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
= {
66 static const stateid_t close_stateid
= {
67 .si_generation
= 0xffffffffU
,
70 static u64 current_sessionid
= 1;
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
74 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
75 #define CLOSE_STATEID(stateid) (!memcmp((stateid), &close_stateid, sizeof(stateid_t)))
77 /* forward declarations */
78 static bool check_for_locks(struct nfs4_file
*fp
, struct nfs4_lockowner
*lowner
);
79 static void nfs4_free_ol_stateid(struct nfs4_stid
*stid
);
84 * Currently used for the del_recall_lru and file hash table. In an
85 * effort to decrease the scope of the client_mutex, this spinlock may
86 * eventually cover more:
88 static DEFINE_SPINLOCK(state_lock
);
90 enum nfsd4_st_mutex_lock_subclass
{
91 OPEN_STATEID_MUTEX
= 0,
92 LOCK_STATEID_MUTEX
= 1,
96 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
97 * the refcount on the open stateid to drop.
99 static DECLARE_WAIT_QUEUE_HEAD(close_wq
);
101 static struct kmem_cache
*client_slab
;
102 static struct kmem_cache
*openowner_slab
;
103 static struct kmem_cache
*lockowner_slab
;
104 static struct kmem_cache
*file_slab
;
105 static struct kmem_cache
*stateid_slab
;
106 static struct kmem_cache
*deleg_slab
;
107 static struct kmem_cache
*odstate_slab
;
109 static void free_session(struct nfsd4_session
*);
111 static const struct nfsd4_callback_ops nfsd4_cb_recall_ops
;
112 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops
;
114 static bool is_session_dead(struct nfsd4_session
*ses
)
116 return ses
->se_flags
& NFS4_SESSION_DEAD
;
119 static __be32
mark_session_dead_locked(struct nfsd4_session
*ses
, int ref_held_by_me
)
121 if (atomic_read(&ses
->se_ref
) > ref_held_by_me
)
122 return nfserr_jukebox
;
123 ses
->se_flags
|= NFS4_SESSION_DEAD
;
127 static bool is_client_expired(struct nfs4_client
*clp
)
129 return clp
->cl_time
== 0;
132 static __be32
get_client_locked(struct nfs4_client
*clp
)
134 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
136 lockdep_assert_held(&nn
->client_lock
);
138 if (is_client_expired(clp
))
139 return nfserr_expired
;
140 atomic_inc(&clp
->cl_refcount
);
144 /* must be called under the client_lock */
146 renew_client_locked(struct nfs4_client
*clp
)
148 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
150 if (is_client_expired(clp
)) {
152 printk("%s: client (clientid %08x/%08x) already expired\n",
154 clp
->cl_clientid
.cl_boot
,
155 clp
->cl_clientid
.cl_id
);
159 dprintk("renewing client (clientid %08x/%08x)\n",
160 clp
->cl_clientid
.cl_boot
,
161 clp
->cl_clientid
.cl_id
);
162 list_move_tail(&clp
->cl_lru
, &nn
->client_lru
);
163 clp
->cl_time
= get_seconds();
166 static void put_client_renew_locked(struct nfs4_client
*clp
)
168 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
170 lockdep_assert_held(&nn
->client_lock
);
172 if (!atomic_dec_and_test(&clp
->cl_refcount
))
174 if (!is_client_expired(clp
))
175 renew_client_locked(clp
);
178 static void put_client_renew(struct nfs4_client
*clp
)
180 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
182 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &nn
->client_lock
))
184 if (!is_client_expired(clp
))
185 renew_client_locked(clp
);
186 spin_unlock(&nn
->client_lock
);
189 static __be32
nfsd4_get_session_locked(struct nfsd4_session
*ses
)
193 if (is_session_dead(ses
))
194 return nfserr_badsession
;
195 status
= get_client_locked(ses
->se_client
);
198 atomic_inc(&ses
->se_ref
);
202 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
204 struct nfs4_client
*clp
= ses
->se_client
;
205 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
207 lockdep_assert_held(&nn
->client_lock
);
209 if (atomic_dec_and_test(&ses
->se_ref
) && is_session_dead(ses
))
211 put_client_renew_locked(clp
);
214 static void nfsd4_put_session(struct nfsd4_session
*ses
)
216 struct nfs4_client
*clp
= ses
->se_client
;
217 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
219 spin_lock(&nn
->client_lock
);
220 nfsd4_put_session_locked(ses
);
221 spin_unlock(&nn
->client_lock
);
224 static struct nfsd4_blocked_lock
*
225 find_blocked_lock(struct nfs4_lockowner
*lo
, struct knfsd_fh
*fh
,
228 struct nfsd4_blocked_lock
*cur
, *found
= NULL
;
230 spin_lock(&nn
->blocked_locks_lock
);
231 list_for_each_entry(cur
, &lo
->lo_blocked
, nbl_list
) {
232 if (fh_match(fh
, &cur
->nbl_fh
)) {
233 list_del_init(&cur
->nbl_list
);
234 list_del_init(&cur
->nbl_lru
);
239 spin_unlock(&nn
->blocked_locks_lock
);
241 locks_delete_block(&found
->nbl_lock
);
245 static struct nfsd4_blocked_lock
*
246 find_or_allocate_block(struct nfs4_lockowner
*lo
, struct knfsd_fh
*fh
,
249 struct nfsd4_blocked_lock
*nbl
;
251 nbl
= find_blocked_lock(lo
, fh
, nn
);
253 nbl
= kmalloc(sizeof(*nbl
), GFP_KERNEL
);
255 fh_copy_shallow(&nbl
->nbl_fh
, fh
);
256 locks_init_lock(&nbl
->nbl_lock
);
257 nfsd4_init_cb(&nbl
->nbl_cb
, lo
->lo_owner
.so_client
,
258 &nfsd4_cb_notify_lock_ops
,
259 NFSPROC4_CLNT_CB_NOTIFY_LOCK
);
266 free_blocked_lock(struct nfsd4_blocked_lock
*nbl
)
268 locks_release_private(&nbl
->nbl_lock
);
273 remove_blocked_locks(struct nfs4_lockowner
*lo
)
275 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
276 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
277 struct nfsd4_blocked_lock
*nbl
;
280 /* Dequeue all blocked locks */
281 spin_lock(&nn
->blocked_locks_lock
);
282 while (!list_empty(&lo
->lo_blocked
)) {
283 nbl
= list_first_entry(&lo
->lo_blocked
,
284 struct nfsd4_blocked_lock
,
286 list_del_init(&nbl
->nbl_list
);
287 list_move(&nbl
->nbl_lru
, &reaplist
);
289 spin_unlock(&nn
->blocked_locks_lock
);
292 while (!list_empty(&reaplist
)) {
293 nbl
= list_first_entry(&reaplist
, struct nfsd4_blocked_lock
,
295 list_del_init(&nbl
->nbl_lru
);
296 locks_delete_block(&nbl
->nbl_lock
);
297 free_blocked_lock(nbl
);
302 nfsd4_cb_notify_lock_done(struct nfsd4_callback
*cb
, struct rpc_task
*task
)
305 * Since this is just an optimization, we don't try very hard if it
306 * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
307 * just quit trying on anything else.
309 switch (task
->tk_status
) {
311 rpc_delay(task
, 1 * HZ
);
319 nfsd4_cb_notify_lock_release(struct nfsd4_callback
*cb
)
321 struct nfsd4_blocked_lock
*nbl
= container_of(cb
,
322 struct nfsd4_blocked_lock
, nbl_cb
);
324 free_blocked_lock(nbl
);
327 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops
= {
328 .done
= nfsd4_cb_notify_lock_done
,
329 .release
= nfsd4_cb_notify_lock_release
,
332 static inline struct nfs4_stateowner
*
333 nfs4_get_stateowner(struct nfs4_stateowner
*sop
)
335 atomic_inc(&sop
->so_count
);
340 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
)
342 return (sop
->so_owner
.len
== owner
->len
) &&
343 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
);
346 static struct nfs4_openowner
*
347 find_openstateowner_str_locked(unsigned int hashval
, struct nfsd4_open
*open
,
348 struct nfs4_client
*clp
)
350 struct nfs4_stateowner
*so
;
352 lockdep_assert_held(&clp
->cl_lock
);
354 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[hashval
],
356 if (!so
->so_is_open_owner
)
358 if (same_owner_str(so
, &open
->op_owner
))
359 return openowner(nfs4_get_stateowner(so
));
364 static struct nfs4_openowner
*
365 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
,
366 struct nfs4_client
*clp
)
368 struct nfs4_openowner
*oo
;
370 spin_lock(&clp
->cl_lock
);
371 oo
= find_openstateowner_str_locked(hashval
, open
, clp
);
372 spin_unlock(&clp
->cl_lock
);
377 opaque_hashval(const void *ptr
, int nbytes
)
379 unsigned char *cptr
= (unsigned char *) ptr
;
389 static void nfsd4_free_file_rcu(struct rcu_head
*rcu
)
391 struct nfs4_file
*fp
= container_of(rcu
, struct nfs4_file
, fi_rcu
);
393 kmem_cache_free(file_slab
, fp
);
397 put_nfs4_file(struct nfs4_file
*fi
)
399 might_lock(&state_lock
);
401 if (refcount_dec_and_lock(&fi
->fi_ref
, &state_lock
)) {
402 hlist_del_rcu(&fi
->fi_hash
);
403 spin_unlock(&state_lock
);
404 WARN_ON_ONCE(!list_empty(&fi
->fi_clnt_odstate
));
405 WARN_ON_ONCE(!list_empty(&fi
->fi_delegations
));
406 call_rcu(&fi
->fi_rcu
, nfsd4_free_file_rcu
);
411 __nfs4_get_fd(struct nfs4_file
*f
, int oflag
)
413 if (f
->fi_fds
[oflag
])
414 return get_file(f
->fi_fds
[oflag
]);
419 find_writeable_file_locked(struct nfs4_file
*f
)
423 lockdep_assert_held(&f
->fi_lock
);
425 ret
= __nfs4_get_fd(f
, O_WRONLY
);
427 ret
= __nfs4_get_fd(f
, O_RDWR
);
432 find_writeable_file(struct nfs4_file
*f
)
436 spin_lock(&f
->fi_lock
);
437 ret
= find_writeable_file_locked(f
);
438 spin_unlock(&f
->fi_lock
);
443 static struct file
*find_readable_file_locked(struct nfs4_file
*f
)
447 lockdep_assert_held(&f
->fi_lock
);
449 ret
= __nfs4_get_fd(f
, O_RDONLY
);
451 ret
= __nfs4_get_fd(f
, O_RDWR
);
456 find_readable_file(struct nfs4_file
*f
)
460 spin_lock(&f
->fi_lock
);
461 ret
= find_readable_file_locked(f
);
462 spin_unlock(&f
->fi_lock
);
468 find_any_file(struct nfs4_file
*f
)
472 spin_lock(&f
->fi_lock
);
473 ret
= __nfs4_get_fd(f
, O_RDWR
);
475 ret
= __nfs4_get_fd(f
, O_WRONLY
);
477 ret
= __nfs4_get_fd(f
, O_RDONLY
);
479 spin_unlock(&f
->fi_lock
);
483 static atomic_long_t num_delegations
;
484 unsigned long max_delegations
;
487 * Open owner state (share locks)
490 /* hash tables for lock and open owners */
491 #define OWNER_HASH_BITS 8
492 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
493 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
495 static unsigned int ownerstr_hashval(struct xdr_netobj
*ownername
)
499 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
500 return ret
& OWNER_HASH_MASK
;
503 /* hash table for nfs4_file */
504 #define FILE_HASH_BITS 8
505 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
507 static unsigned int nfsd_fh_hashval(struct knfsd_fh
*fh
)
509 return jhash2(fh
->fh_base
.fh_pad
, XDR_QUADLEN(fh
->fh_size
), 0);
512 static unsigned int file_hashval(struct knfsd_fh
*fh
)
514 return nfsd_fh_hashval(fh
) & (FILE_HASH_SIZE
- 1);
517 static struct hlist_head file_hashtbl
[FILE_HASH_SIZE
];
520 __nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
522 lockdep_assert_held(&fp
->fi_lock
);
524 if (access
& NFS4_SHARE_ACCESS_WRITE
)
525 atomic_inc(&fp
->fi_access
[O_WRONLY
]);
526 if (access
& NFS4_SHARE_ACCESS_READ
)
527 atomic_inc(&fp
->fi_access
[O_RDONLY
]);
531 nfs4_file_get_access(struct nfs4_file
*fp
, u32 access
)
533 lockdep_assert_held(&fp
->fi_lock
);
535 /* Does this access mode make sense? */
536 if (access
& ~NFS4_SHARE_ACCESS_BOTH
)
539 /* Does it conflict with a deny mode already set? */
540 if ((access
& fp
->fi_share_deny
) != 0)
541 return nfserr_share_denied
;
543 __nfs4_file_get_access(fp
, access
);
547 static __be32
nfs4_file_check_deny(struct nfs4_file
*fp
, u32 deny
)
549 /* Common case is that there is no deny mode. */
551 /* Does this deny mode make sense? */
552 if (deny
& ~NFS4_SHARE_DENY_BOTH
)
555 if ((deny
& NFS4_SHARE_DENY_READ
) &&
556 atomic_read(&fp
->fi_access
[O_RDONLY
]))
557 return nfserr_share_denied
;
559 if ((deny
& NFS4_SHARE_DENY_WRITE
) &&
560 atomic_read(&fp
->fi_access
[O_WRONLY
]))
561 return nfserr_share_denied
;
566 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
568 might_lock(&fp
->fi_lock
);
570 if (atomic_dec_and_lock(&fp
->fi_access
[oflag
], &fp
->fi_lock
)) {
571 struct file
*f1
= NULL
;
572 struct file
*f2
= NULL
;
574 swap(f1
, fp
->fi_fds
[oflag
]);
575 if (atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
576 swap(f2
, fp
->fi_fds
[O_RDWR
]);
577 spin_unlock(&fp
->fi_lock
);
585 static void nfs4_file_put_access(struct nfs4_file
*fp
, u32 access
)
587 WARN_ON_ONCE(access
& ~NFS4_SHARE_ACCESS_BOTH
);
589 if (access
& NFS4_SHARE_ACCESS_WRITE
)
590 __nfs4_file_put_access(fp
, O_WRONLY
);
591 if (access
& NFS4_SHARE_ACCESS_READ
)
592 __nfs4_file_put_access(fp
, O_RDONLY
);
596 * Allocate a new open/delegation state counter. This is needed for
597 * pNFS for proper return on close semantics.
599 * Note that we only allocate it for pNFS-enabled exports, otherwise
600 * all pointers to struct nfs4_clnt_odstate are always NULL.
602 static struct nfs4_clnt_odstate
*
603 alloc_clnt_odstate(struct nfs4_client
*clp
)
605 struct nfs4_clnt_odstate
*co
;
607 co
= kmem_cache_zalloc(odstate_slab
, GFP_KERNEL
);
610 refcount_set(&co
->co_odcount
, 1);
616 hash_clnt_odstate_locked(struct nfs4_clnt_odstate
*co
)
618 struct nfs4_file
*fp
= co
->co_file
;
620 lockdep_assert_held(&fp
->fi_lock
);
621 list_add(&co
->co_perfile
, &fp
->fi_clnt_odstate
);
625 get_clnt_odstate(struct nfs4_clnt_odstate
*co
)
628 refcount_inc(&co
->co_odcount
);
632 put_clnt_odstate(struct nfs4_clnt_odstate
*co
)
634 struct nfs4_file
*fp
;
640 if (refcount_dec_and_lock(&co
->co_odcount
, &fp
->fi_lock
)) {
641 list_del(&co
->co_perfile
);
642 spin_unlock(&fp
->fi_lock
);
644 nfsd4_return_all_file_layouts(co
->co_client
, fp
);
645 kmem_cache_free(odstate_slab
, co
);
649 static struct nfs4_clnt_odstate
*
650 find_or_hash_clnt_odstate(struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*new)
652 struct nfs4_clnt_odstate
*co
;
653 struct nfs4_client
*cl
;
660 spin_lock(&fp
->fi_lock
);
661 list_for_each_entry(co
, &fp
->fi_clnt_odstate
, co_perfile
) {
662 if (co
->co_client
== cl
) {
663 get_clnt_odstate(co
);
669 hash_clnt_odstate_locked(new);
671 spin_unlock(&fp
->fi_lock
);
675 struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
,
676 void (*sc_free
)(struct nfs4_stid
*))
678 struct nfs4_stid
*stid
;
681 stid
= kmem_cache_zalloc(slab
, GFP_KERNEL
);
685 idr_preload(GFP_KERNEL
);
686 spin_lock(&cl
->cl_lock
);
687 new_id
= idr_alloc_cyclic(&cl
->cl_stateids
, stid
, 0, 0, GFP_NOWAIT
);
688 spin_unlock(&cl
->cl_lock
);
693 stid
->sc_free
= sc_free
;
694 stid
->sc_client
= cl
;
695 stid
->sc_stateid
.si_opaque
.so_id
= new_id
;
696 stid
->sc_stateid
.si_opaque
.so_clid
= cl
->cl_clientid
;
697 /* Will be incremented before return to client: */
698 refcount_set(&stid
->sc_count
, 1);
699 spin_lock_init(&stid
->sc_lock
);
702 * It shouldn't be a problem to reuse an opaque stateid value.
703 * I don't think it is for 4.1. But with 4.0 I worry that, for
704 * example, a stray write retransmission could be accepted by
705 * the server when it should have been rejected. Therefore,
706 * adopt a trick from the sctp code to attempt to maximize the
707 * amount of time until an id is reused, by ensuring they always
708 * "increase" (mod INT_MAX):
712 kmem_cache_free(slab
, stid
);
717 * Create a unique stateid_t to represent each COPY.
719 int nfs4_init_cp_state(struct nfsd_net
*nn
, struct nfsd4_copy
*copy
)
723 idr_preload(GFP_KERNEL
);
724 spin_lock(&nn
->s2s_cp_lock
);
725 new_id
= idr_alloc_cyclic(&nn
->s2s_cp_stateids
, copy
, 0, 0, GFP_NOWAIT
);
726 spin_unlock(&nn
->s2s_cp_lock
);
730 copy
->cp_stateid
.si_opaque
.so_id
= new_id
;
731 copy
->cp_stateid
.si_opaque
.so_clid
.cl_boot
= nn
->boot_time
;
732 copy
->cp_stateid
.si_opaque
.so_clid
.cl_id
= nn
->s2s_cp_cl_id
;
736 void nfs4_free_cp_state(struct nfsd4_copy
*copy
)
740 nn
= net_generic(copy
->cp_clp
->net
, nfsd_net_id
);
741 spin_lock(&nn
->s2s_cp_lock
);
742 idr_remove(&nn
->s2s_cp_stateids
, copy
->cp_stateid
.si_opaque
.so_id
);
743 spin_unlock(&nn
->s2s_cp_lock
);
746 static struct nfs4_ol_stateid
* nfs4_alloc_open_stateid(struct nfs4_client
*clp
)
748 struct nfs4_stid
*stid
;
750 stid
= nfs4_alloc_stid(clp
, stateid_slab
, nfs4_free_ol_stateid
);
754 return openlockstateid(stid
);
757 static void nfs4_free_deleg(struct nfs4_stid
*stid
)
759 kmem_cache_free(deleg_slab
, stid
);
760 atomic_long_dec(&num_delegations
);
764 * When we recall a delegation, we should be careful not to hand it
765 * out again straight away.
766 * To ensure this we keep a pair of bloom filters ('new' and 'old')
767 * in which the filehandles of recalled delegations are "stored".
768 * If a filehandle appear in either filter, a delegation is blocked.
769 * When a delegation is recalled, the filehandle is stored in the "new"
771 * Every 30 seconds we swap the filters and clear the "new" one,
772 * unless both are empty of course.
774 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
775 * low 3 bytes as hash-table indices.
777 * 'blocked_delegations_lock', which is always taken in block_delegations(),
778 * is used to manage concurrent access. Testing does not need the lock
779 * except when swapping the two filters.
781 static DEFINE_SPINLOCK(blocked_delegations_lock
);
782 static struct bloom_pair
{
783 int entries
, old_entries
;
785 int new; /* index into 'set' */
786 DECLARE_BITMAP(set
[2], 256);
787 } blocked_delegations
;
789 static int delegation_blocked(struct knfsd_fh
*fh
)
792 struct bloom_pair
*bd
= &blocked_delegations
;
794 if (bd
->entries
== 0)
796 if (seconds_since_boot() - bd
->swap_time
> 30) {
797 spin_lock(&blocked_delegations_lock
);
798 if (seconds_since_boot() - bd
->swap_time
> 30) {
799 bd
->entries
-= bd
->old_entries
;
800 bd
->old_entries
= bd
->entries
;
801 memset(bd
->set
[bd
->new], 0,
804 bd
->swap_time
= seconds_since_boot();
806 spin_unlock(&blocked_delegations_lock
);
808 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
809 if (test_bit(hash
&255, bd
->set
[0]) &&
810 test_bit((hash
>>8)&255, bd
->set
[0]) &&
811 test_bit((hash
>>16)&255, bd
->set
[0]))
814 if (test_bit(hash
&255, bd
->set
[1]) &&
815 test_bit((hash
>>8)&255, bd
->set
[1]) &&
816 test_bit((hash
>>16)&255, bd
->set
[1]))
822 static void block_delegations(struct knfsd_fh
*fh
)
825 struct bloom_pair
*bd
= &blocked_delegations
;
827 hash
= jhash(&fh
->fh_base
, fh
->fh_size
, 0);
829 spin_lock(&blocked_delegations_lock
);
830 __set_bit(hash
&255, bd
->set
[bd
->new]);
831 __set_bit((hash
>>8)&255, bd
->set
[bd
->new]);
832 __set_bit((hash
>>16)&255, bd
->set
[bd
->new]);
833 if (bd
->entries
== 0)
834 bd
->swap_time
= seconds_since_boot();
836 spin_unlock(&blocked_delegations_lock
);
839 static struct nfs4_delegation
*
840 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_file
*fp
,
841 struct svc_fh
*current_fh
,
842 struct nfs4_clnt_odstate
*odstate
)
844 struct nfs4_delegation
*dp
;
847 dprintk("NFSD alloc_init_deleg\n");
848 n
= atomic_long_inc_return(&num_delegations
);
849 if (n
< 0 || n
> max_delegations
)
851 if (delegation_blocked(¤t_fh
->fh_handle
))
853 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
, nfs4_free_deleg
));
858 * delegation seqid's are never incremented. The 4.1 special
859 * meaning of seqid 0 isn't meaningful, really, but let's avoid
860 * 0 anyway just for consistency and use 1:
862 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
863 INIT_LIST_HEAD(&dp
->dl_perfile
);
864 INIT_LIST_HEAD(&dp
->dl_perclnt
);
865 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
866 dp
->dl_clnt_odstate
= odstate
;
867 get_clnt_odstate(odstate
);
868 dp
->dl_type
= NFS4_OPEN_DELEGATE_READ
;
870 nfsd4_init_cb(&dp
->dl_recall
, dp
->dl_stid
.sc_client
,
871 &nfsd4_cb_recall_ops
, NFSPROC4_CLNT_CB_RECALL
);
873 dp
->dl_stid
.sc_file
= fp
;
876 atomic_long_dec(&num_delegations
);
881 nfs4_put_stid(struct nfs4_stid
*s
)
883 struct nfs4_file
*fp
= s
->sc_file
;
884 struct nfs4_client
*clp
= s
->sc_client
;
886 might_lock(&clp
->cl_lock
);
888 if (!refcount_dec_and_lock(&s
->sc_count
, &clp
->cl_lock
)) {
889 wake_up_all(&close_wq
);
892 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
893 spin_unlock(&clp
->cl_lock
);
900 nfs4_inc_and_copy_stateid(stateid_t
*dst
, struct nfs4_stid
*stid
)
902 stateid_t
*src
= &stid
->sc_stateid
;
904 spin_lock(&stid
->sc_lock
);
905 if (unlikely(++src
->si_generation
== 0))
906 src
->si_generation
= 1;
907 memcpy(dst
, src
, sizeof(*dst
));
908 spin_unlock(&stid
->sc_lock
);
911 static void put_deleg_file(struct nfs4_file
*fp
)
913 struct file
*filp
= NULL
;
915 spin_lock(&fp
->fi_lock
);
916 if (--fp
->fi_delegees
== 0)
917 swap(filp
, fp
->fi_deleg_file
);
918 spin_unlock(&fp
->fi_lock
);
924 static void nfs4_unlock_deleg_lease(struct nfs4_delegation
*dp
)
926 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
927 struct file
*filp
= fp
->fi_deleg_file
;
929 WARN_ON_ONCE(!fp
->fi_delegees
);
931 vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&dp
);
935 static void destroy_unhashed_deleg(struct nfs4_delegation
*dp
)
937 put_clnt_odstate(dp
->dl_clnt_odstate
);
938 nfs4_unlock_deleg_lease(dp
);
939 nfs4_put_stid(&dp
->dl_stid
);
942 void nfs4_unhash_stid(struct nfs4_stid
*s
)
948 * nfs4_delegation_exists - Discover if this delegation already exists
949 * @clp: a pointer to the nfs4_client we're granting a delegation to
950 * @fp: a pointer to the nfs4_file we're granting a delegation on
953 * On success: true iff an existing delegation is found
957 nfs4_delegation_exists(struct nfs4_client
*clp
, struct nfs4_file
*fp
)
959 struct nfs4_delegation
*searchdp
= NULL
;
960 struct nfs4_client
*searchclp
= NULL
;
962 lockdep_assert_held(&state_lock
);
963 lockdep_assert_held(&fp
->fi_lock
);
965 list_for_each_entry(searchdp
, &fp
->fi_delegations
, dl_perfile
) {
966 searchclp
= searchdp
->dl_stid
.sc_client
;
967 if (clp
== searchclp
) {
975 * hash_delegation_locked - Add a delegation to the appropriate lists
976 * @dp: a pointer to the nfs4_delegation we are adding.
977 * @fp: a pointer to the nfs4_file we're granting a delegation on
980 * On success: NULL if the delegation was successfully hashed.
982 * On error: -EAGAIN if one was previously granted to this
983 * nfs4_client for this nfs4_file. Delegation is not hashed.
988 hash_delegation_locked(struct nfs4_delegation
*dp
, struct nfs4_file
*fp
)
990 struct nfs4_client
*clp
= dp
->dl_stid
.sc_client
;
992 lockdep_assert_held(&state_lock
);
993 lockdep_assert_held(&fp
->fi_lock
);
995 if (nfs4_delegation_exists(clp
, fp
))
997 refcount_inc(&dp
->dl_stid
.sc_count
);
998 dp
->dl_stid
.sc_type
= NFS4_DELEG_STID
;
999 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
1000 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
1005 unhash_delegation_locked(struct nfs4_delegation
*dp
)
1007 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
1009 lockdep_assert_held(&state_lock
);
1011 if (list_empty(&dp
->dl_perfile
))
1014 dp
->dl_stid
.sc_type
= NFS4_CLOSED_DELEG_STID
;
1015 /* Ensure that deleg break won't try to requeue it */
1017 spin_lock(&fp
->fi_lock
);
1018 list_del_init(&dp
->dl_perclnt
);
1019 list_del_init(&dp
->dl_recall_lru
);
1020 list_del_init(&dp
->dl_perfile
);
1021 spin_unlock(&fp
->fi_lock
);
1025 static void destroy_delegation(struct nfs4_delegation
*dp
)
1029 spin_lock(&state_lock
);
1030 unhashed
= unhash_delegation_locked(dp
);
1031 spin_unlock(&state_lock
);
1033 destroy_unhashed_deleg(dp
);
1036 static void revoke_delegation(struct nfs4_delegation
*dp
)
1038 struct nfs4_client
*clp
= dp
->dl_stid
.sc_client
;
1040 WARN_ON(!list_empty(&dp
->dl_recall_lru
));
1042 if (clp
->cl_minorversion
) {
1043 dp
->dl_stid
.sc_type
= NFS4_REVOKED_DELEG_STID
;
1044 refcount_inc(&dp
->dl_stid
.sc_count
);
1045 spin_lock(&clp
->cl_lock
);
1046 list_add(&dp
->dl_recall_lru
, &clp
->cl_revoked
);
1047 spin_unlock(&clp
->cl_lock
);
1049 destroy_unhashed_deleg(dp
);
1056 static unsigned int clientid_hashval(u32 id
)
1058 return id
& CLIENT_HASH_MASK
;
1061 static unsigned int clientstr_hashval(const char *name
)
1063 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
1067 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1068 * st_{access,deny}_bmap field of the stateid, in order to track not
1069 * only what share bits are currently in force, but also what
1070 * combinations of share bits previous opens have used. This allows us
1071 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1072 * return an error if the client attempt to downgrade to a combination
1073 * of share bits not explicable by closing some of its previous opens.
1075 * XXX: This enforcement is actually incomplete, since we don't keep
1076 * track of access/deny bit combinations; so, e.g., we allow:
1078 * OPEN allow read, deny write
1079 * OPEN allow both, deny none
1080 * DOWNGRADE allow read, deny none
1082 * which we should reject.
1085 bmap_to_share_mode(unsigned long bmap
) {
1087 unsigned int access
= 0;
1089 for (i
= 1; i
< 4; i
++) {
1090 if (test_bit(i
, &bmap
))
1096 /* set share access for a given stateid */
1098 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1100 unsigned char mask
= 1 << access
;
1102 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
1103 stp
->st_access_bmap
|= mask
;
1106 /* clear share access for a given stateid */
1108 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1110 unsigned char mask
= 1 << access
;
1112 WARN_ON_ONCE(access
> NFS4_SHARE_ACCESS_BOTH
);
1113 stp
->st_access_bmap
&= ~mask
;
1116 /* test whether a given stateid has access */
1118 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
1120 unsigned char mask
= 1 << access
;
1122 return (bool)(stp
->st_access_bmap
& mask
);
1125 /* set share deny for a given stateid */
1127 set_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1129 unsigned char mask
= 1 << deny
;
1131 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
1132 stp
->st_deny_bmap
|= mask
;
1135 /* clear share deny for a given stateid */
1137 clear_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1139 unsigned char mask
= 1 << deny
;
1141 WARN_ON_ONCE(deny
> NFS4_SHARE_DENY_BOTH
);
1142 stp
->st_deny_bmap
&= ~mask
;
1145 /* test whether a given stateid is denying specific access */
1147 test_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1149 unsigned char mask
= 1 << deny
;
1151 return (bool)(stp
->st_deny_bmap
& mask
);
1154 static int nfs4_access_to_omode(u32 access
)
1156 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
1157 case NFS4_SHARE_ACCESS_READ
:
1159 case NFS4_SHARE_ACCESS_WRITE
:
1161 case NFS4_SHARE_ACCESS_BOTH
:
1169 * A stateid that had a deny mode associated with it is being released
1170 * or downgraded. Recalculate the deny mode on the file.
1173 recalculate_deny_mode(struct nfs4_file
*fp
)
1175 struct nfs4_ol_stateid
*stp
;
1177 spin_lock(&fp
->fi_lock
);
1178 fp
->fi_share_deny
= 0;
1179 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
)
1180 fp
->fi_share_deny
|= bmap_to_share_mode(stp
->st_deny_bmap
);
1181 spin_unlock(&fp
->fi_lock
);
1185 reset_union_bmap_deny(u32 deny
, struct nfs4_ol_stateid
*stp
)
1188 bool change
= false;
1190 for (i
= 1; i
< 4; i
++) {
1191 if ((i
& deny
) != i
) {
1197 /* Recalculate per-file deny mode if there was a change */
1199 recalculate_deny_mode(stp
->st_stid
.sc_file
);
1202 /* release all access and file references for a given stateid */
1204 release_all_access(struct nfs4_ol_stateid
*stp
)
1207 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
1209 if (fp
&& stp
->st_deny_bmap
!= 0)
1210 recalculate_deny_mode(fp
);
1212 for (i
= 1; i
< 4; i
++) {
1213 if (test_access(i
, stp
))
1214 nfs4_file_put_access(stp
->st_stid
.sc_file
, i
);
1215 clear_access(i
, stp
);
1219 static inline void nfs4_free_stateowner(struct nfs4_stateowner
*sop
)
1221 kfree(sop
->so_owner
.data
);
1222 sop
->so_ops
->so_free(sop
);
1225 static void nfs4_put_stateowner(struct nfs4_stateowner
*sop
)
1227 struct nfs4_client
*clp
= sop
->so_client
;
1229 might_lock(&clp
->cl_lock
);
1231 if (!atomic_dec_and_lock(&sop
->so_count
, &clp
->cl_lock
))
1233 sop
->so_ops
->so_unhash(sop
);
1234 spin_unlock(&clp
->cl_lock
);
1235 nfs4_free_stateowner(sop
);
1238 static bool unhash_ol_stateid(struct nfs4_ol_stateid
*stp
)
1240 struct nfs4_file
*fp
= stp
->st_stid
.sc_file
;
1242 lockdep_assert_held(&stp
->st_stateowner
->so_client
->cl_lock
);
1244 if (list_empty(&stp
->st_perfile
))
1247 spin_lock(&fp
->fi_lock
);
1248 list_del_init(&stp
->st_perfile
);
1249 spin_unlock(&fp
->fi_lock
);
1250 list_del(&stp
->st_perstateowner
);
1254 static void nfs4_free_ol_stateid(struct nfs4_stid
*stid
)
1256 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1258 put_clnt_odstate(stp
->st_clnt_odstate
);
1259 release_all_access(stp
);
1260 if (stp
->st_stateowner
)
1261 nfs4_put_stateowner(stp
->st_stateowner
);
1262 kmem_cache_free(stateid_slab
, stid
);
1265 static void nfs4_free_lock_stateid(struct nfs4_stid
*stid
)
1267 struct nfs4_ol_stateid
*stp
= openlockstateid(stid
);
1268 struct nfs4_lockowner
*lo
= lockowner(stp
->st_stateowner
);
1271 file
= find_any_file(stp
->st_stid
.sc_file
);
1273 filp_close(file
, (fl_owner_t
)lo
);
1274 nfs4_free_ol_stateid(stid
);
1278 * Put the persistent reference to an already unhashed generic stateid, while
1279 * holding the cl_lock. If it's the last reference, then put it onto the
1280 * reaplist for later destruction.
1282 static void put_ol_stateid_locked(struct nfs4_ol_stateid
*stp
,
1283 struct list_head
*reaplist
)
1285 struct nfs4_stid
*s
= &stp
->st_stid
;
1286 struct nfs4_client
*clp
= s
->sc_client
;
1288 lockdep_assert_held(&clp
->cl_lock
);
1290 WARN_ON_ONCE(!list_empty(&stp
->st_locks
));
1292 if (!refcount_dec_and_test(&s
->sc_count
)) {
1293 wake_up_all(&close_wq
);
1297 idr_remove(&clp
->cl_stateids
, s
->sc_stateid
.si_opaque
.so_id
);
1298 list_add(&stp
->st_locks
, reaplist
);
1301 static bool unhash_lock_stateid(struct nfs4_ol_stateid
*stp
)
1303 lockdep_assert_held(&stp
->st_stid
.sc_client
->cl_lock
);
1305 list_del_init(&stp
->st_locks
);
1306 nfs4_unhash_stid(&stp
->st_stid
);
1307 return unhash_ol_stateid(stp
);
1310 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
1312 struct nfs4_client
*clp
= stp
->st_stid
.sc_client
;
1315 spin_lock(&clp
->cl_lock
);
1316 unhashed
= unhash_lock_stateid(stp
);
1317 spin_unlock(&clp
->cl_lock
);
1319 nfs4_put_stid(&stp
->st_stid
);
1322 static void unhash_lockowner_locked(struct nfs4_lockowner
*lo
)
1324 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
1326 lockdep_assert_held(&clp
->cl_lock
);
1328 list_del_init(&lo
->lo_owner
.so_strhash
);
1332 * Free a list of generic stateids that were collected earlier after being
1336 free_ol_stateid_reaplist(struct list_head
*reaplist
)
1338 struct nfs4_ol_stateid
*stp
;
1339 struct nfs4_file
*fp
;
1343 while (!list_empty(reaplist
)) {
1344 stp
= list_first_entry(reaplist
, struct nfs4_ol_stateid
,
1346 list_del(&stp
->st_locks
);
1347 fp
= stp
->st_stid
.sc_file
;
1348 stp
->st_stid
.sc_free(&stp
->st_stid
);
1354 static void release_open_stateid_locks(struct nfs4_ol_stateid
*open_stp
,
1355 struct list_head
*reaplist
)
1357 struct nfs4_ol_stateid
*stp
;
1359 lockdep_assert_held(&open_stp
->st_stid
.sc_client
->cl_lock
);
1361 while (!list_empty(&open_stp
->st_locks
)) {
1362 stp
= list_entry(open_stp
->st_locks
.next
,
1363 struct nfs4_ol_stateid
, st_locks
);
1364 WARN_ON(!unhash_lock_stateid(stp
));
1365 put_ol_stateid_locked(stp
, reaplist
);
1369 static bool unhash_open_stateid(struct nfs4_ol_stateid
*stp
,
1370 struct list_head
*reaplist
)
1374 lockdep_assert_held(&stp
->st_stid
.sc_client
->cl_lock
);
1376 unhashed
= unhash_ol_stateid(stp
);
1377 release_open_stateid_locks(stp
, reaplist
);
1381 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
1383 LIST_HEAD(reaplist
);
1385 spin_lock(&stp
->st_stid
.sc_client
->cl_lock
);
1386 if (unhash_open_stateid(stp
, &reaplist
))
1387 put_ol_stateid_locked(stp
, &reaplist
);
1388 spin_unlock(&stp
->st_stid
.sc_client
->cl_lock
);
1389 free_ol_stateid_reaplist(&reaplist
);
1392 static void unhash_openowner_locked(struct nfs4_openowner
*oo
)
1394 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1396 lockdep_assert_held(&clp
->cl_lock
);
1398 list_del_init(&oo
->oo_owner
.so_strhash
);
1399 list_del_init(&oo
->oo_perclient
);
1402 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
1404 struct nfsd_net
*nn
= net_generic(oo
->oo_owner
.so_client
->net
,
1406 struct nfs4_ol_stateid
*s
;
1408 spin_lock(&nn
->client_lock
);
1409 s
= oo
->oo_last_closed_stid
;
1411 list_del_init(&oo
->oo_close_lru
);
1412 oo
->oo_last_closed_stid
= NULL
;
1414 spin_unlock(&nn
->client_lock
);
1416 nfs4_put_stid(&s
->st_stid
);
1419 static void release_openowner(struct nfs4_openowner
*oo
)
1421 struct nfs4_ol_stateid
*stp
;
1422 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
1423 struct list_head reaplist
;
1425 INIT_LIST_HEAD(&reaplist
);
1427 spin_lock(&clp
->cl_lock
);
1428 unhash_openowner_locked(oo
);
1429 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
1430 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
1431 struct nfs4_ol_stateid
, st_perstateowner
);
1432 if (unhash_open_stateid(stp
, &reaplist
))
1433 put_ol_stateid_locked(stp
, &reaplist
);
1435 spin_unlock(&clp
->cl_lock
);
1436 free_ol_stateid_reaplist(&reaplist
);
1437 release_last_closed_stateid(oo
);
1438 nfs4_put_stateowner(&oo
->oo_owner
);
1442 hash_sessionid(struct nfs4_sessionid
*sessionid
)
1444 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
1446 return sid
->sequence
% SESSION_HASH_SIZE
;
1449 #ifdef CONFIG_SUNRPC_DEBUG
1451 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1453 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
1454 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
1458 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
1464 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1465 * won't be used for replay.
1467 void nfsd4_bump_seqid(struct nfsd4_compound_state
*cstate
, __be32 nfserr
)
1469 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
1471 if (nfserr
== nfserr_replay_me
)
1474 if (!seqid_mutating_err(ntohl(nfserr
))) {
1475 nfsd4_cstate_clear_replay(cstate
);
1480 if (so
->so_is_open_owner
)
1481 release_last_closed_stateid(openowner(so
));
1487 gen_sessionid(struct nfsd4_session
*ses
)
1489 struct nfs4_client
*clp
= ses
->se_client
;
1490 struct nfsd4_sessionid
*sid
;
1492 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
1493 sid
->clientid
= clp
->cl_clientid
;
1494 sid
->sequence
= current_sessionid
++;
1499 * The protocol defines ca_maxresponssize_cached to include the size of
1500 * the rpc header, but all we need to cache is the data starting after
1501 * the end of the initial SEQUENCE operation--the rest we regenerate
1502 * each time. Therefore we can advertise a ca_maxresponssize_cached
1503 * value that is the number of bytes in our cache plus a few additional
1504 * bytes. In order to stay on the safe side, and not promise more than
1505 * we can cache, those additional bytes must be the minimum possible: 24
1506 * bytes of rpc header (xid through accept state, with AUTH_NULL
1507 * verifier), 12 for the compound header (with zero-length tag), and 44
1508 * for the SEQUENCE op response:
1510 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1513 free_session_slots(struct nfsd4_session
*ses
)
1517 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++) {
1518 free_svc_cred(&ses
->se_slots
[i
]->sl_cred
);
1519 kfree(ses
->se_slots
[i
]);
1524 * We don't actually need to cache the rpc and session headers, so we
1525 * can allocate a little less for each slot:
1527 static inline u32
slot_bytes(struct nfsd4_channel_attrs
*ca
)
1531 if (ca
->maxresp_cached
< NFSD_MIN_HDR_SEQ_SZ
)
1534 size
= ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
1535 return size
+ sizeof(struct nfsd4_slot
);
1539 * XXX: If we run out of reserved DRC memory we could (up to a point)
1540 * re-negotiate active sessions and reduce their slot usage to make
1541 * room for new connections. For now we just fail the create session.
1543 static u32
nfsd4_get_drc_mem(struct nfsd4_channel_attrs
*ca
)
1545 u32 slotsize
= slot_bytes(ca
);
1546 u32 num
= ca
->maxreqs
;
1549 spin_lock(&nfsd_drc_lock
);
1550 avail
= min((unsigned long)NFSD_MAX_MEM_PER_SESSION
,
1551 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
1553 * Never use more than a third of the remaining memory,
1554 * unless it's the only way to give this client a slot:
1556 avail
= clamp_t(int, avail
, slotsize
, avail
/3);
1557 num
= min_t(int, num
, avail
/ slotsize
);
1558 nfsd_drc_mem_used
+= num
* slotsize
;
1559 spin_unlock(&nfsd_drc_lock
);
1564 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs
*ca
)
1566 int slotsize
= slot_bytes(ca
);
1568 spin_lock(&nfsd_drc_lock
);
1569 nfsd_drc_mem_used
-= slotsize
* ca
->maxreqs
;
1570 spin_unlock(&nfsd_drc_lock
);
1573 static struct nfsd4_session
*alloc_session(struct nfsd4_channel_attrs
*fattrs
,
1574 struct nfsd4_channel_attrs
*battrs
)
1576 int numslots
= fattrs
->maxreqs
;
1577 int slotsize
= slot_bytes(fattrs
);
1578 struct nfsd4_session
*new;
1581 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
1582 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
1583 mem
= numslots
* sizeof(struct nfsd4_slot
*);
1585 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
1588 /* allocate each struct nfsd4_slot and data cache in one piece */
1589 for (i
= 0; i
< numslots
; i
++) {
1590 new->se_slots
[i
] = kzalloc(slotsize
, GFP_KERNEL
);
1591 if (!new->se_slots
[i
])
1595 memcpy(&new->se_fchannel
, fattrs
, sizeof(struct nfsd4_channel_attrs
));
1596 memcpy(&new->se_bchannel
, battrs
, sizeof(struct nfsd4_channel_attrs
));
1601 kfree(new->se_slots
[i
]);
1606 static void free_conn(struct nfsd4_conn
*c
)
1608 svc_xprt_put(c
->cn_xprt
);
1612 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
1614 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
1615 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
1617 spin_lock(&clp
->cl_lock
);
1618 if (!list_empty(&c
->cn_persession
)) {
1619 list_del(&c
->cn_persession
);
1622 nfsd4_probe_callback(clp
);
1623 spin_unlock(&clp
->cl_lock
);
1626 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
1628 struct nfsd4_conn
*conn
;
1630 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
1633 svc_xprt_get(rqstp
->rq_xprt
);
1634 conn
->cn_xprt
= rqstp
->rq_xprt
;
1635 conn
->cn_flags
= flags
;
1636 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
1640 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1642 conn
->cn_session
= ses
;
1643 list_add(&conn
->cn_persession
, &ses
->se_conns
);
1646 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1648 struct nfs4_client
*clp
= ses
->se_client
;
1650 spin_lock(&clp
->cl_lock
);
1651 __nfsd4_hash_conn(conn
, ses
);
1652 spin_unlock(&clp
->cl_lock
);
1655 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
1657 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
1658 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
1661 static void nfsd4_init_conn(struct svc_rqst
*rqstp
, struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
1665 nfsd4_hash_conn(conn
, ses
);
1666 ret
= nfsd4_register_conn(conn
);
1668 /* oops; xprt is already down: */
1669 nfsd4_conn_lost(&conn
->cn_xpt_user
);
1670 /* We may have gained or lost a callback channel: */
1671 nfsd4_probe_callback_sync(ses
->se_client
);
1674 static struct nfsd4_conn
*alloc_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_create_session
*cses
)
1676 u32 dir
= NFS4_CDFC4_FORE
;
1678 if (cses
->flags
& SESSION4_BACK_CHAN
)
1679 dir
|= NFS4_CDFC4_BACK
;
1680 return alloc_conn(rqstp
, dir
);
1683 /* must be called under client_lock */
1684 static void nfsd4_del_conns(struct nfsd4_session
*s
)
1686 struct nfs4_client
*clp
= s
->se_client
;
1687 struct nfsd4_conn
*c
;
1689 spin_lock(&clp
->cl_lock
);
1690 while (!list_empty(&s
->se_conns
)) {
1691 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
1692 list_del_init(&c
->cn_persession
);
1693 spin_unlock(&clp
->cl_lock
);
1695 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
1698 spin_lock(&clp
->cl_lock
);
1700 spin_unlock(&clp
->cl_lock
);
1703 static void __free_session(struct nfsd4_session
*ses
)
1705 free_session_slots(ses
);
1709 static void free_session(struct nfsd4_session
*ses
)
1711 nfsd4_del_conns(ses
);
1712 nfsd4_put_drc_mem(&ses
->se_fchannel
);
1713 __free_session(ses
);
1716 static void init_session(struct svc_rqst
*rqstp
, struct nfsd4_session
*new, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
1719 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
1721 new->se_client
= clp
;
1724 INIT_LIST_HEAD(&new->se_conns
);
1726 new->se_cb_seq_nr
= 1;
1727 new->se_flags
= cses
->flags
;
1728 new->se_cb_prog
= cses
->callback_prog
;
1729 new->se_cb_sec
= cses
->cb_sec
;
1730 atomic_set(&new->se_ref
, 0);
1731 idx
= hash_sessionid(&new->se_sessionid
);
1732 list_add(&new->se_hash
, &nn
->sessionid_hashtbl
[idx
]);
1733 spin_lock(&clp
->cl_lock
);
1734 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
1735 spin_unlock(&clp
->cl_lock
);
1738 struct sockaddr
*sa
= svc_addr(rqstp
);
1740 * This is a little silly; with sessions there's no real
1741 * use for the callback address. Use the peer address
1742 * as a reasonable default for now, but consider fixing
1743 * the rpc client not to require an address in the
1746 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
1747 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
1751 /* caller must hold client_lock */
1752 static struct nfsd4_session
*
1753 __find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
)
1755 struct nfsd4_session
*elem
;
1757 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1759 lockdep_assert_held(&nn
->client_lock
);
1761 dump_sessionid(__func__
, sessionid
);
1762 idx
= hash_sessionid(sessionid
);
1763 /* Search in the appropriate list */
1764 list_for_each_entry(elem
, &nn
->sessionid_hashtbl
[idx
], se_hash
) {
1765 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1766 NFS4_MAX_SESSIONID_LEN
)) {
1771 dprintk("%s: session not found\n", __func__
);
1775 static struct nfsd4_session
*
1776 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
, struct net
*net
,
1779 struct nfsd4_session
*session
;
1780 __be32 status
= nfserr_badsession
;
1782 session
= __find_in_sessionid_hashtbl(sessionid
, net
);
1785 status
= nfsd4_get_session_locked(session
);
1793 /* caller must hold client_lock */
1795 unhash_session(struct nfsd4_session
*ses
)
1797 struct nfs4_client
*clp
= ses
->se_client
;
1798 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1800 lockdep_assert_held(&nn
->client_lock
);
1802 list_del(&ses
->se_hash
);
1803 spin_lock(&ses
->se_client
->cl_lock
);
1804 list_del(&ses
->se_perclnt
);
1805 spin_unlock(&ses
->se_client
->cl_lock
);
1808 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1810 STALE_CLIENTID(clientid_t
*clid
, struct nfsd_net
*nn
)
1813 * We're assuming the clid was not given out from a boot
1814 * precisely 2^32 (about 136 years) before this one. That seems
1815 * a safe assumption:
1817 if (clid
->cl_boot
== (u32
)nn
->boot_time
)
1819 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1820 clid
->cl_boot
, clid
->cl_id
, nn
->boot_time
);
1825 * XXX Should we use a slab cache ?
1826 * This type of memory management is somewhat inefficient, but we use it
1827 * anyway since SETCLIENTID is not a common operation.
1829 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1831 struct nfs4_client
*clp
;
1834 clp
= kmem_cache_zalloc(client_slab
, GFP_KERNEL
);
1837 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1838 if (clp
->cl_name
.data
== NULL
)
1840 clp
->cl_ownerstr_hashtbl
= kmalloc_array(OWNER_HASH_SIZE
,
1841 sizeof(struct list_head
),
1843 if (!clp
->cl_ownerstr_hashtbl
)
1844 goto err_no_hashtbl
;
1845 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++)
1846 INIT_LIST_HEAD(&clp
->cl_ownerstr_hashtbl
[i
]);
1847 clp
->cl_name
.len
= name
.len
;
1848 INIT_LIST_HEAD(&clp
->cl_sessions
);
1849 idr_init(&clp
->cl_stateids
);
1850 atomic_set(&clp
->cl_refcount
, 0);
1851 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1852 INIT_LIST_HEAD(&clp
->cl_idhash
);
1853 INIT_LIST_HEAD(&clp
->cl_openowners
);
1854 INIT_LIST_HEAD(&clp
->cl_delegations
);
1855 INIT_LIST_HEAD(&clp
->cl_lru
);
1856 INIT_LIST_HEAD(&clp
->cl_revoked
);
1857 #ifdef CONFIG_NFSD_PNFS
1858 INIT_LIST_HEAD(&clp
->cl_lo_states
);
1860 INIT_LIST_HEAD(&clp
->async_copies
);
1861 spin_lock_init(&clp
->async_lock
);
1862 spin_lock_init(&clp
->cl_lock
);
1863 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1866 kfree(clp
->cl_name
.data
);
1868 kmem_cache_free(client_slab
, clp
);
1873 free_client(struct nfs4_client
*clp
)
1875 while (!list_empty(&clp
->cl_sessions
)) {
1876 struct nfsd4_session
*ses
;
1877 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1879 list_del(&ses
->se_perclnt
);
1880 WARN_ON_ONCE(atomic_read(&ses
->se_ref
));
1883 rpc_destroy_wait_queue(&clp
->cl_cb_waitq
);
1884 free_svc_cred(&clp
->cl_cred
);
1885 kfree(clp
->cl_ownerstr_hashtbl
);
1886 kfree(clp
->cl_name
.data
);
1887 idr_destroy(&clp
->cl_stateids
);
1888 kmem_cache_free(client_slab
, clp
);
1891 /* must be called under the client_lock */
1893 unhash_client_locked(struct nfs4_client
*clp
)
1895 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1896 struct nfsd4_session
*ses
;
1898 lockdep_assert_held(&nn
->client_lock
);
1900 /* Mark the client as expired! */
1902 /* Make it invisible */
1903 if (!list_empty(&clp
->cl_idhash
)) {
1904 list_del_init(&clp
->cl_idhash
);
1905 if (test_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
))
1906 rb_erase(&clp
->cl_namenode
, &nn
->conf_name_tree
);
1908 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
1910 list_del_init(&clp
->cl_lru
);
1911 spin_lock(&clp
->cl_lock
);
1912 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1913 list_del_init(&ses
->se_hash
);
1914 spin_unlock(&clp
->cl_lock
);
1918 unhash_client(struct nfs4_client
*clp
)
1920 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1922 spin_lock(&nn
->client_lock
);
1923 unhash_client_locked(clp
);
1924 spin_unlock(&nn
->client_lock
);
1927 static __be32
mark_client_expired_locked(struct nfs4_client
*clp
)
1929 if (atomic_read(&clp
->cl_refcount
))
1930 return nfserr_jukebox
;
1931 unhash_client_locked(clp
);
1936 __destroy_client(struct nfs4_client
*clp
)
1939 struct nfs4_openowner
*oo
;
1940 struct nfs4_delegation
*dp
;
1941 struct list_head reaplist
;
1943 INIT_LIST_HEAD(&reaplist
);
1944 spin_lock(&state_lock
);
1945 while (!list_empty(&clp
->cl_delegations
)) {
1946 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1947 WARN_ON(!unhash_delegation_locked(dp
));
1948 list_add(&dp
->dl_recall_lru
, &reaplist
);
1950 spin_unlock(&state_lock
);
1951 while (!list_empty(&reaplist
)) {
1952 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1953 list_del_init(&dp
->dl_recall_lru
);
1954 destroy_unhashed_deleg(dp
);
1956 while (!list_empty(&clp
->cl_revoked
)) {
1957 dp
= list_entry(clp
->cl_revoked
.next
, struct nfs4_delegation
, dl_recall_lru
);
1958 list_del_init(&dp
->dl_recall_lru
);
1959 nfs4_put_stid(&dp
->dl_stid
);
1961 while (!list_empty(&clp
->cl_openowners
)) {
1962 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1963 nfs4_get_stateowner(&oo
->oo_owner
);
1964 release_openowner(oo
);
1966 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
1967 struct nfs4_stateowner
*so
, *tmp
;
1969 list_for_each_entry_safe(so
, tmp
, &clp
->cl_ownerstr_hashtbl
[i
],
1971 /* Should be no openowners at this point */
1972 WARN_ON_ONCE(so
->so_is_open_owner
);
1973 remove_blocked_locks(lockowner(so
));
1976 nfsd4_return_all_client_layouts(clp
);
1977 nfsd4_shutdown_copy(clp
);
1978 nfsd4_shutdown_callback(clp
);
1979 if (clp
->cl_cb_conn
.cb_xprt
)
1980 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1985 destroy_client(struct nfs4_client
*clp
)
1988 __destroy_client(clp
);
1991 static void expire_client(struct nfs4_client
*clp
)
1994 nfsd4_client_record_remove(clp
);
1995 __destroy_client(clp
);
1998 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
2000 memcpy(target
->cl_verifier
.data
, source
->data
,
2001 sizeof(target
->cl_verifier
.data
));
2004 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
2006 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
2007 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
2010 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
2012 target
->cr_principal
= kstrdup(source
->cr_principal
, GFP_KERNEL
);
2013 target
->cr_raw_principal
= kstrdup(source
->cr_raw_principal
,
2015 target
->cr_targ_princ
= kstrdup(source
->cr_targ_princ
, GFP_KERNEL
);
2016 if ((source
->cr_principal
&& !target
->cr_principal
) ||
2017 (source
->cr_raw_principal
&& !target
->cr_raw_principal
) ||
2018 (source
->cr_targ_princ
&& !target
->cr_targ_princ
))
2021 target
->cr_flavor
= source
->cr_flavor
;
2022 target
->cr_uid
= source
->cr_uid
;
2023 target
->cr_gid
= source
->cr_gid
;
2024 target
->cr_group_info
= source
->cr_group_info
;
2025 get_group_info(target
->cr_group_info
);
2026 target
->cr_gss_mech
= source
->cr_gss_mech
;
2027 if (source
->cr_gss_mech
)
2028 gss_mech_get(source
->cr_gss_mech
);
2033 compare_blob(const struct xdr_netobj
*o1
, const struct xdr_netobj
*o2
)
2035 if (o1
->len
< o2
->len
)
2037 if (o1
->len
> o2
->len
)
2039 return memcmp(o1
->data
, o2
->data
, o1
->len
);
2042 static int same_name(const char *n1
, const char *n2
)
2044 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
2048 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
2050 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
2054 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
2056 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
2059 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
2063 if (g1
->ngroups
!= g2
->ngroups
)
2065 for (i
=0; i
<g1
->ngroups
; i
++)
2066 if (!gid_eq(g1
->gid
[i
], g2
->gid
[i
]))
2072 * RFC 3530 language requires clid_inuse be returned when the
2073 * "principal" associated with a requests differs from that previously
2074 * used. We use uid, gid's, and gss principal string as our best
2075 * approximation. We also don't want to allow non-gss use of a client
2076 * established using gss: in theory cr_principal should catch that
2077 * change, but in practice cr_principal can be null even in the gss case
2078 * since gssd doesn't always pass down a principal string.
2080 static bool is_gss_cred(struct svc_cred
*cr
)
2082 /* Is cr_flavor one of the gss "pseudoflavors"?: */
2083 return (cr
->cr_flavor
> RPC_AUTH_MAXFLAVOR
);
2088 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
2090 if ((is_gss_cred(cr1
) != is_gss_cred(cr2
))
2091 || (!uid_eq(cr1
->cr_uid
, cr2
->cr_uid
))
2092 || (!gid_eq(cr1
->cr_gid
, cr2
->cr_gid
))
2093 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
2095 /* XXX: check that cr_targ_princ fields match ? */
2096 if (cr1
->cr_principal
== cr2
->cr_principal
)
2098 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
2100 return 0 == strcmp(cr1
->cr_principal
, cr2
->cr_principal
);
2103 static bool svc_rqst_integrity_protected(struct svc_rqst
*rqstp
)
2105 struct svc_cred
*cr
= &rqstp
->rq_cred
;
2108 if (!cr
->cr_gss_mech
)
2110 service
= gss_pseudoflavor_to_service(cr
->cr_gss_mech
, cr
->cr_flavor
);
2111 return service
== RPC_GSS_SVC_INTEGRITY
||
2112 service
== RPC_GSS_SVC_PRIVACY
;
2115 bool nfsd4_mach_creds_match(struct nfs4_client
*cl
, struct svc_rqst
*rqstp
)
2117 struct svc_cred
*cr
= &rqstp
->rq_cred
;
2119 if (!cl
->cl_mach_cred
)
2121 if (cl
->cl_cred
.cr_gss_mech
!= cr
->cr_gss_mech
)
2123 if (!svc_rqst_integrity_protected(rqstp
))
2125 if (cl
->cl_cred
.cr_raw_principal
)
2126 return 0 == strcmp(cl
->cl_cred
.cr_raw_principal
,
2127 cr
->cr_raw_principal
);
2128 if (!cr
->cr_principal
)
2130 return 0 == strcmp(cl
->cl_cred
.cr_principal
, cr
->cr_principal
);
2133 static void gen_confirm(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
2138 * This is opaque to client, so no need to byte-swap. Use
2139 * __force to keep sparse happy
2141 verf
[0] = (__force __be32
)get_seconds();
2142 verf
[1] = (__force __be32
)nn
->clverifier_counter
++;
2143 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
2146 static void gen_clid(struct nfs4_client
*clp
, struct nfsd_net
*nn
)
2148 clp
->cl_clientid
.cl_boot
= nn
->boot_time
;
2149 clp
->cl_clientid
.cl_id
= nn
->clientid_counter
++;
2150 gen_confirm(clp
, nn
);
2153 static struct nfs4_stid
*
2154 find_stateid_locked(struct nfs4_client
*cl
, stateid_t
*t
)
2156 struct nfs4_stid
*ret
;
2158 ret
= idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
2159 if (!ret
|| !ret
->sc_type
)
2164 static struct nfs4_stid
*
2165 find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
2167 struct nfs4_stid
*s
;
2169 spin_lock(&cl
->cl_lock
);
2170 s
= find_stateid_locked(cl
, t
);
2172 if (typemask
& s
->sc_type
)
2173 refcount_inc(&s
->sc_count
);
2177 spin_unlock(&cl
->cl_lock
);
2181 static struct nfs4_client
*create_client(struct xdr_netobj name
,
2182 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
2184 struct nfs4_client
*clp
;
2185 struct sockaddr
*sa
= svc_addr(rqstp
);
2187 struct net
*net
= SVC_NET(rqstp
);
2189 clp
= alloc_client(name
);
2193 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
2198 nfsd4_init_cb(&clp
->cl_cb_null
, clp
, NULL
, NFSPROC4_CLNT_CB_NULL
);
2199 clp
->cl_time
= get_seconds();
2200 clear_bit(0, &clp
->cl_cb_slot_busy
);
2201 copy_verf(clp
, verf
);
2202 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
2203 clp
->cl_cb_session
= NULL
;
2209 add_clp_to_name_tree(struct nfs4_client
*new_clp
, struct rb_root
*root
)
2211 struct rb_node
**new = &(root
->rb_node
), *parent
= NULL
;
2212 struct nfs4_client
*clp
;
2215 clp
= rb_entry(*new, struct nfs4_client
, cl_namenode
);
2218 if (compare_blob(&clp
->cl_name
, &new_clp
->cl_name
) > 0)
2219 new = &((*new)->rb_left
);
2221 new = &((*new)->rb_right
);
2224 rb_link_node(&new_clp
->cl_namenode
, parent
, new);
2225 rb_insert_color(&new_clp
->cl_namenode
, root
);
2228 static struct nfs4_client
*
2229 find_clp_in_name_tree(struct xdr_netobj
*name
, struct rb_root
*root
)
2232 struct rb_node
*node
= root
->rb_node
;
2233 struct nfs4_client
*clp
;
2236 clp
= rb_entry(node
, struct nfs4_client
, cl_namenode
);
2237 cmp
= compare_blob(&clp
->cl_name
, name
);
2239 node
= node
->rb_left
;
2241 node
= node
->rb_right
;
2249 add_to_unconfirmed(struct nfs4_client
*clp
)
2251 unsigned int idhashval
;
2252 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2254 lockdep_assert_held(&nn
->client_lock
);
2256 clear_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2257 add_clp_to_name_tree(clp
, &nn
->unconf_name_tree
);
2258 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2259 list_add(&clp
->cl_idhash
, &nn
->unconf_id_hashtbl
[idhashval
]);
2260 renew_client_locked(clp
);
2264 move_to_confirmed(struct nfs4_client
*clp
)
2266 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
2267 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
2269 lockdep_assert_held(&nn
->client_lock
);
2271 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
2272 list_move(&clp
->cl_idhash
, &nn
->conf_id_hashtbl
[idhashval
]);
2273 rb_erase(&clp
->cl_namenode
, &nn
->unconf_name_tree
);
2274 add_clp_to_name_tree(clp
, &nn
->conf_name_tree
);
2275 set_bit(NFSD4_CLIENT_CONFIRMED
, &clp
->cl_flags
);
2276 renew_client_locked(clp
);
2279 static struct nfs4_client
*
2280 find_client_in_id_table(struct list_head
*tbl
, clientid_t
*clid
, bool sessions
)
2282 struct nfs4_client
*clp
;
2283 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
2285 list_for_each_entry(clp
, &tbl
[idhashval
], cl_idhash
) {
2286 if (same_clid(&clp
->cl_clientid
, clid
)) {
2287 if ((bool)clp
->cl_minorversion
!= sessions
)
2289 renew_client_locked(clp
);
2296 static struct nfs4_client
*
2297 find_confirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2299 struct list_head
*tbl
= nn
->conf_id_hashtbl
;
2301 lockdep_assert_held(&nn
->client_lock
);
2302 return find_client_in_id_table(tbl
, clid
, sessions
);
2305 static struct nfs4_client
*
2306 find_unconfirmed_client(clientid_t
*clid
, bool sessions
, struct nfsd_net
*nn
)
2308 struct list_head
*tbl
= nn
->unconf_id_hashtbl
;
2310 lockdep_assert_held(&nn
->client_lock
);
2311 return find_client_in_id_table(tbl
, clid
, sessions
);
2314 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
2316 return clp
->cl_exchange_flags
!= 0;
2319 static struct nfs4_client
*
2320 find_confirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2322 lockdep_assert_held(&nn
->client_lock
);
2323 return find_clp_in_name_tree(name
, &nn
->conf_name_tree
);
2326 static struct nfs4_client
*
2327 find_unconfirmed_client_by_name(struct xdr_netobj
*name
, struct nfsd_net
*nn
)
2329 lockdep_assert_held(&nn
->client_lock
);
2330 return find_clp_in_name_tree(name
, &nn
->unconf_name_tree
);
2334 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
2336 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
2337 struct sockaddr
*sa
= svc_addr(rqstp
);
2338 u32 scopeid
= rpc_get_scope_id(sa
);
2339 unsigned short expected_family
;
2341 /* Currently, we only support tcp and tcp6 for the callback channel */
2342 if (se
->se_callback_netid_len
== 3 &&
2343 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
2344 expected_family
= AF_INET
;
2345 else if (se
->se_callback_netid_len
== 4 &&
2346 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
2347 expected_family
= AF_INET6
;
2351 conn
->cb_addrlen
= rpc_uaddr2sockaddr(clp
->net
, se
->se_callback_addr_val
,
2352 se
->se_callback_addr_len
,
2353 (struct sockaddr
*)&conn
->cb_addr
,
2354 sizeof(conn
->cb_addr
));
2356 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
2359 if (conn
->cb_addr
.ss_family
== AF_INET6
)
2360 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
2362 conn
->cb_prog
= se
->se_callback_prog
;
2363 conn
->cb_ident
= se
->se_callback_ident
;
2364 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
2367 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
2368 conn
->cb_addrlen
= 0;
2369 dprintk("NFSD: this client (clientid %08x/%08x) "
2370 "will not receive delegations\n",
2371 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
2377 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2380 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
2382 struct xdr_buf
*buf
= resp
->xdr
.buf
;
2383 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2386 dprintk("--> %s slot %p\n", __func__
, slot
);
2388 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
2389 slot
->sl_opcnt
= resp
->opcnt
;
2390 slot
->sl_status
= resp
->cstate
.status
;
2391 free_svc_cred(&slot
->sl_cred
);
2392 copy_cred(&slot
->sl_cred
, &resp
->rqstp
->rq_cred
);
2394 if (!nfsd4_cache_this(resp
)) {
2395 slot
->sl_flags
&= ~NFSD4_SLOT_CACHED
;
2398 slot
->sl_flags
|= NFSD4_SLOT_CACHED
;
2400 base
= resp
->cstate
.data_offset
;
2401 slot
->sl_datalen
= buf
->len
- base
;
2402 if (read_bytes_from_xdr_buf(buf
, base
, slot
->sl_data
, slot
->sl_datalen
))
2403 WARN(1, "%s: sessions DRC could not cache compound\n",
2409 * Encode the replay sequence operation from the slot values.
2410 * If cachethis is FALSE encode the uncached rep error on the next
2411 * operation which sets resp->p and increments resp->opcnt for
2412 * nfs4svc_encode_compoundres.
2416 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
2417 struct nfsd4_compoundres
*resp
)
2419 struct nfsd4_op
*op
;
2420 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2422 /* Encode the replayed sequence operation */
2423 op
= &args
->ops
[resp
->opcnt
- 1];
2424 nfsd4_encode_operation(resp
, op
);
2426 if (slot
->sl_flags
& NFSD4_SLOT_CACHED
)
2428 if (args
->opcnt
== 1) {
2430 * The original operation wasn't a solo sequence--we
2431 * always cache those--so this retry must not match the
2434 op
->status
= nfserr_seq_false_retry
;
2436 op
= &args
->ops
[resp
->opcnt
++];
2437 op
->status
= nfserr_retry_uncached_rep
;
2438 nfsd4_encode_operation(resp
, op
);
2444 * The sequence operation is not cached because we can use the slot and
2448 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
2449 struct nfsd4_sequence
*seq
)
2451 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
2452 struct xdr_stream
*xdr
= &resp
->xdr
;
2456 dprintk("--> %s slot %p\n", __func__
, slot
);
2458 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
2462 p
= xdr_reserve_space(xdr
, slot
->sl_datalen
);
2465 return nfserr_serverfault
;
2467 xdr_encode_opaque_fixed(p
, slot
->sl_data
, slot
->sl_datalen
);
2468 xdr_commit_encode(xdr
);
2470 resp
->opcnt
= slot
->sl_opcnt
;
2471 return slot
->sl_status
;
2475 * Set the exchange_id flags returned by the server.
2478 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
2480 #ifdef CONFIG_NFSD_PNFS
2481 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_PNFS_MDS
;
2483 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
2486 /* Referrals are supported, Migration is not. */
2487 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
2489 /* set the wire flags to return to client. */
2490 clid
->flags
= new->cl_exchange_flags
;
2493 static bool client_has_openowners(struct nfs4_client
*clp
)
2495 struct nfs4_openowner
*oo
;
2497 list_for_each_entry(oo
, &clp
->cl_openowners
, oo_perclient
) {
2498 if (!list_empty(&oo
->oo_owner
.so_stateids
))
2504 static bool client_has_state(struct nfs4_client
*clp
)
2506 return client_has_openowners(clp
)
2507 #ifdef CONFIG_NFSD_PNFS
2508 || !list_empty(&clp
->cl_lo_states
)
2510 || !list_empty(&clp
->cl_delegations
)
2511 || !list_empty(&clp
->cl_sessions
)
2512 || !list_empty(&clp
->async_copies
);
2516 nfsd4_exchange_id(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2517 union nfsd4_op_u
*u
)
2519 struct nfsd4_exchange_id
*exid
= &u
->exchange_id
;
2520 struct nfs4_client
*conf
, *new;
2521 struct nfs4_client
*unconf
= NULL
;
2523 char addr_str
[INET6_ADDRSTRLEN
];
2524 nfs4_verifier verf
= exid
->verifier
;
2525 struct sockaddr
*sa
= svc_addr(rqstp
);
2526 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
2527 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2529 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
2530 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2531 "ip_addr=%s flags %x, spa_how %d\n",
2532 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
2533 addr_str
, exid
->flags
, exid
->spa_how
);
2535 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
2536 return nfserr_inval
;
2538 new = create_client(exid
->clname
, rqstp
, &verf
);
2540 return nfserr_jukebox
;
2542 switch (exid
->spa_how
) {
2544 exid
->spo_must_enforce
[0] = 0;
2545 exid
->spo_must_enforce
[1] = (
2546 1 << (OP_BIND_CONN_TO_SESSION
- 32) |
2547 1 << (OP_EXCHANGE_ID
- 32) |
2548 1 << (OP_CREATE_SESSION
- 32) |
2549 1 << (OP_DESTROY_SESSION
- 32) |
2550 1 << (OP_DESTROY_CLIENTID
- 32));
2552 exid
->spo_must_allow
[0] &= (1 << (OP_CLOSE
) |
2553 1 << (OP_OPEN_DOWNGRADE
) |
2555 1 << (OP_DELEGRETURN
));
2557 exid
->spo_must_allow
[1] &= (
2558 1 << (OP_TEST_STATEID
- 32) |
2559 1 << (OP_FREE_STATEID
- 32));
2560 if (!svc_rqst_integrity_protected(rqstp
)) {
2561 status
= nfserr_inval
;
2565 * Sometimes userspace doesn't give us a principal.
2566 * Which is a bug, really. Anyway, we can't enforce
2567 * MACH_CRED in that case, better to give up now:
2569 if (!new->cl_cred
.cr_principal
&&
2570 !new->cl_cred
.cr_raw_principal
) {
2571 status
= nfserr_serverfault
;
2574 new->cl_mach_cred
= true;
2577 default: /* checked by xdr code */
2580 status
= nfserr_encr_alg_unsupp
;
2584 /* Cases below refer to rfc 5661 section 18.35.4: */
2585 spin_lock(&nn
->client_lock
);
2586 conf
= find_confirmed_client_by_name(&exid
->clname
, nn
);
2588 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
2589 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
2592 if (!clp_used_exchangeid(conf
)) { /* buggy client */
2593 status
= nfserr_inval
;
2596 if (!nfsd4_mach_creds_match(conf
, rqstp
)) {
2597 status
= nfserr_wrong_cred
;
2600 if (!creds_match
) { /* case 9 */
2601 status
= nfserr_perm
;
2604 if (!verfs_match
) { /* case 8 */
2605 status
= nfserr_not_same
;
2609 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2612 if (!creds_match
) { /* case 3 */
2613 if (client_has_state(conf
)) {
2614 status
= nfserr_clid_inuse
;
2619 if (verfs_match
) { /* case 2 */
2620 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
2623 /* case 5, client reboot */
2628 if (update
) { /* case 7 */
2629 status
= nfserr_noent
;
2633 unconf
= find_unconfirmed_client_by_name(&exid
->clname
, nn
);
2634 if (unconf
) /* case 4, possible retry or client restart */
2635 unhash_client_locked(unconf
);
2637 /* case 1 (normal case) */
2640 status
= mark_client_expired_locked(conf
);
2644 new->cl_minorversion
= cstate
->minorversion
;
2645 new->cl_spo_must_allow
.u
.words
[0] = exid
->spo_must_allow
[0];
2646 new->cl_spo_must_allow
.u
.words
[1] = exid
->spo_must_allow
[1];
2649 add_to_unconfirmed(new);
2652 exid
->clientid
.cl_boot
= conf
->cl_clientid
.cl_boot
;
2653 exid
->clientid
.cl_id
= conf
->cl_clientid
.cl_id
;
2655 exid
->seqid
= conf
->cl_cs_slot
.sl_seqid
+ 1;
2656 nfsd4_set_ex_flags(conf
, exid
);
2658 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2659 conf
->cl_cs_slot
.sl_seqid
, conf
->cl_exchange_flags
);
2663 spin_unlock(&nn
->client_lock
);
2668 expire_client(unconf
);
2673 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
2675 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
2678 /* The slot is in use, and no response has been sent. */
2680 if (seqid
== slot_seqid
)
2681 return nfserr_jukebox
;
2683 return nfserr_seq_misordered
;
2685 /* Note unsigned 32-bit arithmetic handles wraparound: */
2686 if (likely(seqid
== slot_seqid
+ 1))
2688 if (seqid
== slot_seqid
)
2689 return nfserr_replay_cache
;
2690 return nfserr_seq_misordered
;
2694 * Cache the create session result into the create session single DRC
2695 * slot cache by saving the xdr structure. sl_seqid has been set.
2696 * Do this for solo or embedded create session operations.
2699 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
2700 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
2702 slot
->sl_status
= nfserr
;
2703 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
2707 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
2708 struct nfsd4_clid_slot
*slot
)
2710 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
2711 return slot
->sl_status
;
2714 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2715 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2716 1 + /* MIN tag is length with zero, only length */ \
2717 3 + /* version, opcount, opcode */ \
2718 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2719 /* seqid, slotID, slotID, cache */ \
2720 4 ) * sizeof(__be32))
2722 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2723 2 + /* verifier: AUTH_NULL, length 0 */\
2725 1 + /* MIN tag is length with zero, only length */ \
2726 3 + /* opcount, opcode, opstatus*/ \
2727 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2728 /* seqid, slotID, slotID, slotID, status */ \
2729 5 ) * sizeof(__be32))
2731 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs
*ca
, struct nfsd_net
*nn
)
2733 u32 maxrpc
= nn
->nfsd_serv
->sv_max_mesg
;
2735 if (ca
->maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
)
2736 return nfserr_toosmall
;
2737 if (ca
->maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
)
2738 return nfserr_toosmall
;
2739 ca
->headerpadsz
= 0;
2740 ca
->maxreq_sz
= min_t(u32
, ca
->maxreq_sz
, maxrpc
);
2741 ca
->maxresp_sz
= min_t(u32
, ca
->maxresp_sz
, maxrpc
);
2742 ca
->maxops
= min_t(u32
, ca
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
2743 ca
->maxresp_cached
= min_t(u32
, ca
->maxresp_cached
,
2744 NFSD_SLOT_CACHE_SIZE
+ NFSD_MIN_HDR_SEQ_SZ
);
2745 ca
->maxreqs
= min_t(u32
, ca
->maxreqs
, NFSD_MAX_SLOTS_PER_SESSION
);
2747 * Note decreasing slot size below client's request may make it
2748 * difficult for client to function correctly, whereas
2749 * decreasing the number of slots will (just?) affect
2750 * performance. When short on memory we therefore prefer to
2751 * decrease number of slots instead of their size. Clients that
2752 * request larger slots than they need will get poor results:
2754 ca
->maxreqs
= nfsd4_get_drc_mem(ca
);
2756 return nfserr_jukebox
;
2762 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
2763 * These are based on similar macros in linux/sunrpc/msg_prot.h .
2765 #define RPC_MAX_HEADER_WITH_AUTH_SYS \
2766 (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
2768 #define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
2769 (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
2771 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2772 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
2773 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2774 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
2777 static __be32
check_backchannel_attrs(struct nfsd4_channel_attrs
*ca
)
2779 ca
->headerpadsz
= 0;
2781 if (ca
->maxreq_sz
< NFSD_CB_MAX_REQ_SZ
)
2782 return nfserr_toosmall
;
2783 if (ca
->maxresp_sz
< NFSD_CB_MAX_RESP_SZ
)
2784 return nfserr_toosmall
;
2785 ca
->maxresp_cached
= 0;
2787 return nfserr_toosmall
;
2792 static __be32
nfsd4_check_cb_sec(struct nfsd4_cb_sec
*cbs
)
2794 switch (cbs
->flavor
) {
2800 * GSS case: the spec doesn't allow us to return this
2801 * error. But it also doesn't allow us not to support
2803 * I'd rather this fail hard than return some error the
2804 * client might think it can already handle:
2806 return nfserr_encr_alg_unsupp
;
2811 nfsd4_create_session(struct svc_rqst
*rqstp
,
2812 struct nfsd4_compound_state
*cstate
, union nfsd4_op_u
*u
)
2814 struct nfsd4_create_session
*cr_ses
= &u
->create_session
;
2815 struct sockaddr
*sa
= svc_addr(rqstp
);
2816 struct nfs4_client
*conf
, *unconf
;
2817 struct nfs4_client
*old
= NULL
;
2818 struct nfsd4_session
*new;
2819 struct nfsd4_conn
*conn
;
2820 struct nfsd4_clid_slot
*cs_slot
= NULL
;
2822 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2824 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
2825 return nfserr_inval
;
2826 status
= nfsd4_check_cb_sec(&cr_ses
->cb_sec
);
2829 status
= check_forechannel_attrs(&cr_ses
->fore_channel
, nn
);
2832 status
= check_backchannel_attrs(&cr_ses
->back_channel
);
2834 goto out_release_drc_mem
;
2835 status
= nfserr_jukebox
;
2836 new = alloc_session(&cr_ses
->fore_channel
, &cr_ses
->back_channel
);
2838 goto out_release_drc_mem
;
2839 conn
= alloc_conn_from_crses(rqstp
, cr_ses
);
2841 goto out_free_session
;
2843 spin_lock(&nn
->client_lock
);
2844 unconf
= find_unconfirmed_client(&cr_ses
->clientid
, true, nn
);
2845 conf
= find_confirmed_client(&cr_ses
->clientid
, true, nn
);
2846 WARN_ON_ONCE(conf
&& unconf
);
2849 status
= nfserr_wrong_cred
;
2850 if (!nfsd4_mach_creds_match(conf
, rqstp
))
2852 cs_slot
= &conf
->cl_cs_slot
;
2853 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2855 if (status
== nfserr_replay_cache
)
2856 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
2859 } else if (unconf
) {
2860 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
2861 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
2862 status
= nfserr_clid_inuse
;
2865 status
= nfserr_wrong_cred
;
2866 if (!nfsd4_mach_creds_match(unconf
, rqstp
))
2868 cs_slot
= &unconf
->cl_cs_slot
;
2869 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
2871 /* an unconfirmed replay returns misordered */
2872 status
= nfserr_seq_misordered
;
2875 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
2877 status
= mark_client_expired_locked(old
);
2883 move_to_confirmed(unconf
);
2886 status
= nfserr_stale_clientid
;
2890 /* Persistent sessions are not supported */
2891 cr_ses
->flags
&= ~SESSION4_PERSIST
;
2892 /* Upshifting from TCP to RDMA is not supported */
2893 cr_ses
->flags
&= ~SESSION4_RDMA
;
2895 init_session(rqstp
, new, conf
, cr_ses
);
2896 nfsd4_get_session_locked(new);
2898 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
2899 NFS4_MAX_SESSIONID_LEN
);
2900 cs_slot
->sl_seqid
++;
2901 cr_ses
->seqid
= cs_slot
->sl_seqid
;
2903 /* cache solo and embedded create sessions under the client_lock */
2904 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
2905 spin_unlock(&nn
->client_lock
);
2906 /* init connection and backchannel */
2907 nfsd4_init_conn(rqstp
, conn
, new);
2908 nfsd4_put_session(new);
2913 spin_unlock(&nn
->client_lock
);
2918 __free_session(new);
2919 out_release_drc_mem
:
2920 nfsd4_put_drc_mem(&cr_ses
->fore_channel
);
2924 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
2927 case NFS4_CDFC4_FORE
:
2928 case NFS4_CDFC4_BACK
:
2930 case NFS4_CDFC4_FORE_OR_BOTH
:
2931 case NFS4_CDFC4_BACK_OR_BOTH
:
2932 *dir
= NFS4_CDFC4_BOTH
;
2935 return nfserr_inval
;
2938 __be32
nfsd4_backchannel_ctl(struct svc_rqst
*rqstp
,
2939 struct nfsd4_compound_state
*cstate
,
2940 union nfsd4_op_u
*u
)
2942 struct nfsd4_backchannel_ctl
*bc
= &u
->backchannel_ctl
;
2943 struct nfsd4_session
*session
= cstate
->session
;
2944 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
2947 status
= nfsd4_check_cb_sec(&bc
->bc_cb_sec
);
2950 spin_lock(&nn
->client_lock
);
2951 session
->se_cb_prog
= bc
->bc_cb_program
;
2952 session
->se_cb_sec
= bc
->bc_cb_sec
;
2953 spin_unlock(&nn
->client_lock
);
2955 nfsd4_probe_callback(session
->se_client
);
2960 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
2961 struct nfsd4_compound_state
*cstate
,
2962 union nfsd4_op_u
*u
)
2964 struct nfsd4_bind_conn_to_session
*bcts
= &u
->bind_conn_to_session
;
2966 struct nfsd4_conn
*conn
;
2967 struct nfsd4_session
*session
;
2968 struct net
*net
= SVC_NET(rqstp
);
2969 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
2971 if (!nfsd4_last_compound_op(rqstp
))
2972 return nfserr_not_only_op
;
2973 spin_lock(&nn
->client_lock
);
2974 session
= find_in_sessionid_hashtbl(&bcts
->sessionid
, net
, &status
);
2975 spin_unlock(&nn
->client_lock
);
2977 goto out_no_session
;
2978 status
= nfserr_wrong_cred
;
2979 if (!nfsd4_mach_creds_match(session
->se_client
, rqstp
))
2981 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
2984 conn
= alloc_conn(rqstp
, bcts
->dir
);
2985 status
= nfserr_jukebox
;
2988 nfsd4_init_conn(rqstp
, conn
, session
);
2991 nfsd4_put_session(session
);
2996 static bool nfsd4_compound_in_session(struct nfsd4_compound_state
*cstate
, struct nfs4_sessionid
*sid
)
2998 if (!cstate
->session
)
3000 return !memcmp(sid
, &cstate
->session
->se_sessionid
, sizeof(*sid
));
3004 nfsd4_destroy_session(struct svc_rqst
*r
, struct nfsd4_compound_state
*cstate
,
3005 union nfsd4_op_u
*u
)
3007 struct nfs4_sessionid
*sessionid
= &u
->destroy_session
.sessionid
;
3008 struct nfsd4_session
*ses
;
3010 int ref_held_by_me
= 0;
3011 struct net
*net
= SVC_NET(r
);
3012 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
3014 status
= nfserr_not_only_op
;
3015 if (nfsd4_compound_in_session(cstate
, sessionid
)) {
3016 if (!nfsd4_last_compound_op(r
))
3020 dump_sessionid(__func__
, sessionid
);
3021 spin_lock(&nn
->client_lock
);
3022 ses
= find_in_sessionid_hashtbl(sessionid
, net
, &status
);
3024 goto out_client_lock
;
3025 status
= nfserr_wrong_cred
;
3026 if (!nfsd4_mach_creds_match(ses
->se_client
, r
))
3027 goto out_put_session
;
3028 status
= mark_session_dead_locked(ses
, 1 + ref_held_by_me
);
3030 goto out_put_session
;
3031 unhash_session(ses
);
3032 spin_unlock(&nn
->client_lock
);
3034 nfsd4_probe_callback_sync(ses
->se_client
);
3036 spin_lock(&nn
->client_lock
);
3039 nfsd4_put_session_locked(ses
);
3041 spin_unlock(&nn
->client_lock
);
3046 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
3048 struct nfsd4_conn
*c
;
3050 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
3051 if (c
->cn_xprt
== xpt
) {
3058 static __be32
nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
3060 struct nfs4_client
*clp
= ses
->se_client
;
3061 struct nfsd4_conn
*c
;
3062 __be32 status
= nfs_ok
;
3065 spin_lock(&clp
->cl_lock
);
3066 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
3069 status
= nfserr_conn_not_bound_to_session
;
3070 if (clp
->cl_mach_cred
)
3072 __nfsd4_hash_conn(new, ses
);
3073 spin_unlock(&clp
->cl_lock
);
3074 ret
= nfsd4_register_conn(new);
3076 /* oops; xprt is already down: */
3077 nfsd4_conn_lost(&new->cn_xpt_user
);
3080 spin_unlock(&clp
->cl_lock
);
3085 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
3087 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
3089 return args
->opcnt
> session
->se_fchannel
.maxops
;
3092 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
3093 struct nfsd4_session
*session
)
3095 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
3097 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
3100 static bool replay_matches_cache(struct svc_rqst
*rqstp
,
3101 struct nfsd4_sequence
*seq
, struct nfsd4_slot
*slot
)
3103 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
3105 if ((bool)(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
) !=
3106 (bool)seq
->cachethis
)
3109 * If there's an error than the reply can have fewer ops than
3110 * the call. But if we cached a reply with *more* ops than the
3111 * call you're sending us now, then this new call is clearly not
3112 * really a replay of the old one:
3114 if (slot
->sl_opcnt
< argp
->opcnt
)
3116 /* This is the only check explicitly called by spec: */
3117 if (!same_creds(&rqstp
->rq_cred
, &slot
->sl_cred
))
3120 * There may be more comparisons we could actually do, but the
3121 * spec doesn't require us to catch every case where the calls
3122 * don't match (that would require caching the call as well as
3123 * the reply), so we don't bother.
3129 nfsd4_sequence(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3130 union nfsd4_op_u
*u
)
3132 struct nfsd4_sequence
*seq
= &u
->sequence
;
3133 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
3134 struct xdr_stream
*xdr
= &resp
->xdr
;
3135 struct nfsd4_session
*session
;
3136 struct nfs4_client
*clp
;
3137 struct nfsd4_slot
*slot
;
3138 struct nfsd4_conn
*conn
;
3141 struct net
*net
= SVC_NET(rqstp
);
3142 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
3144 if (resp
->opcnt
!= 1)
3145 return nfserr_sequence_pos
;
3148 * Will be either used or freed by nfsd4_sequence_check_conn
3151 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
3153 return nfserr_jukebox
;
3155 spin_lock(&nn
->client_lock
);
3156 session
= find_in_sessionid_hashtbl(&seq
->sessionid
, net
, &status
);
3158 goto out_no_session
;
3159 clp
= session
->se_client
;
3161 status
= nfserr_too_many_ops
;
3162 if (nfsd4_session_too_many_ops(rqstp
, session
))
3163 goto out_put_session
;
3165 status
= nfserr_req_too_big
;
3166 if (nfsd4_request_too_big(rqstp
, session
))
3167 goto out_put_session
;
3169 status
= nfserr_badslot
;
3170 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
3171 goto out_put_session
;
3173 slot
= session
->se_slots
[seq
->slotid
];
3174 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
3176 /* We do not negotiate the number of slots yet, so set the
3177 * maxslots to the session maxreqs which is used to encode
3178 * sr_highest_slotid and the sr_target_slot id to maxslots */
3179 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
3181 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
3182 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
3183 if (status
== nfserr_replay_cache
) {
3184 status
= nfserr_seq_misordered
;
3185 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
3186 goto out_put_session
;
3187 status
= nfserr_seq_false_retry
;
3188 if (!replay_matches_cache(rqstp
, seq
, slot
))
3189 goto out_put_session
;
3190 cstate
->slot
= slot
;
3191 cstate
->session
= session
;
3193 /* Return the cached reply status and set cstate->status
3194 * for nfsd4_proc_compound processing */
3195 status
= nfsd4_replay_cache_entry(resp
, seq
);
3196 cstate
->status
= nfserr_replay_cache
;
3200 goto out_put_session
;
3202 status
= nfsd4_sequence_check_conn(conn
, session
);
3205 goto out_put_session
;
3207 buflen
= (seq
->cachethis
) ?
3208 session
->se_fchannel
.maxresp_cached
:
3209 session
->se_fchannel
.maxresp_sz
;
3210 status
= (seq
->cachethis
) ? nfserr_rep_too_big_to_cache
:
3212 if (xdr_restrict_buflen(xdr
, buflen
- rqstp
->rq_auth_slack
))
3213 goto out_put_session
;
3214 svc_reserve(rqstp
, buflen
);
3217 /* Success! bump slot seqid */
3218 slot
->sl_seqid
= seq
->seqid
;
3219 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
3221 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
3223 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
3225 cstate
->slot
= slot
;
3226 cstate
->session
= session
;
3230 switch (clp
->cl_cb_state
) {
3232 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
3234 case NFSD4_CB_FAULT
:
3235 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
3238 seq
->status_flags
= 0;
3240 if (!list_empty(&clp
->cl_revoked
))
3241 seq
->status_flags
|= SEQ4_STATUS_RECALLABLE_STATE_REVOKED
;
3245 spin_unlock(&nn
->client_lock
);
3248 nfsd4_put_session_locked(session
);
3249 goto out_no_session
;
3253 nfsd4_sequence_done(struct nfsd4_compoundres
*resp
)
3255 struct nfsd4_compound_state
*cs
= &resp
->cstate
;
3257 if (nfsd4_has_session(cs
)) {
3258 if (cs
->status
!= nfserr_replay_cache
) {
3259 nfsd4_store_cache_entry(resp
);
3260 cs
->slot
->sl_flags
&= ~NFSD4_SLOT_INUSE
;
3262 /* Drop session reference that was taken in nfsd4_sequence() */
3263 nfsd4_put_session(cs
->session
);
3265 put_client_renew(cs
->clp
);
3269 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
,
3270 struct nfsd4_compound_state
*cstate
,
3271 union nfsd4_op_u
*u
)
3273 struct nfsd4_destroy_clientid
*dc
= &u
->destroy_clientid
;
3274 struct nfs4_client
*conf
, *unconf
;
3275 struct nfs4_client
*clp
= NULL
;
3277 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3279 spin_lock(&nn
->client_lock
);
3280 unconf
= find_unconfirmed_client(&dc
->clientid
, true, nn
);
3281 conf
= find_confirmed_client(&dc
->clientid
, true, nn
);
3282 WARN_ON_ONCE(conf
&& unconf
);
3285 if (client_has_state(conf
)) {
3286 status
= nfserr_clientid_busy
;
3289 status
= mark_client_expired_locked(conf
);
3296 status
= nfserr_stale_clientid
;
3299 if (!nfsd4_mach_creds_match(clp
, rqstp
)) {
3301 status
= nfserr_wrong_cred
;
3304 unhash_client_locked(clp
);
3306 spin_unlock(&nn
->client_lock
);
3313 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
,
3314 struct nfsd4_compound_state
*cstate
, union nfsd4_op_u
*u
)
3316 struct nfsd4_reclaim_complete
*rc
= &u
->reclaim_complete
;
3319 if (rc
->rca_one_fs
) {
3320 if (!cstate
->current_fh
.fh_dentry
)
3321 return nfserr_nofilehandle
;
3323 * We don't take advantage of the rca_one_fs case.
3324 * That's OK, it's optional, we can safely ignore it.
3329 status
= nfserr_complete_already
;
3330 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
3331 &cstate
->session
->se_client
->cl_flags
))
3334 status
= nfserr_stale_clientid
;
3335 if (is_client_expired(cstate
->session
->se_client
))
3337 * The following error isn't really legal.
3338 * But we only get here if the client just explicitly
3339 * destroyed the client. Surely it no longer cares what
3340 * error it gets back on an operation for the dead
3346 nfsd4_client_record_create(cstate
->session
->se_client
);
3352 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3353 union nfsd4_op_u
*u
)
3355 struct nfsd4_setclientid
*setclid
= &u
->setclientid
;
3356 struct xdr_netobj clname
= setclid
->se_name
;
3357 nfs4_verifier clverifier
= setclid
->se_verf
;
3358 struct nfs4_client
*conf
, *new;
3359 struct nfs4_client
*unconf
= NULL
;
3361 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3363 new = create_client(clname
, rqstp
, &clverifier
);
3365 return nfserr_jukebox
;
3366 /* Cases below refer to rfc 3530 section 14.2.33: */
3367 spin_lock(&nn
->client_lock
);
3368 conf
= find_confirmed_client_by_name(&clname
, nn
);
3369 if (conf
&& client_has_state(conf
)) {
3371 status
= nfserr_clid_inuse
;
3372 if (clp_used_exchangeid(conf
))
3374 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
3375 char addr_str
[INET6_ADDRSTRLEN
];
3376 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
3378 dprintk("NFSD: setclientid: string in use by client "
3379 "at %s\n", addr_str
);
3383 unconf
= find_unconfirmed_client_by_name(&clname
, nn
);
3385 unhash_client_locked(unconf
);
3386 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
)) {
3387 /* case 1: probable callback update */
3388 copy_clid(new, conf
);
3389 gen_confirm(new, nn
);
3390 } else /* case 4 (new client) or cases 2, 3 (client reboot): */
3392 new->cl_minorversion
= 0;
3393 gen_callback(new, setclid
, rqstp
);
3394 add_to_unconfirmed(new);
3395 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
3396 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
3397 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
3401 spin_unlock(&nn
->client_lock
);
3405 expire_client(unconf
);
3411 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
3412 struct nfsd4_compound_state
*cstate
,
3413 union nfsd4_op_u
*u
)
3415 struct nfsd4_setclientid_confirm
*setclientid_confirm
=
3416 &u
->setclientid_confirm
;
3417 struct nfs4_client
*conf
, *unconf
;
3418 struct nfs4_client
*old
= NULL
;
3419 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
3420 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
3422 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
3424 if (STALE_CLIENTID(clid
, nn
))
3425 return nfserr_stale_clientid
;
3427 spin_lock(&nn
->client_lock
);
3428 conf
= find_confirmed_client(clid
, false, nn
);
3429 unconf
= find_unconfirmed_client(clid
, false, nn
);
3431 * We try hard to give out unique clientid's, so if we get an
3432 * attempt to confirm the same clientid with a different cred,
3433 * the client may be buggy; this should never happen.
3435 * Nevertheless, RFC 7530 recommends INUSE for this case:
3437 status
= nfserr_clid_inuse
;
3438 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
3440 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
3442 /* cases below refer to rfc 3530 section 14.2.34: */
3443 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
3444 if (conf
&& same_verf(&confirm
, &conf
->cl_confirm
)) {
3445 /* case 2: probable retransmit */
3447 } else /* case 4: client hasn't noticed we rebooted yet? */
3448 status
= nfserr_stale_clientid
;
3452 if (conf
) { /* case 1: callback update */
3454 unhash_client_locked(old
);
3455 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
3456 } else { /* case 3: normal case; new or rebooted client */
3457 old
= find_confirmed_client_by_name(&unconf
->cl_name
, nn
);
3459 status
= nfserr_clid_inuse
;
3460 if (client_has_state(old
)
3461 && !same_creds(&unconf
->cl_cred
,
3464 status
= mark_client_expired_locked(old
);
3470 move_to_confirmed(unconf
);
3473 get_client_locked(conf
);
3474 spin_unlock(&nn
->client_lock
);
3475 nfsd4_probe_callback(conf
);
3476 spin_lock(&nn
->client_lock
);
3477 put_client_renew_locked(conf
);
3479 spin_unlock(&nn
->client_lock
);
3485 static struct nfs4_file
*nfsd4_alloc_file(void)
3487 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
3490 /* OPEN Share state helper functions */
3491 static void nfsd4_init_file(struct knfsd_fh
*fh
, unsigned int hashval
,
3492 struct nfs4_file
*fp
)
3494 lockdep_assert_held(&state_lock
);
3496 refcount_set(&fp
->fi_ref
, 1);
3497 spin_lock_init(&fp
->fi_lock
);
3498 INIT_LIST_HEAD(&fp
->fi_stateids
);
3499 INIT_LIST_HEAD(&fp
->fi_delegations
);
3500 INIT_LIST_HEAD(&fp
->fi_clnt_odstate
);
3501 fh_copy_shallow(&fp
->fi_fhandle
, fh
);
3502 fp
->fi_deleg_file
= NULL
;
3503 fp
->fi_had_conflict
= false;
3504 fp
->fi_share_deny
= 0;
3505 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
3506 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
3507 #ifdef CONFIG_NFSD_PNFS
3508 INIT_LIST_HEAD(&fp
->fi_lo_states
);
3509 atomic_set(&fp
->fi_lo_recalls
, 0);
3511 hlist_add_head_rcu(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
3515 nfsd4_free_slabs(void)
3517 kmem_cache_destroy(client_slab
);
3518 kmem_cache_destroy(openowner_slab
);
3519 kmem_cache_destroy(lockowner_slab
);
3520 kmem_cache_destroy(file_slab
);
3521 kmem_cache_destroy(stateid_slab
);
3522 kmem_cache_destroy(deleg_slab
);
3523 kmem_cache_destroy(odstate_slab
);
3527 nfsd4_init_slabs(void)
3529 client_slab
= kmem_cache_create("nfsd4_clients",
3530 sizeof(struct nfs4_client
), 0, 0, NULL
);
3531 if (client_slab
== NULL
)
3533 openowner_slab
= kmem_cache_create("nfsd4_openowners",
3534 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
3535 if (openowner_slab
== NULL
)
3536 goto out_free_client_slab
;
3537 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
3538 sizeof(struct nfs4_lockowner
), 0, 0, NULL
);
3539 if (lockowner_slab
== NULL
)
3540 goto out_free_openowner_slab
;
3541 file_slab
= kmem_cache_create("nfsd4_files",
3542 sizeof(struct nfs4_file
), 0, 0, NULL
);
3543 if (file_slab
== NULL
)
3544 goto out_free_lockowner_slab
;
3545 stateid_slab
= kmem_cache_create("nfsd4_stateids",
3546 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
3547 if (stateid_slab
== NULL
)
3548 goto out_free_file_slab
;
3549 deleg_slab
= kmem_cache_create("nfsd4_delegations",
3550 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
3551 if (deleg_slab
== NULL
)
3552 goto out_free_stateid_slab
;
3553 odstate_slab
= kmem_cache_create("nfsd4_odstate",
3554 sizeof(struct nfs4_clnt_odstate
), 0, 0, NULL
);
3555 if (odstate_slab
== NULL
)
3556 goto out_free_deleg_slab
;
3559 out_free_deleg_slab
:
3560 kmem_cache_destroy(deleg_slab
);
3561 out_free_stateid_slab
:
3562 kmem_cache_destroy(stateid_slab
);
3564 kmem_cache_destroy(file_slab
);
3565 out_free_lockowner_slab
:
3566 kmem_cache_destroy(lockowner_slab
);
3567 out_free_openowner_slab
:
3568 kmem_cache_destroy(openowner_slab
);
3569 out_free_client_slab
:
3570 kmem_cache_destroy(client_slab
);
3572 dprintk("nfsd4: out of memory while initializing nfsv4\n");
3576 static void init_nfs4_replay(struct nfs4_replay
*rp
)
3578 rp
->rp_status
= nfserr_serverfault
;
3580 rp
->rp_buf
= rp
->rp_ibuf
;
3581 mutex_init(&rp
->rp_mutex
);
3584 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state
*cstate
,
3585 struct nfs4_stateowner
*so
)
3587 if (!nfsd4_has_session(cstate
)) {
3588 mutex_lock(&so
->so_replay
.rp_mutex
);
3589 cstate
->replay_owner
= nfs4_get_stateowner(so
);
3593 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state
*cstate
)
3595 struct nfs4_stateowner
*so
= cstate
->replay_owner
;
3598 cstate
->replay_owner
= NULL
;
3599 mutex_unlock(&so
->so_replay
.rp_mutex
);
3600 nfs4_put_stateowner(so
);
3604 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
3606 struct nfs4_stateowner
*sop
;
3608 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
3612 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
3613 if (!sop
->so_owner
.data
) {
3614 kmem_cache_free(slab
, sop
);
3617 sop
->so_owner
.len
= owner
->len
;
3619 INIT_LIST_HEAD(&sop
->so_stateids
);
3620 sop
->so_client
= clp
;
3621 init_nfs4_replay(&sop
->so_replay
);
3622 atomic_set(&sop
->so_count
, 1);
3626 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
3628 lockdep_assert_held(&clp
->cl_lock
);
3630 list_add(&oo
->oo_owner
.so_strhash
,
3631 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
3632 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
3635 static void nfs4_unhash_openowner(struct nfs4_stateowner
*so
)
3637 unhash_openowner_locked(openowner(so
));
3640 static void nfs4_free_openowner(struct nfs4_stateowner
*so
)
3642 struct nfs4_openowner
*oo
= openowner(so
);
3644 kmem_cache_free(openowner_slab
, oo
);
3647 static const struct nfs4_stateowner_operations openowner_ops
= {
3648 .so_unhash
= nfs4_unhash_openowner
,
3649 .so_free
= nfs4_free_openowner
,
3652 static struct nfs4_ol_stateid
*
3653 nfsd4_find_existing_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3655 struct nfs4_ol_stateid
*local
, *ret
= NULL
;
3656 struct nfs4_openowner
*oo
= open
->op_openowner
;
3658 lockdep_assert_held(&fp
->fi_lock
);
3660 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
3661 /* ignore lock owners */
3662 if (local
->st_stateowner
->so_is_open_owner
== 0)
3664 if (local
->st_stateowner
!= &oo
->oo_owner
)
3666 if (local
->st_stid
.sc_type
== NFS4_OPEN_STID
) {
3668 refcount_inc(&ret
->st_stid
.sc_count
);
3676 nfsd4_verify_open_stid(struct nfs4_stid
*s
)
3678 __be32 ret
= nfs_ok
;
3680 switch (s
->sc_type
) {
3684 case NFS4_CLOSED_STID
:
3685 case NFS4_CLOSED_DELEG_STID
:
3686 ret
= nfserr_bad_stateid
;
3688 case NFS4_REVOKED_DELEG_STID
:
3689 ret
= nfserr_deleg_revoked
;
3694 /* Lock the stateid st_mutex, and deal with races with CLOSE */
3696 nfsd4_lock_ol_stateid(struct nfs4_ol_stateid
*stp
)
3700 mutex_lock_nested(&stp
->st_mutex
, LOCK_STATEID_MUTEX
);
3701 ret
= nfsd4_verify_open_stid(&stp
->st_stid
);
3703 mutex_unlock(&stp
->st_mutex
);
3707 static struct nfs4_ol_stateid
*
3708 nfsd4_find_and_lock_existing_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3710 struct nfs4_ol_stateid
*stp
;
3712 spin_lock(&fp
->fi_lock
);
3713 stp
= nfsd4_find_existing_open(fp
, open
);
3714 spin_unlock(&fp
->fi_lock
);
3715 if (!stp
|| nfsd4_lock_ol_stateid(stp
) == nfs_ok
)
3717 nfs4_put_stid(&stp
->st_stid
);
3722 static struct nfs4_openowner
*
3723 alloc_init_open_stateowner(unsigned int strhashval
, struct nfsd4_open
*open
,
3724 struct nfsd4_compound_state
*cstate
)
3726 struct nfs4_client
*clp
= cstate
->clp
;
3727 struct nfs4_openowner
*oo
, *ret
;
3729 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
3732 oo
->oo_owner
.so_ops
= &openowner_ops
;
3733 oo
->oo_owner
.so_is_open_owner
= 1;
3734 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
3736 if (nfsd4_has_session(cstate
))
3737 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3739 oo
->oo_last_closed_stid
= NULL
;
3740 INIT_LIST_HEAD(&oo
->oo_close_lru
);
3741 spin_lock(&clp
->cl_lock
);
3742 ret
= find_openstateowner_str_locked(strhashval
, open
, clp
);
3744 hash_openowner(oo
, clp
, strhashval
);
3747 nfs4_free_stateowner(&oo
->oo_owner
);
3749 spin_unlock(&clp
->cl_lock
);
3753 static struct nfs4_ol_stateid
*
3754 init_open_stateid(struct nfs4_file
*fp
, struct nfsd4_open
*open
)
3757 struct nfs4_openowner
*oo
= open
->op_openowner
;
3758 struct nfs4_ol_stateid
*retstp
= NULL
;
3759 struct nfs4_ol_stateid
*stp
;
3762 /* We are moving these outside of the spinlocks to avoid the warnings */
3763 mutex_init(&stp
->st_mutex
);
3764 mutex_lock_nested(&stp
->st_mutex
, OPEN_STATEID_MUTEX
);
3767 spin_lock(&oo
->oo_owner
.so_client
->cl_lock
);
3768 spin_lock(&fp
->fi_lock
);
3770 retstp
= nfsd4_find_existing_open(fp
, open
);
3774 open
->op_stp
= NULL
;
3775 refcount_inc(&stp
->st_stid
.sc_count
);
3776 stp
->st_stid
.sc_type
= NFS4_OPEN_STID
;
3777 INIT_LIST_HEAD(&stp
->st_locks
);
3778 stp
->st_stateowner
= nfs4_get_stateowner(&oo
->oo_owner
);
3780 stp
->st_stid
.sc_file
= fp
;
3781 stp
->st_access_bmap
= 0;
3782 stp
->st_deny_bmap
= 0;
3783 stp
->st_openstp
= NULL
;
3784 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
3785 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3788 spin_unlock(&fp
->fi_lock
);
3789 spin_unlock(&oo
->oo_owner
.so_client
->cl_lock
);
3791 /* Handle races with CLOSE */
3792 if (nfsd4_lock_ol_stateid(retstp
) != nfs_ok
) {
3793 nfs4_put_stid(&retstp
->st_stid
);
3796 /* To keep mutex tracking happy */
3797 mutex_unlock(&stp
->st_mutex
);
3804 * In the 4.0 case we need to keep the owners around a little while to handle
3805 * CLOSE replay. We still do need to release any file access that is held by
3806 * them before returning however.
3809 move_to_close_lru(struct nfs4_ol_stateid
*s
, struct net
*net
)
3811 struct nfs4_ol_stateid
*last
;
3812 struct nfs4_openowner
*oo
= openowner(s
->st_stateowner
);
3813 struct nfsd_net
*nn
= net_generic(s
->st_stid
.sc_client
->net
,
3816 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
3819 * We know that we hold one reference via nfsd4_close, and another
3820 * "persistent" reference for the client. If the refcount is higher
3821 * than 2, then there are still calls in progress that are using this
3822 * stateid. We can't put the sc_file reference until they are finished.
3823 * Wait for the refcount to drop to 2. Since it has been unhashed,
3824 * there should be no danger of the refcount going back up again at
3827 wait_event(close_wq
, refcount_read(&s
->st_stid
.sc_count
) == 2);
3829 release_all_access(s
);
3830 if (s
->st_stid
.sc_file
) {
3831 put_nfs4_file(s
->st_stid
.sc_file
);
3832 s
->st_stid
.sc_file
= NULL
;
3835 spin_lock(&nn
->client_lock
);
3836 last
= oo
->oo_last_closed_stid
;
3837 oo
->oo_last_closed_stid
= s
;
3838 list_move_tail(&oo
->oo_close_lru
, &nn
->close_lru
);
3839 oo
->oo_time
= get_seconds();
3840 spin_unlock(&nn
->client_lock
);
3842 nfs4_put_stid(&last
->st_stid
);
3845 /* search file_hashtbl[] for file */
3846 static struct nfs4_file
*
3847 find_file_locked(struct knfsd_fh
*fh
, unsigned int hashval
)
3849 struct nfs4_file
*fp
;
3851 hlist_for_each_entry_rcu(fp
, &file_hashtbl
[hashval
], fi_hash
) {
3852 if (fh_match(&fp
->fi_fhandle
, fh
)) {
3853 if (refcount_inc_not_zero(&fp
->fi_ref
))
3861 find_file(struct knfsd_fh
*fh
)
3863 struct nfs4_file
*fp
;
3864 unsigned int hashval
= file_hashval(fh
);
3867 fp
= find_file_locked(fh
, hashval
);
3872 static struct nfs4_file
*
3873 find_or_add_file(struct nfs4_file
*new, struct knfsd_fh
*fh
)
3875 struct nfs4_file
*fp
;
3876 unsigned int hashval
= file_hashval(fh
);
3879 fp
= find_file_locked(fh
, hashval
);
3884 spin_lock(&state_lock
);
3885 fp
= find_file_locked(fh
, hashval
);
3886 if (likely(fp
== NULL
)) {
3887 nfsd4_init_file(fh
, hashval
, new);
3890 spin_unlock(&state_lock
);
3896 * Called to check deny when READ with all zero stateid or
3897 * WRITE with all zero or all one stateid
3900 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
3902 struct nfs4_file
*fp
;
3903 __be32 ret
= nfs_ok
;
3905 fp
= find_file(¤t_fh
->fh_handle
);
3908 /* Check for conflicting share reservations */
3909 spin_lock(&fp
->fi_lock
);
3910 if (fp
->fi_share_deny
& deny_type
)
3911 ret
= nfserr_locked
;
3912 spin_unlock(&fp
->fi_lock
);
3917 static void nfsd4_cb_recall_prepare(struct nfsd4_callback
*cb
)
3919 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3920 struct nfsd_net
*nn
= net_generic(dp
->dl_stid
.sc_client
->net
,
3923 block_delegations(&dp
->dl_stid
.sc_file
->fi_fhandle
);
3926 * We can't do this in nfsd_break_deleg_cb because it is
3927 * already holding inode->i_lock.
3929 * If the dl_time != 0, then we know that it has already been
3930 * queued for a lease break. Don't queue it again.
3932 spin_lock(&state_lock
);
3933 if (dp
->dl_time
== 0) {
3934 dp
->dl_time
= get_seconds();
3935 list_add_tail(&dp
->dl_recall_lru
, &nn
->del_recall_lru
);
3937 spin_unlock(&state_lock
);
3940 static int nfsd4_cb_recall_done(struct nfsd4_callback
*cb
,
3941 struct rpc_task
*task
)
3943 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3945 if (dp
->dl_stid
.sc_type
== NFS4_CLOSED_DELEG_STID
)
3948 switch (task
->tk_status
) {
3952 case -NFS4ERR_BAD_STATEID
:
3954 * Race: client probably got cb_recall before open reply
3955 * granting delegation.
3957 if (dp
->dl_retries
--) {
3958 rpc_delay(task
, 2 * HZ
);
3967 static void nfsd4_cb_recall_release(struct nfsd4_callback
*cb
)
3969 struct nfs4_delegation
*dp
= cb_to_delegation(cb
);
3971 nfs4_put_stid(&dp
->dl_stid
);
3974 static const struct nfsd4_callback_ops nfsd4_cb_recall_ops
= {
3975 .prepare
= nfsd4_cb_recall_prepare
,
3976 .done
= nfsd4_cb_recall_done
,
3977 .release
= nfsd4_cb_recall_release
,
3980 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
3983 * We're assuming the state code never drops its reference
3984 * without first removing the lease. Since we're in this lease
3985 * callback (and since the lease code is serialized by the
3986 * i_lock) we know the server hasn't removed the lease yet, and
3987 * we know it's safe to take a reference.
3989 refcount_inc(&dp
->dl_stid
.sc_count
);
3990 nfsd4_run_cb(&dp
->dl_recall
);
3993 /* Called from break_lease() with i_lock held. */
3995 nfsd_break_deleg_cb(struct file_lock
*fl
)
3998 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
3999 struct nfs4_file
*fp
= dp
->dl_stid
.sc_file
;
4002 * We don't want the locks code to timeout the lease for us;
4003 * we'll remove it ourself if a delegation isn't returned
4006 fl
->fl_break_time
= 0;
4008 spin_lock(&fp
->fi_lock
);
4009 fp
->fi_had_conflict
= true;
4010 nfsd_break_one_deleg(dp
);
4011 spin_unlock(&fp
->fi_lock
);
4016 nfsd_change_deleg_cb(struct file_lock
*onlist
, int arg
,
4017 struct list_head
*dispose
)
4020 return lease_modify(onlist
, arg
, dispose
);
4025 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
4026 .lm_break
= nfsd_break_deleg_cb
,
4027 .lm_change
= nfsd_change_deleg_cb
,
4030 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
4032 if (nfsd4_has_session(cstate
))
4034 if (seqid
== so
->so_seqid
- 1)
4035 return nfserr_replay_me
;
4036 if (seqid
== so
->so_seqid
)
4038 return nfserr_bad_seqid
;
4041 static __be32
lookup_clientid(clientid_t
*clid
,
4042 struct nfsd4_compound_state
*cstate
,
4043 struct nfsd_net
*nn
)
4045 struct nfs4_client
*found
;
4048 found
= cstate
->clp
;
4049 if (!same_clid(&found
->cl_clientid
, clid
))
4050 return nfserr_stale_clientid
;
4054 if (STALE_CLIENTID(clid
, nn
))
4055 return nfserr_stale_clientid
;
4058 * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
4059 * cached already then we know this is for is for v4.0 and "sessions"
4062 WARN_ON_ONCE(cstate
->session
);
4063 spin_lock(&nn
->client_lock
);
4064 found
= find_confirmed_client(clid
, false, nn
);
4066 spin_unlock(&nn
->client_lock
);
4067 return nfserr_expired
;
4069 atomic_inc(&found
->cl_refcount
);
4070 spin_unlock(&nn
->client_lock
);
4072 /* Cache the nfs4_client in cstate! */
4073 cstate
->clp
= found
;
4078 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
4079 struct nfsd4_open
*open
, struct nfsd_net
*nn
)
4081 clientid_t
*clientid
= &open
->op_clientid
;
4082 struct nfs4_client
*clp
= NULL
;
4083 unsigned int strhashval
;
4084 struct nfs4_openowner
*oo
= NULL
;
4087 if (STALE_CLIENTID(&open
->op_clientid
, nn
))
4088 return nfserr_stale_clientid
;
4090 * In case we need it later, after we've already created the
4091 * file and don't want to risk a further failure:
4093 open
->op_file
= nfsd4_alloc_file();
4094 if (open
->op_file
== NULL
)
4095 return nfserr_jukebox
;
4097 status
= lookup_clientid(clientid
, cstate
, nn
);
4102 strhashval
= ownerstr_hashval(&open
->op_owner
);
4103 oo
= find_openstateowner_str(strhashval
, open
, clp
);
4104 open
->op_openowner
= oo
;
4108 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
4109 /* Replace unconfirmed owners without checking for replay. */
4110 release_openowner(oo
);
4111 open
->op_openowner
= NULL
;
4114 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
4119 oo
= alloc_init_open_stateowner(strhashval
, open
, cstate
);
4121 return nfserr_jukebox
;
4122 open
->op_openowner
= oo
;
4124 open
->op_stp
= nfs4_alloc_open_stateid(clp
);
4126 return nfserr_jukebox
;
4128 if (nfsd4_has_session(cstate
) &&
4129 (cstate
->current_fh
.fh_export
->ex_flags
& NFSEXP_PNFS
)) {
4130 open
->op_odstate
= alloc_clnt_odstate(clp
);
4131 if (!open
->op_odstate
)
4132 return nfserr_jukebox
;
4138 static inline __be32
4139 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
4141 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
4142 return nfserr_openmode
;
4147 static int share_access_to_flags(u32 share_access
)
4149 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
4152 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
4154 struct nfs4_stid
*ret
;
4156 ret
= find_stateid_by_type(cl
, s
,
4157 NFS4_DELEG_STID
|NFS4_REVOKED_DELEG_STID
);
4160 return delegstateid(ret
);
4163 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
4165 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
4166 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
4170 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfsd4_open
*open
,
4171 struct nfs4_delegation
**dp
)
4174 __be32 status
= nfserr_bad_stateid
;
4175 struct nfs4_delegation
*deleg
;
4177 deleg
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
4180 if (deleg
->dl_stid
.sc_type
== NFS4_REVOKED_DELEG_STID
) {
4181 nfs4_put_stid(&deleg
->dl_stid
);
4182 if (cl
->cl_minorversion
)
4183 status
= nfserr_deleg_revoked
;
4186 flags
= share_access_to_flags(open
->op_share_access
);
4187 status
= nfs4_check_delegmode(deleg
, flags
);
4189 nfs4_put_stid(&deleg
->dl_stid
);
4194 if (!nfsd4_is_deleg_cur(open
))
4198 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
4202 static inline int nfs4_access_to_access(u32 nfs4_access
)
4206 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
4207 flags
|= NFSD_MAY_READ
;
4208 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
4209 flags
|= NFSD_MAY_WRITE
;
4213 static inline __be32
4214 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
4215 struct nfsd4_open
*open
)
4217 struct iattr iattr
= {
4218 .ia_valid
= ATTR_SIZE
,
4221 if (!open
->op_truncate
)
4223 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
4224 return nfserr_inval
;
4225 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
4228 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
4229 struct svc_fh
*cur_fh
, struct nfs4_ol_stateid
*stp
,
4230 struct nfsd4_open
*open
)
4232 struct file
*filp
= NULL
;
4234 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
4235 int access
= nfs4_access_to_access(open
->op_share_access
);
4236 unsigned char old_access_bmap
, old_deny_bmap
;
4238 spin_lock(&fp
->fi_lock
);
4241 * Are we trying to set a deny mode that would conflict with
4244 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
4245 if (status
!= nfs_ok
) {
4246 spin_unlock(&fp
->fi_lock
);
4250 /* set access to the file */
4251 status
= nfs4_file_get_access(fp
, open
->op_share_access
);
4252 if (status
!= nfs_ok
) {
4253 spin_unlock(&fp
->fi_lock
);
4257 /* Set access bits in stateid */
4258 old_access_bmap
= stp
->st_access_bmap
;
4259 set_access(open
->op_share_access
, stp
);
4261 /* Set new deny mask */
4262 old_deny_bmap
= stp
->st_deny_bmap
;
4263 set_deny(open
->op_share_deny
, stp
);
4264 fp
->fi_share_deny
|= (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
4266 if (!fp
->fi_fds
[oflag
]) {
4267 spin_unlock(&fp
->fi_lock
);
4268 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
, &filp
);
4270 goto out_put_access
;
4271 spin_lock(&fp
->fi_lock
);
4272 if (!fp
->fi_fds
[oflag
]) {
4273 fp
->fi_fds
[oflag
] = filp
;
4277 spin_unlock(&fp
->fi_lock
);
4281 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
4283 goto out_put_access
;
4287 stp
->st_access_bmap
= old_access_bmap
;
4288 nfs4_file_put_access(fp
, open
->op_share_access
);
4289 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap
), stp
);
4294 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
)
4297 unsigned char old_deny_bmap
= stp
->st_deny_bmap
;
4299 if (!test_access(open
->op_share_access
, stp
))
4300 return nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, stp
, open
);
4302 /* test and set deny mode */
4303 spin_lock(&fp
->fi_lock
);
4304 status
= nfs4_file_check_deny(fp
, open
->op_share_deny
);
4305 if (status
== nfs_ok
) {
4306 set_deny(open
->op_share_deny
, stp
);
4307 fp
->fi_share_deny
|=
4308 (open
->op_share_deny
& NFS4_SHARE_DENY_BOTH
);
4310 spin_unlock(&fp
->fi_lock
);
4312 if (status
!= nfs_ok
)
4315 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
4316 if (status
!= nfs_ok
)
4317 reset_union_bmap_deny(old_deny_bmap
, stp
);
4321 /* Should we give out recallable state?: */
4322 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
4324 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
4327 * In the sessions case, since we don't have to establish a
4328 * separate connection for callbacks, we assume it's OK
4329 * until we hear otherwise:
4331 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
4334 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
,
4337 struct file_lock
*fl
;
4339 fl
= locks_alloc_lock();
4342 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
4343 fl
->fl_flags
= FL_DELEG
;
4344 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
4345 fl
->fl_end
= OFFSET_MAX
;
4346 fl
->fl_owner
= (fl_owner_t
)dp
;
4347 fl
->fl_pid
= current
->tgid
;
4348 fl
->fl_file
= dp
->dl_stid
.sc_file
->fi_deleg_file
;
4352 static struct nfs4_delegation
*
4353 nfs4_set_delegation(struct nfs4_client
*clp
, struct svc_fh
*fh
,
4354 struct nfs4_file
*fp
, struct nfs4_clnt_odstate
*odstate
)
4357 struct nfs4_delegation
*dp
;
4359 struct file_lock
*fl
;
4362 * The fi_had_conflict and nfs_get_existing_delegation checks
4363 * here are just optimizations; we'll need to recheck them at
4366 if (fp
->fi_had_conflict
)
4367 return ERR_PTR(-EAGAIN
);
4369 filp
= find_readable_file(fp
);
4371 /* We should always have a readable file here */
4373 return ERR_PTR(-EBADF
);
4375 spin_lock(&state_lock
);
4376 spin_lock(&fp
->fi_lock
);
4377 if (nfs4_delegation_exists(clp
, fp
))
4379 else if (!fp
->fi_deleg_file
) {
4380 fp
->fi_deleg_file
= filp
;
4381 /* increment early to prevent fi_deleg_file from being
4383 fp
->fi_delegees
= 1;
4387 spin_unlock(&fp
->fi_lock
);
4388 spin_unlock(&state_lock
);
4392 return ERR_PTR(status
);
4395 dp
= alloc_init_deleg(clp
, fp
, fh
, odstate
);
4399 fl
= nfs4_alloc_init_lease(dp
, NFS4_OPEN_DELEGATE_READ
);
4401 goto out_clnt_odstate
;
4403 status
= vfs_setlease(fp
->fi_deleg_file
, fl
->fl_type
, &fl
, NULL
);
4405 locks_free_lock(fl
);
4407 goto out_clnt_odstate
;
4409 spin_lock(&state_lock
);
4410 spin_lock(&fp
->fi_lock
);
4411 if (fp
->fi_had_conflict
)
4414 status
= hash_delegation_locked(dp
, fp
);
4415 spin_unlock(&fp
->fi_lock
);
4416 spin_unlock(&state_lock
);
4423 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, NULL
, (void **)&dp
);
4425 put_clnt_odstate(dp
->dl_clnt_odstate
);
4426 nfs4_put_stid(&dp
->dl_stid
);
4429 return ERR_PTR(status
);
4432 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
4434 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4435 if (status
== -EAGAIN
)
4436 open
->op_why_no_deleg
= WND4_CONTENTION
;
4438 open
->op_why_no_deleg
= WND4_RESOURCE
;
4439 switch (open
->op_deleg_want
) {
4440 case NFS4_SHARE_WANT_READ_DELEG
:
4441 case NFS4_SHARE_WANT_WRITE_DELEG
:
4442 case NFS4_SHARE_WANT_ANY_DELEG
:
4444 case NFS4_SHARE_WANT_CANCEL
:
4445 open
->op_why_no_deleg
= WND4_CANCELLED
;
4447 case NFS4_SHARE_WANT_NO_DELEG
:
4454 * Attempt to hand out a delegation.
4456 * Note we don't support write delegations, and won't until the vfs has
4457 * proper support for them.
4460 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
,
4461 struct nfs4_ol_stateid
*stp
)
4463 struct nfs4_delegation
*dp
;
4464 struct nfs4_openowner
*oo
= openowner(stp
->st_stateowner
);
4465 struct nfs4_client
*clp
= stp
->st_stid
.sc_client
;
4469 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
4470 open
->op_recall
= 0;
4471 switch (open
->op_claim_type
) {
4472 case NFS4_OPEN_CLAIM_PREVIOUS
:
4474 open
->op_recall
= 1;
4475 if (open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_READ
)
4478 case NFS4_OPEN_CLAIM_NULL
:
4479 case NFS4_OPEN_CLAIM_FH
:
4481 * Let's not give out any delegations till everyone's
4482 * had the chance to reclaim theirs, *and* until
4483 * NLM locks have all been reclaimed:
4485 if (locks_in_grace(clp
->net
))
4487 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
4490 * Also, if the file was opened for write or
4491 * create, there's a good chance the client's
4492 * about to write to it, resulting in an
4493 * immediate recall (since we don't support
4494 * write delegations):
4496 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
4498 if (open
->op_create
== NFS4_OPEN_CREATE
)
4504 dp
= nfs4_set_delegation(clp
, fh
, stp
->st_stid
.sc_file
, stp
->st_clnt_odstate
);
4508 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
4510 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
4511 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
4512 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_READ
;
4513 nfs4_put_stid(&dp
->dl_stid
);
4516 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE
;
4517 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
4518 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
) {
4519 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4520 open
->op_recall
= 1;
4523 /* 4.1 client asking for a delegation? */
4524 if (open
->op_deleg_want
)
4525 nfsd4_open_deleg_none_ext(open
, status
);
4529 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
4530 struct nfs4_delegation
*dp
)
4532 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
4533 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4534 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4535 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
4536 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
4537 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
4538 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4539 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
4541 /* Otherwise the client must be confused wanting a delegation
4542 * it already has, therefore we don't return
4543 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4548 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
4550 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
4551 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
4552 struct nfs4_file
*fp
= NULL
;
4553 struct nfs4_ol_stateid
*stp
= NULL
;
4554 struct nfs4_delegation
*dp
= NULL
;
4556 bool new_stp
= false;
4559 * Lookup file; if found, lookup stateid and check open request,
4560 * and check for delegations in the process of being recalled.
4561 * If not found, create the nfs4_file struct
4563 fp
= find_or_add_file(open
->op_file
, ¤t_fh
->fh_handle
);
4564 if (fp
!= open
->op_file
) {
4565 status
= nfs4_check_deleg(cl
, open
, &dp
);
4568 stp
= nfsd4_find_and_lock_existing_open(fp
, open
);
4570 open
->op_file
= NULL
;
4571 status
= nfserr_bad_stateid
;
4572 if (nfsd4_is_deleg_cur(open
))
4577 stp
= init_open_stateid(fp
, open
);
4583 * OPEN the file, or upgrade an existing OPEN.
4584 * If truncate fails, the OPEN fails.
4586 * stp is already locked.
4589 /* Stateid was found, this is an OPEN upgrade */
4590 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
4592 mutex_unlock(&stp
->st_mutex
);
4596 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, stp
, open
);
4598 stp
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
4599 release_open_stateid(stp
);
4600 mutex_unlock(&stp
->st_mutex
);
4604 stp
->st_clnt_odstate
= find_or_hash_clnt_odstate(fp
,
4606 if (stp
->st_clnt_odstate
== open
->op_odstate
)
4607 open
->op_odstate
= NULL
;
4610 nfs4_inc_and_copy_stateid(&open
->op_stateid
, &stp
->st_stid
);
4611 mutex_unlock(&stp
->st_mutex
);
4613 if (nfsd4_has_session(&resp
->cstate
)) {
4614 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
4615 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
4616 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
4622 * Attempt to hand out a delegation. No error return, because the
4623 * OPEN succeeds even if we fail.
4625 nfs4_open_delegation(current_fh
, open
, stp
);
4629 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
4630 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
4632 /* 4.1 client trying to upgrade/downgrade delegation? */
4633 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
4634 open
->op_deleg_want
)
4635 nfsd4_deleg_xgrade_none_ext(open
, dp
);
4639 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
4640 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
4642 * To finish the open response, we just need to set the rflags.
4644 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
4645 if (nfsd4_has_session(&resp
->cstate
))
4646 open
->op_rflags
|= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK
;
4647 else if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
))
4648 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
4651 nfs4_put_stid(&dp
->dl_stid
);
4653 nfs4_put_stid(&stp
->st_stid
);
4658 void nfsd4_cleanup_open_state(struct nfsd4_compound_state
*cstate
,
4659 struct nfsd4_open
*open
)
4661 if (open
->op_openowner
) {
4662 struct nfs4_stateowner
*so
= &open
->op_openowner
->oo_owner
;
4664 nfsd4_cstate_assign_replay(cstate
, so
);
4665 nfs4_put_stateowner(so
);
4668 kmem_cache_free(file_slab
, open
->op_file
);
4670 nfs4_put_stid(&open
->op_stp
->st_stid
);
4671 if (open
->op_odstate
)
4672 kmem_cache_free(odstate_slab
, open
->op_odstate
);
4676 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4677 union nfsd4_op_u
*u
)
4679 clientid_t
*clid
= &u
->renew
;
4680 struct nfs4_client
*clp
;
4682 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
4684 dprintk("process_renew(%08x/%08x): starting\n",
4685 clid
->cl_boot
, clid
->cl_id
);
4686 status
= lookup_clientid(clid
, cstate
, nn
);
4690 status
= nfserr_cb_path_down
;
4691 if (!list_empty(&clp
->cl_delegations
)
4692 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
4700 nfsd4_end_grace(struct nfsd_net
*nn
)
4702 /* do nothing if grace period already ended */
4703 if (nn
->grace_ended
)
4706 dprintk("NFSD: end of grace period\n");
4707 nn
->grace_ended
= true;
4709 * If the server goes down again right now, an NFSv4
4710 * client will still be allowed to reclaim after it comes back up,
4711 * even if it hasn't yet had a chance to reclaim state this time.
4714 nfsd4_record_grace_done(nn
);
4716 * At this point, NFSv4 clients can still reclaim. But if the
4717 * server crashes, any that have not yet reclaimed will be out
4718 * of luck on the next boot.
4720 * (NFSv4.1+ clients are considered to have reclaimed once they
4721 * call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
4722 * have reclaimed after their first OPEN.)
4724 locks_end_grace(&nn
->nfsd4_manager
);
4726 * At this point, and once lockd and/or any other containers
4727 * exit their grace period, further reclaims will fail and
4728 * regular locking can resume.
4733 * If we've waited a lease period but there are still clients trying to
4734 * reclaim, wait a little longer to give them a chance to finish.
4736 static bool clients_still_reclaiming(struct nfsd_net
*nn
)
4738 unsigned long now
= get_seconds();
4739 unsigned long double_grace_period_end
= nn
->boot_time
+
4740 2 * nn
->nfsd4_lease
;
4742 if (!nn
->somebody_reclaimed
)
4744 nn
->somebody_reclaimed
= false;
4746 * If we've given them *two* lease times to reclaim, and they're
4747 * still not done, give up:
4749 if (time_after(now
, double_grace_period_end
))
4755 nfs4_laundromat(struct nfsd_net
*nn
)
4757 struct nfs4_client
*clp
;
4758 struct nfs4_openowner
*oo
;
4759 struct nfs4_delegation
*dp
;
4760 struct nfs4_ol_stateid
*stp
;
4761 struct nfsd4_blocked_lock
*nbl
;
4762 struct list_head
*pos
, *next
, reaplist
;
4763 time_t cutoff
= get_seconds() - nn
->nfsd4_lease
;
4764 time_t t
, new_timeo
= nn
->nfsd4_lease
;
4766 dprintk("NFSD: laundromat service - starting\n");
4768 if (clients_still_reclaiming(nn
)) {
4772 nfsd4_end_grace(nn
);
4773 INIT_LIST_HEAD(&reaplist
);
4774 spin_lock(&nn
->client_lock
);
4775 list_for_each_safe(pos
, next
, &nn
->client_lru
) {
4776 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4777 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
4778 t
= clp
->cl_time
- cutoff
;
4779 new_timeo
= min(new_timeo
, t
);
4782 if (mark_client_expired_locked(clp
)) {
4783 dprintk("NFSD: client in use (clientid %08x)\n",
4784 clp
->cl_clientid
.cl_id
);
4787 list_add(&clp
->cl_lru
, &reaplist
);
4789 spin_unlock(&nn
->client_lock
);
4790 list_for_each_safe(pos
, next
, &reaplist
) {
4791 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
4792 dprintk("NFSD: purging unused client (clientid %08x)\n",
4793 clp
->cl_clientid
.cl_id
);
4794 list_del_init(&clp
->cl_lru
);
4797 spin_lock(&state_lock
);
4798 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
4799 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4800 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
4801 t
= dp
->dl_time
- cutoff
;
4802 new_timeo
= min(new_timeo
, t
);
4805 WARN_ON(!unhash_delegation_locked(dp
));
4806 list_add(&dp
->dl_recall_lru
, &reaplist
);
4808 spin_unlock(&state_lock
);
4809 while (!list_empty(&reaplist
)) {
4810 dp
= list_first_entry(&reaplist
, struct nfs4_delegation
,
4812 list_del_init(&dp
->dl_recall_lru
);
4813 revoke_delegation(dp
);
4816 spin_lock(&nn
->client_lock
);
4817 while (!list_empty(&nn
->close_lru
)) {
4818 oo
= list_first_entry(&nn
->close_lru
, struct nfs4_openowner
,
4820 if (time_after((unsigned long)oo
->oo_time
,
4821 (unsigned long)cutoff
)) {
4822 t
= oo
->oo_time
- cutoff
;
4823 new_timeo
= min(new_timeo
, t
);
4826 list_del_init(&oo
->oo_close_lru
);
4827 stp
= oo
->oo_last_closed_stid
;
4828 oo
->oo_last_closed_stid
= NULL
;
4829 spin_unlock(&nn
->client_lock
);
4830 nfs4_put_stid(&stp
->st_stid
);
4831 spin_lock(&nn
->client_lock
);
4833 spin_unlock(&nn
->client_lock
);
4836 * It's possible for a client to try and acquire an already held lock
4837 * that is being held for a long time, and then lose interest in it.
4838 * So, we clean out any un-revisited request after a lease period
4839 * under the assumption that the client is no longer interested.
4841 * RFC5661, sec. 9.6 states that the client must not rely on getting
4842 * notifications and must continue to poll for locks, even when the
4843 * server supports them. Thus this shouldn't lead to clients blocking
4844 * indefinitely once the lock does become free.
4846 BUG_ON(!list_empty(&reaplist
));
4847 spin_lock(&nn
->blocked_locks_lock
);
4848 while (!list_empty(&nn
->blocked_locks_lru
)) {
4849 nbl
= list_first_entry(&nn
->blocked_locks_lru
,
4850 struct nfsd4_blocked_lock
, nbl_lru
);
4851 if (time_after((unsigned long)nbl
->nbl_time
,
4852 (unsigned long)cutoff
)) {
4853 t
= nbl
->nbl_time
- cutoff
;
4854 new_timeo
= min(new_timeo
, t
);
4857 list_move(&nbl
->nbl_lru
, &reaplist
);
4858 list_del_init(&nbl
->nbl_list
);
4860 spin_unlock(&nn
->blocked_locks_lock
);
4862 while (!list_empty(&reaplist
)) {
4863 nbl
= list_first_entry(&reaplist
,
4864 struct nfsd4_blocked_lock
, nbl_lru
);
4865 list_del_init(&nbl
->nbl_lru
);
4866 locks_delete_block(&nbl
->nbl_lock
);
4867 free_blocked_lock(nbl
);
4870 new_timeo
= max_t(time_t, new_timeo
, NFSD_LAUNDROMAT_MINTIMEOUT
);
4874 static struct workqueue_struct
*laundry_wq
;
4875 static void laundromat_main(struct work_struct
*);
4878 laundromat_main(struct work_struct
*laundry
)
4881 struct delayed_work
*dwork
= to_delayed_work(laundry
);
4882 struct nfsd_net
*nn
= container_of(dwork
, struct nfsd_net
,
4885 t
= nfs4_laundromat(nn
);
4886 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
4887 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, t
*HZ
);
4890 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stid
*stp
)
4892 if (!fh_match(&fhp
->fh_handle
, &stp
->sc_file
->fi_fhandle
))
4893 return nfserr_bad_stateid
;
4898 access_permit_read(struct nfs4_ol_stateid
*stp
)
4900 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
4901 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
4902 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
4906 access_permit_write(struct nfs4_ol_stateid
*stp
)
4908 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
4909 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
4913 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
4915 __be32 status
= nfserr_openmode
;
4917 /* For lock stateid's, we test the parent open, not the lock: */
4918 if (stp
->st_openstp
)
4919 stp
= stp
->st_openstp
;
4920 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
4922 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
4929 static inline __be32
4930 check_special_stateids(struct net
*net
, svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
4932 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
4934 else if (opens_in_grace(net
)) {
4935 /* Answer in remaining cases depends on existence of
4936 * conflicting state; so we must wait out the grace period. */
4937 return nfserr_grace
;
4938 } else if (flags
& WR_STATE
)
4939 return nfs4_share_conflict(current_fh
,
4940 NFS4_SHARE_DENY_WRITE
);
4941 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4942 return nfs4_share_conflict(current_fh
,
4943 NFS4_SHARE_DENY_READ
);
4947 * Allow READ/WRITE during grace period on recovered state only for files
4948 * that are not able to provide mandatory locking.
4951 grace_disallows_io(struct net
*net
, struct inode
*inode
)
4953 return opens_in_grace(net
) && mandatory_lock(inode
);
4956 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
4959 * When sessions are used the stateid generation number is ignored
4962 if (has_session
&& in
->si_generation
== 0)
4965 if (in
->si_generation
== ref
->si_generation
)
4968 /* If the client sends us a stateid from the future, it's buggy: */
4969 if (nfsd4_stateid_generation_after(in
, ref
))
4970 return nfserr_bad_stateid
;
4972 * However, we could see a stateid from the past, even from a
4973 * non-buggy client. For example, if the client sends a lock
4974 * while some IO is outstanding, the lock may bump si_generation
4975 * while the IO is still in flight. The client could avoid that
4976 * situation by waiting for responses on all the IO requests,
4977 * but better performance may result in retrying IO that
4978 * receives an old_stateid error if requests are rarely
4979 * reordered in flight:
4981 return nfserr_old_stateid
;
4984 static __be32
nfsd4_stid_check_stateid_generation(stateid_t
*in
, struct nfs4_stid
*s
, bool has_session
)
4988 spin_lock(&s
->sc_lock
);
4989 ret
= nfsd4_verify_open_stid(s
);
4991 ret
= check_stateid_generation(in
, &s
->sc_stateid
, has_session
);
4992 spin_unlock(&s
->sc_lock
);
4996 static __be32
nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid
*ols
)
4998 if (ols
->st_stateowner
->so_is_open_owner
&&
4999 !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
5000 return nfserr_bad_stateid
;
5004 static __be32
nfsd4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
5006 struct nfs4_stid
*s
;
5007 __be32 status
= nfserr_bad_stateid
;
5009 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
) ||
5010 CLOSE_STATEID(stateid
))
5012 /* Client debugging aid. */
5013 if (!same_clid(&stateid
->si_opaque
.so_clid
, &cl
->cl_clientid
)) {
5014 char addr_str
[INET6_ADDRSTRLEN
];
5015 rpc_ntop((struct sockaddr
*)&cl
->cl_addr
, addr_str
,
5017 pr_warn_ratelimited("NFSD: client %s testing state ID "
5018 "with incorrect client ID\n", addr_str
);
5021 spin_lock(&cl
->cl_lock
);
5022 s
= find_stateid_locked(cl
, stateid
);
5025 status
= nfsd4_stid_check_stateid_generation(stateid
, s
, 1);
5028 switch (s
->sc_type
) {
5029 case NFS4_DELEG_STID
:
5032 case NFS4_REVOKED_DELEG_STID
:
5033 status
= nfserr_deleg_revoked
;
5035 case NFS4_OPEN_STID
:
5036 case NFS4_LOCK_STID
:
5037 status
= nfsd4_check_openowner_confirmed(openlockstateid(s
));
5040 printk("unknown stateid type %x\n", s
->sc_type
);
5042 case NFS4_CLOSED_STID
:
5043 case NFS4_CLOSED_DELEG_STID
:
5044 status
= nfserr_bad_stateid
;
5047 spin_unlock(&cl
->cl_lock
);
5052 nfsd4_lookup_stateid(struct nfsd4_compound_state
*cstate
,
5053 stateid_t
*stateid
, unsigned char typemask
,
5054 struct nfs4_stid
**s
, struct nfsd_net
*nn
)
5057 bool return_revoked
= false;
5060 * only return revoked delegations if explicitly asked.
5061 * otherwise we report revoked or bad_stateid status.
5063 if (typemask
& NFS4_REVOKED_DELEG_STID
)
5064 return_revoked
= true;
5065 else if (typemask
& NFS4_DELEG_STID
)
5066 typemask
|= NFS4_REVOKED_DELEG_STID
;
5068 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
) ||
5069 CLOSE_STATEID(stateid
))
5070 return nfserr_bad_stateid
;
5071 status
= lookup_clientid(&stateid
->si_opaque
.so_clid
, cstate
, nn
);
5072 if (status
== nfserr_stale_clientid
) {
5073 if (cstate
->session
)
5074 return nfserr_bad_stateid
;
5075 return nfserr_stale_stateid
;
5079 *s
= find_stateid_by_type(cstate
->clp
, stateid
, typemask
);
5081 return nfserr_bad_stateid
;
5082 if (((*s
)->sc_type
== NFS4_REVOKED_DELEG_STID
) && !return_revoked
) {
5084 if (cstate
->minorversion
)
5085 return nfserr_deleg_revoked
;
5086 return nfserr_bad_stateid
;
5091 static struct file
*
5092 nfs4_find_file(struct nfs4_stid
*s
, int flags
)
5097 switch (s
->sc_type
) {
5098 case NFS4_DELEG_STID
:
5099 if (WARN_ON_ONCE(!s
->sc_file
->fi_deleg_file
))
5101 return get_file(s
->sc_file
->fi_deleg_file
);
5102 case NFS4_OPEN_STID
:
5103 case NFS4_LOCK_STID
:
5104 if (flags
& RD_STATE
)
5105 return find_readable_file(s
->sc_file
);
5107 return find_writeable_file(s
->sc_file
);
5115 nfs4_check_olstateid(struct nfs4_ol_stateid
*ols
, int flags
)
5119 status
= nfsd4_check_openowner_confirmed(ols
);
5122 return nfs4_check_openmode(ols
, flags
);
5126 nfs4_check_file(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct nfs4_stid
*s
,
5127 struct file
**filpp
, bool *tmp_file
, int flags
)
5129 int acc
= (flags
& RD_STATE
) ? NFSD_MAY_READ
: NFSD_MAY_WRITE
;
5133 file
= nfs4_find_file(s
, flags
);
5135 status
= nfsd_permission(rqstp
, fhp
->fh_export
, fhp
->fh_dentry
,
5136 acc
| NFSD_MAY_OWNER_OVERRIDE
);
5144 status
= nfsd_open(rqstp
, fhp
, S_IFREG
, acc
, filpp
);
5156 * Checks for stateid operations
5159 nfs4_preprocess_stateid_op(struct svc_rqst
*rqstp
,
5160 struct nfsd4_compound_state
*cstate
, struct svc_fh
*fhp
,
5161 stateid_t
*stateid
, int flags
, struct file
**filpp
, bool *tmp_file
)
5163 struct inode
*ino
= d_inode(fhp
->fh_dentry
);
5164 struct net
*net
= SVC_NET(rqstp
);
5165 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5166 struct nfs4_stid
*s
= NULL
;
5174 if (grace_disallows_io(net
, ino
))
5175 return nfserr_grace
;
5177 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
5178 status
= check_special_stateids(net
, fhp
, stateid
, flags
);
5182 status
= nfsd4_lookup_stateid(cstate
, stateid
,
5183 NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
,
5187 status
= nfsd4_stid_check_stateid_generation(stateid
, s
,
5188 nfsd4_has_session(cstate
));
5192 switch (s
->sc_type
) {
5193 case NFS4_DELEG_STID
:
5194 status
= nfs4_check_delegmode(delegstateid(s
), flags
);
5196 case NFS4_OPEN_STID
:
5197 case NFS4_LOCK_STID
:
5198 status
= nfs4_check_olstateid(openlockstateid(s
), flags
);
5201 status
= nfserr_bad_stateid
;
5206 status
= nfs4_check_fh(fhp
, s
);
5209 if (!status
&& filpp
)
5210 status
= nfs4_check_file(rqstp
, fhp
, s
, filpp
, tmp_file
, flags
);
5218 * Test if the stateid is valid
5221 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5222 union nfsd4_op_u
*u
)
5224 struct nfsd4_test_stateid
*test_stateid
= &u
->test_stateid
;
5225 struct nfsd4_test_stateid_id
*stateid
;
5226 struct nfs4_client
*cl
= cstate
->session
->se_client
;
5228 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
5229 stateid
->ts_id_status
=
5230 nfsd4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
5236 nfsd4_free_lock_stateid(stateid_t
*stateid
, struct nfs4_stid
*s
)
5238 struct nfs4_ol_stateid
*stp
= openlockstateid(s
);
5241 ret
= nfsd4_lock_ol_stateid(stp
);
5245 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
5249 ret
= nfserr_locks_held
;
5250 if (check_for_locks(stp
->st_stid
.sc_file
,
5251 lockowner(stp
->st_stateowner
)))
5254 release_lock_stateid(stp
);
5258 mutex_unlock(&stp
->st_mutex
);
5265 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5266 union nfsd4_op_u
*u
)
5268 struct nfsd4_free_stateid
*free_stateid
= &u
->free_stateid
;
5269 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
5270 struct nfs4_stid
*s
;
5271 struct nfs4_delegation
*dp
;
5272 struct nfs4_client
*cl
= cstate
->session
->se_client
;
5273 __be32 ret
= nfserr_bad_stateid
;
5275 spin_lock(&cl
->cl_lock
);
5276 s
= find_stateid_locked(cl
, stateid
);
5279 spin_lock(&s
->sc_lock
);
5280 switch (s
->sc_type
) {
5281 case NFS4_DELEG_STID
:
5282 ret
= nfserr_locks_held
;
5284 case NFS4_OPEN_STID
:
5285 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
5288 ret
= nfserr_locks_held
;
5290 case NFS4_LOCK_STID
:
5291 spin_unlock(&s
->sc_lock
);
5292 refcount_inc(&s
->sc_count
);
5293 spin_unlock(&cl
->cl_lock
);
5294 ret
= nfsd4_free_lock_stateid(stateid
, s
);
5296 case NFS4_REVOKED_DELEG_STID
:
5297 spin_unlock(&s
->sc_lock
);
5298 dp
= delegstateid(s
);
5299 list_del_init(&dp
->dl_recall_lru
);
5300 spin_unlock(&cl
->cl_lock
);
5304 /* Default falls through and returns nfserr_bad_stateid */
5306 spin_unlock(&s
->sc_lock
);
5308 spin_unlock(&cl
->cl_lock
);
5316 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
5317 RD_STATE
: WR_STATE
;
5320 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
5322 struct svc_fh
*current_fh
= &cstate
->current_fh
;
5323 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
5326 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
5329 status
= nfsd4_lock_ol_stateid(stp
);
5330 if (status
!= nfs_ok
)
5332 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
5333 if (status
== nfs_ok
)
5334 status
= nfs4_check_fh(current_fh
, &stp
->st_stid
);
5335 if (status
!= nfs_ok
)
5336 mutex_unlock(&stp
->st_mutex
);
5341 * Checks for sequence id mutating operations.
5344 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
5345 stateid_t
*stateid
, char typemask
,
5346 struct nfs4_ol_stateid
**stpp
,
5347 struct nfsd_net
*nn
)
5350 struct nfs4_stid
*s
;
5351 struct nfs4_ol_stateid
*stp
= NULL
;
5353 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
5354 seqid
, STATEID_VAL(stateid
));
5357 status
= nfsd4_lookup_stateid(cstate
, stateid
, typemask
, &s
, nn
);
5360 stp
= openlockstateid(s
);
5361 nfsd4_cstate_assign_replay(cstate
, stp
->st_stateowner
);
5363 status
= nfs4_seqid_op_checks(cstate
, stateid
, seqid
, stp
);
5367 nfs4_put_stid(&stp
->st_stid
);
5371 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
5372 stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
, struct nfsd_net
*nn
)
5375 struct nfs4_openowner
*oo
;
5376 struct nfs4_ol_stateid
*stp
;
5378 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
5379 NFS4_OPEN_STID
, &stp
, nn
);
5382 oo
= openowner(stp
->st_stateowner
);
5383 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
5384 mutex_unlock(&stp
->st_mutex
);
5385 nfs4_put_stid(&stp
->st_stid
);
5386 return nfserr_bad_stateid
;
5393 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5394 union nfsd4_op_u
*u
)
5396 struct nfsd4_open_confirm
*oc
= &u
->open_confirm
;
5398 struct nfs4_openowner
*oo
;
5399 struct nfs4_ol_stateid
*stp
;
5400 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5402 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
5403 cstate
->current_fh
.fh_dentry
);
5405 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
5409 status
= nfs4_preprocess_seqid_op(cstate
,
5410 oc
->oc_seqid
, &oc
->oc_req_stateid
,
5411 NFS4_OPEN_STID
, &stp
, nn
);
5414 oo
= openowner(stp
->st_stateowner
);
5415 status
= nfserr_bad_stateid
;
5416 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
) {
5417 mutex_unlock(&stp
->st_mutex
);
5420 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
5421 nfs4_inc_and_copy_stateid(&oc
->oc_resp_stateid
, &stp
->st_stid
);
5422 mutex_unlock(&stp
->st_mutex
);
5423 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
5424 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
5426 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
5429 nfs4_put_stid(&stp
->st_stid
);
5431 nfsd4_bump_seqid(cstate
, status
);
5435 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
5437 if (!test_access(access
, stp
))
5439 nfs4_file_put_access(stp
->st_stid
.sc_file
, access
);
5440 clear_access(access
, stp
);
5443 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
5445 switch (to_access
) {
5446 case NFS4_SHARE_ACCESS_READ
:
5447 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
5448 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
5450 case NFS4_SHARE_ACCESS_WRITE
:
5451 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
5452 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
5454 case NFS4_SHARE_ACCESS_BOTH
:
5462 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
5463 struct nfsd4_compound_state
*cstate
, union nfsd4_op_u
*u
)
5465 struct nfsd4_open_downgrade
*od
= &u
->open_downgrade
;
5467 struct nfs4_ol_stateid
*stp
;
5468 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5470 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
5471 cstate
->current_fh
.fh_dentry
);
5473 /* We don't yet support WANT bits: */
5474 if (od
->od_deleg_want
)
5475 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
5478 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
5479 &od
->od_stateid
, &stp
, nn
);
5482 status
= nfserr_inval
;
5483 if (!test_access(od
->od_share_access
, stp
)) {
5484 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
5485 stp
->st_access_bmap
, od
->od_share_access
);
5488 if (!test_deny(od
->od_share_deny
, stp
)) {
5489 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
5490 stp
->st_deny_bmap
, od
->od_share_deny
);
5493 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
5494 reset_union_bmap_deny(od
->od_share_deny
, stp
);
5495 nfs4_inc_and_copy_stateid(&od
->od_stateid
, &stp
->st_stid
);
5498 mutex_unlock(&stp
->st_mutex
);
5499 nfs4_put_stid(&stp
->st_stid
);
5501 nfsd4_bump_seqid(cstate
, status
);
5505 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
5507 struct nfs4_client
*clp
= s
->st_stid
.sc_client
;
5509 LIST_HEAD(reaplist
);
5511 spin_lock(&clp
->cl_lock
);
5512 unhashed
= unhash_open_stateid(s
, &reaplist
);
5514 if (clp
->cl_minorversion
) {
5516 put_ol_stateid_locked(s
, &reaplist
);
5517 spin_unlock(&clp
->cl_lock
);
5518 free_ol_stateid_reaplist(&reaplist
);
5520 spin_unlock(&clp
->cl_lock
);
5521 free_ol_stateid_reaplist(&reaplist
);
5523 move_to_close_lru(s
, clp
->net
);
5528 * nfs4_unlock_state() called after encode
5531 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5532 union nfsd4_op_u
*u
)
5534 struct nfsd4_close
*close
= &u
->close
;
5536 struct nfs4_ol_stateid
*stp
;
5537 struct net
*net
= SVC_NET(rqstp
);
5538 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5540 dprintk("NFSD: nfsd4_close on file %pd\n",
5541 cstate
->current_fh
.fh_dentry
);
5543 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
5545 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
5547 nfsd4_bump_seqid(cstate
, status
);
5551 stp
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
5554 * Technically we don't _really_ have to increment or copy it, since
5555 * it should just be gone after this operation and we clobber the
5556 * copied value below, but we continue to do so here just to ensure
5557 * that racing ops see that there was a state change.
5559 nfs4_inc_and_copy_stateid(&close
->cl_stateid
, &stp
->st_stid
);
5561 nfsd4_close_open_stateid(stp
);
5562 mutex_unlock(&stp
->st_mutex
);
5564 /* v4.1+ suggests that we send a special stateid in here, since the
5565 * clients should just ignore this anyway. Since this is not useful
5566 * for v4.0 clients either, we set it to the special close_stateid
5569 * See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
5571 memcpy(&close
->cl_stateid
, &close_stateid
, sizeof(close
->cl_stateid
));
5573 /* put reference from nfs4_preprocess_seqid_op */
5574 nfs4_put_stid(&stp
->st_stid
);
5580 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5581 union nfsd4_op_u
*u
)
5583 struct nfsd4_delegreturn
*dr
= &u
->delegreturn
;
5584 struct nfs4_delegation
*dp
;
5585 stateid_t
*stateid
= &dr
->dr_stateid
;
5586 struct nfs4_stid
*s
;
5588 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
5590 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
5593 status
= nfsd4_lookup_stateid(cstate
, stateid
, NFS4_DELEG_STID
, &s
, nn
);
5596 dp
= delegstateid(s
);
5597 status
= nfsd4_stid_check_stateid_generation(stateid
, &dp
->dl_stid
, nfsd4_has_session(cstate
));
5601 destroy_delegation(dp
);
5603 nfs4_put_stid(&dp
->dl_stid
);
5609 end_offset(u64 start
, u64 len
)
5614 return end
>= start
? end
: NFS4_MAX_UINT64
;
5617 /* last octet in a range */
5619 last_byte_offset(u64 start
, u64 len
)
5625 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
5629 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5630 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5631 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
5632 * locking, this prevents us from being completely protocol-compliant. The
5633 * real solution to this problem is to start using unsigned file offsets in
5634 * the VFS, but this is a very deep change!
5637 nfs4_transform_lock_offset(struct file_lock
*lock
)
5639 if (lock
->fl_start
< 0)
5640 lock
->fl_start
= OFFSET_MAX
;
5641 if (lock
->fl_end
< 0)
5642 lock
->fl_end
= OFFSET_MAX
;
5646 nfsd4_fl_get_owner(fl_owner_t owner
)
5648 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5650 nfs4_get_stateowner(&lo
->lo_owner
);
5655 nfsd4_fl_put_owner(fl_owner_t owner
)
5657 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)owner
;
5660 nfs4_put_stateowner(&lo
->lo_owner
);
5664 nfsd4_lm_notify(struct file_lock
*fl
)
5666 struct nfs4_lockowner
*lo
= (struct nfs4_lockowner
*)fl
->fl_owner
;
5667 struct net
*net
= lo
->lo_owner
.so_client
->net
;
5668 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5669 struct nfsd4_blocked_lock
*nbl
= container_of(fl
,
5670 struct nfsd4_blocked_lock
, nbl_lock
);
5673 /* An empty list means that something else is going to be using it */
5674 spin_lock(&nn
->blocked_locks_lock
);
5675 if (!list_empty(&nbl
->nbl_list
)) {
5676 list_del_init(&nbl
->nbl_list
);
5677 list_del_init(&nbl
->nbl_lru
);
5680 spin_unlock(&nn
->blocked_locks_lock
);
5683 nfsd4_run_cb(&nbl
->nbl_cb
);
5686 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
5687 .lm_notify
= nfsd4_lm_notify
,
5688 .lm_get_owner
= nfsd4_fl_get_owner
,
5689 .lm_put_owner
= nfsd4_fl_put_owner
,
5693 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
5695 struct nfs4_lockowner
*lo
;
5697 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
5698 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
5699 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
5700 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
5701 if (!deny
->ld_owner
.data
)
5702 /* We just don't care that much */
5704 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
5705 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
5708 deny
->ld_owner
.len
= 0;
5709 deny
->ld_owner
.data
= NULL
;
5710 deny
->ld_clientid
.cl_boot
= 0;
5711 deny
->ld_clientid
.cl_id
= 0;
5713 deny
->ld_start
= fl
->fl_start
;
5714 deny
->ld_length
= NFS4_MAX_UINT64
;
5715 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
5716 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
5717 deny
->ld_type
= NFS4_READ_LT
;
5718 if (fl
->fl_type
!= F_RDLCK
)
5719 deny
->ld_type
= NFS4_WRITE_LT
;
5722 static struct nfs4_lockowner
*
5723 find_lockowner_str_locked(struct nfs4_client
*clp
, struct xdr_netobj
*owner
)
5725 unsigned int strhashval
= ownerstr_hashval(owner
);
5726 struct nfs4_stateowner
*so
;
5728 lockdep_assert_held(&clp
->cl_lock
);
5730 list_for_each_entry(so
, &clp
->cl_ownerstr_hashtbl
[strhashval
],
5732 if (so
->so_is_open_owner
)
5734 if (same_owner_str(so
, owner
))
5735 return lockowner(nfs4_get_stateowner(so
));
5740 static struct nfs4_lockowner
*
5741 find_lockowner_str(struct nfs4_client
*clp
, struct xdr_netobj
*owner
)
5743 struct nfs4_lockowner
*lo
;
5745 spin_lock(&clp
->cl_lock
);
5746 lo
= find_lockowner_str_locked(clp
, owner
);
5747 spin_unlock(&clp
->cl_lock
);
5751 static void nfs4_unhash_lockowner(struct nfs4_stateowner
*sop
)
5753 unhash_lockowner_locked(lockowner(sop
));
5756 static void nfs4_free_lockowner(struct nfs4_stateowner
*sop
)
5758 struct nfs4_lockowner
*lo
= lockowner(sop
);
5760 kmem_cache_free(lockowner_slab
, lo
);
5763 static const struct nfs4_stateowner_operations lockowner_ops
= {
5764 .so_unhash
= nfs4_unhash_lockowner
,
5765 .so_free
= nfs4_free_lockowner
,
5769 * Alloc a lock owner structure.
5770 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
5773 * strhashval = ownerstr_hashval
5775 static struct nfs4_lockowner
*
5776 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
,
5777 struct nfs4_ol_stateid
*open_stp
,
5778 struct nfsd4_lock
*lock
)
5780 struct nfs4_lockowner
*lo
, *ret
;
5782 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
5785 INIT_LIST_HEAD(&lo
->lo_blocked
);
5786 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
5787 lo
->lo_owner
.so_is_open_owner
= 0;
5788 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
;
5789 lo
->lo_owner
.so_ops
= &lockowner_ops
;
5790 spin_lock(&clp
->cl_lock
);
5791 ret
= find_lockowner_str_locked(clp
, &lock
->lk_new_owner
);
5793 list_add(&lo
->lo_owner
.so_strhash
,
5794 &clp
->cl_ownerstr_hashtbl
[strhashval
]);
5797 nfs4_free_stateowner(&lo
->lo_owner
);
5799 spin_unlock(&clp
->cl_lock
);
5803 static struct nfs4_ol_stateid
*
5804 find_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
)
5806 struct nfs4_ol_stateid
*lst
;
5807 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5809 lockdep_assert_held(&clp
->cl_lock
);
5811 list_for_each_entry(lst
, &lo
->lo_owner
.so_stateids
, st_perstateowner
) {
5812 if (lst
->st_stid
.sc_type
!= NFS4_LOCK_STID
)
5814 if (lst
->st_stid
.sc_file
== fp
) {
5815 refcount_inc(&lst
->st_stid
.sc_count
);
5822 static struct nfs4_ol_stateid
*
5823 init_lock_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_lockowner
*lo
,
5824 struct nfs4_file
*fp
, struct inode
*inode
,
5825 struct nfs4_ol_stateid
*open_stp
)
5827 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
5828 struct nfs4_ol_stateid
*retstp
;
5830 mutex_init(&stp
->st_mutex
);
5831 mutex_lock_nested(&stp
->st_mutex
, OPEN_STATEID_MUTEX
);
5833 spin_lock(&clp
->cl_lock
);
5834 spin_lock(&fp
->fi_lock
);
5835 retstp
= find_lock_stateid(lo
, fp
);
5839 refcount_inc(&stp
->st_stid
.sc_count
);
5840 stp
->st_stid
.sc_type
= NFS4_LOCK_STID
;
5841 stp
->st_stateowner
= nfs4_get_stateowner(&lo
->lo_owner
);
5843 stp
->st_stid
.sc_file
= fp
;
5844 stp
->st_access_bmap
= 0;
5845 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
5846 stp
->st_openstp
= open_stp
;
5847 list_add(&stp
->st_locks
, &open_stp
->st_locks
);
5848 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
5849 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
5851 spin_unlock(&fp
->fi_lock
);
5852 spin_unlock(&clp
->cl_lock
);
5854 if (nfsd4_lock_ol_stateid(retstp
) != nfs_ok
) {
5855 nfs4_put_stid(&retstp
->st_stid
);
5858 /* To keep mutex tracking happy */
5859 mutex_unlock(&stp
->st_mutex
);
5865 static struct nfs4_ol_stateid
*
5866 find_or_create_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fi
,
5867 struct inode
*inode
, struct nfs4_ol_stateid
*ost
,
5870 struct nfs4_stid
*ns
= NULL
;
5871 struct nfs4_ol_stateid
*lst
;
5872 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5873 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
5876 spin_lock(&clp
->cl_lock
);
5877 lst
= find_lock_stateid(lo
, fi
);
5878 spin_unlock(&clp
->cl_lock
);
5880 if (nfsd4_lock_ol_stateid(lst
) == nfs_ok
)
5882 nfs4_put_stid(&lst
->st_stid
);
5884 ns
= nfs4_alloc_stid(clp
, stateid_slab
, nfs4_free_lock_stateid
);
5888 lst
= init_lock_stateid(openlockstateid(ns
), lo
, fi
, inode
, ost
);
5889 if (lst
== openlockstateid(ns
))
5898 check_lock_length(u64 offset
, u64 length
)
5900 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
5901 (length
> ~offset
)));
5904 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
5906 struct nfs4_file
*fp
= lock_stp
->st_stid
.sc_file
;
5908 lockdep_assert_held(&fp
->fi_lock
);
5910 if (test_access(access
, lock_stp
))
5912 __nfs4_file_get_access(fp
, access
);
5913 set_access(access
, lock_stp
);
5917 lookup_or_create_lock_state(struct nfsd4_compound_state
*cstate
,
5918 struct nfs4_ol_stateid
*ost
,
5919 struct nfsd4_lock
*lock
,
5920 struct nfs4_ol_stateid
**plst
, bool *new)
5923 struct nfs4_file
*fi
= ost
->st_stid
.sc_file
;
5924 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
5925 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
5926 struct inode
*inode
= d_inode(cstate
->current_fh
.fh_dentry
);
5927 struct nfs4_lockowner
*lo
;
5928 struct nfs4_ol_stateid
*lst
;
5929 unsigned int strhashval
;
5931 lo
= find_lockowner_str(cl
, &lock
->lk_new_owner
);
5933 strhashval
= ownerstr_hashval(&lock
->lk_new_owner
);
5934 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
5936 return nfserr_jukebox
;
5938 /* with an existing lockowner, seqids must be the same */
5939 status
= nfserr_bad_seqid
;
5940 if (!cstate
->minorversion
&&
5941 lock
->lk_new_lock_seqid
!= lo
->lo_owner
.so_seqid
)
5945 lst
= find_or_create_lock_stateid(lo
, fi
, inode
, ost
, new);
5947 status
= nfserr_jukebox
;
5954 nfs4_put_stateowner(&lo
->lo_owner
);
5962 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
5963 union nfsd4_op_u
*u
)
5965 struct nfsd4_lock
*lock
= &u
->lock
;
5966 struct nfs4_openowner
*open_sop
= NULL
;
5967 struct nfs4_lockowner
*lock_sop
= NULL
;
5968 struct nfs4_ol_stateid
*lock_stp
= NULL
;
5969 struct nfs4_ol_stateid
*open_stp
= NULL
;
5970 struct nfs4_file
*fp
;
5971 struct file
*filp
= NULL
;
5972 struct nfsd4_blocked_lock
*nbl
= NULL
;
5973 struct file_lock
*file_lock
= NULL
;
5974 struct file_lock
*conflock
= NULL
;
5979 unsigned char fl_type
;
5980 unsigned int fl_flags
= FL_POSIX
;
5981 struct net
*net
= SVC_NET(rqstp
);
5982 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
5984 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5985 (long long) lock
->lk_offset
,
5986 (long long) lock
->lk_length
);
5988 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
5989 return nfserr_inval
;
5991 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
5992 S_IFREG
, NFSD_MAY_LOCK
))) {
5993 dprintk("NFSD: nfsd4_lock: permission denied!\n");
5997 if (lock
->lk_is_new
) {
5998 if (nfsd4_has_session(cstate
))
5999 /* See rfc 5661 18.10.3: given clientid is ignored: */
6000 memcpy(&lock
->lk_new_clientid
,
6001 &cstate
->session
->se_client
->cl_clientid
,
6002 sizeof(clientid_t
));
6004 status
= nfserr_stale_clientid
;
6005 if (STALE_CLIENTID(&lock
->lk_new_clientid
, nn
))
6008 /* validate and update open stateid and open seqid */
6009 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
6010 lock
->lk_new_open_seqid
,
6011 &lock
->lk_new_open_stateid
,
6015 mutex_unlock(&open_stp
->st_mutex
);
6016 open_sop
= openowner(open_stp
->st_stateowner
);
6017 status
= nfserr_bad_stateid
;
6018 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
6019 &lock
->lk_new_clientid
))
6021 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
6024 status
= nfs4_preprocess_seqid_op(cstate
,
6025 lock
->lk_old_lock_seqid
,
6026 &lock
->lk_old_lock_stateid
,
6027 NFS4_LOCK_STID
, &lock_stp
, nn
);
6031 lock_sop
= lockowner(lock_stp
->st_stateowner
);
6033 lkflg
= setlkflg(lock
->lk_type
);
6034 status
= nfs4_check_openmode(lock_stp
, lkflg
);
6038 status
= nfserr_grace
;
6039 if (locks_in_grace(net
) && !lock
->lk_reclaim
)
6041 status
= nfserr_no_grace
;
6042 if (!locks_in_grace(net
) && lock
->lk_reclaim
)
6045 fp
= lock_stp
->st_stid
.sc_file
;
6046 switch (lock
->lk_type
) {
6048 if (nfsd4_has_session(cstate
))
6049 fl_flags
|= FL_SLEEP
;
6052 spin_lock(&fp
->fi_lock
);
6053 filp
= find_readable_file_locked(fp
);
6055 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
6056 spin_unlock(&fp
->fi_lock
);
6059 case NFS4_WRITEW_LT
:
6060 if (nfsd4_has_session(cstate
))
6061 fl_flags
|= FL_SLEEP
;
6064 spin_lock(&fp
->fi_lock
);
6065 filp
= find_writeable_file_locked(fp
);
6067 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
6068 spin_unlock(&fp
->fi_lock
);
6072 status
= nfserr_inval
;
6077 status
= nfserr_openmode
;
6081 nbl
= find_or_allocate_block(lock_sop
, &fp
->fi_fhandle
, nn
);
6083 dprintk("NFSD: %s: unable to allocate block!\n", __func__
);
6084 status
= nfserr_jukebox
;
6088 file_lock
= &nbl
->nbl_lock
;
6089 file_lock
->fl_type
= fl_type
;
6090 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(&lock_sop
->lo_owner
));
6091 file_lock
->fl_pid
= current
->tgid
;
6092 file_lock
->fl_file
= filp
;
6093 file_lock
->fl_flags
= fl_flags
;
6094 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
6095 file_lock
->fl_start
= lock
->lk_offset
;
6096 file_lock
->fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
6097 nfs4_transform_lock_offset(file_lock
);
6099 conflock
= locks_alloc_lock();
6101 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
6102 status
= nfserr_jukebox
;
6106 if (fl_flags
& FL_SLEEP
) {
6107 nbl
->nbl_time
= jiffies
;
6108 spin_lock(&nn
->blocked_locks_lock
);
6109 list_add_tail(&nbl
->nbl_list
, &lock_sop
->lo_blocked
);
6110 list_add_tail(&nbl
->nbl_lru
, &nn
->blocked_locks_lru
);
6111 spin_unlock(&nn
->blocked_locks_lock
);
6114 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, conflock
);
6116 case 0: /* success! */
6117 nfs4_inc_and_copy_stateid(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
);
6119 if (lock
->lk_reclaim
)
6120 nn
->somebody_reclaimed
= true;
6122 case FILE_LOCK_DEFERRED
:
6125 case -EAGAIN
: /* conflock holds conflicting lock */
6126 status
= nfserr_denied
;
6127 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
6128 nfs4_set_lock_denied(conflock
, &lock
->lk_denied
);
6131 status
= nfserr_deadlock
;
6134 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
6135 status
= nfserrno(err
);
6140 /* dequeue it if we queued it before */
6141 if (fl_flags
& FL_SLEEP
) {
6142 spin_lock(&nn
->blocked_locks_lock
);
6143 list_del_init(&nbl
->nbl_list
);
6144 list_del_init(&nbl
->nbl_lru
);
6145 spin_unlock(&nn
->blocked_locks_lock
);
6147 free_blocked_lock(nbl
);
6152 /* Bump seqid manually if the 4.0 replay owner is openowner */
6153 if (cstate
->replay_owner
&&
6154 cstate
->replay_owner
!= &lock_sop
->lo_owner
&&
6155 seqid_mutating_err(ntohl(status
)))
6156 lock_sop
->lo_owner
.so_seqid
++;
6159 * If this is a new, never-before-used stateid, and we are
6160 * returning an error, then just go ahead and release it.
6163 release_lock_stateid(lock_stp
);
6165 mutex_unlock(&lock_stp
->st_mutex
);
6167 nfs4_put_stid(&lock_stp
->st_stid
);
6170 nfs4_put_stid(&open_stp
->st_stid
);
6171 nfsd4_bump_seqid(cstate
, status
);
6173 locks_free_lock(conflock
);
6178 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
6179 * so we do a temporary open here just to get an open file to pass to
6180 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
6183 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
6186 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
6188 err
= nfserrno(vfs_test_lock(file
, lock
));
6198 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
6199 union nfsd4_op_u
*u
)
6201 struct nfsd4_lockt
*lockt
= &u
->lockt
;
6202 struct file_lock
*file_lock
= NULL
;
6203 struct nfs4_lockowner
*lo
= NULL
;
6205 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
6207 if (locks_in_grace(SVC_NET(rqstp
)))
6208 return nfserr_grace
;
6210 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
6211 return nfserr_inval
;
6213 if (!nfsd4_has_session(cstate
)) {
6214 status
= lookup_clientid(&lockt
->lt_clientid
, cstate
, nn
);
6219 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
6222 file_lock
= locks_alloc_lock();
6224 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
6225 status
= nfserr_jukebox
;
6229 switch (lockt
->lt_type
) {
6232 file_lock
->fl_type
= F_RDLCK
;
6235 case NFS4_WRITEW_LT
:
6236 file_lock
->fl_type
= F_WRLCK
;
6239 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
6240 status
= nfserr_inval
;
6244 lo
= find_lockowner_str(cstate
->clp
, &lockt
->lt_owner
);
6246 file_lock
->fl_owner
= (fl_owner_t
)lo
;
6247 file_lock
->fl_pid
= current
->tgid
;
6248 file_lock
->fl_flags
= FL_POSIX
;
6250 file_lock
->fl_start
= lockt
->lt_offset
;
6251 file_lock
->fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
6253 nfs4_transform_lock_offset(file_lock
);
6255 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, file_lock
);
6259 if (file_lock
->fl_type
!= F_UNLCK
) {
6260 status
= nfserr_denied
;
6261 nfs4_set_lock_denied(file_lock
, &lockt
->lt_denied
);
6265 nfs4_put_stateowner(&lo
->lo_owner
);
6267 locks_free_lock(file_lock
);
6272 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
6273 union nfsd4_op_u
*u
)
6275 struct nfsd4_locku
*locku
= &u
->locku
;
6276 struct nfs4_ol_stateid
*stp
;
6277 struct file
*filp
= NULL
;
6278 struct file_lock
*file_lock
= NULL
;
6281 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
6283 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
6284 (long long) locku
->lu_offset
,
6285 (long long) locku
->lu_length
);
6287 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
6288 return nfserr_inval
;
6290 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
6291 &locku
->lu_stateid
, NFS4_LOCK_STID
,
6295 filp
= find_any_file(stp
->st_stid
.sc_file
);
6297 status
= nfserr_lock_range
;
6300 file_lock
= locks_alloc_lock();
6302 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
6303 status
= nfserr_jukebox
;
6307 file_lock
->fl_type
= F_UNLCK
;
6308 file_lock
->fl_owner
= (fl_owner_t
)lockowner(nfs4_get_stateowner(stp
->st_stateowner
));
6309 file_lock
->fl_pid
= current
->tgid
;
6310 file_lock
->fl_file
= filp
;
6311 file_lock
->fl_flags
= FL_POSIX
;
6312 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
6313 file_lock
->fl_start
= locku
->lu_offset
;
6315 file_lock
->fl_end
= last_byte_offset(locku
->lu_offset
,
6317 nfs4_transform_lock_offset(file_lock
);
6319 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, NULL
);
6321 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
6324 nfs4_inc_and_copy_stateid(&locku
->lu_stateid
, &stp
->st_stid
);
6328 mutex_unlock(&stp
->st_mutex
);
6329 nfs4_put_stid(&stp
->st_stid
);
6331 nfsd4_bump_seqid(cstate
, status
);
6333 locks_free_lock(file_lock
);
6337 status
= nfserrno(err
);
6343 * true: locks held by lockowner
6344 * false: no locks held by lockowner
6347 check_for_locks(struct nfs4_file
*fp
, struct nfs4_lockowner
*lowner
)
6349 struct file_lock
*fl
;
6351 struct file
*filp
= find_any_file(fp
);
6352 struct inode
*inode
;
6353 struct file_lock_context
*flctx
;
6356 /* Any valid lock stateid should have some sort of access */
6361 inode
= locks_inode(filp
);
6362 flctx
= inode
->i_flctx
;
6364 if (flctx
&& !list_empty_careful(&flctx
->flc_posix
)) {
6365 spin_lock(&flctx
->flc_lock
);
6366 list_for_each_entry(fl
, &flctx
->flc_posix
, fl_list
) {
6367 if (fl
->fl_owner
== (fl_owner_t
)lowner
) {
6372 spin_unlock(&flctx
->flc_lock
);
6379 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
6380 struct nfsd4_compound_state
*cstate
,
6381 union nfsd4_op_u
*u
)
6383 struct nfsd4_release_lockowner
*rlockowner
= &u
->release_lockowner
;
6384 clientid_t
*clid
= &rlockowner
->rl_clientid
;
6385 struct nfs4_stateowner
*sop
;
6386 struct nfs4_lockowner
*lo
= NULL
;
6387 struct nfs4_ol_stateid
*stp
;
6388 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
6389 unsigned int hashval
= ownerstr_hashval(owner
);
6391 struct nfsd_net
*nn
= net_generic(SVC_NET(rqstp
), nfsd_net_id
);
6392 struct nfs4_client
*clp
;
6393 LIST_HEAD (reaplist
);
6395 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
6396 clid
->cl_boot
, clid
->cl_id
);
6398 status
= lookup_clientid(clid
, cstate
, nn
);
6403 /* Find the matching lock stateowner */
6404 spin_lock(&clp
->cl_lock
);
6405 list_for_each_entry(sop
, &clp
->cl_ownerstr_hashtbl
[hashval
],
6408 if (sop
->so_is_open_owner
|| !same_owner_str(sop
, owner
))
6411 /* see if there are still any locks associated with it */
6412 lo
= lockowner(sop
);
6413 list_for_each_entry(stp
, &sop
->so_stateids
, st_perstateowner
) {
6414 if (check_for_locks(stp
->st_stid
.sc_file
, lo
)) {
6415 status
= nfserr_locks_held
;
6416 spin_unlock(&clp
->cl_lock
);
6421 nfs4_get_stateowner(sop
);
6425 spin_unlock(&clp
->cl_lock
);
6429 unhash_lockowner_locked(lo
);
6430 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
6431 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
6432 struct nfs4_ol_stateid
,
6434 WARN_ON(!unhash_lock_stateid(stp
));
6435 put_ol_stateid_locked(stp
, &reaplist
);
6437 spin_unlock(&clp
->cl_lock
);
6438 free_ol_stateid_reaplist(&reaplist
);
6439 remove_blocked_locks(lo
);
6440 nfs4_put_stateowner(&lo
->lo_owner
);
6445 static inline struct nfs4_client_reclaim
*
6448 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
6452 nfs4_has_reclaimed_state(const char *name
, struct nfsd_net
*nn
)
6454 struct nfs4_client_reclaim
*crp
;
6456 crp
= nfsd4_find_reclaim_client(name
, nn
);
6457 return (crp
&& crp
->cr_clp
);
6461 * failure => all reset bets are off, nfserr_no_grace...
6463 struct nfs4_client_reclaim
*
6464 nfs4_client_to_reclaim(const char *name
, struct nfsd_net
*nn
)
6466 unsigned int strhashval
;
6467 struct nfs4_client_reclaim
*crp
;
6469 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
6470 crp
= alloc_reclaim();
6472 strhashval
= clientstr_hashval(name
);
6473 INIT_LIST_HEAD(&crp
->cr_strhash
);
6474 list_add(&crp
->cr_strhash
, &nn
->reclaim_str_hashtbl
[strhashval
]);
6475 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
6477 nn
->reclaim_str_hashtbl_size
++;
6483 nfs4_remove_reclaim_record(struct nfs4_client_reclaim
*crp
, struct nfsd_net
*nn
)
6485 list_del(&crp
->cr_strhash
);
6487 nn
->reclaim_str_hashtbl_size
--;
6491 nfs4_release_reclaim(struct nfsd_net
*nn
)
6493 struct nfs4_client_reclaim
*crp
= NULL
;
6496 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
6497 while (!list_empty(&nn
->reclaim_str_hashtbl
[i
])) {
6498 crp
= list_entry(nn
->reclaim_str_hashtbl
[i
].next
,
6499 struct nfs4_client_reclaim
, cr_strhash
);
6500 nfs4_remove_reclaim_record(crp
, nn
);
6503 WARN_ON_ONCE(nn
->reclaim_str_hashtbl_size
);
6507 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
6508 struct nfs4_client_reclaim
*
6509 nfsd4_find_reclaim_client(const char *recdir
, struct nfsd_net
*nn
)
6511 unsigned int strhashval
;
6512 struct nfs4_client_reclaim
*crp
= NULL
;
6514 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir
);
6516 strhashval
= clientstr_hashval(recdir
);
6517 list_for_each_entry(crp
, &nn
->reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
6518 if (same_name(crp
->cr_recdir
, recdir
)) {
6526 * Called from OPEN. Look for clientid in reclaim list.
6529 nfs4_check_open_reclaim(clientid_t
*clid
,
6530 struct nfsd4_compound_state
*cstate
,
6531 struct nfsd_net
*nn
)
6535 /* find clientid in conf_id_hashtbl */
6536 status
= lookup_clientid(clid
, cstate
, nn
);
6538 return nfserr_reclaim_bad
;
6540 if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
, &cstate
->clp
->cl_flags
))
6541 return nfserr_no_grace
;
6543 if (nfsd4_client_record_check(cstate
->clp
))
6544 return nfserr_reclaim_bad
;
6549 #ifdef CONFIG_NFSD_FAULT_INJECTION
6551 put_client(struct nfs4_client
*clp
)
6553 atomic_dec(&clp
->cl_refcount
);
6556 static struct nfs4_client
*
6557 nfsd_find_client(struct sockaddr_storage
*addr
, size_t addr_size
)
6559 struct nfs4_client
*clp
;
6560 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6563 if (!nfsd_netns_ready(nn
))
6566 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6567 if (memcmp(&clp
->cl_addr
, addr
, addr_size
) == 0)
6574 nfsd_inject_print_clients(void)
6576 struct nfs4_client
*clp
;
6578 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6580 char buf
[INET6_ADDRSTRLEN
];
6582 if (!nfsd_netns_ready(nn
))
6585 spin_lock(&nn
->client_lock
);
6586 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6587 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
6588 pr_info("NFS Client: %s\n", buf
);
6591 spin_unlock(&nn
->client_lock
);
6597 nfsd_inject_forget_client(struct sockaddr_storage
*addr
, size_t addr_size
)
6600 struct nfs4_client
*clp
;
6601 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6604 if (!nfsd_netns_ready(nn
))
6607 spin_lock(&nn
->client_lock
);
6608 clp
= nfsd_find_client(addr
, addr_size
);
6610 if (mark_client_expired_locked(clp
) == nfs_ok
)
6615 spin_unlock(&nn
->client_lock
);
6624 nfsd_inject_forget_clients(u64 max
)
6627 struct nfs4_client
*clp
, *next
;
6628 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6630 LIST_HEAD(reaplist
);
6632 if (!nfsd_netns_ready(nn
))
6635 spin_lock(&nn
->client_lock
);
6636 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
6637 if (mark_client_expired_locked(clp
) == nfs_ok
) {
6638 list_add(&clp
->cl_lru
, &reaplist
);
6639 if (max
!= 0 && ++count
>= max
)
6643 spin_unlock(&nn
->client_lock
);
6645 list_for_each_entry_safe(clp
, next
, &reaplist
, cl_lru
)
6651 static void nfsd_print_count(struct nfs4_client
*clp
, unsigned int count
,
6654 char buf
[INET6_ADDRSTRLEN
];
6655 rpc_ntop((struct sockaddr
*)&clp
->cl_addr
, buf
, sizeof(buf
));
6656 printk(KERN_INFO
"NFS Client: %s has %u %s\n", buf
, count
, type
);
6660 nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid
*lst
,
6661 struct list_head
*collect
)
6663 struct nfs4_client
*clp
= lst
->st_stid
.sc_client
;
6664 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6670 lockdep_assert_held(&nn
->client_lock
);
6671 atomic_inc(&clp
->cl_refcount
);
6672 list_add(&lst
->st_locks
, collect
);
6675 static u64
nfsd_foreach_client_lock(struct nfs4_client
*clp
, u64 max
,
6676 struct list_head
*collect
,
6677 bool (*func
)(struct nfs4_ol_stateid
*))
6679 struct nfs4_openowner
*oop
;
6680 struct nfs4_ol_stateid
*stp
, *st_next
;
6681 struct nfs4_ol_stateid
*lst
, *lst_next
;
6684 spin_lock(&clp
->cl_lock
);
6685 list_for_each_entry(oop
, &clp
->cl_openowners
, oo_perclient
) {
6686 list_for_each_entry_safe(stp
, st_next
,
6687 &oop
->oo_owner
.so_stateids
, st_perstateowner
) {
6688 list_for_each_entry_safe(lst
, lst_next
,
6689 &stp
->st_locks
, st_locks
) {
6692 nfsd_inject_add_lock_to_list(lst
,
6697 * Despite the fact that these functions deal
6698 * with 64-bit integers for "count", we must
6699 * ensure that it doesn't blow up the
6700 * clp->cl_refcount. Throw a warning if we
6701 * start to approach INT_MAX here.
6703 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6710 spin_unlock(&clp
->cl_lock
);
6716 nfsd_collect_client_locks(struct nfs4_client
*clp
, struct list_head
*collect
,
6719 return nfsd_foreach_client_lock(clp
, max
, collect
, unhash_lock_stateid
);
6723 nfsd_print_client_locks(struct nfs4_client
*clp
)
6725 u64 count
= nfsd_foreach_client_lock(clp
, 0, NULL
, NULL
);
6726 nfsd_print_count(clp
, count
, "locked files");
6731 nfsd_inject_print_locks(void)
6733 struct nfs4_client
*clp
;
6735 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6738 if (!nfsd_netns_ready(nn
))
6741 spin_lock(&nn
->client_lock
);
6742 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6743 count
+= nfsd_print_client_locks(clp
);
6744 spin_unlock(&nn
->client_lock
);
6750 nfsd_reap_locks(struct list_head
*reaplist
)
6752 struct nfs4_client
*clp
;
6753 struct nfs4_ol_stateid
*stp
, *next
;
6755 list_for_each_entry_safe(stp
, next
, reaplist
, st_locks
) {
6756 list_del_init(&stp
->st_locks
);
6757 clp
= stp
->st_stid
.sc_client
;
6758 nfs4_put_stid(&stp
->st_stid
);
6764 nfsd_inject_forget_client_locks(struct sockaddr_storage
*addr
, size_t addr_size
)
6766 unsigned int count
= 0;
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_collect_client_locks(clp
, &reaplist
, 0);
6779 spin_unlock(&nn
->client_lock
);
6780 nfsd_reap_locks(&reaplist
);
6785 nfsd_inject_forget_locks(u64 max
)
6788 struct nfs4_client
*clp
;
6789 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6791 LIST_HEAD(reaplist
);
6793 if (!nfsd_netns_ready(nn
))
6796 spin_lock(&nn
->client_lock
);
6797 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6798 count
+= nfsd_collect_client_locks(clp
, &reaplist
, max
- count
);
6799 if (max
!= 0 && count
>= max
)
6802 spin_unlock(&nn
->client_lock
);
6803 nfsd_reap_locks(&reaplist
);
6808 nfsd_foreach_client_openowner(struct nfs4_client
*clp
, u64 max
,
6809 struct list_head
*collect
,
6810 void (*func
)(struct nfs4_openowner
*))
6812 struct nfs4_openowner
*oop
, *next
;
6813 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6817 lockdep_assert_held(&nn
->client_lock
);
6819 spin_lock(&clp
->cl_lock
);
6820 list_for_each_entry_safe(oop
, next
, &clp
->cl_openowners
, oo_perclient
) {
6824 atomic_inc(&clp
->cl_refcount
);
6825 list_add(&oop
->oo_perclient
, collect
);
6830 * Despite the fact that these functions deal with
6831 * 64-bit integers for "count", we must ensure that
6832 * it doesn't blow up the clp->cl_refcount. Throw a
6833 * warning if we start to approach INT_MAX here.
6835 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6839 spin_unlock(&clp
->cl_lock
);
6845 nfsd_print_client_openowners(struct nfs4_client
*clp
)
6847 u64 count
= nfsd_foreach_client_openowner(clp
, 0, NULL
, NULL
);
6849 nfsd_print_count(clp
, count
, "openowners");
6854 nfsd_collect_client_openowners(struct nfs4_client
*clp
,
6855 struct list_head
*collect
, u64 max
)
6857 return nfsd_foreach_client_openowner(clp
, max
, collect
,
6858 unhash_openowner_locked
);
6862 nfsd_inject_print_openowners(void)
6864 struct nfs4_client
*clp
;
6866 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6869 if (!nfsd_netns_ready(nn
))
6872 spin_lock(&nn
->client_lock
);
6873 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
6874 count
+= nfsd_print_client_openowners(clp
);
6875 spin_unlock(&nn
->client_lock
);
6881 nfsd_reap_openowners(struct list_head
*reaplist
)
6883 struct nfs4_client
*clp
;
6884 struct nfs4_openowner
*oop
, *next
;
6886 list_for_each_entry_safe(oop
, next
, reaplist
, oo_perclient
) {
6887 list_del_init(&oop
->oo_perclient
);
6888 clp
= oop
->oo_owner
.so_client
;
6889 release_openowner(oop
);
6895 nfsd_inject_forget_client_openowners(struct sockaddr_storage
*addr
,
6898 unsigned int count
= 0;
6899 struct nfs4_client
*clp
;
6900 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6902 LIST_HEAD(reaplist
);
6904 if (!nfsd_netns_ready(nn
))
6907 spin_lock(&nn
->client_lock
);
6908 clp
= nfsd_find_client(addr
, addr_size
);
6910 count
= nfsd_collect_client_openowners(clp
, &reaplist
, 0);
6911 spin_unlock(&nn
->client_lock
);
6912 nfsd_reap_openowners(&reaplist
);
6917 nfsd_inject_forget_openowners(u64 max
)
6920 struct nfs4_client
*clp
;
6921 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6923 LIST_HEAD(reaplist
);
6925 if (!nfsd_netns_ready(nn
))
6928 spin_lock(&nn
->client_lock
);
6929 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
6930 count
+= nfsd_collect_client_openowners(clp
, &reaplist
,
6932 if (max
!= 0 && count
>= max
)
6935 spin_unlock(&nn
->client_lock
);
6936 nfsd_reap_openowners(&reaplist
);
6940 static u64
nfsd_find_all_delegations(struct nfs4_client
*clp
, u64 max
,
6941 struct list_head
*victims
)
6943 struct nfs4_delegation
*dp
, *next
;
6944 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6948 lockdep_assert_held(&nn
->client_lock
);
6950 spin_lock(&state_lock
);
6951 list_for_each_entry_safe(dp
, next
, &clp
->cl_delegations
, dl_perclnt
) {
6954 * It's not safe to mess with delegations that have a
6955 * non-zero dl_time. They might have already been broken
6956 * and could be processed by the laundromat outside of
6957 * the state_lock. Just leave them be.
6959 if (dp
->dl_time
!= 0)
6962 atomic_inc(&clp
->cl_refcount
);
6963 WARN_ON(!unhash_delegation_locked(dp
));
6964 list_add(&dp
->dl_recall_lru
, victims
);
6968 * Despite the fact that these functions deal with
6969 * 64-bit integers for "count", we must ensure that
6970 * it doesn't blow up the clp->cl_refcount. Throw a
6971 * warning if we start to approach INT_MAX here.
6973 WARN_ON_ONCE(count
== (INT_MAX
/ 2));
6977 spin_unlock(&state_lock
);
6982 nfsd_print_client_delegations(struct nfs4_client
*clp
)
6984 u64 count
= nfsd_find_all_delegations(clp
, 0, NULL
);
6986 nfsd_print_count(clp
, count
, "delegations");
6991 nfsd_inject_print_delegations(void)
6993 struct nfs4_client
*clp
;
6995 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
6998 if (!nfsd_netns_ready(nn
))
7001 spin_lock(&nn
->client_lock
);
7002 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
)
7003 count
+= nfsd_print_client_delegations(clp
);
7004 spin_unlock(&nn
->client_lock
);
7010 nfsd_forget_delegations(struct list_head
*reaplist
)
7012 struct nfs4_client
*clp
;
7013 struct nfs4_delegation
*dp
, *next
;
7015 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
7016 list_del_init(&dp
->dl_recall_lru
);
7017 clp
= dp
->dl_stid
.sc_client
;
7018 revoke_delegation(dp
);
7024 nfsd_inject_forget_client_delegations(struct sockaddr_storage
*addr
,
7028 struct nfs4_client
*clp
;
7029 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
7031 LIST_HEAD(reaplist
);
7033 if (!nfsd_netns_ready(nn
))
7036 spin_lock(&nn
->client_lock
);
7037 clp
= nfsd_find_client(addr
, addr_size
);
7039 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
7040 spin_unlock(&nn
->client_lock
);
7042 nfsd_forget_delegations(&reaplist
);
7047 nfsd_inject_forget_delegations(u64 max
)
7050 struct nfs4_client
*clp
;
7051 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
7053 LIST_HEAD(reaplist
);
7055 if (!nfsd_netns_ready(nn
))
7058 spin_lock(&nn
->client_lock
);
7059 list_for_each_entry(clp
, &nn
->client_lru
, cl_lru
) {
7060 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
7061 if (max
!= 0 && count
>= max
)
7064 spin_unlock(&nn
->client_lock
);
7065 nfsd_forget_delegations(&reaplist
);
7070 nfsd_recall_delegations(struct list_head
*reaplist
)
7072 struct nfs4_client
*clp
;
7073 struct nfs4_delegation
*dp
, *next
;
7075 list_for_each_entry_safe(dp
, next
, reaplist
, dl_recall_lru
) {
7076 list_del_init(&dp
->dl_recall_lru
);
7077 clp
= dp
->dl_stid
.sc_client
;
7079 * We skipped all entries that had a zero dl_time before,
7080 * so we can now reset the dl_time back to 0. If a delegation
7081 * break comes in now, then it won't make any difference since
7082 * we're recalling it either way.
7084 spin_lock(&state_lock
);
7086 spin_unlock(&state_lock
);
7087 nfsd_break_one_deleg(dp
);
7093 nfsd_inject_recall_client_delegations(struct sockaddr_storage
*addr
,
7097 struct nfs4_client
*clp
;
7098 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
7100 LIST_HEAD(reaplist
);
7102 if (!nfsd_netns_ready(nn
))
7105 spin_lock(&nn
->client_lock
);
7106 clp
= nfsd_find_client(addr
, addr_size
);
7108 count
= nfsd_find_all_delegations(clp
, 0, &reaplist
);
7109 spin_unlock(&nn
->client_lock
);
7111 nfsd_recall_delegations(&reaplist
);
7116 nfsd_inject_recall_delegations(u64 max
)
7119 struct nfs4_client
*clp
, *next
;
7120 struct nfsd_net
*nn
= net_generic(current
->nsproxy
->net_ns
,
7122 LIST_HEAD(reaplist
);
7124 if (!nfsd_netns_ready(nn
))
7127 spin_lock(&nn
->client_lock
);
7128 list_for_each_entry_safe(clp
, next
, &nn
->client_lru
, cl_lru
) {
7129 count
+= nfsd_find_all_delegations(clp
, max
- count
, &reaplist
);
7130 if (max
!= 0 && ++count
>= max
)
7133 spin_unlock(&nn
->client_lock
);
7134 nfsd_recall_delegations(&reaplist
);
7137 #endif /* CONFIG_NFSD_FAULT_INJECTION */
7140 * Since the lifetime of a delegation isn't limited to that of an open, a
7141 * client may quite reasonably hang on to a delegation as long as it has
7142 * the inode cached. This becomes an obvious problem the first time a
7143 * client's inode cache approaches the size of the server's total memory.
7145 * For now we avoid this problem by imposing a hard limit on the number
7146 * of delegations, which varies according to the server's memory size.
7149 set_max_delegations(void)
7152 * Allow at most 4 delegations per megabyte of RAM. Quick
7153 * estimates suggest that in the worst case (where every delegation
7154 * is for a different inode), a delegation could take about 1.5K,
7155 * giving a worst case usage of about 6% of memory.
7157 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
7160 static int nfs4_state_create_net(struct net
*net
)
7162 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
7165 nn
->conf_id_hashtbl
= kmalloc_array(CLIENT_HASH_SIZE
,
7166 sizeof(struct list_head
),
7168 if (!nn
->conf_id_hashtbl
)
7170 nn
->unconf_id_hashtbl
= kmalloc_array(CLIENT_HASH_SIZE
,
7171 sizeof(struct list_head
),
7173 if (!nn
->unconf_id_hashtbl
)
7175 nn
->sessionid_hashtbl
= kmalloc_array(SESSION_HASH_SIZE
,
7176 sizeof(struct list_head
),
7178 if (!nn
->sessionid_hashtbl
)
7181 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
7182 INIT_LIST_HEAD(&nn
->conf_id_hashtbl
[i
]);
7183 INIT_LIST_HEAD(&nn
->unconf_id_hashtbl
[i
]);
7185 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
7186 INIT_LIST_HEAD(&nn
->sessionid_hashtbl
[i
]);
7187 nn
->conf_name_tree
= RB_ROOT
;
7188 nn
->unconf_name_tree
= RB_ROOT
;
7189 nn
->boot_time
= get_seconds();
7190 nn
->grace_ended
= false;
7191 nn
->nfsd4_manager
.block_opens
= true;
7192 INIT_LIST_HEAD(&nn
->nfsd4_manager
.list
);
7193 INIT_LIST_HEAD(&nn
->client_lru
);
7194 INIT_LIST_HEAD(&nn
->close_lru
);
7195 INIT_LIST_HEAD(&nn
->del_recall_lru
);
7196 spin_lock_init(&nn
->client_lock
);
7197 spin_lock_init(&nn
->s2s_cp_lock
);
7198 idr_init(&nn
->s2s_cp_stateids
);
7200 spin_lock_init(&nn
->blocked_locks_lock
);
7201 INIT_LIST_HEAD(&nn
->blocked_locks_lru
);
7203 INIT_DELAYED_WORK(&nn
->laundromat_work
, laundromat_main
);
7209 kfree(nn
->unconf_id_hashtbl
);
7211 kfree(nn
->conf_id_hashtbl
);
7217 nfs4_state_destroy_net(struct net
*net
)
7220 struct nfs4_client
*clp
= NULL
;
7221 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
7223 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
7224 while (!list_empty(&nn
->conf_id_hashtbl
[i
])) {
7225 clp
= list_entry(nn
->conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
7226 destroy_client(clp
);
7230 WARN_ON(!list_empty(&nn
->blocked_locks_lru
));
7232 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
7233 while (!list_empty(&nn
->unconf_id_hashtbl
[i
])) {
7234 clp
= list_entry(nn
->unconf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
7235 destroy_client(clp
);
7239 kfree(nn
->sessionid_hashtbl
);
7240 kfree(nn
->unconf_id_hashtbl
);
7241 kfree(nn
->conf_id_hashtbl
);
7246 nfs4_state_start_net(struct net
*net
)
7248 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
7251 ret
= nfs4_state_create_net(net
);
7254 locks_start_grace(net
, &nn
->nfsd4_manager
);
7255 nfsd4_client_tracking_init(net
);
7256 printk(KERN_INFO
"NFSD: starting %ld-second grace period (net %x)\n",
7257 nn
->nfsd4_grace
, net
->ns
.inum
);
7258 queue_delayed_work(laundry_wq
, &nn
->laundromat_work
, nn
->nfsd4_grace
* HZ
);
7262 /* initialization to perform when the nfsd service is started: */
7265 nfs4_state_start(void)
7269 laundry_wq
= alloc_workqueue("%s", WQ_UNBOUND
, 0, "nfsd4");
7270 if (laundry_wq
== NULL
) {
7274 ret
= nfsd4_create_callback_queue();
7276 goto out_free_laundry
;
7278 set_max_delegations();
7282 destroy_workqueue(laundry_wq
);
7288 nfs4_state_shutdown_net(struct net
*net
)
7290 struct nfs4_delegation
*dp
= NULL
;
7291 struct list_head
*pos
, *next
, reaplist
;
7292 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
7294 cancel_delayed_work_sync(&nn
->laundromat_work
);
7295 locks_end_grace(&nn
->nfsd4_manager
);
7297 INIT_LIST_HEAD(&reaplist
);
7298 spin_lock(&state_lock
);
7299 list_for_each_safe(pos
, next
, &nn
->del_recall_lru
) {
7300 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
7301 WARN_ON(!unhash_delegation_locked(dp
));
7302 list_add(&dp
->dl_recall_lru
, &reaplist
);
7304 spin_unlock(&state_lock
);
7305 list_for_each_safe(pos
, next
, &reaplist
) {
7306 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
7307 list_del_init(&dp
->dl_recall_lru
);
7308 destroy_unhashed_deleg(dp
);
7311 nfsd4_client_tracking_exit(net
);
7312 nfs4_state_destroy_net(net
);
7316 nfs4_state_shutdown(void)
7318 destroy_workqueue(laundry_wq
);
7319 nfsd4_destroy_callback_queue();
7323 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
7325 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
7326 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
7330 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
7332 if (cstate
->minorversion
) {
7333 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
7334 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
7339 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
7341 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
7345 * functions to set current state id
7348 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
,
7349 union nfsd4_op_u
*u
)
7351 put_stateid(cstate
, &u
->open_downgrade
.od_stateid
);
7355 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
,
7356 union nfsd4_op_u
*u
)
7358 put_stateid(cstate
, &u
->open
.op_stateid
);
7362 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
,
7363 union nfsd4_op_u
*u
)
7365 put_stateid(cstate
, &u
->close
.cl_stateid
);
7369 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
,
7370 union nfsd4_op_u
*u
)
7372 put_stateid(cstate
, &u
->lock
.lk_resp_stateid
);
7376 * functions to consume current state id
7380 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
,
7381 union nfsd4_op_u
*u
)
7383 get_stateid(cstate
, &u
->open_downgrade
.od_stateid
);
7387 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
,
7388 union nfsd4_op_u
*u
)
7390 get_stateid(cstate
, &u
->delegreturn
.dr_stateid
);
7394 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
,
7395 union nfsd4_op_u
*u
)
7397 get_stateid(cstate
, &u
->free_stateid
.fr_stateid
);
7401 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
,
7402 union nfsd4_op_u
*u
)
7404 get_stateid(cstate
, &u
->setattr
.sa_stateid
);
7408 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
,
7409 union nfsd4_op_u
*u
)
7411 get_stateid(cstate
, &u
->close
.cl_stateid
);
7415 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
,
7416 union nfsd4_op_u
*u
)
7418 get_stateid(cstate
, &u
->locku
.lu_stateid
);
7422 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
,
7423 union nfsd4_op_u
*u
)
7425 get_stateid(cstate
, &u
->read
.rd_stateid
);
7429 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
,
7430 union nfsd4_op_u
*u
)
7432 get_stateid(cstate
, &u
->write
.wr_stateid
);