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/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease
= 90; /* default lease time */
49 time_t nfsd4_grace
= 90;
50 static time_t boot_time
;
51 static u32 current_ownerid
= 1;
52 static u32 current_fileid
= 1;
53 static u32 current_delegid
= 1;
54 static stateid_t zerostateid
; /* bits all 0 */
55 static stateid_t onestateid
; /* bits all 1 */
56 static u64 current_sessionid
= 1;
58 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
61 /* forward declarations */
62 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
63 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
64 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
65 static void nfs4_set_recdir(char *recdir
);
69 /* Currently used for almost all code touching nfsv4 state: */
70 static DEFINE_MUTEX(client_mutex
);
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
77 static DEFINE_SPINLOCK(recall_lock
);
79 static struct kmem_cache
*stateowner_slab
= NULL
;
80 static struct kmem_cache
*file_slab
= NULL
;
81 static struct kmem_cache
*stateid_slab
= NULL
;
82 static struct kmem_cache
*deleg_slab
= NULL
;
87 mutex_lock(&client_mutex
);
91 nfs4_unlock_state(void)
93 mutex_unlock(&client_mutex
);
97 opaque_hashval(const void *ptr
, int nbytes
)
99 unsigned char *cptr
= (unsigned char *) ptr
;
109 static struct list_head del_recall_lru
;
112 put_nfs4_file(struct nfs4_file
*fi
)
114 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
115 list_del(&fi
->fi_hash
);
116 spin_unlock(&recall_lock
);
118 kmem_cache_free(file_slab
, fi
);
123 get_nfs4_file(struct nfs4_file
*fi
)
125 atomic_inc(&fi
->fi_ref
);
128 static int num_delegations
;
129 unsigned int max_delegations
;
132 * Open owner state (share locks)
135 /* hash tables for nfs4_stateowner */
136 #define OWNER_HASH_BITS 8
137 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140 #define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142 #define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
146 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
148 /* hash table for nfs4_file */
149 #define FILE_HASH_BITS 8
150 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152 /* hash table for (open)nfs4_stateid */
153 #define STATEID_HASH_BITS 10
154 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157 #define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159 #define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
163 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
165 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
167 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
168 atomic_inc(&fp
->fi_access
[oflag
]);
171 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
173 if (oflag
== O_RDWR
) {
174 __nfs4_file_get_access(fp
, O_RDONLY
);
175 __nfs4_file_get_access(fp
, O_WRONLY
);
177 __nfs4_file_get_access(fp
, oflag
);
180 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
182 if (fp
->fi_fds
[oflag
]) {
183 fput(fp
->fi_fds
[oflag
]);
184 fp
->fi_fds
[oflag
] = NULL
;
188 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
190 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
191 nfs4_file_put_fd(fp
, O_RDWR
);
192 nfs4_file_put_fd(fp
, oflag
);
196 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
198 if (oflag
== O_RDWR
) {
199 __nfs4_file_put_access(fp
, O_RDONLY
);
200 __nfs4_file_put_access(fp
, O_WRONLY
);
202 __nfs4_file_put_access(fp
, oflag
);
205 static struct nfs4_delegation
*
206 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
208 struct nfs4_delegation
*dp
;
209 struct nfs4_file
*fp
= stp
->st_file
;
211 dprintk("NFSD alloc_init_deleg\n");
213 * Major work on the lease subsystem (for example, to support
214 * calbacks on stat) will be required before we can support
215 * write delegations properly.
217 if (type
!= NFS4_OPEN_DELEGATE_READ
)
219 if (fp
->fi_had_conflict
)
221 if (num_delegations
> max_delegations
)
223 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
227 INIT_LIST_HEAD(&dp
->dl_perfile
);
228 INIT_LIST_HEAD(&dp
->dl_perclnt
);
229 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
234 dp
->dl_stateid
.si_boot
= boot_time
;
235 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
236 dp
->dl_stateid
.si_fileid
= 0;
237 dp
->dl_stateid
.si_generation
= 0;
238 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
240 atomic_set(&dp
->dl_count
, 1);
241 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
246 nfs4_put_delegation(struct nfs4_delegation
*dp
)
248 if (atomic_dec_and_test(&dp
->dl_count
)) {
249 dprintk("NFSD: freeing dp %p\n",dp
);
250 put_nfs4_file(dp
->dl_file
);
251 kmem_cache_free(deleg_slab
, dp
);
256 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
258 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
259 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
261 fp
->fi_deleg_file
= NULL
;
265 /* Called under the state lock. */
267 unhash_delegation(struct nfs4_delegation
*dp
)
269 list_del_init(&dp
->dl_perclnt
);
270 spin_lock(&recall_lock
);
271 list_del_init(&dp
->dl_perfile
);
272 list_del_init(&dp
->dl_recall_lru
);
273 spin_unlock(&recall_lock
);
274 nfs4_put_deleg_lease(dp
->dl_file
);
275 nfs4_put_delegation(dp
);
282 /* client_lock protects the client lru list and session hash table */
283 static DEFINE_SPINLOCK(client_lock
);
285 /* Hash tables for nfs4_clientid state */
286 #define CLIENT_HASH_BITS 4
287 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
288 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
290 #define clientid_hashval(id) \
291 ((id) & CLIENT_HASH_MASK)
292 #define clientstr_hashval(name) \
293 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
295 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
296 * used in reboot/reset lease grace period processing
298 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
299 * setclientid_confirmed info.
301 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
304 * client_lru holds client queue ordered by nfs4_client.cl_time
307 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
308 * for last close replay.
310 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
311 static int reclaim_str_hashtbl_size
= 0;
312 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
313 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
314 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
315 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
316 static struct list_head client_lru
;
317 static struct list_head close_lru
;
320 * We store the NONE, READ, WRITE, and BOTH bits separately in the
321 * st_{access,deny}_bmap field of the stateid, in order to track not
322 * only what share bits are currently in force, but also what
323 * combinations of share bits previous opens have used. This allows us
324 * to enforce the recommendation of rfc 3530 14.2.19 that the server
325 * return an error if the client attempt to downgrade to a combination
326 * of share bits not explicable by closing some of its previous opens.
328 * XXX: This enforcement is actually incomplete, since we don't keep
329 * track of access/deny bit combinations; so, e.g., we allow:
331 * OPEN allow read, deny write
332 * OPEN allow both, deny none
333 * DOWNGRADE allow read, deny none
335 * which we should reject.
338 set_access(unsigned int *access
, unsigned long bmap
) {
342 for (i
= 1; i
< 4; i
++) {
343 if (test_bit(i
, &bmap
))
349 set_deny(unsigned int *deny
, unsigned long bmap
) {
353 for (i
= 0; i
< 4; i
++) {
354 if (test_bit(i
, &bmap
))
360 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
361 unsigned int access
, deny
;
363 set_access(&access
, stp
->st_access_bmap
);
364 set_deny(&deny
, stp
->st_deny_bmap
);
365 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
370 static int nfs4_access_to_omode(u32 access
)
372 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
373 case NFS4_SHARE_ACCESS_READ
:
375 case NFS4_SHARE_ACCESS_WRITE
:
377 case NFS4_SHARE_ACCESS_BOTH
:
383 static int nfs4_access_bmap_to_omode(struct nfs4_stateid
*stp
)
387 set_access(&access
, stp
->st_access_bmap
);
388 return nfs4_access_to_omode(access
);
391 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
393 list_del(&stp
->st_hash
);
394 list_del(&stp
->st_perfile
);
395 list_del(&stp
->st_perstateowner
);
398 static void free_generic_stateid(struct nfs4_stateid
*stp
)
400 int oflag
= nfs4_access_bmap_to_omode(stp
);
402 nfs4_file_put_access(stp
->st_file
, oflag
);
403 put_nfs4_file(stp
->st_file
);
404 kmem_cache_free(stateid_slab
, stp
);
407 static void release_lock_stateid(struct nfs4_stateid
*stp
)
411 unhash_generic_stateid(stp
);
412 file
= find_any_file(stp
->st_file
);
414 locks_remove_posix(file
, (fl_owner_t
)stp
->st_stateowner
);
415 free_generic_stateid(stp
);
418 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
420 struct nfs4_stateid
*stp
;
422 list_del(&sop
->so_idhash
);
423 list_del(&sop
->so_strhash
);
424 list_del(&sop
->so_perstateid
);
425 while (!list_empty(&sop
->so_stateids
)) {
426 stp
= list_first_entry(&sop
->so_stateids
,
427 struct nfs4_stateid
, st_perstateowner
);
428 release_lock_stateid(stp
);
432 static void release_lockowner(struct nfs4_stateowner
*sop
)
434 unhash_lockowner(sop
);
435 nfs4_put_stateowner(sop
);
439 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
441 struct nfs4_stateowner
*lock_sop
;
443 while (!list_empty(&open_stp
->st_lockowners
)) {
444 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
445 struct nfs4_stateowner
, so_perstateid
);
446 /* list_del(&open_stp->st_lockowners); */
447 BUG_ON(lock_sop
->so_is_open_owner
);
448 release_lockowner(lock_sop
);
452 static void release_open_stateid(struct nfs4_stateid
*stp
)
454 unhash_generic_stateid(stp
);
455 release_stateid_lockowners(stp
);
456 free_generic_stateid(stp
);
459 static void unhash_openowner(struct nfs4_stateowner
*sop
)
461 struct nfs4_stateid
*stp
;
463 list_del(&sop
->so_idhash
);
464 list_del(&sop
->so_strhash
);
465 list_del(&sop
->so_perclient
);
466 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
467 while (!list_empty(&sop
->so_stateids
)) {
468 stp
= list_first_entry(&sop
->so_stateids
,
469 struct nfs4_stateid
, st_perstateowner
);
470 release_open_stateid(stp
);
474 static void release_openowner(struct nfs4_stateowner
*sop
)
476 unhash_openowner(sop
);
477 list_del(&sop
->so_close_lru
);
478 nfs4_put_stateowner(sop
);
481 #define SESSION_HASH_SIZE 512
482 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
485 hash_sessionid(struct nfs4_sessionid
*sessionid
)
487 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
489 return sid
->sequence
% SESSION_HASH_SIZE
;
493 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
495 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
496 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
500 gen_sessionid(struct nfsd4_session
*ses
)
502 struct nfs4_client
*clp
= ses
->se_client
;
503 struct nfsd4_sessionid
*sid
;
505 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
506 sid
->clientid
= clp
->cl_clientid
;
507 sid
->sequence
= current_sessionid
++;
512 * The protocol defines ca_maxresponssize_cached to include the size of
513 * the rpc header, but all we need to cache is the data starting after
514 * the end of the initial SEQUENCE operation--the rest we regenerate
515 * each time. Therefore we can advertise a ca_maxresponssize_cached
516 * value that is the number of bytes in our cache plus a few additional
517 * bytes. In order to stay on the safe side, and not promise more than
518 * we can cache, those additional bytes must be the minimum possible: 24
519 * bytes of rpc header (xid through accept state, with AUTH_NULL
520 * verifier), 12 for the compound header (with zero-length tag), and 44
521 * for the SEQUENCE op response:
523 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
526 free_session_slots(struct nfsd4_session
*ses
)
530 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
531 kfree(ses
->se_slots
[i
]);
535 * We don't actually need to cache the rpc and session headers, so we
536 * can allocate a little less for each slot:
538 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
540 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
543 static int nfsd4_sanitize_slot_size(u32 size
)
545 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
546 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
552 * XXX: If we run out of reserved DRC memory we could (up to a point)
553 * re-negotiate active sessions and reduce their slot usage to make
554 * rooom for new connections. For now we just fail the create session.
556 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
560 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
562 spin_lock(&nfsd_drc_lock
);
563 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
564 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
565 num
= min_t(int, num
, avail
/ slotsize
);
566 nfsd_drc_mem_used
+= num
* slotsize
;
567 spin_unlock(&nfsd_drc_lock
);
572 static void nfsd4_put_drc_mem(int slotsize
, int num
)
574 spin_lock(&nfsd_drc_lock
);
575 nfsd_drc_mem_used
-= slotsize
* num
;
576 spin_unlock(&nfsd_drc_lock
);
579 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
581 struct nfsd4_session
*new;
584 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
585 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
586 mem
= numslots
* sizeof(struct nfsd4_slot
*);
588 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
591 /* allocate each struct nfsd4_slot and data cache in one piece */
592 for (i
= 0; i
< numslots
; i
++) {
593 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
594 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
595 if (!new->se_slots
[i
])
601 kfree(new->se_slots
[i
]);
606 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
608 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
610 new->maxreqs
= numslots
;
611 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
612 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
613 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
614 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
615 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
618 static void free_conn(struct nfsd4_conn
*c
)
620 svc_xprt_put(c
->cn_xprt
);
624 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
626 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
627 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
629 spin_lock(&clp
->cl_lock
);
630 if (!list_empty(&c
->cn_persession
)) {
631 list_del(&c
->cn_persession
);
634 spin_unlock(&clp
->cl_lock
);
635 nfsd4_probe_callback(clp
);
638 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
640 struct nfsd4_conn
*conn
;
642 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
645 svc_xprt_get(rqstp
->rq_xprt
);
646 conn
->cn_xprt
= rqstp
->rq_xprt
;
647 conn
->cn_flags
= flags
;
648 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
652 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
654 conn
->cn_session
= ses
;
655 list_add(&conn
->cn_persession
, &ses
->se_conns
);
658 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
660 struct nfs4_client
*clp
= ses
->se_client
;
662 spin_lock(&clp
->cl_lock
);
663 __nfsd4_hash_conn(conn
, ses
);
664 spin_unlock(&clp
->cl_lock
);
667 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
669 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
670 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
673 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
675 struct nfsd4_conn
*conn
;
678 conn
= alloc_conn(rqstp
, dir
);
680 return nfserr_jukebox
;
681 nfsd4_hash_conn(conn
, ses
);
682 ret
= nfsd4_register_conn(conn
);
684 /* oops; xprt is already down: */
685 nfsd4_conn_lost(&conn
->cn_xpt_user
);
689 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
691 u32 dir
= NFS4_CDFC4_FORE
;
693 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
694 dir
|= NFS4_CDFC4_BACK
;
696 return nfsd4_new_conn(rqstp
, ses
, dir
);
699 /* must be called under client_lock */
700 static void nfsd4_del_conns(struct nfsd4_session
*s
)
702 struct nfs4_client
*clp
= s
->se_client
;
703 struct nfsd4_conn
*c
;
705 spin_lock(&clp
->cl_lock
);
706 while (!list_empty(&s
->se_conns
)) {
707 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
708 list_del_init(&c
->cn_persession
);
709 spin_unlock(&clp
->cl_lock
);
711 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
714 spin_lock(&clp
->cl_lock
);
716 spin_unlock(&clp
->cl_lock
);
719 void free_session(struct kref
*kref
)
721 struct nfsd4_session
*ses
;
724 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
725 nfsd4_del_conns(ses
);
726 spin_lock(&nfsd_drc_lock
);
727 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
728 nfsd_drc_mem_used
-= mem
;
729 spin_unlock(&nfsd_drc_lock
);
730 free_session_slots(ses
);
734 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
736 struct nfsd4_session
*new;
737 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
738 int numslots
, slotsize
;
743 * Note decreasing slot size below client's request may
744 * make it difficult for client to function correctly, whereas
745 * decreasing the number of slots will (just?) affect
746 * performance. When short on memory we therefore prefer to
747 * decrease number of slots instead of their size.
749 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
750 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
754 new = alloc_session(slotsize
, numslots
);
756 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
759 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
761 new->se_client
= clp
;
764 INIT_LIST_HEAD(&new->se_conns
);
766 new->se_cb_seq_nr
= 1;
767 new->se_flags
= cses
->flags
;
768 new->se_cb_prog
= cses
->callback_prog
;
769 kref_init(&new->se_ref
);
770 idx
= hash_sessionid(&new->se_sessionid
);
771 spin_lock(&client_lock
);
772 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
773 spin_lock(&clp
->cl_lock
);
774 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
775 spin_unlock(&clp
->cl_lock
);
776 spin_unlock(&client_lock
);
778 status
= nfsd4_new_conn_from_crses(rqstp
, new);
779 /* whoops: benny points out, status is ignored! (err, or bogus) */
781 free_session(&new->se_ref
);
784 if (cses
->flags
& SESSION4_BACK_CHAN
) {
785 struct sockaddr
*sa
= svc_addr(rqstp
);
787 * This is a little silly; with sessions there's no real
788 * use for the callback address. Use the peer address
789 * as a reasonable default for now, but consider fixing
790 * the rpc client not to require an address in the
793 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
794 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
796 nfsd4_probe_callback(clp
);
800 /* caller must hold client_lock */
801 static struct nfsd4_session
*
802 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
804 struct nfsd4_session
*elem
;
807 dump_sessionid(__func__
, sessionid
);
808 idx
= hash_sessionid(sessionid
);
809 /* Search in the appropriate list */
810 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
811 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
812 NFS4_MAX_SESSIONID_LEN
)) {
817 dprintk("%s: session not found\n", __func__
);
821 /* caller must hold client_lock */
823 unhash_session(struct nfsd4_session
*ses
)
825 list_del(&ses
->se_hash
);
826 spin_lock(&ses
->se_client
->cl_lock
);
827 list_del(&ses
->se_perclnt
);
828 spin_unlock(&ses
->se_client
->cl_lock
);
831 /* must be called under the client_lock */
833 renew_client_locked(struct nfs4_client
*clp
)
835 if (is_client_expired(clp
)) {
836 dprintk("%s: client (clientid %08x/%08x) already expired\n",
838 clp
->cl_clientid
.cl_boot
,
839 clp
->cl_clientid
.cl_id
);
844 * Move client to the end to the LRU list.
846 dprintk("renewing client (clientid %08x/%08x)\n",
847 clp
->cl_clientid
.cl_boot
,
848 clp
->cl_clientid
.cl_id
);
849 list_move_tail(&clp
->cl_lru
, &client_lru
);
850 clp
->cl_time
= get_seconds();
854 renew_client(struct nfs4_client
*clp
)
856 spin_lock(&client_lock
);
857 renew_client_locked(clp
);
858 spin_unlock(&client_lock
);
861 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
863 STALE_CLIENTID(clientid_t
*clid
)
865 if (clid
->cl_boot
== boot_time
)
867 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
868 clid
->cl_boot
, clid
->cl_id
, boot_time
);
873 * XXX Should we use a slab cache ?
874 * This type of memory management is somewhat inefficient, but we use it
875 * anyway since SETCLIENTID is not a common operation.
877 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
879 struct nfs4_client
*clp
;
881 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
884 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
885 if (clp
->cl_name
.data
== NULL
) {
889 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
890 clp
->cl_name
.len
= name
.len
;
895 free_client(struct nfs4_client
*clp
)
897 while (!list_empty(&clp
->cl_sessions
)) {
898 struct nfsd4_session
*ses
;
899 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
901 list_del(&ses
->se_perclnt
);
902 nfsd4_put_session(ses
);
904 if (clp
->cl_cred
.cr_group_info
)
905 put_group_info(clp
->cl_cred
.cr_group_info
);
906 kfree(clp
->cl_principal
);
907 kfree(clp
->cl_name
.data
);
912 release_session_client(struct nfsd4_session
*session
)
914 struct nfs4_client
*clp
= session
->se_client
;
916 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
918 if (is_client_expired(clp
)) {
920 session
->se_client
= NULL
;
922 renew_client_locked(clp
);
923 spin_unlock(&client_lock
);
926 /* must be called under the client_lock */
928 unhash_client_locked(struct nfs4_client
*clp
)
930 struct nfsd4_session
*ses
;
932 mark_client_expired(clp
);
933 list_del(&clp
->cl_lru
);
934 spin_lock(&clp
->cl_lock
);
935 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
936 list_del_init(&ses
->se_hash
);
937 spin_unlock(&clp
->cl_lock
);
941 expire_client(struct nfs4_client
*clp
)
943 struct nfs4_stateowner
*sop
;
944 struct nfs4_delegation
*dp
;
945 struct list_head reaplist
;
947 INIT_LIST_HEAD(&reaplist
);
948 spin_lock(&recall_lock
);
949 while (!list_empty(&clp
->cl_delegations
)) {
950 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
951 list_del_init(&dp
->dl_perclnt
);
952 list_move(&dp
->dl_recall_lru
, &reaplist
);
954 spin_unlock(&recall_lock
);
955 while (!list_empty(&reaplist
)) {
956 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
957 list_del_init(&dp
->dl_recall_lru
);
958 unhash_delegation(dp
);
960 while (!list_empty(&clp
->cl_openowners
)) {
961 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
962 release_openowner(sop
);
964 nfsd4_shutdown_callback(clp
);
965 if (clp
->cl_cb_conn
.cb_xprt
)
966 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
967 list_del(&clp
->cl_idhash
);
968 list_del(&clp
->cl_strhash
);
969 spin_lock(&client_lock
);
970 unhash_client_locked(clp
);
971 if (atomic_read(&clp
->cl_refcount
) == 0)
973 spin_unlock(&client_lock
);
976 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
978 memcpy(target
->cl_verifier
.data
, source
->data
,
979 sizeof(target
->cl_verifier
.data
));
982 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
984 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
985 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
988 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
990 target
->cr_uid
= source
->cr_uid
;
991 target
->cr_gid
= source
->cr_gid
;
992 target
->cr_group_info
= source
->cr_group_info
;
993 get_group_info(target
->cr_group_info
);
996 static int same_name(const char *n1
, const char *n2
)
998 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1002 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1004 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1008 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1010 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1013 /* XXX what about NGROUP */
1015 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1017 return cr1
->cr_uid
== cr2
->cr_uid
;
1020 static void gen_clid(struct nfs4_client
*clp
)
1022 static u32 current_clientid
= 1;
1024 clp
->cl_clientid
.cl_boot
= boot_time
;
1025 clp
->cl_clientid
.cl_id
= current_clientid
++;
1028 static void gen_confirm(struct nfs4_client
*clp
)
1033 p
= (u32
*)clp
->cl_confirm
.data
;
1034 *p
++ = get_seconds();
1038 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1039 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1041 struct nfs4_client
*clp
;
1042 struct sockaddr
*sa
= svc_addr(rqstp
);
1045 clp
= alloc_client(name
);
1049 INIT_LIST_HEAD(&clp
->cl_sessions
);
1051 princ
= svc_gss_principal(rqstp
);
1053 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1054 if (clp
->cl_principal
== NULL
) {
1060 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1061 atomic_set(&clp
->cl_refcount
, 0);
1062 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1063 INIT_LIST_HEAD(&clp
->cl_idhash
);
1064 INIT_LIST_HEAD(&clp
->cl_strhash
);
1065 INIT_LIST_HEAD(&clp
->cl_openowners
);
1066 INIT_LIST_HEAD(&clp
->cl_delegations
);
1067 INIT_LIST_HEAD(&clp
->cl_lru
);
1068 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1069 spin_lock_init(&clp
->cl_lock
);
1070 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1071 clp
->cl_time
= get_seconds();
1072 clear_bit(0, &clp
->cl_cb_slot_busy
);
1073 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1074 copy_verf(clp
, verf
);
1075 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1076 clp
->cl_flavor
= rqstp
->rq_flavor
;
1077 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1079 clp
->cl_cb_session
= NULL
;
1083 static int check_name(struct xdr_netobj name
)
1087 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
1088 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
1095 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1097 unsigned int idhashval
;
1099 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1100 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1101 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1106 move_to_confirmed(struct nfs4_client
*clp
)
1108 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1109 unsigned int strhashval
;
1111 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1112 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1113 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1114 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1118 static struct nfs4_client
*
1119 find_confirmed_client(clientid_t
*clid
)
1121 struct nfs4_client
*clp
;
1122 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1124 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1125 if (same_clid(&clp
->cl_clientid
, clid
))
1131 static struct nfs4_client
*
1132 find_unconfirmed_client(clientid_t
*clid
)
1134 struct nfs4_client
*clp
;
1135 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1137 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1138 if (same_clid(&clp
->cl_clientid
, clid
))
1144 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1146 return clp
->cl_exchange_flags
!= 0;
1149 static struct nfs4_client
*
1150 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1152 struct nfs4_client
*clp
;
1154 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1155 if (same_name(clp
->cl_recdir
, dname
))
1161 static struct nfs4_client
*
1162 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1164 struct nfs4_client
*clp
;
1166 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1167 if (same_name(clp
->cl_recdir
, dname
))
1173 static void rpc_svcaddr2sockaddr(struct sockaddr
*sa
, unsigned short family
, union svc_addr_u
*svcaddr
)
1177 ((struct sockaddr_in
*)sa
)->sin_family
= AF_INET
;
1178 ((struct sockaddr_in
*)sa
)->sin_addr
= svcaddr
->addr
;
1181 ((struct sockaddr_in6
*)sa
)->sin6_family
= AF_INET6
;
1182 ((struct sockaddr_in6
*)sa
)->sin6_addr
= svcaddr
->addr6
;
1188 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1190 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1191 struct sockaddr
*sa
= svc_addr(rqstp
);
1192 u32 scopeid
= rpc_get_scope_id(sa
);
1193 unsigned short expected_family
;
1195 /* Currently, we only support tcp and tcp6 for the callback channel */
1196 if (se
->se_callback_netid_len
== 3 &&
1197 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1198 expected_family
= AF_INET
;
1199 else if (se
->se_callback_netid_len
== 4 &&
1200 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1201 expected_family
= AF_INET6
;
1205 conn
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
1206 se
->se_callback_addr_len
,
1207 (struct sockaddr
*)&conn
->cb_addr
,
1208 sizeof(conn
->cb_addr
));
1210 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1213 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1214 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1216 conn
->cb_prog
= se
->se_callback_prog
;
1217 conn
->cb_ident
= se
->se_callback_ident
;
1218 rpc_svcaddr2sockaddr((struct sockaddr
*)&conn
->cb_saddr
, expected_family
, &rqstp
->rq_daddr
);
1221 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1222 conn
->cb_addrlen
= 0;
1223 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1224 "will not receive delegations\n",
1225 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1231 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1234 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1236 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1239 dprintk("--> %s slot %p\n", __func__
, slot
);
1241 slot
->sl_opcnt
= resp
->opcnt
;
1242 slot
->sl_status
= resp
->cstate
.status
;
1244 if (nfsd4_not_cached(resp
)) {
1245 slot
->sl_datalen
= 0;
1248 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1249 base
= (char *)resp
->cstate
.datap
-
1250 (char *)resp
->xbuf
->head
[0].iov_base
;
1251 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1253 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1258 * Encode the replay sequence operation from the slot values.
1259 * If cachethis is FALSE encode the uncached rep error on the next
1260 * operation which sets resp->p and increments resp->opcnt for
1261 * nfs4svc_encode_compoundres.
1265 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1266 struct nfsd4_compoundres
*resp
)
1268 struct nfsd4_op
*op
;
1269 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1271 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1272 resp
->opcnt
, resp
->cstate
.slot
->sl_cachethis
);
1274 /* Encode the replayed sequence operation */
1275 op
= &args
->ops
[resp
->opcnt
- 1];
1276 nfsd4_encode_operation(resp
, op
);
1278 /* Return nfserr_retry_uncached_rep in next operation. */
1279 if (args
->opcnt
> 1 && slot
->sl_cachethis
== 0) {
1280 op
= &args
->ops
[resp
->opcnt
++];
1281 op
->status
= nfserr_retry_uncached_rep
;
1282 nfsd4_encode_operation(resp
, op
);
1288 * The sequence operation is not cached because we can use the slot and
1292 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1293 struct nfsd4_sequence
*seq
)
1295 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1298 dprintk("--> %s slot %p\n", __func__
, slot
);
1300 /* Either returns 0 or nfserr_retry_uncached */
1301 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1302 if (status
== nfserr_retry_uncached_rep
)
1305 /* The sequence operation has been encoded, cstate->datap set. */
1306 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1308 resp
->opcnt
= slot
->sl_opcnt
;
1309 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1310 status
= slot
->sl_status
;
1316 * Set the exchange_id flags returned by the server.
1319 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1321 /* pNFS is not supported */
1322 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1324 /* Referrals are supported, Migration is not. */
1325 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1327 /* set the wire flags to return to client. */
1328 clid
->flags
= new->cl_exchange_flags
;
1332 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1333 struct nfsd4_compound_state
*cstate
,
1334 struct nfsd4_exchange_id
*exid
)
1336 struct nfs4_client
*unconf
, *conf
, *new;
1338 unsigned int strhashval
;
1339 char dname
[HEXDIR_LEN
];
1340 char addr_str
[INET6_ADDRSTRLEN
];
1341 nfs4_verifier verf
= exid
->verifier
;
1342 struct sockaddr
*sa
= svc_addr(rqstp
);
1344 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1345 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1346 "ip_addr=%s flags %x, spa_how %d\n",
1347 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1348 addr_str
, exid
->flags
, exid
->spa_how
);
1350 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1351 return nfserr_inval
;
1353 /* Currently only support SP4_NONE */
1354 switch (exid
->spa_how
) {
1358 return nfserr_serverfault
;
1360 BUG(); /* checked by xdr code */
1362 return nfserr_serverfault
; /* no excuse :-/ */
1365 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1370 strhashval
= clientstr_hashval(dname
);
1375 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1377 if (!clp_used_exchangeid(conf
)) {
1378 status
= nfserr_clid_inuse
; /* XXX: ? */
1381 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1382 /* 18.35.4 case 8 */
1383 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1384 status
= nfserr_not_same
;
1387 /* Client reboot: destroy old state */
1388 expire_client(conf
);
1391 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1392 /* 18.35.4 case 9 */
1393 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1394 status
= nfserr_perm
;
1397 expire_client(conf
);
1401 * Set bit when the owner id and verifier map to an already
1402 * confirmed client id (18.35.3).
1404 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1407 * Falling into 18.35.4 case 2, possible router replay.
1408 * Leave confirmed record intact and return same result.
1410 copy_verf(conf
, &verf
);
1415 /* 18.35.4 case 7 */
1416 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1417 status
= nfserr_noent
;
1421 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1424 * Possible retry or client restart. Per 18.35.4 case 4,
1425 * a new unconfirmed record should be generated regardless
1426 * of whether any properties have changed.
1428 expire_client(unconf
);
1433 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1435 status
= nfserr_jukebox
;
1440 add_to_unconfirmed(new, strhashval
);
1442 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1443 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1446 nfsd4_set_ex_flags(new, exid
);
1448 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1449 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1453 nfs4_unlock_state();
1455 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1460 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1462 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1465 /* The slot is in use, and no response has been sent. */
1467 if (seqid
== slot_seqid
)
1468 return nfserr_jukebox
;
1470 return nfserr_seq_misordered
;
1473 if (likely(seqid
== slot_seqid
+ 1))
1476 if (seqid
== slot_seqid
)
1477 return nfserr_replay_cache
;
1479 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1481 /* Misordered replay or misordered new request */
1482 return nfserr_seq_misordered
;
1486 * Cache the create session result into the create session single DRC
1487 * slot cache by saving the xdr structure. sl_seqid has been set.
1488 * Do this for solo or embedded create session operations.
1491 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1492 struct nfsd4_clid_slot
*slot
, int nfserr
)
1494 slot
->sl_status
= nfserr
;
1495 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1499 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1500 struct nfsd4_clid_slot
*slot
)
1502 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1503 return slot
->sl_status
;
1507 nfsd4_create_session(struct svc_rqst
*rqstp
,
1508 struct nfsd4_compound_state
*cstate
,
1509 struct nfsd4_create_session
*cr_ses
)
1511 struct sockaddr
*sa
= svc_addr(rqstp
);
1512 struct nfs4_client
*conf
, *unconf
;
1513 struct nfsd4_session
*new;
1514 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1515 bool confirm_me
= false;
1518 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1519 return nfserr_inval
;
1522 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1523 conf
= find_confirmed_client(&cr_ses
->clientid
);
1526 cs_slot
= &conf
->cl_cs_slot
;
1527 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1528 if (status
== nfserr_replay_cache
) {
1529 dprintk("Got a create_session replay! seqid= %d\n",
1531 /* Return the cached reply status */
1532 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1534 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1535 status
= nfserr_seq_misordered
;
1536 dprintk("Sequence misordered!\n");
1537 dprintk("Expected seqid= %d but got seqid= %d\n",
1538 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1541 } else if (unconf
) {
1542 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1543 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1544 status
= nfserr_clid_inuse
;
1548 cs_slot
= &unconf
->cl_cs_slot
;
1549 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1551 /* an unconfirmed replay returns misordered */
1552 status
= nfserr_seq_misordered
;
1559 status
= nfserr_stale_clientid
;
1564 * XXX: we should probably set this at creation time, and check
1565 * for consistent minorversion use throughout:
1567 conf
->cl_minorversion
= 1;
1569 * We do not support RDMA or persistent sessions
1571 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1572 cr_ses
->flags
&= ~SESSION4_RDMA
;
1574 status
= nfserr_jukebox
;
1575 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1579 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1580 NFS4_MAX_SESSIONID_LEN
);
1581 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1582 sizeof(struct nfsd4_channel_attrs
));
1583 cs_slot
->sl_seqid
++;
1584 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1586 /* cache solo and embedded create sessions under the state lock */
1587 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1589 move_to_confirmed(conf
);
1591 nfs4_unlock_state();
1592 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1596 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1598 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1599 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1601 return argp
->opcnt
== resp
->opcnt
;
1604 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1607 case NFS4_CDFC4_FORE
:
1608 case NFS4_CDFC4_BACK
:
1610 case NFS4_CDFC4_FORE_OR_BOTH
:
1611 case NFS4_CDFC4_BACK_OR_BOTH
:
1612 *dir
= NFS4_CDFC4_BOTH
;
1615 return nfserr_inval
;
1618 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1619 struct nfsd4_compound_state
*cstate
,
1620 struct nfsd4_bind_conn_to_session
*bcts
)
1624 if (!nfsd4_last_compound_op(rqstp
))
1625 return nfserr_not_only_op
;
1626 spin_lock(&client_lock
);
1627 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1628 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1629 * client_lock iself: */
1630 if (cstate
->session
) {
1631 nfsd4_get_session(cstate
->session
);
1632 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1634 spin_unlock(&client_lock
);
1635 if (!cstate
->session
)
1636 return nfserr_badsession
;
1638 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1640 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1644 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1648 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1652 nfsd4_destroy_session(struct svc_rqst
*r
,
1653 struct nfsd4_compound_state
*cstate
,
1654 struct nfsd4_destroy_session
*sessionid
)
1656 struct nfsd4_session
*ses
;
1657 u32 status
= nfserr_badsession
;
1660 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1661 * - Should we return nfserr_back_chan_busy if waiting for
1662 * callbacks on to-be-destroyed session?
1663 * - Do we need to clear any callback info from previous session?
1666 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1667 if (!nfsd4_last_compound_op(r
))
1668 return nfserr_not_only_op
;
1670 dump_sessionid(__func__
, &sessionid
->sessionid
);
1671 spin_lock(&client_lock
);
1672 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1674 spin_unlock(&client_lock
);
1678 unhash_session(ses
);
1679 spin_unlock(&client_lock
);
1682 nfsd4_probe_callback_sync(ses
->se_client
);
1683 nfs4_unlock_state();
1685 nfsd4_del_conns(ses
);
1687 nfsd4_put_session(ses
);
1690 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1694 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1696 struct nfsd4_conn
*c
;
1698 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1699 if (c
->cn_xprt
== xpt
) {
1706 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1708 struct nfs4_client
*clp
= ses
->se_client
;
1709 struct nfsd4_conn
*c
;
1712 spin_lock(&clp
->cl_lock
);
1713 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1715 spin_unlock(&clp
->cl_lock
);
1719 __nfsd4_hash_conn(new, ses
);
1720 spin_unlock(&clp
->cl_lock
);
1721 ret
= nfsd4_register_conn(new);
1723 /* oops; xprt is already down: */
1724 nfsd4_conn_lost(&new->cn_xpt_user
);
1728 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1730 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1732 return args
->opcnt
> session
->se_fchannel
.maxops
;
1736 nfsd4_sequence(struct svc_rqst
*rqstp
,
1737 struct nfsd4_compound_state
*cstate
,
1738 struct nfsd4_sequence
*seq
)
1740 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1741 struct nfsd4_session
*session
;
1742 struct nfsd4_slot
*slot
;
1743 struct nfsd4_conn
*conn
;
1746 if (resp
->opcnt
!= 1)
1747 return nfserr_sequence_pos
;
1750 * Will be either used or freed by nfsd4_sequence_check_conn
1753 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1755 return nfserr_jukebox
;
1757 spin_lock(&client_lock
);
1758 status
= nfserr_badsession
;
1759 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1763 status
= nfserr_too_many_ops
;
1764 if (nfsd4_session_too_many_ops(rqstp
, session
))
1767 status
= nfserr_badslot
;
1768 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1771 slot
= session
->se_slots
[seq
->slotid
];
1772 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1774 /* We do not negotiate the number of slots yet, so set the
1775 * maxslots to the session maxreqs which is used to encode
1776 * sr_highest_slotid and the sr_target_slot id to maxslots */
1777 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1779 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1780 if (status
== nfserr_replay_cache
) {
1781 cstate
->slot
= slot
;
1782 cstate
->session
= session
;
1783 /* Return the cached reply status and set cstate->status
1784 * for nfsd4_proc_compound processing */
1785 status
= nfsd4_replay_cache_entry(resp
, seq
);
1786 cstate
->status
= nfserr_replay_cache
;
1792 nfsd4_sequence_check_conn(conn
, session
);
1795 /* Success! bump slot seqid */
1796 slot
->sl_inuse
= true;
1797 slot
->sl_seqid
= seq
->seqid
;
1798 slot
->sl_cachethis
= seq
->cachethis
;
1800 cstate
->slot
= slot
;
1801 cstate
->session
= session
;
1804 /* Hold a session reference until done processing the compound. */
1805 if (cstate
->session
) {
1806 struct nfs4_client
*clp
= session
->se_client
;
1808 nfsd4_get_session(cstate
->session
);
1809 atomic_inc(&clp
->cl_refcount
);
1810 if (clp
->cl_cb_state
== NFSD4_CB_DOWN
)
1811 seq
->status_flags
|= SEQ4_STATUS_CB_PATH_DOWN
;
1814 spin_unlock(&client_lock
);
1815 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1820 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
1824 if (rc
->rca_one_fs
) {
1825 if (!cstate
->current_fh
.fh_dentry
)
1826 return nfserr_nofilehandle
;
1828 * We don't take advantage of the rca_one_fs case.
1829 * That's OK, it's optional, we can safely ignore it.
1835 status
= nfserr_complete_already
;
1836 if (cstate
->session
->se_client
->cl_firststate
)
1839 status
= nfserr_stale_clientid
;
1840 if (is_client_expired(cstate
->session
->se_client
))
1842 * The following error isn't really legal.
1843 * But we only get here if the client just explicitly
1844 * destroyed the client. Surely it no longer cares what
1845 * error it gets back on an operation for the dead
1851 nfsd4_create_clid_dir(cstate
->session
->se_client
);
1853 nfs4_unlock_state();
1858 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1859 struct nfsd4_setclientid
*setclid
)
1861 struct xdr_netobj clname
= {
1862 .len
= setclid
->se_namelen
,
1863 .data
= setclid
->se_name
,
1865 nfs4_verifier clverifier
= setclid
->se_verf
;
1866 unsigned int strhashval
;
1867 struct nfs4_client
*conf
, *unconf
, *new;
1869 char dname
[HEXDIR_LEN
];
1871 if (!check_name(clname
))
1872 return nfserr_inval
;
1874 status
= nfs4_make_rec_clidname(dname
, &clname
);
1879 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1880 * We get here on a DRC miss.
1883 strhashval
= clientstr_hashval(dname
);
1886 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1888 /* RFC 3530 14.2.33 CASE 0: */
1889 status
= nfserr_clid_inuse
;
1890 if (clp_used_exchangeid(conf
))
1892 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1893 char addr_str
[INET6_ADDRSTRLEN
];
1894 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1896 dprintk("NFSD: setclientid: string in use by client "
1897 "at %s\n", addr_str
);
1902 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1903 * has a description of SETCLIENTID request processing consisting
1904 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1906 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1907 status
= nfserr_resource
;
1910 * RFC 3530 14.2.33 CASE 4:
1911 * placed first, because it is the normal case
1914 expire_client(unconf
);
1915 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1919 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1921 * RFC 3530 14.2.33 CASE 1:
1922 * probable callback update
1925 /* Note this is removing unconfirmed {*x***},
1926 * which is stronger than RFC recommended {vxc**}.
1927 * This has the advantage that there is at most
1928 * one {*x***} in either list at any time.
1930 expire_client(unconf
);
1932 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1935 copy_clid(new, conf
);
1936 } else if (!unconf
) {
1938 * RFC 3530 14.2.33 CASE 2:
1939 * probable client reboot; state will be removed if
1942 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1948 * RFC 3530 14.2.33 CASE 3:
1949 * probable client reboot; state will be removed if
1952 expire_client(unconf
);
1953 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1959 * XXX: we should probably set this at creation time, and check
1960 * for consistent minorversion use throughout:
1962 new->cl_minorversion
= 0;
1963 gen_callback(new, setclid
, rqstp
);
1964 add_to_unconfirmed(new, strhashval
);
1965 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1966 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1967 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1970 nfs4_unlock_state();
1976 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1977 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1978 * bullets, labeled as CASE1 - CASE4 below.
1981 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1982 struct nfsd4_compound_state
*cstate
,
1983 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1985 struct sockaddr
*sa
= svc_addr(rqstp
);
1986 struct nfs4_client
*conf
, *unconf
;
1987 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1988 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1991 if (STALE_CLIENTID(clid
))
1992 return nfserr_stale_clientid
;
1994 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1995 * We get here on a DRC miss.
2000 conf
= find_confirmed_client(clid
);
2001 unconf
= find_unconfirmed_client(clid
);
2003 status
= nfserr_clid_inuse
;
2004 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
2006 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
2010 * section 14.2.34 of RFC 3530 has a description of
2011 * SETCLIENTID_CONFIRM request processing consisting
2012 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2014 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
2016 * RFC 3530 14.2.34 CASE 1:
2019 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
2020 status
= nfserr_clid_inuse
;
2022 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2023 nfsd4_probe_callback(conf
);
2024 expire_client(unconf
);
2028 } else if (conf
&& !unconf
) {
2030 * RFC 3530 14.2.34 CASE 2:
2031 * probable retransmitted request; play it safe and
2034 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2035 status
= nfserr_clid_inuse
;
2038 } else if (!conf
&& unconf
2039 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
2041 * RFC 3530 14.2.34 CASE 3:
2042 * Normal case; new or rebooted client:
2044 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
2045 status
= nfserr_clid_inuse
;
2048 clientstr_hashval(unconf
->cl_recdir
);
2049 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
2052 nfsd4_remove_clid_dir(conf
);
2053 expire_client(conf
);
2055 move_to_confirmed(unconf
);
2057 nfsd4_probe_callback(conf
);
2060 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
2061 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
2064 * RFC 3530 14.2.34 CASE 4:
2065 * Client probably hasn't noticed that we rebooted yet.
2067 status
= nfserr_stale_clientid
;
2069 /* check that we have hit one of the cases...*/
2070 status
= nfserr_clid_inuse
;
2073 nfs4_unlock_state();
2077 /* OPEN Share state helper functions */
2078 static inline struct nfs4_file
*
2079 alloc_init_file(struct inode
*ino
)
2081 struct nfs4_file
*fp
;
2082 unsigned int hashval
= file_hashval(ino
);
2084 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2086 atomic_set(&fp
->fi_ref
, 1);
2087 INIT_LIST_HEAD(&fp
->fi_hash
);
2088 INIT_LIST_HEAD(&fp
->fi_stateids
);
2089 INIT_LIST_HEAD(&fp
->fi_delegations
);
2090 fp
->fi_inode
= igrab(ino
);
2091 fp
->fi_id
= current_fileid
++;
2092 fp
->fi_had_conflict
= false;
2093 fp
->fi_lease
= NULL
;
2094 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2095 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2096 spin_lock(&recall_lock
);
2097 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2098 spin_unlock(&recall_lock
);
2105 nfsd4_free_slab(struct kmem_cache
**slab
)
2109 kmem_cache_destroy(*slab
);
2114 nfsd4_free_slabs(void)
2116 nfsd4_free_slab(&stateowner_slab
);
2117 nfsd4_free_slab(&file_slab
);
2118 nfsd4_free_slab(&stateid_slab
);
2119 nfsd4_free_slab(&deleg_slab
);
2123 nfsd4_init_slabs(void)
2125 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
2126 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
2127 if (stateowner_slab
== NULL
)
2129 file_slab
= kmem_cache_create("nfsd4_files",
2130 sizeof(struct nfs4_file
), 0, 0, NULL
);
2131 if (file_slab
== NULL
)
2133 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2134 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
2135 if (stateid_slab
== NULL
)
2137 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2138 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2139 if (deleg_slab
== NULL
)
2144 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2149 nfs4_free_stateowner(struct kref
*kref
)
2151 struct nfs4_stateowner
*sop
=
2152 container_of(kref
, struct nfs4_stateowner
, so_ref
);
2153 kfree(sop
->so_owner
.data
);
2154 kmem_cache_free(stateowner_slab
, sop
);
2157 static inline struct nfs4_stateowner
*
2158 alloc_stateowner(struct xdr_netobj
*owner
)
2160 struct nfs4_stateowner
*sop
;
2162 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
2163 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
2164 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
2165 sop
->so_owner
.len
= owner
->len
;
2166 kref_init(&sop
->so_ref
);
2169 kmem_cache_free(stateowner_slab
, sop
);
2174 static struct nfs4_stateowner
*
2175 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2176 struct nfs4_stateowner
*sop
;
2177 struct nfs4_replay
*rp
;
2178 unsigned int idhashval
;
2180 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
2182 idhashval
= ownerid_hashval(current_ownerid
);
2183 INIT_LIST_HEAD(&sop
->so_idhash
);
2184 INIT_LIST_HEAD(&sop
->so_strhash
);
2185 INIT_LIST_HEAD(&sop
->so_perclient
);
2186 INIT_LIST_HEAD(&sop
->so_stateids
);
2187 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
2188 INIT_LIST_HEAD(&sop
->so_close_lru
);
2190 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
2191 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2192 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
2193 sop
->so_is_open_owner
= 1;
2194 sop
->so_id
= current_ownerid
++;
2195 sop
->so_client
= clp
;
2196 sop
->so_seqid
= open
->op_seqid
;
2197 sop
->so_confirmed
= 0;
2198 rp
= &sop
->so_replay
;
2199 rp
->rp_status
= nfserr_serverfault
;
2201 rp
->rp_buf
= rp
->rp_ibuf
;
2206 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2207 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2208 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
2210 INIT_LIST_HEAD(&stp
->st_hash
);
2211 INIT_LIST_HEAD(&stp
->st_perstateowner
);
2212 INIT_LIST_HEAD(&stp
->st_lockowners
);
2213 INIT_LIST_HEAD(&stp
->st_perfile
);
2214 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
2215 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
2216 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2217 stp
->st_stateowner
= sop
;
2220 stp
->st_stateid
.si_boot
= boot_time
;
2221 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
2222 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
2223 stp
->st_stateid
.si_generation
= 0;
2224 stp
->st_access_bmap
= 0;
2225 stp
->st_deny_bmap
= 0;
2226 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
2227 &stp
->st_access_bmap
);
2228 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2229 stp
->st_openstp
= NULL
;
2233 move_to_close_lru(struct nfs4_stateowner
*sop
)
2235 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
2237 list_move_tail(&sop
->so_close_lru
, &close_lru
);
2238 sop
->so_time
= get_seconds();
2242 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2245 return (sop
->so_owner
.len
== owner
->len
) &&
2246 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2247 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2250 static struct nfs4_stateowner
*
2251 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2253 struct nfs4_stateowner
*so
= NULL
;
2255 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2256 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
2262 /* search file_hashtbl[] for file */
2263 static struct nfs4_file
*
2264 find_file(struct inode
*ino
)
2266 unsigned int hashval
= file_hashval(ino
);
2267 struct nfs4_file
*fp
;
2269 spin_lock(&recall_lock
);
2270 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2271 if (fp
->fi_inode
== ino
) {
2273 spin_unlock(&recall_lock
);
2277 spin_unlock(&recall_lock
);
2281 static inline int access_valid(u32 x
, u32 minorversion
)
2283 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
2285 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
2287 x
&= ~NFS4_SHARE_ACCESS_MASK
;
2288 if (minorversion
&& x
) {
2289 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
2291 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
2293 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
2300 static inline int deny_valid(u32 x
)
2302 /* Note: unlike access bits, deny bits may be zero. */
2303 return x
<= NFS4_SHARE_DENY_BOTH
;
2307 * Called to check deny when READ with all zero stateid or
2308 * WRITE with all zero or all one stateid
2311 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2313 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2314 struct nfs4_file
*fp
;
2315 struct nfs4_stateid
*stp
;
2318 dprintk("NFSD: nfs4_share_conflict\n");
2320 fp
= find_file(ino
);
2323 ret
= nfserr_locked
;
2324 /* Search for conflicting share reservations */
2325 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2326 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2327 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2337 nfs4_file_downgrade(struct nfs4_file
*fp
, unsigned int share_access
)
2339 if (share_access
& NFS4_SHARE_ACCESS_WRITE
)
2340 nfs4_file_put_access(fp
, O_WRONLY
);
2341 if (share_access
& NFS4_SHARE_ACCESS_READ
)
2342 nfs4_file_put_access(fp
, O_RDONLY
);
2345 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2347 /* We're assuming the state code never drops its reference
2348 * without first removing the lease. Since we're in this lease
2349 * callback (and since the lease code is serialized by the kernel
2350 * lock) we know the server hasn't removed the lease yet, we know
2351 * it's safe to take a reference: */
2352 atomic_inc(&dp
->dl_count
);
2354 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2356 /* only place dl_time is set. protected by lock_flocks*/
2357 dp
->dl_time
= get_seconds();
2359 nfsd4_cb_recall(dp
);
2362 /* Called from break_lease() with lock_flocks() held. */
2363 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2365 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2366 struct nfs4_delegation
*dp
;
2369 /* We assume break_lease is only called once per lease: */
2370 BUG_ON(fp
->fi_had_conflict
);
2372 * We don't want the locks code to timeout the lease for us;
2373 * we'll remove it ourself if a delegation isn't returned
2376 fl
->fl_break_time
= 0;
2378 spin_lock(&recall_lock
);
2379 fp
->fi_had_conflict
= true;
2380 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2381 nfsd_break_one_deleg(dp
);
2382 spin_unlock(&recall_lock
);
2386 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2389 return lease_modify(onlist
, arg
);
2394 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2395 .fl_break
= nfsd_break_deleg_cb
,
2396 .fl_change
= nfsd_change_deleg_cb
,
2401 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2402 struct nfsd4_open
*open
)
2404 clientid_t
*clientid
= &open
->op_clientid
;
2405 struct nfs4_client
*clp
= NULL
;
2406 unsigned int strhashval
;
2407 struct nfs4_stateowner
*sop
= NULL
;
2409 if (!check_name(open
->op_owner
))
2410 return nfserr_inval
;
2412 if (STALE_CLIENTID(&open
->op_clientid
))
2413 return nfserr_stale_clientid
;
2415 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2416 sop
= find_openstateowner_str(strhashval
, open
);
2417 open
->op_stateowner
= sop
;
2419 /* Make sure the client's lease hasn't expired. */
2420 clp
= find_confirmed_client(clientid
);
2422 return nfserr_expired
;
2425 /* When sessions are used, skip open sequenceid processing */
2426 if (nfsd4_has_session(cstate
))
2428 if (!sop
->so_confirmed
) {
2429 /* Replace unconfirmed owners without checking for replay. */
2430 clp
= sop
->so_client
;
2431 release_openowner(sop
);
2432 open
->op_stateowner
= NULL
;
2435 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2436 if (sop
->so_replay
.rp_buflen
)
2437 return nfserr_replay_me
;
2438 /* The original OPEN failed so spectacularly
2439 * that we don't even have replay data saved!
2440 * Therefore, we have no choice but to continue
2441 * processing this OPEN; presumably, we'll
2442 * fail again for the same reason.
2444 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2447 if (open
->op_seqid
!= sop
->so_seqid
)
2448 return nfserr_bad_seqid
;
2450 if (open
->op_stateowner
== NULL
) {
2451 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2453 return nfserr_resource
;
2454 open
->op_stateowner
= sop
;
2456 list_del_init(&sop
->so_close_lru
);
2457 renew_client(sop
->so_client
);
2461 static inline __be32
2462 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2464 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2465 return nfserr_openmode
;
2470 static struct nfs4_delegation
*
2471 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2473 struct nfs4_delegation
*dp
;
2475 spin_lock(&recall_lock
);
2476 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2477 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
) {
2478 spin_unlock(&recall_lock
);
2481 spin_unlock(&recall_lock
);
2485 int share_access_to_flags(u32 share_access
)
2487 share_access
&= ~NFS4_SHARE_WANT_MASK
;
2489 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2493 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2494 struct nfs4_delegation
**dp
)
2497 __be32 status
= nfserr_bad_stateid
;
2499 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2502 flags
= share_access_to_flags(open
->op_share_access
);
2503 status
= nfs4_check_delegmode(*dp
, flags
);
2507 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2511 open
->op_stateowner
->so_confirmed
= 1;
2516 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2518 struct nfs4_stateid
*local
;
2519 __be32 status
= nfserr_share_denied
;
2520 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2522 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2523 /* ignore lock owners */
2524 if (local
->st_stateowner
->so_is_open_owner
== 0)
2526 /* remember if we have seen this open owner */
2527 if (local
->st_stateowner
== sop
)
2529 /* check for conflicting share reservations */
2530 if (!test_share(local
, open
))
2538 static inline struct nfs4_stateid
*
2539 nfs4_alloc_stateid(void)
2541 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2544 static inline int nfs4_access_to_access(u32 nfs4_access
)
2548 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2549 flags
|= NFSD_MAY_READ
;
2550 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2551 flags
|= NFSD_MAY_WRITE
;
2555 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
2556 *fp
, struct svc_fh
*cur_fh
, u32 nfs4_access
)
2559 int oflag
= nfs4_access_to_omode(nfs4_access
);
2560 int access
= nfs4_access_to_access(nfs4_access
);
2562 if (!fp
->fi_fds
[oflag
]) {
2563 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2564 &fp
->fi_fds
[oflag
]);
2568 nfs4_file_get_access(fp
, oflag
);
2574 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2575 struct nfs4_file
*fp
, struct svc_fh
*cur_fh
,
2576 struct nfsd4_open
*open
)
2578 struct nfs4_stateid
*stp
;
2581 stp
= nfs4_alloc_stateid();
2583 return nfserr_resource
;
2585 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
->op_share_access
);
2587 kmem_cache_free(stateid_slab
, stp
);
2594 static inline __be32
2595 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2596 struct nfsd4_open
*open
)
2598 struct iattr iattr
= {
2599 .ia_valid
= ATTR_SIZE
,
2602 if (!open
->op_truncate
)
2604 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2605 return nfserr_inval
;
2606 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2610 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2612 u32 op_share_access
= open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
;
2616 new_access
= !test_bit(op_share_access
, &stp
->st_access_bmap
);
2618 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, op_share_access
);
2622 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2625 int oflag
= nfs4_access_to_omode(new_access
);
2626 nfs4_file_put_access(fp
, oflag
);
2630 /* remember the open */
2631 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2632 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2639 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2641 open
->op_stateowner
->so_confirmed
= 1;
2642 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2645 /* Should we give out recallable state?: */
2646 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2648 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2651 * In the sessions case, since we don't have to establish a
2652 * separate connection for callbacks, we assume it's OK
2653 * until we hear otherwise:
2655 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2658 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2660 struct file_lock
*fl
;
2662 fl
= locks_alloc_lock();
2665 locks_init_lock(fl
);
2666 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2667 fl
->fl_flags
= FL_LEASE
;
2668 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2669 fl
->fl_end
= OFFSET_MAX
;
2670 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2671 fl
->fl_pid
= current
->tgid
;
2675 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2677 struct nfs4_file
*fp
= dp
->dl_file
;
2678 struct file_lock
*fl
;
2681 fl
= nfs4_alloc_init_lease(dp
, flag
);
2684 fl
->fl_file
= find_readable_file(fp
);
2685 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2686 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2688 list_del_init(&dp
->dl_perclnt
);
2689 locks_free_lock(fl
);
2693 fp
->fi_deleg_file
= fl
->fl_file
;
2694 get_file(fp
->fi_deleg_file
);
2695 atomic_set(&fp
->fi_delegees
, 1);
2696 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2700 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2702 struct nfs4_file
*fp
= dp
->dl_file
;
2705 return nfs4_setlease(dp
, flag
);
2706 spin_lock(&recall_lock
);
2707 if (fp
->fi_had_conflict
) {
2708 spin_unlock(&recall_lock
);
2711 atomic_inc(&fp
->fi_delegees
);
2712 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2713 spin_unlock(&recall_lock
);
2714 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2719 * Attempt to hand out a delegation.
2722 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2724 struct nfs4_delegation
*dp
;
2725 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2727 int status
, flag
= 0;
2729 cb_up
= nfsd4_cb_channel_good(sop
->so_client
);
2730 flag
= NFS4_OPEN_DELEGATE_NONE
;
2731 open
->op_recall
= 0;
2732 switch (open
->op_claim_type
) {
2733 case NFS4_OPEN_CLAIM_PREVIOUS
:
2735 open
->op_recall
= 1;
2736 flag
= open
->op_delegate_type
;
2737 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2740 case NFS4_OPEN_CLAIM_NULL
:
2741 /* Let's not give out any delegations till everyone's
2742 * had the chance to reclaim theirs.... */
2743 if (locks_in_grace())
2745 if (!cb_up
|| !sop
->so_confirmed
)
2747 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2748 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2750 flag
= NFS4_OPEN_DELEGATE_READ
;
2756 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2759 status
= nfs4_set_delegation(dp
, flag
);
2763 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2765 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2766 STATEID_VAL(&dp
->dl_stateid
));
2768 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2769 && flag
== NFS4_OPEN_DELEGATE_NONE
2770 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2771 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2772 open
->op_delegate_type
= flag
;
2775 nfs4_put_delegation(dp
);
2777 flag
= NFS4_OPEN_DELEGATE_NONE
;
2782 * called with nfs4_lock_state() held.
2785 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2787 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2788 struct nfs4_file
*fp
= NULL
;
2789 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2790 struct nfs4_stateid
*stp
= NULL
;
2791 struct nfs4_delegation
*dp
= NULL
;
2794 status
= nfserr_inval
;
2795 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2796 || !deny_valid(open
->op_share_deny
))
2799 * Lookup file; if found, lookup stateid and check open request,
2800 * and check for delegations in the process of being recalled.
2801 * If not found, create the nfs4_file struct
2803 fp
= find_file(ino
);
2805 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2807 status
= nfs4_check_deleg(fp
, open
, &dp
);
2811 status
= nfserr_bad_stateid
;
2812 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2814 status
= nfserr_resource
;
2815 fp
= alloc_init_file(ino
);
2821 * OPEN the file, or upgrade an existing OPEN.
2822 * If truncate fails, the OPEN fails.
2825 /* Stateid was found, this is an OPEN upgrade */
2826 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
2829 update_stateid(&stp
->st_stateid
);
2831 status
= nfs4_new_open(rqstp
, &stp
, fp
, current_fh
, open
);
2834 init_stateid(stp
, fp
, open
);
2835 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2837 release_open_stateid(stp
);
2840 if (nfsd4_has_session(&resp
->cstate
))
2841 update_stateid(&stp
->st_stateid
);
2843 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2845 if (nfsd4_has_session(&resp
->cstate
))
2846 open
->op_stateowner
->so_confirmed
= 1;
2849 * Attempt to hand out a delegation. No error return, because the
2850 * OPEN succeeds even if we fail.
2852 nfs4_open_delegation(current_fh
, open
, stp
);
2856 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2857 STATEID_VAL(&stp
->st_stateid
));
2861 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2862 nfs4_set_claim_prev(open
);
2864 * To finish the open response, we just need to set the rflags.
2866 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2867 if (!open
->op_stateowner
->so_confirmed
&&
2868 !nfsd4_has_session(&resp
->cstate
))
2869 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2875 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2878 struct nfs4_client
*clp
;
2882 dprintk("process_renew(%08x/%08x): starting\n",
2883 clid
->cl_boot
, clid
->cl_id
);
2884 status
= nfserr_stale_clientid
;
2885 if (STALE_CLIENTID(clid
))
2887 clp
= find_confirmed_client(clid
);
2888 status
= nfserr_expired
;
2890 /* We assume the client took too long to RENEW. */
2891 dprintk("nfsd4_renew: clientid not found!\n");
2895 status
= nfserr_cb_path_down
;
2896 if (!list_empty(&clp
->cl_delegations
)
2897 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
2901 nfs4_unlock_state();
2905 struct lock_manager nfsd4_manager
= {
2909 nfsd4_end_grace(void)
2911 dprintk("NFSD: end of grace period\n");
2912 nfsd4_recdir_purge_old();
2913 locks_end_grace(&nfsd4_manager
);
2915 * Now that every NFSv4 client has had the chance to recover and
2916 * to see the (possibly new, possibly shorter) lease time, we
2917 * can safely set the next grace time to the current lease time:
2919 nfsd4_grace
= nfsd4_lease
;
2923 nfs4_laundromat(void)
2925 struct nfs4_client
*clp
;
2926 struct nfs4_stateowner
*sop
;
2927 struct nfs4_delegation
*dp
;
2928 struct list_head
*pos
, *next
, reaplist
;
2929 time_t cutoff
= get_seconds() - nfsd4_lease
;
2930 time_t t
, clientid_val
= nfsd4_lease
;
2931 time_t u
, test_val
= nfsd4_lease
;
2935 dprintk("NFSD: laundromat service - starting\n");
2936 if (locks_in_grace())
2938 INIT_LIST_HEAD(&reaplist
);
2939 spin_lock(&client_lock
);
2940 list_for_each_safe(pos
, next
, &client_lru
) {
2941 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2942 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2943 t
= clp
->cl_time
- cutoff
;
2944 if (clientid_val
> t
)
2948 if (atomic_read(&clp
->cl_refcount
)) {
2949 dprintk("NFSD: client in use (clientid %08x)\n",
2950 clp
->cl_clientid
.cl_id
);
2953 unhash_client_locked(clp
);
2954 list_add(&clp
->cl_lru
, &reaplist
);
2956 spin_unlock(&client_lock
);
2957 list_for_each_safe(pos
, next
, &reaplist
) {
2958 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2959 dprintk("NFSD: purging unused client (clientid %08x)\n",
2960 clp
->cl_clientid
.cl_id
);
2961 nfsd4_remove_clid_dir(clp
);
2964 spin_lock(&recall_lock
);
2965 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2966 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2967 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2968 u
= dp
->dl_time
- cutoff
;
2973 list_move(&dp
->dl_recall_lru
, &reaplist
);
2975 spin_unlock(&recall_lock
);
2976 list_for_each_safe(pos
, next
, &reaplist
) {
2977 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2978 list_del_init(&dp
->dl_recall_lru
);
2979 unhash_delegation(dp
);
2981 test_val
= nfsd4_lease
;
2982 list_for_each_safe(pos
, next
, &close_lru
) {
2983 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2984 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2985 u
= sop
->so_time
- cutoff
;
2990 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2992 release_openowner(sop
);
2994 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2995 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2996 nfs4_unlock_state();
2997 return clientid_val
;
3000 static struct workqueue_struct
*laundry_wq
;
3001 static void laundromat_main(struct work_struct
*);
3002 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3005 laundromat_main(struct work_struct
*not_used
)
3009 t
= nfs4_laundromat();
3010 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3011 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3014 static struct nfs4_stateowner
*
3015 search_close_lru(u32 st_id
, int flags
)
3017 struct nfs4_stateowner
*local
= NULL
;
3019 if (flags
& CLOSE_STATE
) {
3020 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
3021 if (local
->so_id
== st_id
)
3029 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
3031 return fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
;
3035 STALE_STATEID(stateid_t
*stateid
)
3037 if (stateid
->si_boot
== boot_time
)
3039 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3040 STATEID_VAL(stateid
));
3045 access_permit_read(unsigned long access_bmap
)
3047 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
3048 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
3049 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
3053 access_permit_write(unsigned long access_bmap
)
3055 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
3056 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
3060 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
3062 __be32 status
= nfserr_openmode
;
3064 /* For lock stateid's, we test the parent open, not the lock: */
3065 if (stp
->st_openstp
)
3066 stp
= stp
->st_openstp
;
3067 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
3069 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
3076 static inline __be32
3077 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3079 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3081 else if (locks_in_grace()) {
3082 /* Answer in remaining cases depends on existance of
3083 * conflicting state; so we must wait out the grace period. */
3084 return nfserr_grace
;
3085 } else if (flags
& WR_STATE
)
3086 return nfs4_share_conflict(current_fh
,
3087 NFS4_SHARE_DENY_WRITE
);
3088 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3089 return nfs4_share_conflict(current_fh
,
3090 NFS4_SHARE_DENY_READ
);
3094 * Allow READ/WRITE during grace period on recovered state only for files
3095 * that are not able to provide mandatory locking.
3098 grace_disallows_io(struct inode
*inode
)
3100 return locks_in_grace() && mandatory_lock(inode
);
3103 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
3106 * When sessions are used the stateid generation number is ignored
3109 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
3112 /* If the client sends us a stateid from the future, it's buggy: */
3113 if (in
->si_generation
> ref
->si_generation
)
3114 return nfserr_bad_stateid
;
3116 * The following, however, can happen. For example, if the
3117 * client sends an open and some IO at the same time, the open
3118 * may bump si_generation while the IO is still in flight.
3119 * Thanks to hard links and renames, the client never knows what
3120 * file an open will affect. So it could avoid that situation
3121 * only by serializing all opens and IO from the same open
3122 * owner. To recover from the old_stateid error, the client
3123 * will just have to retry the IO:
3125 if (in
->si_generation
< ref
->si_generation
)
3126 return nfserr_old_stateid
;
3131 static int is_delegation_stateid(stateid_t
*stateid
)
3133 return stateid
->si_fileid
== 0;
3137 * Checks for stateid operations
3140 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3141 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3143 struct nfs4_stateid
*stp
= NULL
;
3144 struct nfs4_delegation
*dp
= NULL
;
3145 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3146 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3152 if (grace_disallows_io(ino
))
3153 return nfserr_grace
;
3155 if (nfsd4_has_session(cstate
))
3156 flags
|= HAS_SESSION
;
3158 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3159 return check_special_stateids(current_fh
, stateid
, flags
);
3161 status
= nfserr_stale_stateid
;
3162 if (STALE_STATEID(stateid
))
3166 * We assume that any stateid that has the current boot time,
3167 * but that we can't find, is expired:
3169 status
= nfserr_expired
;
3170 if (is_delegation_stateid(stateid
)) {
3171 dp
= find_delegation_stateid(ino
, stateid
);
3174 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
3178 status
= nfs4_check_delegmode(dp
, flags
);
3181 renew_client(dp
->dl_client
);
3183 *filpp
= dp
->dl_file
->fi_deleg_file
;
3186 } else { /* open or lock stateid */
3187 stp
= find_stateid(stateid
, flags
);
3190 status
= nfserr_bad_stateid
;
3191 if (nfs4_check_fh(current_fh
, stp
))
3193 if (!stp
->st_stateowner
->so_confirmed
)
3195 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
3199 status
= nfs4_check_openmode(stp
, flags
);
3202 renew_client(stp
->st_stateowner
->so_client
);
3204 if (flags
& RD_STATE
)
3205 *filpp
= find_readable_file(stp
->st_file
);
3207 *filpp
= find_writeable_file(stp
->st_file
);
3218 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3219 RD_STATE
: WR_STATE
;
3223 * Checks for sequence id mutating operations.
3226 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3227 stateid_t
*stateid
, int flags
,
3228 struct nfs4_stateowner
**sopp
,
3229 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
3231 struct nfs4_stateid
*stp
;
3232 struct nfs4_stateowner
*sop
;
3233 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3236 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3237 seqid
, STATEID_VAL(stateid
));
3242 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
3243 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3244 return nfserr_bad_stateid
;
3247 if (STALE_STATEID(stateid
))
3248 return nfserr_stale_stateid
;
3250 if (nfsd4_has_session(cstate
))
3251 flags
|= HAS_SESSION
;
3254 * We return BAD_STATEID if filehandle doesn't match stateid,
3255 * the confirmed flag is incorrecly set, or the generation
3256 * number is incorrect.
3258 stp
= find_stateid(stateid
, flags
);
3261 * Also, we should make sure this isn't just the result of
3264 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
3265 /* It's not stale; let's assume it's expired: */
3267 return nfserr_expired
;
3273 *sopp
= sop
= stp
->st_stateowner
;
3276 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
3277 struct nfs4_client
*clp
= sop
->so_client
;
3281 lkflg
= setlkflg(lock
->lk_type
);
3283 if (lock
->lk_is_new
) {
3284 if (!sop
->so_is_open_owner
)
3285 return nfserr_bad_stateid
;
3286 if (!(flags
& HAS_SESSION
) &&
3287 !same_clid(&clp
->cl_clientid
, lockclid
))
3288 return nfserr_bad_stateid
;
3289 /* stp is the open stateid */
3290 status
= nfs4_check_openmode(stp
, lkflg
);
3294 /* stp is the lock stateid */
3295 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
3301 if (nfs4_check_fh(current_fh
, stp
)) {
3302 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3303 return nfserr_bad_stateid
;
3307 * We now validate the seqid and stateid generation numbers.
3308 * For the moment, we ignore the possibility of
3309 * generation number wraparound.
3311 if (!(flags
& HAS_SESSION
) && seqid
!= sop
->so_seqid
)
3314 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3315 dprintk("NFSD: preprocess_seqid_op: expected"
3316 " unconfirmed stateowner!\n");
3317 return nfserr_bad_stateid
;
3319 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3320 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3321 " confirmed yet!\n");
3322 return nfserr_bad_stateid
;
3324 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, flags
);
3327 renew_client(sop
->so_client
);
3331 if (seqid
== sop
->so_seqid
- 1) {
3332 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3333 /* indicate replay to calling function */
3334 return nfserr_replay_me
;
3336 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3337 sop
->so_seqid
, seqid
);
3339 return nfserr_bad_seqid
;
3343 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3344 struct nfsd4_open_confirm
*oc
)
3347 struct nfs4_stateowner
*sop
;
3348 struct nfs4_stateid
*stp
;
3350 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3351 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3352 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3354 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3360 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3361 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3362 CONFIRM
| OPEN_STATE
,
3363 &oc
->oc_stateowner
, &stp
, NULL
)))
3366 sop
= oc
->oc_stateowner
;
3367 sop
->so_confirmed
= 1;
3368 update_stateid(&stp
->st_stateid
);
3369 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3370 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3371 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stateid
));
3373 nfsd4_create_clid_dir(sop
->so_client
);
3375 if (oc
->oc_stateowner
) {
3376 nfs4_get_stateowner(oc
->oc_stateowner
);
3377 cstate
->replay_owner
= oc
->oc_stateowner
;
3379 nfs4_unlock_state();
3385 * unset all bits in union bitmap (bmap) that
3386 * do not exist in share (from successful OPEN_DOWNGRADE)
3389 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3392 for (i
= 1; i
< 4; i
++) {
3393 if ((i
& access
) != i
)
3394 __clear_bit(i
, bmap
);
3399 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3402 for (i
= 0; i
< 4; i
++) {
3403 if ((i
& deny
) != i
)
3404 __clear_bit(i
, bmap
);
3409 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3410 struct nfsd4_compound_state
*cstate
,
3411 struct nfsd4_open_downgrade
*od
)
3414 struct nfs4_stateid
*stp
;
3415 unsigned int share_access
;
3417 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3418 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3419 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3421 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3422 || !deny_valid(od
->od_share_deny
))
3423 return nfserr_inval
;
3426 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3430 &od
->od_stateowner
, &stp
, NULL
)))
3433 status
= nfserr_inval
;
3434 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3435 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3436 stp
->st_access_bmap
, od
->od_share_access
);
3439 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3440 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3441 stp
->st_deny_bmap
, od
->od_share_deny
);
3444 set_access(&share_access
, stp
->st_access_bmap
);
3445 nfs4_file_downgrade(stp
->st_file
, share_access
& ~od
->od_share_access
);
3447 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3448 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3450 update_stateid(&stp
->st_stateid
);
3451 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3454 if (od
->od_stateowner
) {
3455 nfs4_get_stateowner(od
->od_stateowner
);
3456 cstate
->replay_owner
= od
->od_stateowner
;
3458 nfs4_unlock_state();
3463 * nfs4_unlock_state() called after encode
3466 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3467 struct nfsd4_close
*close
)
3470 struct nfs4_stateid
*stp
;
3472 dprintk("NFSD: nfsd4_close on file %.*s\n",
3473 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3474 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3477 /* check close_lru for replay */
3478 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3481 OPEN_STATE
| CLOSE_STATE
,
3482 &close
->cl_stateowner
, &stp
, NULL
)))
3485 update_stateid(&stp
->st_stateid
);
3486 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3488 /* release_stateid() calls nfsd_close() if needed */
3489 release_open_stateid(stp
);
3491 /* place unused nfs4_stateowners on so_close_lru list to be
3492 * released by the laundromat service after the lease period
3493 * to enable us to handle CLOSE replay
3495 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3496 move_to_close_lru(close
->cl_stateowner
);
3498 if (close
->cl_stateowner
) {
3499 nfs4_get_stateowner(close
->cl_stateowner
);
3500 cstate
->replay_owner
= close
->cl_stateowner
;
3502 nfs4_unlock_state();
3507 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3508 struct nfsd4_delegreturn
*dr
)
3510 struct nfs4_delegation
*dp
;
3511 stateid_t
*stateid
= &dr
->dr_stateid
;
3512 struct inode
*inode
;
3516 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3518 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3520 if (nfsd4_has_session(cstate
))
3521 flags
|= HAS_SESSION
;
3523 status
= nfserr_bad_stateid
;
3524 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3526 status
= nfserr_stale_stateid
;
3527 if (STALE_STATEID(stateid
))
3529 status
= nfserr_bad_stateid
;
3530 if (!is_delegation_stateid(stateid
))
3532 status
= nfserr_expired
;
3533 dp
= find_delegation_stateid(inode
, stateid
);
3536 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3539 renew_client(dp
->dl_client
);
3541 unhash_delegation(dp
);
3543 nfs4_unlock_state();
3550 * Lock owner state (byte-range locks)
3552 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3553 #define LOCK_HASH_BITS 8
3554 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3555 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3558 end_offset(u64 start
, u64 len
)
3563 return end
>= start
? end
: NFS4_MAX_UINT64
;
3566 /* last octet in a range */
3568 last_byte_offset(u64 start
, u64 len
)
3574 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3577 #define lockownerid_hashval(id) \
3578 ((id) & LOCK_HASH_MASK)
3580 static inline unsigned int
3581 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3582 struct xdr_netobj
*ownername
)
3584 return (file_hashval(inode
) + cl_id
3585 + opaque_hashval(ownername
->data
, ownername
->len
))
3589 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3590 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3591 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3593 static struct nfs4_stateid
*
3594 find_stateid(stateid_t
*stid
, int flags
)
3596 struct nfs4_stateid
*local
;
3597 u32 st_id
= stid
->si_stateownerid
;
3598 u32 f_id
= stid
->si_fileid
;
3599 unsigned int hashval
;
3601 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3602 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3603 hashval
= stateid_hashval(st_id
, f_id
);
3604 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3605 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3606 (local
->st_stateid
.si_fileid
== f_id
))
3611 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3612 hashval
= stateid_hashval(st_id
, f_id
);
3613 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3614 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3615 (local
->st_stateid
.si_fileid
== f_id
))
3622 static struct nfs4_delegation
*
3623 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3625 struct nfs4_file
*fp
;
3626 struct nfs4_delegation
*dl
;
3628 dprintk("NFSD: %s: stateid=" STATEID_FMT
"\n", __func__
,
3631 fp
= find_file(ino
);
3634 dl
= find_delegation_file(fp
, stid
);
3640 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3641 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3642 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3643 * locking, this prevents us from being completely protocol-compliant. The
3644 * real solution to this problem is to start using unsigned file offsets in
3645 * the VFS, but this is a very deep change!
3648 nfs4_transform_lock_offset(struct file_lock
*lock
)
3650 if (lock
->fl_start
< 0)
3651 lock
->fl_start
= OFFSET_MAX
;
3652 if (lock
->fl_end
< 0)
3653 lock
->fl_end
= OFFSET_MAX
;
3656 /* Hack!: For now, we're defining this just so we can use a pointer to it
3657 * as a unique cookie to identify our (NFSv4's) posix locks. */
3658 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3662 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3664 struct nfs4_stateowner
*sop
;
3666 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3667 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3668 kref_get(&sop
->so_ref
);
3670 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3672 deny
->ld_sop
= NULL
;
3673 deny
->ld_clientid
.cl_boot
= 0;
3674 deny
->ld_clientid
.cl_id
= 0;
3676 deny
->ld_start
= fl
->fl_start
;
3677 deny
->ld_length
= NFS4_MAX_UINT64
;
3678 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3679 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3680 deny
->ld_type
= NFS4_READ_LT
;
3681 if (fl
->fl_type
!= F_RDLCK
)
3682 deny
->ld_type
= NFS4_WRITE_LT
;
3685 static struct nfs4_stateowner
*
3686 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3687 struct xdr_netobj
*owner
)
3689 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3690 struct nfs4_stateowner
*op
;
3692 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3693 if (same_owner_str(op
, owner
, clid
))
3700 * Alloc a lock owner structure.
3701 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3704 * strhashval = lock_ownerstr_hashval
3707 static struct nfs4_stateowner
*
3708 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3709 struct nfs4_stateowner
*sop
;
3710 struct nfs4_replay
*rp
;
3711 unsigned int idhashval
;
3713 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3715 idhashval
= lockownerid_hashval(current_ownerid
);
3716 INIT_LIST_HEAD(&sop
->so_idhash
);
3717 INIT_LIST_HEAD(&sop
->so_strhash
);
3718 INIT_LIST_HEAD(&sop
->so_perclient
);
3719 INIT_LIST_HEAD(&sop
->so_stateids
);
3720 INIT_LIST_HEAD(&sop
->so_perstateid
);
3721 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3723 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3724 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3725 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3726 sop
->so_is_open_owner
= 0;
3727 sop
->so_id
= current_ownerid
++;
3728 sop
->so_client
= clp
;
3729 /* It is the openowner seqid that will be incremented in encode in the
3730 * case of new lockowners; so increment the lock seqid manually: */
3731 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3732 sop
->so_confirmed
= 1;
3733 rp
= &sop
->so_replay
;
3734 rp
->rp_status
= nfserr_serverfault
;
3736 rp
->rp_buf
= rp
->rp_ibuf
;
3740 static struct nfs4_stateid
*
3741 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3743 struct nfs4_stateid
*stp
;
3744 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3746 stp
= nfs4_alloc_stateid();
3749 INIT_LIST_HEAD(&stp
->st_hash
);
3750 INIT_LIST_HEAD(&stp
->st_perfile
);
3751 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3752 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3753 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3754 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3755 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3756 stp
->st_stateowner
= sop
;
3759 stp
->st_stateid
.si_boot
= boot_time
;
3760 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3761 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3762 stp
->st_stateid
.si_generation
= 0;
3763 stp
->st_access_bmap
= 0;
3764 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3765 stp
->st_openstp
= open_stp
;
3772 check_lock_length(u64 offset
, u64 length
)
3774 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3775 LOFF_OVERFLOW(offset
, length
)));
3778 static void get_lock_access(struct nfs4_stateid
*lock_stp
, u32 access
)
3780 struct nfs4_file
*fp
= lock_stp
->st_file
;
3781 int oflag
= nfs4_access_to_omode(access
);
3783 if (test_bit(access
, &lock_stp
->st_access_bmap
))
3785 nfs4_file_get_access(fp
, oflag
);
3786 __set_bit(access
, &lock_stp
->st_access_bmap
);
3793 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3794 struct nfsd4_lock
*lock
)
3796 struct nfs4_stateowner
*open_sop
= NULL
;
3797 struct nfs4_stateowner
*lock_sop
= NULL
;
3798 struct nfs4_stateid
*lock_stp
;
3799 struct nfs4_file
*fp
;
3800 struct file
*filp
= NULL
;
3801 struct file_lock file_lock
;
3802 struct file_lock conflock
;
3804 unsigned int strhashval
;
3807 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3808 (long long) lock
->lk_offset
,
3809 (long long) lock
->lk_length
);
3811 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3812 return nfserr_inval
;
3814 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3815 S_IFREG
, NFSD_MAY_LOCK
))) {
3816 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3822 if (lock
->lk_is_new
) {
3824 * Client indicates that this is a new lockowner.
3825 * Use open owner and open stateid to create lock owner and
3828 struct nfs4_stateid
*open_stp
= NULL
;
3830 status
= nfserr_stale_clientid
;
3831 if (!nfsd4_has_session(cstate
) &&
3832 STALE_CLIENTID(&lock
->lk_new_clientid
))
3835 /* validate and update open stateid and open seqid */
3836 status
= nfs4_preprocess_seqid_op(cstate
,
3837 lock
->lk_new_open_seqid
,
3838 &lock
->lk_new_open_stateid
,
3840 &lock
->lk_replay_owner
, &open_stp
,
3844 open_sop
= lock
->lk_replay_owner
;
3845 /* create lockowner and lock stateid */
3846 fp
= open_stp
->st_file
;
3847 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3848 open_sop
->so_client
->cl_clientid
.cl_id
,
3849 &lock
->v
.new.owner
);
3850 /* XXX: Do we need to check for duplicate stateowners on
3851 * the same file, or should they just be allowed (and
3852 * create new stateids)? */
3853 status
= nfserr_resource
;
3854 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3855 open_sop
->so_client
, open_stp
, lock
);
3856 if (lock_sop
== NULL
)
3858 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3859 if (lock_stp
== NULL
)
3862 /* lock (lock owner + lock stateid) already exists */
3863 status
= nfs4_preprocess_seqid_op(cstate
,
3864 lock
->lk_old_lock_seqid
,
3865 &lock
->lk_old_lock_stateid
,
3867 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3870 lock_sop
= lock
->lk_replay_owner
;
3871 fp
= lock_stp
->st_file
;
3873 /* lock->lk_replay_owner and lock_stp have been created or found */
3875 status
= nfserr_grace
;
3876 if (locks_in_grace() && !lock
->lk_reclaim
)
3878 status
= nfserr_no_grace
;
3879 if (!locks_in_grace() && lock
->lk_reclaim
)
3882 locks_init_lock(&file_lock
);
3883 switch (lock
->lk_type
) {
3886 filp
= find_readable_file(lock_stp
->st_file
);
3888 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
3889 file_lock
.fl_type
= F_RDLCK
;
3892 case NFS4_WRITEW_LT
:
3893 filp
= find_writeable_file(lock_stp
->st_file
);
3895 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
3896 file_lock
.fl_type
= F_WRLCK
;
3899 status
= nfserr_inval
;
3903 status
= nfserr_openmode
;
3906 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3907 file_lock
.fl_pid
= current
->tgid
;
3908 file_lock
.fl_file
= filp
;
3909 file_lock
.fl_flags
= FL_POSIX
;
3910 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3912 file_lock
.fl_start
= lock
->lk_offset
;
3913 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3914 nfs4_transform_lock_offset(&file_lock
);
3917 * Try to lock the file in the VFS.
3918 * Note: locks.c uses the BKL to protect the inode's lock list.
3921 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
3923 case 0: /* success! */
3924 update_stateid(&lock_stp
->st_stateid
);
3925 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3929 case (EAGAIN
): /* conflock holds conflicting lock */
3930 status
= nfserr_denied
;
3931 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3932 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3935 status
= nfserr_deadlock
;
3938 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3939 status
= nfserr_resource
;
3943 if (status
&& lock
->lk_is_new
&& lock_sop
)
3944 release_lockowner(lock_sop
);
3945 if (lock
->lk_replay_owner
) {
3946 nfs4_get_stateowner(lock
->lk_replay_owner
);
3947 cstate
->replay_owner
= lock
->lk_replay_owner
;
3949 nfs4_unlock_state();
3954 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3955 * so we do a temporary open here just to get an open file to pass to
3956 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3959 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3964 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3967 err
= vfs_test_lock(file
, lock
);
3976 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3977 struct nfsd4_lockt
*lockt
)
3979 struct inode
*inode
;
3980 struct file_lock file_lock
;
3984 if (locks_in_grace())
3985 return nfserr_grace
;
3987 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3988 return nfserr_inval
;
3990 lockt
->lt_stateowner
= NULL
;
3993 status
= nfserr_stale_clientid
;
3994 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3997 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3998 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3999 if (status
== nfserr_symlink
)
4000 status
= nfserr_inval
;
4004 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4005 locks_init_lock(&file_lock
);
4006 switch (lockt
->lt_type
) {
4009 file_lock
.fl_type
= F_RDLCK
;
4012 case NFS4_WRITEW_LT
:
4013 file_lock
.fl_type
= F_WRLCK
;
4016 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4017 status
= nfserr_inval
;
4021 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
4022 &lockt
->lt_clientid
, &lockt
->lt_owner
);
4023 if (lockt
->lt_stateowner
)
4024 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
4025 file_lock
.fl_pid
= current
->tgid
;
4026 file_lock
.fl_flags
= FL_POSIX
;
4028 file_lock
.fl_start
= lockt
->lt_offset
;
4029 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4031 nfs4_transform_lock_offset(&file_lock
);
4034 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4036 status
= nfserrno(error
);
4039 if (file_lock
.fl_type
!= F_UNLCK
) {
4040 status
= nfserr_denied
;
4041 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4044 nfs4_unlock_state();
4049 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4050 struct nfsd4_locku
*locku
)
4052 struct nfs4_stateid
*stp
;
4053 struct file
*filp
= NULL
;
4054 struct file_lock file_lock
;
4058 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4059 (long long) locku
->lu_offset
,
4060 (long long) locku
->lu_length
);
4062 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4063 return nfserr_inval
;
4067 if ((status
= nfs4_preprocess_seqid_op(cstate
,
4071 &locku
->lu_stateowner
, &stp
, NULL
)))
4074 filp
= find_any_file(stp
->st_file
);
4076 status
= nfserr_lock_range
;
4080 locks_init_lock(&file_lock
);
4081 file_lock
.fl_type
= F_UNLCK
;
4082 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
4083 file_lock
.fl_pid
= current
->tgid
;
4084 file_lock
.fl_file
= filp
;
4085 file_lock
.fl_flags
= FL_POSIX
;
4086 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4087 file_lock
.fl_start
= locku
->lu_offset
;
4089 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4090 nfs4_transform_lock_offset(&file_lock
);
4093 * Try to unlock the file in the VFS.
4095 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4097 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4101 * OK, unlock succeeded; the only thing left to do is update the stateid.
4103 update_stateid(&stp
->st_stateid
);
4104 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
4107 if (locku
->lu_stateowner
) {
4108 nfs4_get_stateowner(locku
->lu_stateowner
);
4109 cstate
->replay_owner
= locku
->lu_stateowner
;
4111 nfs4_unlock_state();
4115 status
= nfserrno(err
);
4121 * 1: locks held by lockowner
4122 * 0: no locks held by lockowner
4125 check_for_locks(struct nfs4_file
*filp
, struct nfs4_stateowner
*lowner
)
4127 struct file_lock
**flpp
;
4128 struct inode
*inode
= filp
->fi_inode
;
4132 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4133 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4144 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4145 struct nfsd4_compound_state
*cstate
,
4146 struct nfsd4_release_lockowner
*rlockowner
)
4148 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4149 struct nfs4_stateowner
*sop
;
4150 struct nfs4_stateid
*stp
;
4151 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4152 struct list_head matches
;
4156 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4157 clid
->cl_boot
, clid
->cl_id
);
4159 /* XXX check for lease expiration */
4161 status
= nfserr_stale_clientid
;
4162 if (STALE_CLIENTID(clid
))
4167 status
= nfserr_locks_held
;
4168 /* XXX: we're doing a linear search through all the lockowners.
4169 * Yipes! For now we'll just hope clients aren't really using
4170 * release_lockowner much, but eventually we have to fix these
4171 * data structures. */
4172 INIT_LIST_HEAD(&matches
);
4173 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4174 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
4175 if (!same_owner_str(sop
, owner
, clid
))
4177 list_for_each_entry(stp
, &sop
->so_stateids
,
4179 if (check_for_locks(stp
->st_file
, sop
))
4181 /* Note: so_perclient unused for lockowners,
4182 * so it's OK to fool with here. */
4183 list_add(&sop
->so_perclient
, &matches
);
4187 /* Clients probably won't expect us to return with some (but not all)
4188 * of the lockowner state released; so don't release any until all
4189 * have been checked. */
4191 while (!list_empty(&matches
)) {
4192 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
4194 /* unhash_stateowner deletes so_perclient only
4195 * for openowners. */
4196 list_del(&sop
->so_perclient
);
4197 release_lockowner(sop
);
4200 nfs4_unlock_state();
4204 static inline struct nfs4_client_reclaim
*
4207 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4211 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4213 unsigned int strhashval
= clientstr_hashval(name
);
4214 struct nfs4_client
*clp
;
4216 clp
= find_confirmed_client_by_str(name
, strhashval
);
4221 * failure => all reset bets are off, nfserr_no_grace...
4224 nfs4_client_to_reclaim(const char *name
)
4226 unsigned int strhashval
;
4227 struct nfs4_client_reclaim
*crp
= NULL
;
4229 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4230 crp
= alloc_reclaim();
4233 strhashval
= clientstr_hashval(name
);
4234 INIT_LIST_HEAD(&crp
->cr_strhash
);
4235 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4236 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4237 reclaim_str_hashtbl_size
++;
4242 nfs4_release_reclaim(void)
4244 struct nfs4_client_reclaim
*crp
= NULL
;
4247 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4248 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4249 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4250 struct nfs4_client_reclaim
, cr_strhash
);
4251 list_del(&crp
->cr_strhash
);
4253 reclaim_str_hashtbl_size
--;
4256 BUG_ON(reclaim_str_hashtbl_size
);
4260 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4261 static struct nfs4_client_reclaim
*
4262 nfs4_find_reclaim_client(clientid_t
*clid
)
4264 unsigned int strhashval
;
4265 struct nfs4_client
*clp
;
4266 struct nfs4_client_reclaim
*crp
= NULL
;
4269 /* find clientid in conf_id_hashtbl */
4270 clp
= find_confirmed_client(clid
);
4274 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4275 clp
->cl_name
.len
, clp
->cl_name
.data
,
4278 /* find clp->cl_name in reclaim_str_hashtbl */
4279 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4280 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4281 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4289 * Called from OPEN. Look for clientid in reclaim list.
4292 nfs4_check_open_reclaim(clientid_t
*clid
)
4294 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4297 /* initialization to perform at module load time: */
4300 nfs4_state_init(void)
4304 status
= nfsd4_init_slabs();
4307 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4308 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4309 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4310 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4311 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4312 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4314 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4315 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4316 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4317 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4319 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4320 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4321 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
4323 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
4324 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4325 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
4327 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4328 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4329 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4331 memset(&onestateid
, ~0, sizeof(stateid_t
));
4332 INIT_LIST_HEAD(&close_lru
);
4333 INIT_LIST_HEAD(&client_lru
);
4334 INIT_LIST_HEAD(&del_recall_lru
);
4335 reclaim_str_hashtbl_size
= 0;
4340 nfsd4_load_reboot_recovery_data(void)
4345 nfsd4_init_recdir(user_recovery_dirname
);
4346 status
= nfsd4_recdir_load();
4347 nfs4_unlock_state();
4349 printk("NFSD: Failure reading reboot recovery data\n");
4353 * Since the lifetime of a delegation isn't limited to that of an open, a
4354 * client may quite reasonably hang on to a delegation as long as it has
4355 * the inode cached. This becomes an obvious problem the first time a
4356 * client's inode cache approaches the size of the server's total memory.
4358 * For now we avoid this problem by imposing a hard limit on the number
4359 * of delegations, which varies according to the server's memory size.
4362 set_max_delegations(void)
4365 * Allow at most 4 delegations per megabyte of RAM. Quick
4366 * estimates suggest that in the worst case (where every delegation
4367 * is for a different inode), a delegation could take about 1.5K,
4368 * giving a worst case usage of about 6% of memory.
4370 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4373 /* initialization to perform when the nfsd service is started: */
4376 __nfs4_state_start(void)
4380 boot_time
= get_seconds();
4381 locks_start_grace(&nfsd4_manager
);
4382 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4384 ret
= set_callback_cred();
4387 laundry_wq
= create_singlethread_workqueue("nfsd4");
4388 if (laundry_wq
== NULL
)
4390 ret
= nfsd4_create_callback_queue();
4392 goto out_free_laundry
;
4393 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4394 set_max_delegations();
4397 destroy_workqueue(laundry_wq
);
4402 nfs4_state_start(void)
4404 nfsd4_load_reboot_recovery_data();
4405 return __nfs4_state_start();
4409 __nfs4_state_shutdown(void)
4412 struct nfs4_client
*clp
= NULL
;
4413 struct nfs4_delegation
*dp
= NULL
;
4414 struct list_head
*pos
, *next
, reaplist
;
4416 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4417 while (!list_empty(&conf_id_hashtbl
[i
])) {
4418 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4421 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4422 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4426 INIT_LIST_HEAD(&reaplist
);
4427 spin_lock(&recall_lock
);
4428 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4429 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4430 list_move(&dp
->dl_recall_lru
, &reaplist
);
4432 spin_unlock(&recall_lock
);
4433 list_for_each_safe(pos
, next
, &reaplist
) {
4434 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4435 list_del_init(&dp
->dl_recall_lru
);
4436 unhash_delegation(dp
);
4439 nfsd4_shutdown_recdir();
4443 nfs4_state_shutdown(void)
4445 cancel_delayed_work_sync(&laundromat_work
);
4446 destroy_workqueue(laundry_wq
);
4447 locks_end_grace(&nfsd4_manager
);
4449 nfs4_release_reclaim();
4450 __nfs4_state_shutdown();
4451 nfs4_unlock_state();
4452 nfsd4_destroy_callback_queue();
4456 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4457 * accessed when nfsd is starting.
4460 nfs4_set_recdir(char *recdir
)
4462 strcpy(user_recovery_dirname
, recdir
);
4466 * Change the NFSv4 recovery directory to recdir.
4469 nfs4_reset_recoverydir(char *recdir
)
4474 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4478 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4479 nfs4_set_recdir(recdir
);
4487 nfs4_recoverydir(void)
4489 return user_recovery_dirname
;