn
, session
);
2876 nfsd4_put_session(session
);
2881 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
2885 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
2889 nfsd4_destroy_session(struct svc_rqst
*r
,
2890 struct nfsd4_compound_state
*cstate
,
2891 struct nfsd4_destroy_session
*sessionid
)
2893 struct nfsd4_session
*ses
;
2895 int ref_held_by_me
= 0;
2896 struct net
*net
= SVC_NET(r
);
2897 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2899 status
= nfserr_not_only_op
;
2900 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
2901 if (!nfsd4_last_compound_op(r
))
2905 dump_sessionid(__func__
, &sessionid
->sessionid
);
2906 spin_lock(&nn
->client_lock
);
2907 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
, net
, &status
);
2909 goto out_client_lock
;
2910 status
= nfserr_wrong_cred
;
2911 if (!nfsd4_mach_creds_match(ses
->se_client
, r
))
2912 goto out_put_session
;
2913 status
= mark_session_dead_locked(ses
, 1 + ref_held_by_me
);
2915 goto out_put_session
;
2916 unhash_session(ses
);
2917 spin_unlock(&nn
->client_lock
);
2919 nfsd4_probe_callback_sync(ses
->se_client
);
2921 spin_lock(&nn
->client_lock
);
2924 nfsd4_put_session_locked(ses
);
2926 spin_unlock(&nn
->client_lock
);
2931 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
2933 struct nfsd4_conn
*c
;
2935 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
2936 if (c
->cn_xprt
== xpt
) {
2943 static __be32
nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
2945 struct nfs4_client
*clp
= ses
->se_client
;
2946 struct nfsd4_conn
*c
;
2947 __be32 status
= nfs_ok
;
2950 spin_lock(&clp
->cl_lock
);
2951 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
2954 status
= nfserr_conn_not_bound_to_session
;
2955 if (clp
->cl_mach_cred
)
2957 __nfsd4_hash_conn(new, ses
);
2958 spin_unlock(&clp
->cl_lock
);
2959 ret
= nfsd4_register_conn(new);
2961 /* oops; xprt is already down: */
2962 nfsd4_conn_lost(&new->cn_xpt_user
);
2965 spin_unlock(&clp
->cl_lock
);
2970 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
2972 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
2974 return args
->opcnt
> session
->se_fchannel
.maxops
;
2977 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
2978 struct nfsd4_session
*session
)
2980 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
2982 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
2986 nfsd4_sequence(struct svc_rqst
*rqstp
,
2987 struct nfsd4_compound_state
*cstate
,
2988 struct nfsd4_sequence
*seq
)
2990 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2991 struct xdr_stream
*xdr
= &resp
->xdr
;
2992 struct nfsd4_session
*session
;
2993 struct nfs4_client
*clp
;
2994 struct nfsd4_slot
*slot
;
2995 struct nfsd4_conn
*conn
;
2998 struct net
*net
= SVC_NET(rqstp
);
2999 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
3001 if (resp
->opcnt
!= 1)
3002 return nfserr_sequence_pos
;
3005 * Will be either used or freed by nfsd4_sequence_check_conn
3008 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
3010 return nfserr_jukebox
;
3012 spin_lock(&nn
->client_lock
);
3013 session
= find_in_sessionid_hashtbl(&seq
->sessionid
, net
, &status
);
3015 goto out_no_session
;
3016 clp
= session
->se_client
;
3018 status
= nfserr_too_many_ops
;
3019 if (nfsd4_session_too_many_ops(rqstp
, session
))
3020 goto out_put_session
;
3022 status
= nfserr_req_too_big
;
3023 if (nfsd4_request_too_big(rqstp
, session
))
3024 goto out_put_session
;
3026 status
= nfserr_badslot
;
3027 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
3028 goto out_put_session
;
3030 slot
= session
->se_slots
[seq
->slotid
];
3031 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
3033 /* We do not negotiate the number of slots yet, so set the
3034 * maxslots to the session maxreqs which is used to encode
3035 * sr_highest_slotid and the sr_target_slot id to maxslots */
3036 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
3038 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
3039 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
3040 if (status
== nfserr_replay_cache
) {
3041 status
= nfserr_seq_misordered
;
3042 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
3043 goto out_put_session
;
3044 cstate
->slot
= slot
;
3045 cstate
->session
= session
;
3047 /* Return the cached reply status and set cstate->status
3048 * for nfsd4_proc_compound processing */
3049 status
= nfsd4_replay_cache_entry(resp
, seq
);
3050 cstate
->status
= nfserr_replay_cache
;
3054 goto out_put_session
;
3056 status
= nfsd4_sequence_check_conn(conn
, session
);
3059 goto out_put_session
;
3061 buflen
= (seq
->cachethis
) ?
3062 session
->se_fchannel
.maxresp_cached
:
3063 session
->se_fchannel
.maxresp_sz
;
3064 status
= (seq
->cachethis
) ? nfserr_rep_too_big_to_cache
:
3066 if (xdr_restrict_buflen(xdr
, buflen
- rqstp
->rq_auth_slack
))
3067 goto out_put_session
;
3068 svc_reserve(rqstp
, buflen
);
3071 /* Success! bump slot seqid */
3072 slot
->sl_seqid
= seq
->seqid
;
3073 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
3075 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
3077 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
3079 cstate
->slot
= slot
;
3080 cstate
->session
= session
;
3084 switch (clp
->cl_cb_state
) {
3086 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
3088 case NFSD4_CB_FAULT
:
3089 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
3092 seq
->status_flags
= 0;
3094 if (!list_empty(&clp
->cl_revoked
))
3095 seq
->status_flags
|= SEQ4_STATUS_RECALLABLE_STATE_REVOKED
;
3099 spin_unlock(&nn
->client_lock
);
3102 nfsd4_put_session_locked(session
);
3103 goto out_no_session
;
3107 nfsd4_sequence_done(struct nfsd4_compoundres
*resp
)
3109 struct nfsd4_compound_state
*cs
= &resp
->cstate
;
3111 if (nfsd4_has_session(cs
)) {
3112 if (cs
->status
!= nfserr_replay_cache
) {
3113 nfsd4_store_cache_entry(resp
);
3114 cs
->slot
->sl_flags
&= ~NFSD4_SLOT_INUSE
;
3116 /* Drop session reference that was taken in nfsd4_sequence() */
3117 nfsd4_put_session(cs
->session
);
3119 put_client_renew(cs
->clp
);
3123 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
3125 struct nfs4_client
*conf
, *unconf
;
3126 struct nfs4_client
*clp
= NULL
;
3128 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3130 spin_lock(&nn
->client_lock
);
3131 unconf
= find_unconfirmed_client(&dc
->clientid
, true, nn
);
3132 conf
= find_confirmed_client(&dc
->clientid
, true, nn
);
3133 WARN_ON_ONCE(conf
&& unconf
);
3136 if (client_has_state(conf
)) {
3137 status
= nfserr_clientid_busy
;
3140 status
= mark_client_expired_locked(conf
);
3147 status
= nfserr_stale_clientid
;
3150 if (!nfsd4_mach_creds_match(clp
, rqstp
)) {
3152 status
= nfserr_wrong_cred
;
3155 unhash_client_locked(clp
);
3157 spin_unlock(&nn
->client_lock
);
3164 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
3168 if (rc
->rca_one_fs
) {
3169 if (!cstate
->current_fh
.fh_dentry
)
3170 return nfserr_nofilehandle
;
3172 * We don't take advantage of the rca_one_fs case.
3173 * That's OK, it's optional, we can safely ignore it.
3178 status
= nfserr_complete_already
;
3179 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
3180 &cstate
->session
->se_client
->cl_flags
))
3183 status
= nfserr_stale_clientid
;
3184 if (is_client_expired(cstate
->session
->se_client
))
3186 * The following error isn't really legal.
3187 * But we only get here if the client just explicitly
3188 * destroyed the client. Surely it no longer cares what
3189 * error it gets back on an operation for the dead
3195 nfsd4_client_record_create(cstate
->session
->se_client
);
3201 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3202 struct nfsd4_setclientid
*setclid
)
3204 struct xdr_netobj clname
= setclid
->se_name
;
3205 nfs4_verifier clverifier
= setclid
->se_verf
;
3206 struct nfs4_client
*conf
, *new;
3207 struct nfs4_client
*unconf
= NULL
;
3209 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3211 new = create_client(clname
, rqstp
, &clverifier
);
3213 return nfserr_jukebox
;
3214 /* Cases below refer to rfc 3530 section 14.2.33: */
3215 spin_lock(&nn
->client_lock
);
3216 conf
= find_confirmed_client_by_name(&clname
, nn
);
3217 if (conf
&& client_has_state(conf
)) {
3219 status
= nfserr_clid_inuse
;
3220 if (clp_used_exchangeid(conf
))
3222 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
3223 char addr_str
[INET6_ADDRSTRLEN
];
3224 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
3226 dprintk("NFSD: setclientid: string in use by client "
3227 "at %s\n", addr_str
);
3231 unconf
= find_unconfirmed_client_by_name(&clname
, nn
);
3233 unhash_client_locked(unconf
);
3234 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
)) {
3235 /* case 1: probable callback update */
3236 copy_clid(new, conf
);
3237 gen_confirm(new, nn
);
3238 } else /* case 4 (new client) or cases 2, 3 (client reboot): */
3240 new->cl_minorversion
= 0;
3241 gen_callback(new, setclid
, rqstp
);
3242 add_to_unconfirmed(new);
3243 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
3244 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
3245 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
3249 spin_unlock(&nn
->client_lock
);
3253 expire_client(unconf
);
3259 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
3260 struct nfsd4_compound_state
*cstate
,
3261 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
3263 struct nfs4_client
*conf
, *unconf
;
3264 struct nfs4_client
*old
= NULL
;
3265 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
3266 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
3268 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3270 if (STALE_CLIENTID(clid
, nn
))
3271 return nfserr_stale_clientid
;
3273 spin_lock(&nn
->client_lock
);
3274 conf
= find_confirmed_client(clid
, false, nn
);
3275 unconf
= find_unconfirmed_client(clid
, false, nn
);
3277 * We try hard to give out unique clientid's, so if we get an
3278 * attempt to confirm the same clientid with a different cred,
3279 * the client may be buggy; this should never happen.
3281 * Nevertheless, RFC 7530 recommends INUSE for this case:
3283 status
= nfserr_clid_inuse
;
3284 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
3286 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
3288 /* cases below refer to rfc 3530 section 14.2.34: */
3289 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
3290 if (conf
&& same_verf(&confirm
, &conf
->cl_confirm
)) {
3291 /* case 2: probable retransmit */
3293 } else /* case 4: client hasn't noticed we rebooted yet? */
3294 status
= nfserr_stale_clientid
;
3298 if (conf
) { /* case 1: callback update */
3300 unhash_client_locked(old
);
3301 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
3302 } else { /* case 3: normal case; new or rebooted client */
3303 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
3305 status
= nfserr_clid_inuse
;
3306 if (client_has_state(old
)
3307 && !same_creds(&unconf
->cl_cred
,
3310 status
= mark_client_expired_locked(old
);
3316 move_to_confirmed(unconf
);
3319 get_client_locked(conf
);
3320 spin_unlock(&nn
->client_lock
);
3321 nfsd4_probe_callback(conf
);
3322 spin_lock(&nn
->client_lock
);
3323 put_client_renew_locked(conf
);
3325 spin_unlock(&nn
->client_lock
);
3331 static struct nfs4_file
*nfsd4_alloc_file(void)
3333 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
3336 /* OPEN Share state helper functions */
3337 static void nfsd4_init_file(struct knfsd_fh
*fh
, unsigned int hashval
,
3338 struct nfs4_file
*fp
)
3340 lockdep_assert_held(&state_lock
);
3342 atomic_set(&fp
->fi_ref
, 1);
3343 spin_lock_init(&fp
->fi_lock
);
3344 INIT_LIST_HEAD(&fp
->fi_stateids
);
3345 INIT_LIST_HEAD(&fp
->fi_delegations
);
3346 INIT_LIST_HEAD(&fp
->fi_clnt_odstate
);
3347 fh_copy_shallow(&fp
->fi_fhandle
, fh
);
3348 fp
->fi_deleg_file
= NULL
;
3349 fp
->fi_had_conflict
= false;
3350 fp
->fi_share_deny
= 0;
3351 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
3352 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
3353 #ifdef CONFIG_NFSD_PNFS
3354 INIT_LIST_HEAD(&fp
->fi_lo_states
);
3355 atomic_set(&fp
->fi_lo_recalls
, 0);
3357 hlist_add_head_rcu(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
3361 nfsd4_free_slabs(void)
3363 kmem_cache_destroy(odstate_slab
);
3364 kmem_cache_destroy(openowner_slab
);
3365 kmem_cache_destroy(lockowner_slab
);
3366 kmem_cache_destroy(file_slab
);
3367 kmem_cache_destroy(stateid_slab
);
3368 kmem_cache_destroy(deleg_slab
);
3372 nfsd4_init_slabs(void)
3374 openowner_slab
= kmem_cache_create("nfsd4_openowners",
3375 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
3376 if (openowner_slab
== NULL
)
3378 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
3379 sizeof(struct nfs4_lockowner
), 0, 0, NULL
);
3380 if (lockowner_slab
== NULL
)
3381 goto out_free_openowner_slab
;
3382 file_slab
= kmem_cache_create("nfsd4_files",
3383 sizeof(struct nfs4_file
), 0, 0, NULL
);
3384 if (file_slab
== NULL
)
3385 goto out_free_lockowner_slab
;
3386 stateid_slab
= kmem_cache_create("nfsd4_stateids",
3387 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
3388 if (stateid_slab
== NULL
)
3389 goto out_free_file_slab
;
3390 deleg_slab
= kmem_cache_create("nfsd4_delegations",
3391 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
3392 if (deleg_slab
== NULL
)
3393 goto out_free_stateid_slab
;
3394 odstate_slab
= kmem_cache_create("nfsd4_odstate",
3395 sizeof(struct nfs4_clnt_odstate
), 0, 0, NULL
);
3396 if (odstate_slab
== NULL
)
3397 goto out_free_deleg_slab
;
3400 out_free_deleg_slab
:
3401 kmem_cache_destroy(deleg_slab
);
3402 out_free_stateid_slab
:
3403 kmem_cache_destroy(stateid_slab
);
3405 kmem_cache_destroy(file_slab
);
3406 out_free_lockowner_slab
:
3407 kmem_cache_destroy(lockowner_slab
);
3408 out_free_openowner_slab
:
3409 kmem_cache_destroy(openowner_slab
);
3411 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3415 static void init_nfs4_replay(struct nfs4_replay
*rp
)
3417 rp
->rp_status
= nfserr_serverfault
;
3419 rp
->rp_buf
= rp
->rp_ibuf
;
3420 mutex_init(&rp
->rp_mutex
);
3423 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state
*cstate
,
3424 struct nfs4_stateowner
*so
)
3426 if (!nfsd4_has_session(cstate
)) {
3427 mutex_lock(&so
->so_replay
.rp_mutex
);
3428 cstate
->replay_owner
= nfs4_get_stateowner(so
);
3432 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state
*cstate
)
3434 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
3437 cstate
->replay_owner
= NULL
;
3438 mutex_unlock(&so
->so_replay
.rp_mutex
);
3439 nfs4_put_stateowner(so
);
3443 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
3445 struct nfs4_stateowner
*sop
;
3447 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
3451 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
3452 if (!sop
->so_owner
.data
) {
3453 kmem_cache_free(slab
, sop
);
3456 sop
->so_owner
.len
= owner
->len
;
3458 INIT_LIST_HEAD(&sop
->so_stateids
);
3459 sop
->so_client
= clp
;
3460 init_nfs4_replay(&sop
->so_replay
);
3461 atomic_set(&sop
->so_count
, 1);
3465 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
3467 lockdep_assert_held(&clp
->cl_lock
);
3469 list_add(&oo
->oo_owner
.so_strhash
,
3470 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
3471 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
3474 static void nfs4_unhash_openowner(struct nfs4_stateowner
*so
)
3476 unhash_openowner_locked(openowner(so
));
3479 static void nfs4_free_openowner(struct nfs4_stateowner
*so
)
3481 struct nfs4_openowner
*oo
= openowner(so
);
3483 kmem_cache_free(openowner_slab
, oo
);
3486 static const struct nfs4_stateowner_operations openowner_ops
= {
3487 .so_unhash
= nfs4_unhash_openowner
,
3488 .so_free
= nfs4_free_openowner
,
3491 static struct nfs4_ol_stateid
*
3492 nfsd4_find_existing_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3494 struct nfs4_ol_stateid
*local
, *ret
= NULL
;
3495 struct nfs4_openowner
*oo
= open
->op_openowner
;
3497 lockdep_assert_held(&fp
->fi_lock
);
3499 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
3500 /* ignore lock owners */
3501 if (local
->st_stateowner
->so_is_open_owner
== 0)
3503 if (local
->st_stateowner
== &oo
->oo_owner
) {
3505 atomic_inc(&ret
->st_stid
.sc_count
);
3512 static struct nfs4_openowner
*
3513 alloc_init_open_stateowner(unsigned int strhashval
, struct nfsd4_open
*open
,
3514 struct nfsd4_compound_state
*cstate
)
3516 struct nfs4_client
*clp
= cstate
->clp
;
3517 struct nfs4_openowner
*oo
, *ret
;
3519 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
3522 oo
->oo_owner
.so_ops
= &openowner_ops
;
3523 oo
->oo_owner
.so_is_open_owner
= 1;
3524 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
3526 if (nfsd4_has_session(cstate
))
3527 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3529 oo
->oo_last_closed_stid
= NULL
;
3530 INIT_LIST_HEAD(&oo
->oo_close_lru
);
3531 spin_lock(&clp
->cl_lock
);
3532 ret
= find_openstateowner_str_locked(strhashval
, open
, clp
);
3534 hash_openowner(oo
, clp
, strhashval
);
3537 nfs4_free_stateowner(&oo
->oo_owner
);
3539 spin_unlock(&clp
->cl_lock
);
3543 static struct nfs4_ol_stateid
*
3544 init_open_stateid(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3547 struct nfs4_openowner
*oo
= open
->op_openowner
;
3548 struct nfs4_ol_stateid
*retstp
= NULL
;
3549 struct nfs4_ol_stateid
*stp
;
3552 /* We are moving these outside of the spinlocks to avoid the warnings */
3553 mutex_init(&stp
->st_mutex
);
3554 mutex_lock(&stp
->st_mutex
);
3556 spin_lock(&oo
->oo_owner
.so_client
->cl_lock
);
3557 spin_lock(&fp
->fi_lock
);
3559 retstp
= nfsd4_find_existing_open(fp
, open
);
3563 open
->op_stp
= NULL
;
3564 atomic_inc(&stp
->st_stid
.sc_count
);
3565 stp
->st_stid
.sc_type
= NFS4_OPEN_STID
;
3566 INIT_LIST_HEAD(&stp
->st_locks
);
3567 stp
->st_stateowner
= nfs4_get_stateowner(&oo
->oo_owner
);
3569 stp
->st_stid
.sc_file
= fp
;
3570 stp
->st_access_bmap
= 0;
3571 stp
->st_deny_bmap
= 0;
3572 stp
->st_openstp
= NULL
;
3573 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
3574 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3577 spin_unlock(&fp
->fi_lock
);
3578 spin_unlock(&oo
->oo_owner
.so_client
->cl_lock
);
3580 mutex_lock(&retstp
->st_mutex
);
3581 /* To keep mutex tracking happy */
3582 mutex_unlock(&stp
->st_mutex
);
3589 * In the 4.0 case we need to keep the owners around a little while to handle
3590 * CLOSE replay. We still do need to release any file access that is held by
3591 * them before returning however.
3594 move_to_close_lru(struct nfs4_ol_stateid
*s
, struct net
*net
)
3596 struct nfs4_ol_stateid
*last
;
3597 struct nfs4_openowner
*oo
= openowner(s
->st_stateowner
);
3598 struct nfsd_net
*nn
= net_generic(s
->st_stid
.sc_client
->net
,
3601 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
3604 * We know that we hold one reference via nfsd4_close, and another
3605 * "persistent" reference for the client. If the refcount is higher
3606 * than 2, then there are still calls in progress that are using this
3607 * stateid. We can't put the sc_file reference until they are finished.
3608 * Wait for the refcount to drop to 2. Since it has been unhashed,
3609 * there should be no danger of the refcount going back up again at
3612 wait_event(close_wq
, atomic_read(&s
->st_stid
.sc_count
) == 2);
3614 release_all_access(s
);
3615 if (s
->st_stid
.sc_file
) {
3616 put_nfs4_file(s
->st_stid
.sc_file
);
3617 s
->st_stid
.sc_file
= NULL
;
3620 spin_lock(&nn
->client_lock
);
3621 last
= oo
->oo_last_closed_stid
;
3622 oo
->oo_last_closed_stid
= s
;
3623 list_move_tail(&oo
->oo_close_lru
, &nn
->close_lru
);
3624 oo
->oo_time
= get_seconds();
3625 spin_unlock(&nn
->client_lock
);
3627 nfs4_put_stid(&last
->st_stid
);
3630 /* search file_hashtbl[] for file */
3631 static struct nfs4_file
*
3632 find_file_locked(struct knfsd_fh
*fh
, unsigned int hashval
)
3634 struct nfs4_file
*fp
;
3636 hlist_for_each_entry_rcu(fp
, &file_hashtbl
[hashval
], fi_hash
) {
3637 if (fh_match(&fp
->fi_fhandle
, fh
)) {
3638 if (atomic_inc_not_zero(&fp
->fi_ref
))
3646 find_file(struct knfsd_fh
*fh
)
3648 struct nfs4_file
*fp
;
3649 unsigned int hashval
= file_hashval(fh
);
3652 fp
= find_file_locked(fh
, hashval
);
3657 static struct nfs4_file
*
3658 find_or_add_file(struct nfs4_file
*new, struct knfsd_fh
*fh
)
3660 struct nfs4_file
*fp
;
3661 unsigned int hashval
= file_hashval(fh
);
3664 fp
= find_file_locked(fh
, hashval
);
3669 spin_lock(&state_lock
);
3670 fp
= find_file_locked(fh
, hashval
);
3671 if (likely(fp
== NULL
)) {
3672 nfsd4_init_file(fh
, hashval
, new);
3675 spin_unlock(&state_lock
);
3681 * Called to check deny when READ with all zero stateid or
3682 * WRITE with all zero or all one stateid
3685 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
3687 struct nfs4_file
*fp
;
3688 __be32 ret
= nfs_ok
;
3690 fp
= find_file(¤t_fh
->fh_handle
);
3693 /* Check for conflicting share reservations */
3694 spin_lock(&fp
->fi_lock
);
3695 if (fp
->fi_share_deny
& deny_type
)
3696 ret
= nfserr_locked
;
3697 spin_unlock(&fp
->fi_lock
);
3702 static void nfsd4_cb_recall_prepare(struct nfsd4_callback
*cb
)
3704 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3705 struct nfsd_net
*nn
= net_generic(dp
->dl_stid
.sc_client
->net
,
3708 block_delegations(&dp
->dl_stid
.sc_file
->fi_fhandle
);
3711 * We can't do this in nfsd_break_deleg_cb because it is
3712 * already holding inode->i_lock.
3714 * If the dl_time != 0, then we know that it has already been
3715 * queued for a lease break. Don't queue it again.
3717 spin_lock(&state_lock
);
3718 if (dp
->dl_time
== 0) {
3719 dp
->dl_time
= get_seconds();
3720 list_add_tail(&dp
->dl_recall_lru
, &nn
->del_recall_lru
);
3722 spin_unlock(&state_lock
);
3725 static int nfsd4_cb_recall_done(struct nfsd4_callback
*cb
,
3726 struct rpc_task
*task
)
3728 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3730 if (dp
->dl_stid
.sc_type
== NFS4_CLOSED_DELEG_STID
)
3733 switch (task
->tk_status
) {
3737 case -NFS4ERR_BAD_STATEID
:
3739 * Race: client probably got cb_recall before open reply
3740 * granting delegation.
3742 if (dp
->dl_retries
--) {
3743 rpc_delay(task
, 2 * HZ
);
3752 static void nfsd4_cb_recall_release(struct nfsd4_callback
*cb
)
3754 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3756 nfs4_put_stid(&dp
->dl_stid
);
3759 static const struct nfsd4_callback_ops nfsd4_cb_recall_ops
= {
3760 .prepare
= nfsd4_cb_recall_prepare
,
3761 .done
= nfsd4_cb_recall_done
,
3762 .release
= nfsd4_cb_recall_release
,
3765 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
3768 * We're assuming the state code never drops its reference
3769 * without first removing the lease. Since we're in this lease
3770 * callback (and since the lease code is serialized by the kernel
3771 * lock) we know the server hasn't removed the lease yet, we know
3772 * it's safe to take a reference.
3774 atomic_inc(&dp
->dl_stid
.sc_count
);
3775 nfsd4_run_cb(&dp
->dl_recall
);
3778 /* Called from break_lease() with i_lock held. */
3780 nfsd_break_deleg_cb(struct file_lock
*fl
)
3783 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
3784 struct nfs4_delegation
*dp
;
3787 WARN(1, "(%p)->fl_owner NULL\n", fl
);
3790 if (fp
->fi_had_conflict
) {
3791 WARN(1, "duplicate break on %p\n", fp
);
3795 * We don't want the locks code to timeout the lease for us;
3796 * we'll remove it ourself if a delegation isn't returned
3799 fl
->fl_break_time
= 0;
3801 spin_lock(&fp
->fi_lock
);
3802 fp
->fi_had_conflict
= true;
3804 * If there are no delegations on the list, then return true
3805 * so that the lease code will go ahead and delete it.
3807 if (list_empty(&fp
->fi_delegations
))
3810 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
3811 nfsd_break_one_deleg(dp
);
3812 spin_unlock(&fp
->fi_lock
);
3817 nfsd_change_deleg_cb(struct file_lock
*onlist
, int arg
,
3818 struct list_head
*dispose
)
3821 return lease_modify(onlist
, arg
, dispose
);
3826 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
3827 .lm_break
= nfsd_break_deleg_cb
,
3828 .lm_change
= nfsd_change_deleg_cb
,
3831 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
3833 if (nfsd4_has_session(cstate
))
3835 if (seqid
== so
->so_seqid
- 1)
3836 return nfserr_replay_me
;
3837 if (seqid
== so
->so_seqid
)
3839 return nfserr_bad_seqid
;
3842 static __be32
lookup_clientid(clientid_t
*clid
,
3843 struct nfsd4_compound_state
*cstate
,
3844 struct nfsd_net
*nn
)
3846 struct nfs4_client
*found
;
3849 found
= cstate
->clp
;
3850 if (!same_clid(&found
->cl_clientid
, clid
))
3851 return nfserr_stale_clientid
;
3855 if (STALE_CLIENTID(clid
, nn
))
3856 return nfserr_stale_clientid
;
3859 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3860 * cached already then we know this is for is for v4.0 and "sessions"
3863 WARN_ON_ONCE(cstate
->session
);
3864 spin_lock(&nn
->client_lock
);
3865 found
= find_confirmed_client(clid
, false, nn
);
3867 spin_unlock(&nn
->client_lock
);
3868 return nfserr_expired
;
3870 atomic_inc(&found
->cl_refcount
);
3871 spin_unlock(&nn
->client_lock
);
3873 /* Cache the nfs4_client in cstate! */
3874 cstate
->clp
= found
;
3879 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
3880 struct nfsd4_open
*open
, struct nfsd_net
*nn
)
3882 clientid_t
*clientid
= &open
->op_clientid
;
3883 struct nfs4_client
*clp
= NULL
;
3884 unsigned int strhashval
;
3885 struct nfs4_openowner
*oo
= NULL
;
3888 if (STALE_CLIENTID(&open
->op_clientid
, nn
))
3889 return nfserr_stale_clientid
;
3891 * In case we need it later, after we've already created the
3892 * file and don't want to risk a further failure:
3894 open
->op_file
= nfsd4_alloc_file();
3895 if (open
->op_file
== NULL
)
3896 return nfserr_jukebox
;
3898 status
= lookup_clientid(clientid
, cstate
, nn
);
3903 strhashval
= ownerstr_hashval(&open
->op_owner
);
3904 oo
= find_openstateowner_str(strhashval
, open
, clp
);
3905 open
->op_openowner
= oo
;
3909 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
3910 /* Replace unconfirmed owners without checking for replay. */
3911 release_openowner(oo
);
3912 open
->op_openowner
= NULL
;
3915 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
3920 oo
= alloc_init_open_stateowner(strhashval
, open
, cstate
);
3922 return nfserr_jukebox
;
3923 open
->op_openowner
= oo
;
3925 open
->op_stp
= nfs4_alloc_open_stateid(clp
);
3927 return nfserr_jukebox
;
3929 if (nfsd4_has_session(cstate
) &&
3930 (cstate
->current_fh
.fh_export
->ex_flags
& NFSEXP_PNFS
)) {
3931 open
->op_odstate
= alloc_clnt_odstate(clp
);
3932 if (!open
->op_odstate
)
3933 return nfserr_jukebox
;
3939 static inline __be32
3940 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
3942 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
3943 return nfserr_openmode
;
3948 static int share_access_to_flags(u32 share_access
)
3950 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
3953 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
3955 struct nfs4_stid
*ret
;
3957 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
3960 return delegstateid(ret
);
3963 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
3965 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
3966 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
3970 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfsd4_open
*open
,
3971 struct nfs4_delegation
**dp
)
3974 __be32 status
= nfserr_bad_stateid
;
3975 struct nfs4_delegation
*deleg
;
3977 deleg
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
3980 flags
= share_access_to_flags(open
->op_share_access
);
3981 status
= nfs4_check_delegmode(deleg
, flags
);
3983 nfs4_put_stid(&deleg
->dl_stid
);
3988 if (!nfsd4_is_deleg_cur(open
))
3992 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3996 static inline int nfs4_access_to_access(u32 nfs4_access
)
4000 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
4001 flags
|= NFSD_MAY_READ
;
4002 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
4003 flags
|= NFSD_MAY_WRITE
;
4007 static inline __be32
4008 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
4009 struct nfsd4_open
*open
)
4011 struct iattr iattr
= {
4012 .ia_valid
= ATTR_SIZE
,
4015 if (!open
->op_truncate
)
4017 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
4018 return nfserr_inval
;
4019 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
4022 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
4023 struct svc_fh
*cur_fh
, struct nfs4_ol_stateid
*stp
,
4024 struct nfsd4_open
*open
)
4026 struct file
*filp
= NULL
;
4028 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
4029 int access
= nfs4_access_to_access(open
->op_share_access
);
4030 unsigned char old_access_bmap
, old_deny_bmap
;
4032 spin_lock(&fp
->fi_lock
);
4035 * Are we trying to set a deny mode that would conflict with
4038 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
4039 if (status
!= nfs_ok
) {
4040 spin_unlock(&fp
->fi_lock
);
4044 /* set access to the file */
4045 status
= nfs4_file_get_access(fp
, open
->op_share_access
);
4046 if (status
!= nfs_ok
) {
4047 spin_unlock(&fp
->fi_lock
);
4051 /* Set access bits in stateid */
4052 old_access_bmap
= stp
->st_access_bmap
;
4053 set_access(open
->op_share_access
, stp
);
4055 /* Set new deny mask */
4056 old_deny_bmap
= stp
->st_deny_bmap
;
4057 set_deny(open
->op_share_deny
, stp
);
4058 fp
->fi_share_deny
|= (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
4060 if (!fp
->fi_fds
[oflag
]) {
4061 spin_unlock(&fp
->fi_lock
);
4062 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
, &filp
);
4064 goto out_put_access
;
4065 spin_lock(&fp
->fi_lock
);
4066 if (!fp
->fi_fds
[oflag
]) {
4067 fp
->fi_fds
[oflag
] = filp
;
4071 spin_unlock(&fp
->fi_lock
);
4075 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
4077 goto out_put_access
;
4081 stp
->st_access_bmap
= old_access_bmap
;
4082 nfs4_file_put_access(fp
, open
->op_share_access
);
4083 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap
), stp
);
4088 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
)
4091 unsigned char old_deny_bmap
= stp
->st_deny_bmap
;
4093 if (!test_access(open
->op_share_access
, stp
))
4094 return nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, stp
, open
);
4096 /* test and set deny mode */
4097 spin_lock(&fp
->fi_lock
);
4098 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
4099 if (status
== nfs_ok
) {
4100 set_deny(open
->op_share_deny
, stp
);
4101 fp
->fi_share_deny
|=
4102 (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
4104 spin_unlock(&fp
->fi_lock
);
4106 if (status
!= nfs_ok
)
4109 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
4110 if (status
!= nfs_ok
)
4111 reset_union_bmap_deny(old_deny_bmap
, stp
);
4115 /* Should we give out recallable state?: */
4116 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
4118 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
4121 * In the sessions case, since we don't have to establish a
4122 * separate connection for callbacks, we assume it's OK
4123 * until we hear otherwise:
4125 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
4128 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_file
*fp
, int flag
)
4130 struct file_lock
*fl
;
4132 fl
= locks_alloc_lock();
4135 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
4136 fl
->fl_flags
= FL_DELEG
;
4137 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
4138 fl
->fl_end
= OFFSET_MAX
;
4139 fl
->fl_owner
= (fl_owner_t
)fp
;
4140 fl
->fl_pid
= current
->tgid
;
4145 * nfs4_setlease - Obtain a delegation by requesting lease from vfs layer
4146 * @dp: a pointer to the nfs4_delegation we're adding.
4149 * On success: Return code will be 0 on success.
4151 * On error: -EAGAIN if there was an existing delegation.
4152 * nonzero if there is an error in other cases.
4156 static int nfs4_setlease(struct nfs4_delegation
*dp
)
4158 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
4159 struct file_lock
*fl
;
4163 fl
= nfs4_alloc_init_lease(fp
, NFS4_OPEN_DELEGATE_READ
);
4166 filp
= find_readable_file(fp
);
4168 /* We should always have a readable file here */
4170 locks_free_lock(fl
);
4174 status
= vfs_setlease(filp
, fl
->fl_type
, &fl
, NULL
);
4176 locks_free_lock(fl
);
4179 spin_lock(&state_lock
);
4180 spin_lock(&fp
->fi_lock
);
4181 /* Did the lease get broken before we took the lock? */
4183 if (fp
->fi_had_conflict
)
4186 if (fp
->fi_deleg_file
) {
4187 status
= hash_delegation_locked(dp
, fp
);
4190 fp
->fi_deleg_file
= filp
;
4191 fp
->fi_delegees
= 0;
4192 status
= hash_delegation_locked(dp
, fp
);
4193 spin_unlock(&fp
->fi_lock
);
4194 spin_unlock(&state_lock
);
4196 /* Should never happen, this is a new fi_deleg_file */
4202 spin_unlock(&fp
->fi_lock
);
4203 spin_unlock(&state_lock
);
4209 static struct nfs4_delegation
*
4210 nfs4_set_delegation(struct nfs4_client
*clp
, struct svc_fh
*fh
,
4211 struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*odstate
)
4214 struct nfs4_delegation
*dp
;
4216 if (fp
->fi_had_conflict
)
4217 return ERR_PTR(-EAGAIN
);
4219 spin_lock(&state_lock
);
4220 spin_lock(&fp
->fi_lock
);
4221 status
= nfs4_get_existing_delegation(clp
, fp
);
4222 spin_unlock(&fp
->fi_lock
);
4223 spin_unlock(&state_lock
);
4226 return ERR_PTR(status
);
4228 dp
= alloc_init_deleg(clp
, fh
, odstate
);
4230 return ERR_PTR(-ENOMEM
);
4233 spin_lock(&state_lock
);
4234 spin_lock(&fp
->fi_lock
);
4235 dp
->dl_stid
.sc_file
= fp
;
4236 if (!fp
->fi_deleg_file
) {
4237 spin_unlock(&fp
->fi_lock
);
4238 spin_unlock(&state_lock
);
4239 status
= nfs4_setlease(dp
);
4242 if (fp
->fi_had_conflict
) {
4246 status
= hash_delegation_locked(dp
, fp
);
4248 spin_unlock(&fp
->fi_lock
);
4249 spin_unlock(&state_lock
);
4252 put_clnt_odstate(dp
->dl_clnt_odstate
);
4253 nfs4_put_stid(&dp
->dl_stid
);
4254 return ERR_PTR(status
);
4259 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
4261 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4262 if (status
== -EAGAIN
)
4263 open
->op_why_no_deleg
= WND4_CONTENTION
;
4265 open
->op_why_no_deleg
= WND4_RESOURCE
;
4266 switch (open
->op_deleg_want
) {
4267 case NFS4_SHARE_WANT_READ_DELEG
:
4268 case NFS4_SHARE_WANT_WRITE_DELEG
:
4269 case NFS4_SHARE_WANT_ANY_DELEG
:
4271 case NFS4_SHARE_WANT_CANCEL
:
4272 open
->op_why_no_deleg
= WND4_CANCELLED
;
4274 case NFS4_SHARE_WANT_NO_DELEG
:
4281 * Attempt to hand out a delegation.
4283 * Note we don't support write delegations, and won't until the vfs has
4284 * proper support for them.
4287 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
,
4288 struct nfs4_ol_stateid
*stp
)
4290 struct nfs4_delegation
*dp
;
4291 struct nfs4_openowner
*oo
= openowner(stp
->st_stateowner
);
4292 struct nfs4_client
*clp
= stp
->st_stid
.sc_client
;
4296 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
4297 open
->op_recall
= 0;
4298 switch (open
->op_claim_type
) {
4299 case NFS4_OPEN_CLAIM_PREVIOUS
:
4301 open
->op_recall
= 1;
4302 if (open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_READ
)
4305 case NFS4_OPEN_CLAIM_NULL
:
4306 case NFS4_OPEN_CLAIM_FH
:
4308 * Let's not give out any delegations till everyone's
4309 * had the chance to reclaim theirs, *and* until
4310 * NLM locks have all been reclaimed:
4312 if (locks_in_grace(clp
->net
))
4314 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
4317 * Also, if the file was opened for write or
4318 * create, there's a good chance the client's
4319 * about to write to it, resulting in an
4320 * immediate recall (since we don't support
4321 * write delegations):
4323 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
4325 if (open
->op_create
== NFS4_OPEN_CREATE
)
4331 dp
= nfs4_set_delegation(clp
, fh
, stp
->st_stid
.sc_file
, stp
->st_clnt_odstate
);
4335 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
4337 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
4338 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
4339 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_READ
;
4340 nfs4_put_stid(&dp
->dl_stid
);
4343 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE
;
4344 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
4345 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
) {
4346 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4347 open
->op_recall
= 1;
4350 /* 4.1 client asking for a delegation? */
4351 if (open
->op_deleg_want
)
4352 nfsd4_open_deleg_none_ext(open
, status
);
4356 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
4357 struct nfs4_delegation
*dp
)
4359 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
4360 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4361 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4362 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
4363 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
4364 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4365 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4366 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
4368 /* Otherwise the client must be confused wanting a delegation
4369 * it already has, therefore we don't return
4370 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4375 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
4377 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
4378 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
4379 struct nfs4_file
*fp
= NULL
;
4380 struct nfs4_ol_stateid
*stp
= NULL
;
4381 struct nfs4_delegation
*dp
= NULL
;
4385 * Lookup file; if found, lookup stateid and check open request,
4386 * and check for delegations in the process of being recalled.
4387 * If not found, create the nfs4_file struct
4389 fp
= find_or_add_file(open
->op_file
, ¤t_fh
->fh_handle
);
4390 if (fp
!= open
->op_file
) {
4391 status
= nfs4_check_deleg(cl
, open
, &dp
);
4394 spin_lock(&fp
->fi_lock
);
4395 stp
= nfsd4_find_existing_open(fp
, open
);
4396 spin_unlock(&fp
->fi_lock
);
4398 open
->op_file
= NULL
;
4399 status
= nfserr_bad_stateid
;
4400 if (nfsd4_is_deleg_cur(open
))
4405 * OPEN the file, or upgrade an existing OPEN.
4406 * If truncate fails, the OPEN fails.
4409 /* Stateid was found, this is an OPEN upgrade */
4410 mutex_lock(&stp
->st_mutex
);
4411 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
4413 mutex_unlock(&stp
->st_mutex
);
4417 /* stp is returned locked. */
4418 stp
= init_open_stateid(fp
, open
);
4419 /* See if we lost the race to some other thread */
4420 if (stp
->st_access_bmap
!= 0) {
4421 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
,
4424 mutex_unlock(&stp
->st_mutex
);
4429 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, stp
, open
);
4431 mutex_unlock(&stp
->st_mutex
);
4432 release_open_stateid(stp
);
4436 stp
->st_clnt_odstate
= find_or_hash_clnt_odstate(fp
,
4438 if (stp
->st_clnt_odstate
== open
->op_odstate
)
4439 open
->op_odstate
= NULL
;
4442 nfs4_inc_and_copy_stateid(&open
->op_stateid
, &stp
->st_stid
);
4443 mutex_unlock(&stp
->st_mutex
);
4445 if (nfsd4_has_session(&resp
->cstate
)) {
4446 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
4447 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4448 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
4454 * Attempt to hand out a delegation. No error return, because the
4455 * OPEN succeeds even if we fail.
4457 nfs4_open_delegation(current_fh
, open
, stp
);
4461 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
4462 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
4464 /* 4.1 client trying to upgrade/downgrade delegation? */
4465 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
4466 open
->op_deleg_want
)
4467 nfsd4_deleg_xgrade_none_ext(open
, dp
);
4471 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
4472 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
4474 * To finish the open response, we just need to set the rflags.
4476 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
4477 if (nfsd4_has_session(&resp
->cstate
))
4478 open
->op_rflags
|= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK
;
4479 else if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
))
4480 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
4483 nfs4_put_stid(&dp
->dl_stid
);
4485 nfs4_put_stid(&stp
->st_stid
);
4490 void nfsd4_cleanup_open_state(struct nfsd4_compound_state
*cstate
,
4491 struct nfsd4_open
*open
)
4493 if (open
->op_openowner
) {
4494 struct nfs4_stateowner
*so
= &open
->op_openowner
->oo_owner
;
4496 nfsd4_cstate_assign_replay(cstate
, so
);
4497 nfs4_put_stateowner(so
);
4500 kmem_cache_free(file_slab
, open
->op_file
);
4502 nfs4_put_stid(&open
->op_stp
->st_stid
);
4503 if (open
->op_odstate
)
4504 kmem_cache_free(odstate_slab
, open
->op_odstate
);
4508 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4511 struct nfs4_client
*clp
;
4513 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
4515 dprintk("process_renew(%08x/%08x): starting\n",
4516 clid
->cl_boot
, clid
->cl_id
);
4517 status
= lookup_clientid(clid
, cstate
, nn
);
4521 status
= nfserr_cb_path_down
;
4522 if (!list_empty(&clp
->cl_delegations
)
4523 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
4531 nfsd4_end_grace(struct nfsd_net
*nn
)
4533 /* do nothing if grace period already ended */
4534 if (nn
->grace_ended
)
4537 dprintk("NFSD: end of grace period\n");
4538 nn
->grace_ended
= true;
4540 * If the server goes down again right now, an NFSv4
4541 * client will still be allowed to reclaim after it comes back up,
4542 * even if it hasn't yet had a chance to reclaim state this time.
4545 nfsd4_record_grace_done(nn
);
4547 * At this point, NFSv4 clients can still reclaim. But if the
4548 * server crashes, any that have not yet reclaimed will be out
4549 * of luck on the next boot.
4551 * (NFSv4.1+ clients are considered to have reclaimed once they
4552 * call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
4553 * have reclaimed after their first OPEN.)
4555 locks_end_grace(&nn
->nfsd4_manager
);
4557 * At this point, and once lockd and/or any other containers
4558 * exit their grace period, further reclaims will fail and
4559 * regular locking can resume.
4564 nfs4_laundromat(struct nfsd_net
*nn
)
4566 struct nfs4_client
*clp
;
4567 struct nfs4_openowner
*oo
;
4568 struct nfs4_delegation
*dp
;
4569 struct nfs4_ol_stateid
*stp
;
4570 struct nfsd4_blocked_lock
*nbl
;
4571 struct list_head
*pos
, *next
, reaplist
;
4572 time_t cutoff
= get_seconds() - nn
->nfsd4_lease
;
4573 time_t t
, new_timeo
= nn
->nfsd4_lease
;
4575 dprintk("NFSD: laundromat service - starting\n");
4576 nfsd4_end_grace(nn
);
4577 INIT_LIST_HEAD(&reaplist
);
4578 spin_lock(&nn
->client_lock
);
4579 list_for_each_safe(pos
, next
, &nn
->client_lru
) {
4580 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4581 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
4582 t
= clp
->cl_time
- cutoff
;
4583 new_timeo
= min(new_timeo
, t
);
4586 if (mark_client_expired_locked(clp
)) {
4587 dprintk("NFSD: client in use (clientid %08x)\n",
4588 clp
->cl_clientid
.cl_id
);
4591 list_add(&clp
->cl_lru
, &reaplist
);
4593 spin_unlock(&nn
->client_lock
);
4594 list_for_each_safe(pos
, next
, &reaplist
) {
4595 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4596 dprintk("NFSD: purging unused client (clientid %08x)\n",
4597 clp
->cl_clientid
.cl_id
);
4598 list_del_init(&clp
->cl_lru
);
4601 spin_lock(&state_lock
);
4602 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
4603 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4604 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
4605 t
= dp
->dl_time
- cutoff
;
4606 new_timeo
= min(new_timeo
, t
);
4609 WARN_ON(!unhash_delegation_locked(dp
));
4610 list_add(&dp
->dl_recall_lru
, &reaplist
);
4612 spin_unlock(&state_lock
);
4613 while (!list_empty(&reaplist
)) {
4614 dp
= list_first_entry(&reaplist
, struct nfs4_delegation
,
4616 list_del_init(&dp
->dl_recall_lru
);
4617 revoke_delegation(dp
);
4620 spin_lock(&nn
->client_lock
);
4621 while (!list_empty(&nn
->close_lru
)) {
4622 oo
= list_first_entry(&nn
->close_lru
, struct nfs4_openowner
,
4624 if (time_after((unsigned long)oo
->oo_time
,
4625 (unsigned long)cutoff
)) {
4626 t
= oo
->oo_time
- cutoff
;
4627 new_timeo
= min(new_timeo
, t
);
4630 list_del_init(&oo
->oo_close_lru
);
4631 stp
= oo
->oo_last_closed_stid
;
4632 oo
->oo_last_closed_stid
= NULL
;
4633 spin_unlock(&nn
->client_lock
);
4634 nfs4_put_stid(&stp
->st_stid
);
4635 spin_lock(&nn
->client_lock
);
4637 spin_unlock(&nn
->client_lock
);
4640 * It's possible for a client to try and acquire an already held lock
4641 * that is being held for a long time, and then lose interest in it.
4642 * So, we clean out any un-revisited request after a lease period
4643 * under the assumption that the client is no longer interested.
4645 * RFC5661, sec. 9.6 states that the client must not rely on getting
4646 * notifications and must continue to poll for locks, even when the
4647 * server supports them. Thus this shouldn't lead to clients blocking
4648 * indefinitely once the lock does become free.
4650 BUG_ON(!list_empty(&reaplist
));
4651 spin_lock(&nn
->blocked_locks_lock
);
4652 while (!list_empty(&nn
->blocked_locks_lru
)) {
4653 nbl
= list_first_entry(&nn
->blocked_locks_lru
,
4654 struct nfsd4_blocked_lock
, nbl_lru
);
4655 if (time_after((unsigned long)nbl
->nbl_time
,
4656 (unsigned long)cutoff
)) {
4657 t
= nbl
->nbl_time
- cutoff
;
4658 new_timeo
= min(new_timeo
, t
);
4661 list_move(&nbl
->nbl_lru
, &reaplist
);
4662 list_del_init(&nbl
->nbl_list
);
4664 spin_unlock(&nn
->blocked_locks_lock
);
4666 while (!list_empty(&reaplist
)) {
4667 nbl
= list_first_entry(&nn
->blocked_locks_lru
,
4668 struct nfsd4_blocked_lock
, nbl_lru
);
4669 list_del_init(&nbl
->nbl_lru
);
4670 posix_unblock_lock(&nbl
->nbl_lock
);
4671 free_blocked_lock(nbl
);
4674 new_timeo
= max_t(time_t, new_timeo
, NFSD_LAUNDROMAT_MINTIMEOUT
);
4678 static struct workqueue_struct
*laundry_wq
;
4679 static void laundromat_main(struct work_struct
*);
4682 laundromat_main(struct work_struct
*laundry
)
4685 struct delayed_work
*dwork
= to_delayed_work(laundry
);
4686 struct nfsd_net
*nn
= container_of(dwork
, struct nfsd_net
,
4689 t
= nfs4_laundromat(nn
);
4690 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
4691 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, t
*HZ
);
4694 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stid
*stp
)
4696 if (!fh_match(&fhp
->fh_handle
, &stp
->sc_file
->fi_fhandle
))
4697 return nfserr_bad_stateid
;
4702 access_permit_read(struct nfs4_ol_stateid
*stp
)
4704 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
4705 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
4706 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
4710 access_permit_write(struct nfs4_ol_stateid
*stp
)
4712 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
4713 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
4717 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
4719 __be32 status
= nfserr_openmode
;
4721 /* For lock stateid's, we test the parent open, not the lock: */
4722 if (stp
->st_openstp
)
4723 stp
= stp
->st_openstp
;
4724 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
4726 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
4733 static inline __be32
4734 check_special_stateids(struct net
*net
, svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
4736 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
4738 else if (opens_in_grace(net
)) {
4739 /* Answer in remaining cases depends on existence of
4740 * conflicting state; so we must wait out the grace period. */
4741 return nfserr_grace
;
4742 } else if (flags
& WR_STATE
)
4743 return nfs4_share_conflict(current_fh
,
4744 NFS4_SHARE_DENY_WRITE
);
4745 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4746 return nfs4_share_conflict(current_fh
,
4747 NFS4_SHARE_DENY_READ
);
4751 * Allow READ/WRITE during grace period on recovered state only for files
4752 * that are not able to provide mandatory locking.
4755 grace_disallows_io(struct net
*net
, struct inode
*inode
)
4757 return opens_in_grace(net
) && mandatory_lock(inode
);
4760 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
4763 * When sessions are used the stateid generation number is ignored
4766 if (has_session
&& in
->si_generation
== 0)
4769 if (in
->si_generation
== ref
->si_generation
)
4772 /* If the client sends us a stateid from the future, it's buggy: */
4773 if (nfsd4_stateid_generation_after(in
, ref
))
4774 return nfserr_bad_stateid
;
4776 * However, we could see a stateid from the past, even from a
4777 * non-buggy client. For example, if the client sends a lock
4778 * while some IO is outstanding, the lock may bump si_generation
4779 * while the IO is still in flight. The client could avoid that
4780 * situation by waiting for responses on all the IO requests,
4781 * but better performance may result in retrying IO that
4782 * receives an old_stateid error if requests are rarely
4783 * reordered in flight:
4785 return nfserr_old_stateid
;
4788 static __be32
nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid
*ols
)
4790 if (ols
->st_stateowner
->so_is_open_owner
&&
4791 !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
4792 return nfserr_bad_stateid
;
4796 static __be32
nfsd4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
4798 struct nfs4_stid
*s
;
4799 __be32 status
= nfserr_bad_stateid
;
4801 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
4803 /* Client debugging aid. */
4804 if (!same_clid(&stateid
->si_opaque
.so_clid
, &cl
->cl_clientid
)) {
4805 char addr_str
[INET6_ADDRSTRLEN
];
4806 rpc_ntop((struct sockaddr
*)&cl
->cl_addr
, addr_str
,
4808 pr_warn_ratelimited("NFSD: client %s testing state ID "
4809 "with incorrect client ID\n", addr_str
);
4812 spin_lock(&cl
->cl_lock
);
4813 s
= find_stateid_locked(cl
, stateid
);
4816 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
4819 switch (s
->sc_type
) {
4820 case NFS4_DELEG_STID
:
4823 case NFS4_REVOKED_DELEG_STID
:
4824 status
= nfserr_deleg_revoked
;
4826 case NFS4_OPEN_STID
:
4827 case NFS4_LOCK_STID
:
4828 status
= nfsd4_check_openowner_confirmed(openlockstateid(s
));
4831 printk("unknown stateid type %x\n", s
->sc_type
);
4833 case NFS4_CLOSED_STID
:
4834 case NFS4_CLOSED_DELEG_STID
:
4835 status
= nfserr_bad_stateid
;
4838 spin_unlock(&cl
->cl_lock
);
4843 nfsd4_lookup_stateid(struct nfsd4_compound_state
*cstate
,
4844 stateid_t
*stateid
, unsigned char typemask
,
4845 struct nfs4_stid
**s
, struct nfsd_net
*nn
)
4849 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
4850 return nfserr_bad_stateid
;
4851 status
= lookup_clientid(&stateid
->si_opaque
.so_clid
, cstate
, nn
);
4852 if (status
== nfserr_stale_clientid
) {
4853 if (cstate
->session
)
4854 return nfserr_bad_stateid
;
4855 return nfserr_stale_stateid
;
4859 *s
= find_stateid_by_type(cstate
->clp
, stateid
, typemask
);
4861 return nfserr_bad_stateid
;
4865 static struct file
*
4866 nfs4_find_file(struct nfs4_stid
*s
, int flags
)
4871 switch (s
->sc_type
) {
4872 case NFS4_DELEG_STID
:
4873 if (WARN_ON_ONCE(!s
->sc_file
->fi_deleg_file
))
4875 return get_file(s
->sc_file
->fi_deleg_file
);
4876 case NFS4_OPEN_STID
:
4877 case NFS4_LOCK_STID
:
4878 if (flags
& RD_STATE
)
4879 return find_readable_file(s
->sc_file
);
4881 return find_writeable_file(s
->sc_file
);
4889 nfs4_check_olstateid(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*ols
, int flags
)
4893 status
= nfsd4_check_openowner_confirmed(ols
);
4896 return nfs4_check_openmode(ols
, flags
);
4900 nfs4_check_file(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct nfs4_stid
*s
,
4901 struct file
**filpp
, bool *tmp_file
, int flags
)
4903 int acc
= (flags
& RD_STATE
) ? NFSD_MAY_READ
: NFSD_MAY_WRITE
;
4907 file
= nfs4_find_file(s
, flags
);
4909 status
= nfsd_permission(rqstp
, fhp
->fh_export
, fhp
->fh_dentry
,
4910 acc
| NFSD_MAY_OWNER_OVERRIDE
);
4918 status
= nfsd_open(rqstp
, fhp
, S_IFREG
, acc
, filpp
);
4930 * Checks for stateid operations
4933 nfs4_preprocess_stateid_op(struct svc_rqst
*rqstp
,
4934 struct nfsd4_compound_state
*cstate
, struct svc_fh
*fhp
,
4935 stateid_t
*stateid
, int flags
, struct file
**filpp
, bool *tmp_file
)
4937 struct inode
*ino
= d_inode(fhp
->fh_dentry
);
4938 struct net
*net
= SVC_NET(rqstp
);
4939 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4940 struct nfs4_stid
*s
= NULL
;
4948 if (grace_disallows_io(net
, ino
))
4949 return nfserr_grace
;
4951 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
4952 status
= check_special_stateids(net
, fhp
, stateid
, flags
);
4956 status
= nfsd4_lookup_stateid(cstate
, stateid
,
4957 NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
,
4961 status
= check_stateid_generation(stateid
, &s
->sc_stateid
,
4962 nfsd4_has_session(cstate
));
4966 switch (s
->sc_type
) {
4967 case NFS4_DELEG_STID
:
4968 status
= nfs4_check_delegmode(delegstateid(s
), flags
);
4970 case NFS4_OPEN_STID
:
4971 case NFS4_LOCK_STID
:
4972 status
= nfs4_check_olstateid(fhp
, openlockstateid(s
), flags
);
4975 status
= nfserr_bad_stateid
;
4980 status
= nfs4_check_fh(fhp
, s
);
4983 if (!status
&& filpp
)
4984 status
= nfs4_check_file(rqstp
, fhp
, s
, filpp
, tmp_file
, flags
);
4992 * Test if the stateid is valid
4995 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4996 struct nfsd4_test_stateid
*test_stateid
)
4998 struct nfsd4_test_stateid_id
*stateid
;
4999 struct nfs4_client
*cl
= cstate
->session
->se_client
;
5001 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
5002 stateid
->ts_id_status
=
5003 nfsd4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
5009 nfsd4_free_lock_stateid(stateid_t
*stateid
, struct nfs4_stid
*s
)
5011 struct nfs4_ol_stateid
*stp
= openlockstateid(s
);
5014 mutex_lock(&stp
->st_mutex
);
5016 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
5020 ret
= nfserr_locks_held
;
5021 if (check_for_locks(stp
->st_stid
.sc_file
,
5022 lockowner(stp
->st_stateowner
)))
5025 release_lock_stateid(stp
);
5029 mutex_unlock(&stp
->st_mutex
);
5035 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5036 struct nfsd4_free_stateid
*free_stateid
)
5038 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
5039 struct nfs4_stid
*s
;
5040 struct nfs4_delegation
*dp
;
5041 struct nfs4_client
*cl
= cstate
->session
->se_client
;
5042 __be32 ret
= nfserr_bad_stateid
;
5044 spin_lock(&cl
->cl_lock
);
5045 s
= find_stateid_locked(cl
, stateid
);
5048 switch (s
->sc_type
) {
5049 case NFS4_DELEG_STID
:
5050 ret
= nfserr_locks_held
;
5052 case NFS4_OPEN_STID
:
5053 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
5056 ret
= nfserr_locks_held
;
5058 case NFS4_LOCK_STID
:
5059 atomic_inc(&s
->sc_count
);
5060 spin_unlock(&cl
->cl_lock
);
5061 ret
= nfsd4_free_lock_stateid(stateid
, s
);
5063 case NFS4_REVOKED_DELEG_STID
:
5064 dp
= delegstateid(s
);
5065 list_del_init(&dp
->dl_recall_lru
);
5066 spin_unlock(&cl
->cl_lock
);
5070 /* Default falls through and returns nfserr_bad_stateid */
5073 spin_unlock(&cl
->cl_lock
);
5081 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
5082 RD_STATE
: WR_STATE
;
5085 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
5087 struct svc_fh
*current_fh
= &cstate
->current_fh
;
5088 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
5091 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
5094 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
5095 || stp
->st_stid
.sc_type
== NFS4_REVOKED_DELEG_STID
)
5097 * "Closed" stateid's exist *only* to return
5098 * nfserr_replay_me from the previous step, and
5099 * revoked delegations are kept only for free_stateid.
5101 return nfserr_bad_stateid
;
5102 mutex_lock(&stp
->st_mutex
);
5103 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
5104 if (status
== nfs_ok
)
5105 status
= nfs4_check_fh(current_fh
, &stp
->st_stid
);
5106 if (status
!= nfs_ok
)
5107 mutex_unlock(&stp
->st_mutex
);
5112 * Checks for sequence id mutating operations.
5115 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
5116 stateid_t
*stateid
, char typemask
,
5117 struct nfs4_ol_stateid
**stpp
,
5118 struct nfsd_net
*nn
)
5121 struct nfs4_stid
*s
;
5122 struct nfs4_ol_stateid
*stp
= NULL
;
5124 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
5125 seqid
, STATEID_VAL(stateid
));
5128 status
= nfsd4_lookup_stateid(cstate
, stateid
, typemask
, &s
, nn
);
5131 stp
= openlockstateid(s
);
5132 nfsd4_cstate_assign_replay(cstate
, stp
->st_stateowner
);
5134 status
= nfs4_seqid_op_checks(cstate
, stateid
, seqid
, stp
);
5138 nfs4_put_stid(&stp
->st_stid
);
5142 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
5143 stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
, struct nfsd_net
*nn
)
5146 struct nfs4_openowner
*oo
;
5147 struct nfs4_ol_stateid
*stp
;
5149 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
5150 NFS4_OPEN_STID
, &stp
, nn
);
5153 oo
= openowner(stp
->st_stateowner
);
5154 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
5155 mutex_unlock(&stp
->st_mutex
);
5156 nfs4_put_stid(&stp
->st_stid
);
5157 return nfserr_bad_stateid
;
5164 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5165 struct nfsd4_open_confirm
*oc
)
5168 struct nfs4_openowner
*oo
;
5169 struct nfs4_ol_stateid
*stp
;
5170 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5172 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
5173 cstate
->current_fh
.fh_dentry
);
5175 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
5179 status
= nfs4_preprocess_seqid_op(cstate
,
5180 oc
->oc_seqid
, &oc
->oc_req_stateid
,
5181 NFS4_OPEN_STID
, &stp
, nn
);
5184 oo
= openowner(stp
->st_stateowner
);
5185 status
= nfserr_bad_stateid
;
5186 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
) {
5187 mutex_unlock(&stp
->st_mutex
);
5190 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
5191 nfs4_inc_and_copy_stateid(&oc
->oc_resp_stateid
, &stp
->st_stid
);
5192 mutex_unlock(&stp
->st_mutex
);
5193 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
5194 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
5196 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
5199 nfs4_put_stid(&stp
->st_stid
);
5201 nfsd4_bump_seqid(cstate
, status
);
5205 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
5207 if (!test_access(access
, stp
))
5209 nfs4_file_put_access(stp
->st_stid
.sc_file
, access
);
5210 clear_access(access
, stp
);
5213 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
5215 switch (to_access
) {
5216 case NFS4_SHARE_ACCESS_READ
:
5217 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
5218 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
5220 case NFS4_SHARE_ACCESS_WRITE
:
5221 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
5222 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
5224 case NFS4_SHARE_ACCESS_BOTH
:
5232 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
5233 struct nfsd4_compound_state
*cstate
,
5234 struct nfsd4_open_downgrade
*od
)
5237 struct nfs4_ol_stateid
*stp
;
5238 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5240 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
5241 cstate
->current_fh
.fh_dentry
);
5243 /* We don't yet support WANT bits: */
5244 if (od
->od_deleg_want
)
5245 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
5248 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
5249 &od
->od_stateid
, &stp
, nn
);
5252 status
= nfserr_inval
;
5253 if (!test_access(od
->od_share_access
, stp
)) {
5254 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
5255 stp
->st_access_bmap
, od
->od_share_access
);
5258 if (!test_deny(od
->od_share_deny
, stp
)) {
5259 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
5260 stp
->st_deny_bmap
, od
->od_share_deny
);
5263 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
5264 reset_union_bmap_deny(od
->od_share_deny
, stp
);
5265 nfs4_inc_and_copy_stateid(&od
->od_stateid
, &stp
->st_stid
);
5268 mutex_unlock(&stp
->st_mutex
);
5269 nfs4_put_stid(&stp
->st_stid
);
5271 nfsd4_bump_seqid(cstate
, status
);
5275 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
5277 struct nfs4_client
*clp
= s
->st_stid
.sc_client
;
5279 LIST_HEAD(reaplist
);
5281 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
5282 spin_lock(&clp
->cl_lock
);
5283 unhashed
= unhash_open_stateid(s
, &reaplist
);
5285 if (clp
->cl_minorversion
) {
5287 put_ol_stateid_locked(s
, &reaplist
);
5288 spin_unlock(&clp
->cl_lock
);
5289 free_ol_stateid_reaplist(&reaplist
);
5291 spin_unlock(&clp
->cl_lock
);
5292 free_ol_stateid_reaplist(&reaplist
);
5294 move_to_close_lru(s
, clp
->net
);
5299 * nfs4_unlock_state() called after encode
5302 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5303 struct nfsd4_close
*close
)
5306 struct nfs4_ol_stateid
*stp
;
5307 struct net
*net
= SVC_NET(rqstp
);
5308 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5310 dprintk("NFSD: nfsd4_close on file %pd\n",
5311 cstate
->current_fh
.fh_dentry
);
5313 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
5315 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
5317 nfsd4_bump_seqid(cstate
, status
);
5320 nfs4_inc_and_copy_stateid(&close
->cl_stateid
, &stp
->st_stid
);
5321 mutex_unlock(&stp
->st_mutex
);
5323 nfsd4_close_open_stateid(stp
);
5325 /* put reference from nfs4_preprocess_seqid_op */
5326 nfs4_put_stid(&stp
->st_stid
);
5332 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5333 struct nfsd4_delegreturn
*dr
)
5335 struct nfs4_delegation
*dp
;
5336 stateid_t
*stateid
= &dr
->dr_stateid
;
5337 struct nfs4_stid
*s
;
5339 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5341 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
5344 status
= nfsd4_lookup_stateid(cstate
, stateid
, NFS4_DELEG_STID
, &s
, nn
);
5347 dp
= delegstateid(s
);
5348 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
5352 destroy_delegation(dp
);
5354 nfs4_put_stid(&dp
->dl_stid
);
5360 end_offset(u64 start
, u64 len
)
5365 return end
>= start
? end
: NFS4_MAX_UINT64
;
5368 /* last octet in a range */
5370 last_byte_offset(u64 start
, u64 len
)
5376 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
5380 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5381 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5382 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
5383 * locking, this prevents us from being completely protocol-compliant. The
5384 * real solution to this problem is to start using unsigned file offsets in
5385 * the VFS, but this is a very deep change!
5388 nfs4_transform_lock_offset(struct file_lock
*lock
)
5390 if (lock
->fl_start
< 0)
5391 lock
->fl_start
= OFFSET_MAX
;
5392 if (lock
->fl_end
< 0)
5393 lock
->fl_end
= OFFSET_MAX
;
5397 nfsd4_fl_get_owner(fl_owner_t owner
)
5399 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5401 nfs4_get_stateowner(&lo
->lo_owner
);
5406 nfsd4_fl_put_owner(fl_owner_t owner
)
5408 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5411 nfs4_put_stateowner(&lo
->lo_owner
);
5415 nfsd4_lm_notify(struct file_lock
*fl
)
5417 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)fl
->fl_owner
;
5418 struct net
*net
= lo
->lo_owner
.so_client
->net
;
5419 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5420 struct nfsd4_blocked_lock
*nbl
= container_of(fl
,
5421 struct nfsd4_blocked_lock
, nbl_lock
);
5424 /* An empty list means that something else is going to be using it */
5425 spin_lock(&nn
->blocked_locks_lock
);
5426 if (!list_empty(&nbl
->nbl_list
)) {
5427 list_del_init(&nbl
->nbl_list
);
5428 list_del_init(&nbl
->nbl_lru
);
5431 spin_unlock(&nn
->blocked_locks_lock
);
5434 nfsd4_run_cb(&nbl
->nbl_cb
);
5437 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
5438 .lm_notify
= nfsd4_lm_notify
,
5439 .lm_get_owner
= nfsd4_fl_get_owner
,
5440 .lm_put_owner
= nfsd4_fl_put_owner
,
5444 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
5446 struct nfs4_lockowner
*lo
;
5448 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
5449 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
5450 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
5451 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
5452 if (!deny
->ld_owner
.data
)
5453 /* We just don't care that much */
5455 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
5456 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
5459 deny
->ld_owner
.len
= 0;
5460 deny
->ld_owner
.data
= NULL
;
5461 deny
->ld_clientid
.cl_boot
= 0;
5462 deny
->ld_clientid
.cl_id
= 0;
5464 deny
->ld_start
= fl
->fl_start
;
5465 deny
->ld_length
= NFS4_MAX_UINT64
;
5466 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
5467 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
5468 deny
->ld_type
= NFS4_READ_LT
;
5469 if (fl
->fl_type
!= F_RDLCK
)
5470 deny
->ld_type
= NFS4_WRITE_LT
;
5473 static struct nfs4_lockowner
*
5474 find_lockowner_str_locked(struct nfs4_client
*clp
, struct xdr_netobj
*owner
)
5476 unsigned int strhashval
= ownerstr_hashval(owner
);
5477 struct nfs4_stateowner
*so
;
5479 lockdep_assert_held(&clp
->cl_lock
);
5481 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[strhashval
],
5483 if (so
->so_is_open_owner
)
5485 if (same_owner_str(so
, owner
))
5486 return lockowner(nfs4_get_stateowner(so
));
5491 static struct nfs4_lockowner
*
5492 find_lockowner_str(struct nfs4_client
*clp
, struct xdr_netobj
*owner
)
5494 struct nfs4_lockowner
*lo
;
5496 spin_lock(&clp
->cl_lock
);
5497 lo
= find_lockowner_str_locked(clp
, owner
);
5498 spin_unlock(&clp
->cl_lock
);
5502 static void nfs4_unhash_lockowner(struct nfs4_stateowner
*sop
)
5504 unhash_lockowner_locked(lockowner(sop
));
5507 static void nfs4_free_lockowner(struct nfs4_stateowner
*sop
)
5509 struct nfs4_lockowner
*lo
= lockowner(sop
);
5511 kmem_cache_free(lockowner_slab
, lo
);
5514 static const struct nfs4_stateowner_operations lockowner_ops
= {
5515 .so_unhash
= nfs4_unhash_lockowner
,
5516 .so_free
= nfs4_free_lockowner
,
5520 * Alloc a lock owner structure.
5521 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
5524 * strhashval = ownerstr_hashval
5526 static struct nfs4_lockowner
*
5527 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
,
5528 struct nfs4_ol_stateid
*open_stp
,
5529 struct nfsd4_lock
*lock
)
5531 struct nfs4_lockowner
*lo
, *ret
;
5533 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
5536 INIT_LIST_HEAD(&lo
->lo_blocked
);
5537 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
5538 lo
->lo_owner
.so_is_open_owner
= 0;
5539 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
;
5540 lo
->lo_owner
.so_ops
= &lockowner_ops
;
5541 spin_lock(&clp
->cl_lock
);
5542 ret
= find_lockowner_str_locked(clp
, &lock
->lk_new_owner
);
5544 list_add(&lo
->lo_owner
.so_strhash
,
5545 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
5548 nfs4_free_stateowner(&lo
->lo_owner
);
5550 spin_unlock(&clp
->cl_lock
);
5555 init_lock_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_lockowner
*lo
,
5556 struct nfs4_file
*fp
, struct inode
*inode
,
5557 struct nfs4_ol_stateid
*open_stp
)
5559 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5561 lockdep_assert_held(&clp
->cl_lock
);
5563 atomic_inc(&stp
->st_stid
.sc_count
);
5564 stp
->st_stid
.sc_type
= NFS4_LOCK_STID
;
5565 stp
->st_stateowner
= nfs4_get_stateowner(&lo
->lo_owner
);
5567 stp
->st_stid
.sc_file
= fp
;
5568 stp
->st_access_bmap
= 0;
5569 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
5570 stp
->st_openstp
= open_stp
;
5571 mutex_init(&stp
->st_mutex
);
5572 list_add(&stp
->st_locks
, &open_stp
->st_locks
);
5573 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
5574 spin_lock(&fp
->fi_lock
);
5575 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
5576 spin_unlock(&fp
->fi_lock
);
5579 static struct nfs4_ol_stateid
*
5580 find_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
)
5582 struct nfs4_ol_stateid
*lst
;
5583 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5585 lockdep_assert_held(&clp
->cl_lock
);
5587 list_for_each_entry(lst
, &lo
->lo_owner
.so_stateids
, st_perstateowner
) {
5588 if (lst
->st_stid
.sc_file
== fp
) {
5589 atomic_inc(&lst
->st_stid
.sc_count
);
5596 static struct nfs4_ol_stateid
*
5597 find_or_create_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fi
,
5598 struct inode
*inode
, struct nfs4_ol_stateid
*ost
,
5601 struct nfs4_stid
*ns
= NULL
;
5602 struct nfs4_ol_stateid
*lst
;
5603 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5604 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
5606 spin_lock(&clp
->cl_lock
);
5607 lst
= find_lock_stateid(lo
, fi
);
5609 spin_unlock(&clp
->cl_lock
);
5610 ns
= nfs4_alloc_stid(clp
, stateid_slab
, nfs4_free_lock_stateid
);
5614 spin_lock(&clp
->cl_lock
);
5615 lst
= find_lock_stateid(lo
, fi
);
5617 lst
= openlockstateid(ns
);
5618 init_lock_stateid(lst
, lo
, fi
, inode
, ost
);
5623 spin_unlock(&clp
->cl_lock
);
5630 check_lock_length(u64 offset
, u64 length
)
5632 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
5633 (length
> ~offset
)));
5636 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
5638 struct nfs4_file
*fp
= lock_stp
->st_stid
.sc_file
;
5640 lockdep_assert_held(&fp
->fi_lock
);
5642 if (test_access(access
, lock_stp
))
5644 __nfs4_file_get_access(fp
, access
);
5645 set_access(access
, lock_stp
);
5649 lookup_or_create_lock_state(struct nfsd4_compound_state
*cstate
,
5650 struct nfs4_ol_stateid
*ost
,
5651 struct nfsd4_lock
*lock
,
5652 struct nfs4_ol_stateid
**plst
, bool *new)
5655 struct nfs4_file
*fi
= ost
->st_stid
.sc_file
;
5656 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5657 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
5658 struct inode
*inode
= d_inode(cstate
->current_fh
.fh_dentry
);
5659 struct nfs4_lockowner
*lo
;
5660 struct nfs4_ol_stateid
*lst
;
5661 unsigned int strhashval
;
5664 lo
= find_lockowner_str(cl
, &lock
->lk_new_owner
);
5666 strhashval
= ownerstr_hashval(&lock
->lk_new_owner
);
5667 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
5669 return nfserr_jukebox
;
5671 /* with an existing lockowner, seqids must be the same */
5672 status
= nfserr_bad_seqid
;
5673 if (!cstate
->minorversion
&&
5674 lock
->lk_new_lock_seqid
!= lo
->lo_owner
.so_seqid
)
5679 lst
= find_or_create_lock_stateid(lo
, fi
, inode
, ost
, new);
5681 status
= nfserr_jukebox
;
5685 mutex_lock(&lst
->st_mutex
);
5687 /* See if it's still hashed to avoid race with FREE_STATEID */
5688 spin_lock(&cl
->cl_lock
);
5689 hashed
= !list_empty(&lst
->st_perfile
);
5690 spin_unlock(&cl
->cl_lock
);
5693 mutex_unlock(&lst
->st_mutex
);
5694 nfs4_put_stid(&lst
->st_stid
);
5700 nfs4_put_stateowner(&lo
->lo_owner
);
5708 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5709 struct nfsd4_lock
*lock
)
5711 struct nfs4_openowner
*open_sop
= NULL
;
5712 struct nfs4_lockowner
*lock_sop
= NULL
;
5713 struct nfs4_ol_stateid
*lock_stp
= NULL
;
5714 struct nfs4_ol_stateid
*open_stp
= NULL
;
5715 struct nfs4_file
*fp
;
5716 struct file
*filp
= NULL
;
5717 struct nfsd4_blocked_lock
*nbl
= NULL
;
5718 struct file_lock
*file_lock
= NULL
;
5719 struct file_lock
*conflock
= NULL
;
5724 unsigned char fl_type
;
5725 unsigned int fl_flags
= FL_POSIX
;
5726 struct net
*net
= SVC_NET(rqstp
);
5727 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5729 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5730 (long long) lock
->lk_offset
,
5731 (long long) lock
->lk_length
);
5733 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
5734 return nfserr_inval
;
5736 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
5737 S_IFREG
, NFSD_MAY_LOCK
))) {
5738 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5742 if (lock
->lk_is_new
) {
5743 if (nfsd4_has_session(cstate
))
5744 /* See rfc 5661 18.10.3: given clientid is ignored: */
5745 memcpy(&lock
->lk_new_clientid
,
5746 &cstate
->session
->se_client
->cl_clientid
,
5747 sizeof(clientid_t
));
5749 status
= nfserr_stale_clientid
;
5750 if (STALE_CLIENTID(&lock
->lk_new_clientid
, nn
))
5753 /* validate and update open stateid and open seqid */
5754 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
5755 lock
->lk_new_open_seqid
,
5756 &lock
->lk_new_open_stateid
,
5760 mutex_unlock(&open_stp
->st_mutex
);
5761 open_sop
= openowner(open_stp
->st_stateowner
);
5762 status
= nfserr_bad_stateid
;
5763 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
5764 &lock
->lk_new_clientid
))
5766 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
5769 status
= nfs4_preprocess_seqid_op(cstate
,
5770 lock
->lk_old_lock_seqid
,
5771 &lock
->lk_old_lock_stateid
,
5772 NFS4_LOCK_STID
, &lock_stp
, nn
);
5776 lock_sop
= lockowner(lock_stp
->st_stateowner
);
5778 lkflg
= setlkflg(lock
->lk_type
);
5779 status
= nfs4_check_openmode(lock_stp
, lkflg
);
5783 status
= nfserr_grace
;
5784 if (locks_in_grace(net
) && !lock
->lk_reclaim
)
5786 status
= nfserr_no_grace
;
5787 if (!locks_in_grace(net
) && lock
->lk_reclaim
)
5790 fp
= lock_stp
->st_stid
.sc_file
;
5791 switch (lock
->lk_type
) {
5793 if (nfsd4_has_session(cstate
))
5794 fl_flags
|= FL_SLEEP
;
5797 spin_lock(&fp
->fi_lock
);
5798 filp
= find_readable_file_locked(fp
);
5800 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
5801 spin_unlock(&fp
->fi_lock
);
5804 case NFS4_WRITEW_LT
:
5805 if (nfsd4_has_session(cstate
))
5806 fl_flags
|= FL_SLEEP
;
5809 spin_lock(&fp
->fi_lock
);
5810 filp
= find_writeable_file_locked(fp
);
5812 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
5813 spin_unlock(&fp
->fi_lock
);
5817 status
= nfserr_inval
;
5822 status
= nfserr_openmode
;
5826 nbl
= find_or_allocate_block(lock_sop
, &fp
->fi_fhandle
, nn
);
5828 dprintk("NFSD: %s: unable to allocate block!\n", __func__
);
5829 status
= nfserr_jukebox
;
5833 file_lock
= &nbl
->nbl_lock
;
5834 file_lock
->fl_type
= fl_type
;
5835 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(&lock_sop
->lo_owner
));
5836 file_lock
->fl_pid
= current
->tgid
;
5837 file_lock
->fl_file
= filp
;
5838 file_lock
->fl_flags
= fl_flags
;
5839 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
5840 file_lock
->fl_start
= lock
->lk_offset
;
5841 file_lock
->fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
5842 nfs4_transform_lock_offset(file_lock
);
5844 conflock
= locks_alloc_lock();
5846 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5847 status
= nfserr_jukebox
;
5851 if (fl_flags
& FL_SLEEP
) {
5852 nbl
->nbl_time
= jiffies
;
5853 spin_lock(&nn
->blocked_locks_lock
);
5854 list_add_tail(&nbl
->nbl_list
, &lock_sop
->lo_blocked
);
5855 list_add_tail(&nbl
->nbl_lru
, &nn
->blocked_locks_lru
);
5856 spin_unlock(&nn
->blocked_locks_lock
);
5859 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, conflock
);
5861 case 0: /* success! */
5862 nfs4_inc_and_copy_stateid(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
);
5865 case FILE_LOCK_DEFERRED
:
5868 case -EAGAIN
: /* conflock holds conflicting lock */
5869 status
= nfserr_denied
;
5870 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5871 nfs4_set_lock_denied(conflock
, &lock
->lk_denied
);
5874 status
= nfserr_deadlock
;
5877 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
5878 status
= nfserrno(err
);
5883 /* dequeue it if we queued it before */
5884 if (fl_flags
& FL_SLEEP
) {
5885 spin_lock(&nn
->blocked_locks_lock
);
5886 list_del_init(&nbl
->nbl_list
);
5887 list_del_init(&nbl
->nbl_lru
);
5888 spin_unlock(&nn
->blocked_locks_lock
);
5890 free_blocked_lock(nbl
);
5895 /* Bump seqid manually if the 4.0 replay owner is openowner */
5896 if (cstate
->replay_owner
&&
5897 cstate
->replay_owner
!= &lock_sop
->lo_owner
&&
5898 seqid_mutating_err(ntohl(status
)))
5899 lock_sop
->lo_owner
.so_seqid
++;
5901 mutex_unlock(&lock_stp
->st_mutex
);
5904 * If this is a new, never-before-used stateid, and we are
5905 * returning an error, then just go ahead and release it.
5908 release_lock_stateid(lock_stp
);
5910 nfs4_put_stid(&lock_stp
->st_stid
);
5913 nfs4_put_stid(&open_stp
->st_stid
);
5914 nfsd4_bump_seqid(cstate
, status
);
5916 locks_free_lock(conflock
);
5921 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5922 * so we do a temporary open here just to get an open file to pass to
5923 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
5926 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
5929 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
5931 err
= nfserrno(vfs_test_lock(file
, lock
));
5941 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5942 struct nfsd4_lockt
*lockt
)
5944 struct file_lock
*file_lock
= NULL
;
5945 struct nfs4_lockowner
*lo
= NULL
;
5947 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5949 if (locks_in_grace(SVC_NET(rqstp
)))
5950 return nfserr_grace
;
5952 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
5953 return nfserr_inval
;
5955 if (!nfsd4_has_session(cstate
)) {
5956 status
= lookup_clientid(&lockt
->lt_clientid
, cstate
, nn
);
5961 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
5964 file_lock
= locks_alloc_lock();
5966 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
5967 status
= nfserr_jukebox
;
5971 switch (lockt
->lt_type
) {
5974 file_lock
->fl_type
= F_RDLCK
;
5977 case NFS4_WRITEW_LT
:
5978 file_lock
->fl_type
= F_WRLCK
;
5981 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5982 status
= nfserr_inval
;
5986 lo
= find_lockowner_str(cstate
->clp
, &lockt
->lt_owner
);
5988 file_lock
->fl_owner
= (fl_owner_t
)lo
;
5989 file_lock
->fl_pid
= current
->tgid
;
5990 file_lock
->fl_flags
= FL_POSIX
;
5992 file_lock
->fl_start
= lockt
->lt_offset
;
5993 file_lock
->fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
5995 nfs4_transform_lock_offset(file_lock
);
5997 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, file_lock
);
6001 if (file_lock
->fl_type
!= F_UNLCK
) {
6002 status
= nfserr_denied
;
6003 nfs4_set_lock_denied(file_lock
, &lockt
->lt_denied
);
6007 nfs4_put_stateowner(&lo
->lo_owner
);
6009 locks_free_lock(file_lock
);
6014 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
6015 struct nfsd4_locku
*locku
)
6017 struct nfs4_ol_stateid
*stp
;
6018 struct file
*filp
= NULL
;
6019 struct file_lock
*file_lock
= NULL
;
6022 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
6024 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
6025 (long long) locku
->lu_offset
,
6026 (long long) locku
->lu_length
);
6028 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
6029 return nfserr_inval
;
6031 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
6032 &locku
->lu_stateid
, NFS4_LOCK_STID
,
6036 filp
= find_any_file(stp
->st_stid
.sc_file
);
6038 status
= nfserr_lock_range
;
6041 file_lock
= locks_alloc_lock();
6043 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
6044 status
= nfserr_jukebox
;
6048 file_lock
->fl_type
= F_UNLCK
;
6049 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(stp
->st_stateowner
));
6050 file_lock
->fl_pid
= current
->tgid
;
6051 file_lock
->fl_file
= filp
;
6052 file_lock
->fl_flags
= FL_POSIX
;
6053 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
6054 file_lock
->fl_start
= locku
->lu_offset
;
6056 file_lock
->fl_end
= last_byte_offset(locku
->lu_offset
,
6058 nfs4_transform_lock_offset(file_lock
);
6060 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, NULL
);
6062 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
6065 nfs4_inc_and_copy_stateid(&locku
->lu_stateid
, &stp
->st_stid
);
6069 mutex_unlock(&stp
->st_mutex
);
6070 nfs4_put_stid(&stp
->st_stid
);
6072 nfsd4_bump_seqid(cstate
, status
);
6074 locks_free_lock(file_lock
);
6078 status
= nfserrno(err
);
6084 * true: locks held by lockowner
6085 * false: no locks held by lockowner
6088 check_for_locks(struct nfs4_file
*fp
, struct nfs4_lockowner
*lowner
)
6090 struct file_lock
*fl
;
6092 struct file
*filp
= find_any_file(fp
);
6093 struct inode
*inode
;
6094 struct file_lock_context
*flctx
;
6097 /* Any valid lock stateid should have some sort of access */
6102 inode
= file_inode(filp
);
6103 flctx
= inode
->i_flctx
;
6105 if (flctx
&& !list_empty_careful(&flctx
->flc_posix
)) {
6106 spin_lock(&flctx
->flc_lock
);
6107 list_for_each_entry(fl
, &flctx
->flc_posix
, fl_list
) {
6108 if (fl
->fl_owner
== (fl_owner_t
)lowner
) {
6113 spin_unlock(&flctx
->flc_lock
);
6120 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
6121 struct nfsd4_compound_state
*cstate
,
6122 struct nfsd4_release_lockowner
*rlockowner
)
6124 clientid_t
*clid
= &rlockowner
->rl_clientid
;
6125 struct nfs4_stateowner
*sop
;
6126 struct nfs4_lockowner
*lo
= NULL
;
6127 struct nfs4_ol_stateid
*stp
;
6128 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
6129 unsigned int hashval
= ownerstr_hashval(owner
);
6131 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
6132 struct nfs4_client
*clp
;
6133 LIST_HEAD (reaplist
);
6135 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
6136 clid
->cl_boot
, clid
->cl_id
);
6138 status
= lookup_clientid(clid
, cstate
, nn
);
6143 /* Find the matching lock stateowner */
6144 spin_lock(&clp
->cl_lock
);
6145 list_for_each_entry(sop
, &clp
->cl_ownerstr_hashtbl
[hashval
],
6148 if (sop
->so_is_open_owner
|| !same_owner_str(sop
, owner
))
6151 /* see if there are still any locks associated with it */
6152 lo
= lockowner(sop
);
6153 list_for_each_entry(stp
, &sop
->so_stateids
, st_perstateowner
) {
6154 if (check_for_locks(stp
->st_stid
.sc_file
, lo
)) {
6155 status
= nfserr_locks_held
;
6156 spin_unlock(&clp
->cl_lock
);
6161 nfs4_get_stateowner(sop
);
6165 spin_unlock(&clp
->cl_lock
);
6169 unhash_lockowner_locked(lo
);
6170 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
6171 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
6172 struct nfs4_ol_stateid
,
6174 WARN_ON(!unhash_lock_stateid(stp
));
6175 put_ol_stateid_locked(stp
, &reaplist
);
6177 spin_unlock(&clp
->cl_lock
);
6178 free_ol_stateid_reaplist(&reaplist
);
6179 nfs4_put_stateowner(&lo
->lo_owner
);
6184 static inline struct nfs4_client_reclaim
*
6187 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
6191 nfs4_has_reclaimed_state(const char *name
, struct nfsd_net
*nn
)
6193 struct nfs4_client_reclaim
*crp
;
6195 crp
= nfsd4_find_reclaim_client(name
, nn
);
6196 return (crp
&& crp
->cr_clp
);
6200 * failure => all reset bets are off, nfserr_no_grace...
6202 struct nfs4_client_reclaim
*
6203 nfs4_client_to_reclaim(const char *name
, struct nfsd_net
*nn
)
6205 unsigned int strhashval
;
6206 struct nfs4_client_reclaim
*crp
;
6208 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
6209 crp
= alloc_reclaim();
6211 strhashval
= clientstr_hashval(name
);
6212 INIT_LIST_HEAD(&crp
->cr_strhash
);
6213 list_add(&crp
->cr_strhash
, &nn
->reclaim_str_hashtbl
[strhashval
]);
6214 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
6216 nn
->reclaim_str_hashtbl_size
++;
6222 nfs4_remove_reclaim_record(struct nfs4_client_reclaim
*crp
, struct nfsd_net
*nn
)
6224 list_del(&crp
->cr_strhash
);
6226 nn
->reclaim_str_hashtbl_size
--;
6230 nfs4_release_reclaim(struct nfsd_net
*nn
)
6232 struct nfs4_client_reclaim
*crp
= NULL
;
6235 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6236 while (!list_empty(&nn
->reclaim_str_hashtbl
[i
])) {
6237 crp
= list_entry(nn
->reclaim_str_hashtbl
[i
].next
,
6238 struct nfs4_client_reclaim
, cr_strhash
);
6239 nfs4_remove_reclaim_record(crp
, nn
);
6242 WARN_ON_ONCE(nn
->reclaim_str_hashtbl_size
);
6246 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
6247 struct nfs4_client_reclaim
*
6248 nfsd4_find_reclaim_client(const char *recdir
, struct nfsd_net
*nn
)
6250 unsigned int strhashval
;
6251 struct nfs4_client_reclaim
*crp
= NULL
;
6253 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir
);
6255 strhashval
= clientstr_hashval(recdir
);
6256 list_for_each_entry(crp
, &nn
->reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
6257 if (same_name(crp
->cr_recdir
, recdir
)) {
6265 * Called from OPEN. Look for clientid in reclaim list.
6268 nfs4_check_open_reclaim(clientid_t
*clid
,
6269 struct nfsd4_compound_state
*cstate
,
6270 struct nfsd_net
*nn
)
6274 /* find clientid in conf_id_hashtbl */
6275 status
= lookup_clientid(clid
, cstate
, nn
);
6277 return nfserr_reclaim_bad
;
6279 if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
, &cstate
->clp
->cl_flags
))
6280 return nfserr_no_grace
;
6282 if (nfsd4_client_record_check(cstate
->clp
))
6283 return nfserr_reclaim_bad
;
6288 #ifdef CONFIG_NFSD_FAULT_INJECTION
6290 put_client(struct nfs4_client
*clp
)
6292 atomic_dec(&clp
->cl_refcount
);
6295 static struct nfs4_client
*
6296 nfsd_find_client(struct sockaddr_storage
*addr
, size_t addr_size
)
6298 struct nfs4_client
*clp
;
6299 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6302 if (!nfsd_netns_ready(nn
))
6305 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6306 if (memcmp(&clp
->cl_addr
, addr
, addr_size
) == 0)
6313 nfsd_inject_print_clients(void)
6315 struct nfs4_client
*clp
;
6317 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6319 char buf
[INET6_ADDRSTRLEN
];
6321 if (!nfsd_netns_ready(nn
))
6324 spin_lock(&nn
->client_lock
);
6325 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6326 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
6327 pr_info("NFS Client: %s\n", buf
);
6330 spin_unlock(&nn
->client_lock
);
6336 nfsd_inject_forget_client(struct sockaddr_storage
*addr
, size_t addr_size
)
6339 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 clp
= nfsd_find_client(addr
, addr_size
);
6349 if (mark_client_expired_locked(clp
) == nfs_ok
)
6354 spin_unlock(&nn
->client_lock
);
6363 nfsd_inject_forget_clients(u64 max
)
6366 struct nfs4_client
*clp
, *next
;
6367 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6369 LIST_HEAD(reaplist
);
6371 if (!nfsd_netns_ready(nn
))
6374 spin_lock(&nn
->client_lock
);
6375 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
6376 if (mark_client_expired_locked(clp
) == nfs_ok
) {
6377 list_add(&clp
->cl_lru
, &reaplist
);
6378 if (max
!= 0 && ++count
>= max
)
6382 spin_unlock(&nn
->client_lock
);
6384 list_for_each_entry_safe(clp
, next
, &reaplist
, cl_lru
)
6390 static void nfsd_print_count(struct nfs4_client
*clp
, unsigned int count
,
6393 char buf
[INET6_ADDRSTRLEN
];
6394 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
6395 printk(KERN_INFO
"NFS Client: %s has %u %s\n", buf
, count
, type
);
6399 nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid
*lst
,
6400 struct list_head
*collect
)
6402 struct nfs4_client
*clp
= lst
->st_stid
.sc_client
;
6403 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6409 lockdep_assert_held(&nn
->client_lock
);
6410 atomic_inc(&clp
->cl_refcount
);
6411 list_add(&lst
->st_locks
, collect
);
6414 static u64
nfsd_foreach_client_lock(struct nfs4_client
*clp
, u64 max
,
6415 struct list_head
*collect
,
6416 bool (*func
)(struct nfs4_ol_stateid
*))
6418 struct nfs4_openowner
*oop
;
6419 struct nfs4_ol_stateid
*stp
, *st_next
;
6420 struct nfs4_ol_stateid
*lst
, *lst_next
;
6423 spin_lock(&clp
->cl_lock
);
6424 list_for_each_entry(oop
, &clp
->cl_openowners
, oo_perclient
) {
6425 list_for_each_entry_safe(stp
, st_next
,
6426 &oop
->oo_owner
.so_stateids
, st_perstateowner
) {
6427 list_for_each_entry_safe(lst
, lst_next
,
6428 &stp
->st_locks
, st_locks
) {
6431 nfsd_inject_add_lock_to_list(lst
,
6436 * Despite the fact that these functions deal
6437 * with 64-bit integers for "count", we must
6438 * ensure that it doesn't blow up the
6439 * clp->cl_refcount. Throw a warning if we
6440 * start to approach INT_MAX here.
6442 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6449 spin_unlock(&clp
->cl_lock
);
6455 nfsd_collect_client_locks(struct nfs4_client
*clp
, struct list_head
*collect
,
6458 return nfsd_foreach_client_lock(clp
, max
, collect
, unhash_lock_stateid
);
6462 nfsd_print_client_locks(struct nfs4_client
*clp
)
6464 u64 count
= nfsd_foreach_client_lock(clp
, 0, NULL
, NULL
);
6465 nfsd_print_count(clp
, count
, "locked files");
6470 nfsd_inject_print_locks(void)
6472 struct nfs4_client
*clp
;
6474 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6477 if (!nfsd_netns_ready(nn
))
6480 spin_lock(&nn
->client_lock
);
6481 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6482 count
+= nfsd_print_client_locks(clp
);
6483 spin_unlock(&nn
->client_lock
);
6489 nfsd_reap_locks(struct list_head
*reaplist
)
6491 struct nfs4_client
*clp
;
6492 struct nfs4_ol_stateid
*stp
, *next
;
6494 list_for_each_entry_safe(stp
, next
, reaplist
, st_locks
) {
6495 list_del_init(&stp
->st_locks
);
6496 clp
= stp
->st_stid
.sc_client
;
6497 nfs4_put_stid(&stp
->st_stid
);
6503 nfsd_inject_forget_client_locks(struct sockaddr_storage
*addr
, size_t addr_size
)
6505 unsigned int count
= 0;
6506 struct nfs4_client
*clp
;
6507 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6509 LIST_HEAD(reaplist
);
6511 if (!nfsd_netns_ready(nn
))
6514 spin_lock(&nn
->client_lock
);
6515 clp
= nfsd_find_client(addr
, addr_size
);
6517 count
= nfsd_collect_client_locks(clp
, &reaplist
, 0);
6518 spin_unlock(&nn
->client_lock
);
6519 nfsd_reap_locks(&reaplist
);
6524 nfsd_inject_forget_locks(u64 max
)
6527 struct nfs4_client
*clp
;
6528 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6530 LIST_HEAD(reaplist
);
6532 if (!nfsd_netns_ready(nn
))
6535 spin_lock(&nn
->client_lock
);
6536 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6537 count
+= nfsd_collect_client_locks(clp
, &reaplist
, max
- count
);
6538 if (max
!= 0 && count
>= max
)
6541 spin_unlock(&nn
->client_lock
);
6542 nfsd_reap_locks(&reaplist
);
6547 nfsd_foreach_client_openowner(struct nfs4_client
*clp
, u64 max
,
6548 struct list_head
*collect
,
6549 void (*func
)(struct nfs4_openowner
*))
6551 struct nfs4_openowner
*oop
, *next
;
6552 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6556 lockdep_assert_held(&nn
->client_lock
);
6558 spin_lock(&clp
->cl_lock
);
6559 list_for_each_entry_safe(oop
, next
, &clp
->cl_openowners
, oo_perclient
) {
6563 atomic_inc(&clp
->cl_refcount
);
6564 list_add(&oop
->oo_perclient
, collect
);
6569 * Despite the fact that these functions deal with
6570 * 64-bit integers for "count", we must ensure that
6571 * it doesn't blow up the clp->cl_refcount. Throw a
6572 * warning if we start to approach INT_MAX here.
6574 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6578 spin_unlock(&clp
->cl_lock
);
6584 nfsd_print_client_openowners(struct nfs4_client
*clp
)
6586 u64 count
= nfsd_foreach_client_openowner(clp
, 0, NULL
, NULL
);
6588 nfsd_print_count(clp
, count
, "openowners");
6593 nfsd_collect_client_openowners(struct nfs4_client
*clp
,
6594 struct list_head
*collect
, u64 max
)
6596 return nfsd_foreach_client_openowner(clp
, max
, collect
,
6597 unhash_openowner_locked
);
6601 nfsd_inject_print_openowners(void)
6603 struct nfs4_client
*clp
;
6605 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6608 if (!nfsd_netns_ready(nn
))
6611 spin_lock(&nn
->client_lock
);
6612 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6613 count
+= nfsd_print_client_openowners(clp
);
6614 spin_unlock(&nn
->client_lock
);
6620 nfsd_reap_openowners(struct list_head
*reaplist
)
6622 struct nfs4_client
*clp
;
6623 struct nfs4_openowner
*oop
, *next
;
6625 list_for_each_entry_safe(oop
, next
, reaplist
, oo_perclient
) {
6626 list_del_init(&oop
->oo_perclient
);
6627 clp
= oop
->oo_owner
.so_client
;
6628 release_openowner(oop
);
6634 nfsd_inject_forget_client_openowners(struct sockaddr_storage
*addr
,
6637 unsigned int count
= 0;
6638 struct nfs4_client
*clp
;
6639 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6641 LIST_HEAD(reaplist
);
6643 if (!nfsd_netns_ready(nn
))
6646 spin_lock(&nn
->client_lock
);
6647 clp
= nfsd_find_client(addr
, addr_size
);
6649 count
= nfsd_collect_client_openowners(clp
, &reaplist
, 0);
6650 spin_unlock(&nn
->client_lock
);
6651 nfsd_reap_openowners(&reaplist
);
6656 nfsd_inject_forget_openowners(u64 max
)
6659 struct nfs4_client
*clp
;
6660 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6662 LIST_HEAD(reaplist
);
6664 if (!nfsd_netns_ready(nn
))
6667 spin_lock(&nn
->client_lock
);
6668 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6669 count
+= nfsd_collect_client_openowners(clp
, &reaplist
,
6671 if (max
!= 0 && count
>= max
)
6674 spin_unlock(&nn
->client_lock
);
6675 nfsd_reap_openowners(&reaplist
);
6679 static u64
nfsd_find_all_delegations(struct nfs4_client
*clp
, u64 max
,
6680 struct list_head
*victims
)
6682 struct nfs4_delegation
*dp
, *next
;
6683 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6687 lockdep_assert_held(&nn
->client_lock
);
6689 spin_lock(&state_lock
);
6690 list_for_each_entry_safe(dp
, next
, &clp
->cl_delegations
, dl_perclnt
) {
6693 * It's not safe to mess with delegations that have a
6694 * non-zero dl_time. They might have already been broken
6695 * and could be processed by the laundromat outside of
6696 * the state_lock. Just leave them be.
6698 if (dp
->dl_time
!= 0)
6701 atomic_inc(&clp
->cl_refcount
);
6702 WARN_ON(!unhash_delegation_locked(dp
));
6703 list_add(&dp
->dl_recall_lru
, victims
);
6707 * Despite the fact that these functions deal with
6708 * 64-bit integers for "count", we must ensure that
6709 * it doesn't blow up the clp->cl_refcount. Throw a
6710 * warning if we start to approach INT_MAX here.
6712 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6716 spin_unlock(&state_lock
);
6721 nfsd_print_client_delegations(struct nfs4_client
*clp
)
6723 u64 count
= nfsd_find_all_delegations(clp
, 0, NULL
);
6725 nfsd_print_count(clp
, count
, "delegations");
6730 nfsd_inject_print_delegations(void)
6732 struct nfs4_client
*clp
;
6734 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6737 if (!nfsd_netns_ready(nn
))
6740 spin_lock(&nn
->client_lock
);
6741 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6742 count
+= nfsd_print_client_delegations(clp
);
6743 spin_unlock(&nn
->client_lock
);
6749 nfsd_forget_delegations(struct list_head
*reaplist
)
6751 struct nfs4_client
*clp
;
6752 struct nfs4_delegation
*dp
, *next
;
6754 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
6755 list_del_init(&dp
->dl_recall_lru
);
6756 clp
= dp
->dl_stid
.sc_client
;
6757 revoke_delegation(dp
);
6763 nfsd_inject_forget_client_delegations(struct sockaddr_storage
*addr
,
6767 struct nfs4_client
*clp
;
6768 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6770 LIST_HEAD(reaplist
);
6772 if (!nfsd_netns_ready(nn
))
6775 spin_lock(&nn
->client_lock
);
6776 clp
= nfsd_find_client(addr
, addr_size
);
6778 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
6779 spin_unlock(&nn
->client_lock
);
6781 nfsd_forget_delegations(&reaplist
);
6786 nfsd_inject_forget_delegations(u64 max
)
6789 struct nfs4_client
*clp
;
6790 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6792 LIST_HEAD(reaplist
);
6794 if (!nfsd_netns_ready(nn
))
6797 spin_lock(&nn
->client_lock
);
6798 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6799 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
6800 if (max
!= 0 && count
>= max
)
6803 spin_unlock(&nn
->client_lock
);
6804 nfsd_forget_delegations(&reaplist
);
6809 nfsd_recall_delegations(struct list_head
*reaplist
)
6811 struct nfs4_client
*clp
;
6812 struct nfs4_delegation
*dp
, *next
;
6814 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
6815 list_del_init(&dp
->dl_recall_lru
);
6816 clp
= dp
->dl_stid
.sc_client
;
6818 * We skipped all entries that had a zero dl_time before,
6819 * so we can now reset the dl_time back to 0. If a delegation
6820 * break comes in now, then it won't make any difference since
6821 * we're recalling it either way.
6823 spin_lock(&state_lock
);
6825 spin_unlock(&state_lock
);
6826 nfsd_break_one_deleg(dp
);
6832 nfsd_inject_recall_client_delegations(struct sockaddr_storage
*addr
,
6836 struct nfs4_client
*clp
;
6837 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6839 LIST_HEAD(reaplist
);
6841 if (!nfsd_netns_ready(nn
))
6844 spin_lock(&nn
->client_lock
);
6845 clp
= nfsd_find_client(addr
, addr_size
);
6847 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
6848 spin_unlock(&nn
->client_lock
);
6850 nfsd_recall_delegations(&reaplist
);
6855 nfsd_inject_recall_delegations(u64 max
)
6858 struct nfs4_client
*clp
, *next
;
6859 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6861 LIST_HEAD(reaplist
);
6863 if (!nfsd_netns_ready(nn
))
6866 spin_lock(&nn
->client_lock
);
6867 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
6868 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
6869 if (max
!= 0 && ++count
>= max
)
6872 spin_unlock(&nn
->client_lock
);
6873 nfsd_recall_delegations(&reaplist
);
6876 #endif /* CONFIG_NFSD_FAULT_INJECTION */
6879 * Since the lifetime of a delegation isn't limited to that of an open, a
6880 * client may quite reasonably hang on to a delegation as long as it has
6881 * the inode cached. This becomes an obvious problem the first time a
6882 * client's inode cache approaches the size of the server's total memory.
6884 * For now we avoid this problem by imposing a hard limit on the number
6885 * of delegations, which varies according to the server's memory size.
6888 set_max_delegations(void)
6891 * Allow at most 4 delegations per megabyte of RAM. Quick
6892 * estimates suggest that in the worst case (where every delegation
6893 * is for a different inode), a delegation could take about 1.5K,
6894 * giving a worst case usage of about 6% of memory.
6896 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
6899 static int nfs4_state_create_net(struct net
*net
)
6901 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6904 nn
->conf_id_hashtbl
= kmalloc(sizeof(struct list_head
) *
6905 CLIENT_HASH_SIZE
, GFP_KERNEL
);
6906 if (!nn
->conf_id_hashtbl
)
6908 nn
->unconf_id_hashtbl
= kmalloc(sizeof(struct list_head
) *
6909 CLIENT_HASH_SIZE
, GFP_KERNEL
);
6910 if (!nn
->unconf_id_hashtbl
)
6912 nn
->sessionid_hashtbl
= kmalloc(sizeof(struct list_head
) *
6913 SESSION_HASH_SIZE
, GFP_KERNEL
);
6914 if (!nn
->sessionid_hashtbl
)
6917 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6918 INIT_LIST_HEAD(&nn
->conf_id_hashtbl
[i
]);
6919 INIT_LIST_HEAD(&nn
->unconf_id_hashtbl
[i
]);
6921 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
6922 INIT_LIST_HEAD(&nn
->sessionid_hashtbl
[i
]);
6923 nn
->conf_name_tree
= RB_ROOT
;
6924 nn
->unconf_name_tree
= RB_ROOT
;
6925 INIT_LIST_HEAD(&nn
->client_lru
);
6926 INIT_LIST_HEAD(&nn
->close_lru
);
6927 INIT_LIST_HEAD(&nn
->del_recall_lru
);
6928 spin_lock_init(&nn
->client_lock
);
6930 spin_lock_init(&nn
->blocked_locks_lock
);
6931 INIT_LIST_HEAD(&nn
->blocked_locks_lru
);
6933 INIT_DELAYED_WORK(&nn
->laundromat_work
, laundromat_main
);
6939 kfree(nn
->unconf_id_hashtbl
);
6941 kfree(nn
->conf_id_hashtbl
);
6947 nfs4_state_destroy_net(struct net
*net
)
6950 struct nfs4_client
*clp
= NULL
;
6951 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6953 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6954 while (!list_empty(&nn
->conf_id_hashtbl
[i
])) {
6955 clp
= list_entry(nn
->conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
6956 destroy_client(clp
);
6960 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6961 while (!list_empty(&nn
->unconf_id_hashtbl
[i
])) {
6962 clp
= list_entry(nn
->unconf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
6963 destroy_client(clp
);
6967 kfree(nn
->sessionid_hashtbl
);
6968 kfree(nn
->unconf_id_hashtbl
);
6969 kfree(nn
->conf_id_hashtbl
);
6974 nfs4_state_start_net(struct net
*net
)
6976 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
6979 ret
= nfs4_state_create_net(net
);
6982 nn
->boot_time
= get_seconds();
6983 nn
->grace_ended
= false;
6984 nn
->nfsd4_manager
.block_opens
= true;
6985 locks_start_grace(net
, &nn
->nfsd4_manager
);
6986 nfsd4_client_tracking_init(net
);
6987 printk(KERN_INFO
"NFSD: starting %ld-second grace period (net %p)\n",
6988 nn
->nfsd4_grace
, net
);
6989 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, nn
->nfsd4_grace
* HZ
);
6993 /* initialization to perform when the nfsd service is started: */
6996 nfs4_state_start(void)
7000 ret
= set_callback_cred();
7004 laundry_wq
= alloc_workqueue("%s", WQ_UNBOUND
, 0, "nfsd4");
7005 if (laundry_wq
== NULL
) {
7007 goto out_cleanup_cred
;
7009 ret
= nfsd4_create_callback_queue();
7011 goto out_free_laundry
;
7013 set_max_delegations();
7017 destroy_workqueue(laundry_wq
);
7019 cleanup_callback_cred();
7024 nfs4_state_shutdown_net(struct net
*net
)
7026 struct nfs4_delegation
*dp
= NULL
;
7027 struct list_head
*pos
, *next
, reaplist
;
7028 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
7029 struct nfsd4_blocked_lock
*nbl
;
7031 cancel_delayed_work_sync(&nn
->laundromat_work
);
7032 locks_end_grace(&nn
->nfsd4_manager
);
7034 INIT_LIST_HEAD(&reaplist
);
7035 spin_lock(&state_lock
);
7036 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
7037 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
7038 WARN_ON(!unhash_delegation_locked(dp
));
7039 list_add(&dp
->dl_recall_lru
, &reaplist
);
7041 spin_unlock(&state_lock
);
7042 list_for_each_safe(pos
, next
, &reaplist
) {
7043 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
7044 list_del_init(&dp
->dl_recall_lru
);
7045 put_clnt_odstate(dp
->dl_clnt_odstate
);
7046 nfs4_put_deleg_lease(dp
->dl_stid
.sc_file
);
7047 nfs4_put_stid(&dp
->dl_stid
);
7050 BUG_ON(!list_empty(&reaplist
));
7051 spin_lock(&nn
->blocked_locks_lock
);
7052 while (!list_empty(&nn
->blocked_locks_lru
)) {
7053 nbl
= list_first_entry(&nn
->blocked_locks_lru
,
7054 struct nfsd4_blocked_lock
, nbl_lru
);
7055 list_move(&nbl
->nbl_lru
, &reaplist
);
7056 list_del_init(&nbl
->nbl_list
);
7058 spin_unlock(&nn
->blocked_locks_lock
);
7060 while (!list_empty(&reaplist
)) {
7061 nbl
= list_first_entry(&nn
->blocked_locks_lru
,
7062 struct nfsd4_blocked_lock
, nbl_lru
);
7063 list_del_init(&nbl
->nbl_lru
);
7064 posix_unblock_lock(&nbl
->nbl_lock
);
7065 free_blocked_lock(nbl
);
7068 nfsd4_client_tracking_exit(net
);
7069 nfs4_state_destroy_net(net
);
7073 nfs4_state_shutdown(void)
7075 destroy_workqueue(laundry_wq
);
7076 nfsd4_destroy_callback_queue();
7077 cleanup_callback_cred();
7081 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
7083 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
7084 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
7088 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
7090 if (cstate
->minorversion
) {
7091 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
7092 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
7097 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
7099 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
7103 * functions to set current state id
7106 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
7108 put_stateid(cstate
, &odp
->od_stateid
);
7112 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
7114 put_stateid(cstate
, &open
->op_stateid
);
7118 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
7120 put_stateid(cstate
, &close
->cl_stateid
);
7124 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
7126 put_stateid(cstate
, &lock
->lk_resp_stateid
);
7130 * functions to consume current state id
7134 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
7136 get_stateid(cstate
, &odp
->od_stateid
);
7140 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
7142 get_stateid(cstate
, &drp
->dr_stateid
);
7146 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
7148 get_stateid(cstate
, &fsp
->fr_stateid
);
7152 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
7154 get_stateid(cstate
, &setattr
->sa_stateid
);
7158 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
7160 get_stateid(cstate
, &close
->cl_stateid
);
7164 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
7166 get_stateid(cstate
, &locku
->lu_stateid
);
7170 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
7172 get_stateid(cstate
, &read
->rd_stateid
);
7176 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
7178 get_stateid(cstate
, &write
->wr_stateid
);