2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 #include <linux/sunrpc/svcauth_gss.h>
59 #define NFSDDBG_FACILITY NFSDDBG_PROC
62 static time_t lease_time
= 90; /* default lease time */
63 static time_t user_lease_time
= 90;
64 static time_t boot_time
;
65 static u32 current_ownerid
= 1;
66 static u32 current_fileid
= 1;
67 static u32 current_delegid
= 1;
69 static stateid_t zerostateid
; /* bits all 0 */
70 static stateid_t onestateid
; /* bits all 1 */
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
77 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
78 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
79 static void nfs4_set_recdir(char *recdir
);
83 /* Currently used for almost all code touching nfsv4 state: */
84 static DEFINE_MUTEX(client_mutex
);
87 * Currently used for the del_recall_lru and file hash table. In an
88 * effort to decrease the scope of the client_mutex, this spinlock may
89 * eventually cover more:
91 static DEFINE_SPINLOCK(recall_lock
);
93 static struct kmem_cache
*stateowner_slab
= NULL
;
94 static struct kmem_cache
*file_slab
= NULL
;
95 static struct kmem_cache
*stateid_slab
= NULL
;
96 static struct kmem_cache
*deleg_slab
= NULL
;
101 mutex_lock(&client_mutex
);
105 nfs4_unlock_state(void)
107 mutex_unlock(&client_mutex
);
111 opaque_hashval(const void *ptr
, int nbytes
)
113 unsigned char *cptr
= (unsigned char *) ptr
;
123 static struct list_head del_recall_lru
;
126 put_nfs4_file(struct nfs4_file
*fi
)
128 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
129 list_del(&fi
->fi_hash
);
130 spin_unlock(&recall_lock
);
132 kmem_cache_free(file_slab
, fi
);
137 get_nfs4_file(struct nfs4_file
*fi
)
139 atomic_inc(&fi
->fi_ref
);
142 static int num_delegations
;
143 unsigned int max_delegations
;
146 * Open owner state (share locks)
149 /* hash tables for nfs4_stateowner */
150 #define OWNER_HASH_BITS 8
151 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
152 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
154 #define ownerid_hashval(id) \
155 ((id) & OWNER_HASH_MASK)
156 #define ownerstr_hashval(clientid, ownername) \
157 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
159 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
160 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
162 /* hash table for nfs4_file */
163 #define FILE_HASH_BITS 8
164 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
165 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
166 /* hash table for (open)nfs4_stateid */
167 #define STATEID_HASH_BITS 10
168 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
169 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
171 #define file_hashval(x) \
172 hash_ptr(x, FILE_HASH_BITS)
173 #define stateid_hashval(owner_id, file_id) \
174 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
176 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
177 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
179 static struct nfs4_delegation
*
180 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
182 struct nfs4_delegation
*dp
;
183 struct nfs4_file
*fp
= stp
->st_file
;
184 struct nfs4_callback
*cb
= &stp
->st_stateowner
->so_client
->cl_callback
;
186 dprintk("NFSD alloc_init_deleg\n");
187 if (fp
->fi_had_conflict
)
189 if (num_delegations
> max_delegations
)
191 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
195 INIT_LIST_HEAD(&dp
->dl_perfile
);
196 INIT_LIST_HEAD(&dp
->dl_perclnt
);
197 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
202 get_file(stp
->st_vfs_file
);
203 dp
->dl_vfs_file
= stp
->st_vfs_file
;
205 dp
->dl_recall
.cbr_dp
= NULL
;
206 dp
->dl_recall
.cbr_ident
= cb
->cb_ident
;
207 dp
->dl_recall
.cbr_trunc
= 0;
208 dp
->dl_stateid
.si_boot
= boot_time
;
209 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
210 dp
->dl_stateid
.si_fileid
= 0;
211 dp
->dl_stateid
.si_generation
= 0;
212 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
214 atomic_set(&dp
->dl_count
, 1);
215 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
216 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
221 nfs4_put_delegation(struct nfs4_delegation
*dp
)
223 if (atomic_dec_and_test(&dp
->dl_count
)) {
224 dprintk("NFSD: freeing dp %p\n",dp
);
225 put_nfs4_file(dp
->dl_file
);
226 kmem_cache_free(deleg_slab
, dp
);
231 /* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
237 nfs4_close_delegation(struct nfs4_delegation
*dp
)
239 struct file
*filp
= dp
->dl_vfs_file
;
241 dprintk("NFSD: close_delegation dp %p\n",dp
);
242 dp
->dl_vfs_file
= NULL
;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
246 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
250 /* Called under the state lock. */
252 unhash_delegation(struct nfs4_delegation
*dp
)
254 list_del_init(&dp
->dl_perfile
);
255 list_del_init(&dp
->dl_perclnt
);
256 spin_lock(&recall_lock
);
257 list_del_init(&dp
->dl_recall_lru
);
258 spin_unlock(&recall_lock
);
259 nfs4_close_delegation(dp
);
260 nfs4_put_delegation(dp
);
267 /* Hash tables for nfs4_clientid state */
268 #define CLIENT_HASH_BITS 4
269 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
272 #define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
274 #define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
286 * client_lru holds client queue ordered by nfs4_client.cl_time
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
292 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
293 static int reclaim_str_hashtbl_size
= 0;
294 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
295 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
296 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
297 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
298 static struct list_head client_lru
;
299 static struct list_head close_lru
;
301 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
303 list_del(&stp
->st_hash
);
304 list_del(&stp
->st_perfile
);
305 list_del(&stp
->st_perstateowner
);
308 static void free_generic_stateid(struct nfs4_stateid
*stp
)
310 put_nfs4_file(stp
->st_file
);
311 kmem_cache_free(stateid_slab
, stp
);
314 static void release_lock_stateid(struct nfs4_stateid
*stp
)
316 unhash_generic_stateid(stp
);
317 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
318 free_generic_stateid(stp
);
321 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
323 struct nfs4_stateid
*stp
;
325 list_del(&sop
->so_idhash
);
326 list_del(&sop
->so_strhash
);
327 list_del(&sop
->so_perstateid
);
328 while (!list_empty(&sop
->so_stateids
)) {
329 stp
= list_first_entry(&sop
->so_stateids
,
330 struct nfs4_stateid
, st_perstateowner
);
331 release_lock_stateid(stp
);
335 static void release_lockowner(struct nfs4_stateowner
*sop
)
337 unhash_lockowner(sop
);
338 nfs4_put_stateowner(sop
);
342 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
344 struct nfs4_stateowner
*lock_sop
;
346 while (!list_empty(&open_stp
->st_lockowners
)) {
347 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
348 struct nfs4_stateowner
, so_perstateid
);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop
->so_is_open_owner
);
351 release_lockowner(lock_sop
);
355 static void release_open_stateid(struct nfs4_stateid
*stp
)
357 unhash_generic_stateid(stp
);
358 release_stateid_lockowners(stp
);
359 nfsd_close(stp
->st_vfs_file
);
360 free_generic_stateid(stp
);
363 static void unhash_openowner(struct nfs4_stateowner
*sop
)
365 struct nfs4_stateid
*stp
;
367 list_del(&sop
->so_idhash
);
368 list_del(&sop
->so_strhash
);
369 list_del(&sop
->so_perclient
);
370 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
371 while (!list_empty(&sop
->so_stateids
)) {
372 stp
= list_first_entry(&sop
->so_stateids
,
373 struct nfs4_stateid
, st_perstateowner
);
374 release_open_stateid(stp
);
378 static void release_openowner(struct nfs4_stateowner
*sop
)
380 unhash_openowner(sop
);
381 list_del(&sop
->so_close_lru
);
382 nfs4_put_stateowner(sop
);
385 static DEFINE_SPINLOCK(sessionid_lock
);
386 #define SESSION_HASH_SIZE 512
387 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
390 hash_sessionid(struct nfs4_sessionid
*sessionid
)
392 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
394 return sid
->sequence
% SESSION_HASH_SIZE
;
398 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
400 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
404 /* caller must hold sessionid_lock */
405 static struct nfsd4_session
*
406 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
408 struct nfsd4_session
*elem
;
411 dump_sessionid(__func__
, sessionid
);
412 idx
= hash_sessionid(sessionid
);
413 dprintk("%s: idx is %d\n", __func__
, idx
);
414 /* Search in the appropriate list */
415 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
416 dump_sessionid("list traversal", &elem
->se_sessionid
);
417 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
418 NFS4_MAX_SESSIONID_LEN
)) {
423 dprintk("%s: session not found\n", __func__
);
427 /* caller must hold sessionid_lock */
429 unhash_session(struct nfsd4_session
*ses
)
431 list_del(&ses
->se_hash
);
432 list_del(&ses
->se_perclnt
);
436 release_session(struct nfsd4_session
*ses
)
438 spin_lock(&sessionid_lock
);
440 spin_unlock(&sessionid_lock
);
441 nfsd4_put_session(ses
);
445 free_session(struct kref
*kref
)
447 struct nfsd4_session
*ses
;
449 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
450 kfree(ses
->se_slots
);
455 renew_client(struct nfs4_client
*clp
)
458 * Move client to the end to the LRU list.
460 dprintk("renewing client (clientid %08x/%08x)\n",
461 clp
->cl_clientid
.cl_boot
,
462 clp
->cl_clientid
.cl_id
);
463 list_move_tail(&clp
->cl_lru
, &client_lru
);
464 clp
->cl_time
= get_seconds();
467 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
469 STALE_CLIENTID(clientid_t
*clid
)
471 if (clid
->cl_boot
== boot_time
)
473 dprintk("NFSD stale clientid (%08x/%08x)\n",
474 clid
->cl_boot
, clid
->cl_id
);
479 * XXX Should we use a slab cache ?
480 * This type of memory management is somewhat inefficient, but we use it
481 * anyway since SETCLIENTID is not a common operation.
483 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
485 struct nfs4_client
*clp
;
487 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
490 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
491 if (clp
->cl_name
.data
== NULL
) {
495 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
496 clp
->cl_name
.len
= name
.len
;
501 shutdown_callback_client(struct nfs4_client
*clp
)
503 struct rpc_clnt
*clnt
= clp
->cl_callback
.cb_client
;
507 * Callback threads take a reference on the client, so there
508 * should be no outstanding callbacks at this point.
510 clp
->cl_callback
.cb_client
= NULL
;
511 rpc_shutdown_client(clnt
);
516 free_client(struct nfs4_client
*clp
)
518 shutdown_callback_client(clp
);
519 if (clp
->cl_cred
.cr_group_info
)
520 put_group_info(clp
->cl_cred
.cr_group_info
);
521 kfree(clp
->cl_principal
);
522 kfree(clp
->cl_name
.data
);
527 put_nfs4_client(struct nfs4_client
*clp
)
529 if (atomic_dec_and_test(&clp
->cl_count
))
534 expire_client(struct nfs4_client
*clp
)
536 struct nfs4_stateowner
*sop
;
537 struct nfs4_delegation
*dp
;
538 struct list_head reaplist
;
540 dprintk("NFSD: expire_client cl_count %d\n",
541 atomic_read(&clp
->cl_count
));
543 INIT_LIST_HEAD(&reaplist
);
544 spin_lock(&recall_lock
);
545 while (!list_empty(&clp
->cl_delegations
)) {
546 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
547 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
549 list_del_init(&dp
->dl_perclnt
);
550 list_move(&dp
->dl_recall_lru
, &reaplist
);
552 spin_unlock(&recall_lock
);
553 while (!list_empty(&reaplist
)) {
554 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
555 list_del_init(&dp
->dl_recall_lru
);
556 unhash_delegation(dp
);
558 list_del(&clp
->cl_idhash
);
559 list_del(&clp
->cl_strhash
);
560 list_del(&clp
->cl_lru
);
561 while (!list_empty(&clp
->cl_openowners
)) {
562 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
563 release_openowner(sop
);
565 while (!list_empty(&clp
->cl_sessions
)) {
566 struct nfsd4_session
*ses
;
567 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
569 release_session(ses
);
571 put_nfs4_client(clp
);
574 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
576 struct nfs4_client
*clp
;
578 clp
= alloc_client(name
);
581 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
582 atomic_set(&clp
->cl_count
, 1);
583 atomic_set(&clp
->cl_callback
.cb_set
, 0);
584 INIT_LIST_HEAD(&clp
->cl_idhash
);
585 INIT_LIST_HEAD(&clp
->cl_strhash
);
586 INIT_LIST_HEAD(&clp
->cl_openowners
);
587 INIT_LIST_HEAD(&clp
->cl_delegations
);
588 INIT_LIST_HEAD(&clp
->cl_sessions
);
589 INIT_LIST_HEAD(&clp
->cl_lru
);
593 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
595 memcpy(target
->cl_verifier
.data
, source
->data
,
596 sizeof(target
->cl_verifier
.data
));
599 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
601 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
602 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
605 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
607 target
->cr_uid
= source
->cr_uid
;
608 target
->cr_gid
= source
->cr_gid
;
609 target
->cr_group_info
= source
->cr_group_info
;
610 get_group_info(target
->cr_group_info
);
613 static int same_name(const char *n1
, const char *n2
)
615 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
619 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
621 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
625 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
627 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
630 /* XXX what about NGROUP */
632 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
634 return cr1
->cr_uid
== cr2
->cr_uid
;
637 static void gen_clid(struct nfs4_client
*clp
)
639 static u32 current_clientid
= 1;
641 clp
->cl_clientid
.cl_boot
= boot_time
;
642 clp
->cl_clientid
.cl_id
= current_clientid
++;
645 static void gen_confirm(struct nfs4_client
*clp
)
650 p
= (u32
*)clp
->cl_confirm
.data
;
651 *p
++ = get_seconds();
655 static int check_name(struct xdr_netobj name
)
659 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
660 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
667 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
669 unsigned int idhashval
;
671 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
672 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
673 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
674 list_add_tail(&clp
->cl_lru
, &client_lru
);
675 clp
->cl_time
= get_seconds();
679 move_to_confirmed(struct nfs4_client
*clp
)
681 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
682 unsigned int strhashval
;
684 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
685 list_del_init(&clp
->cl_strhash
);
686 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
687 strhashval
= clientstr_hashval(clp
->cl_recdir
);
688 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
692 static struct nfs4_client
*
693 find_confirmed_client(clientid_t
*clid
)
695 struct nfs4_client
*clp
;
696 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
698 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
699 if (same_clid(&clp
->cl_clientid
, clid
))
705 static struct nfs4_client
*
706 find_unconfirmed_client(clientid_t
*clid
)
708 struct nfs4_client
*clp
;
709 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
711 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
712 if (same_clid(&clp
->cl_clientid
, clid
))
719 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
720 * parameter. Matching is based on the fact the at least one of the
721 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
723 * FIXME: we need to unify the clientid namespaces for nfsv4.x
724 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
725 * and SET_CLIENTID{,_CONFIRM}
728 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
730 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
731 return use_exchange_id
== has_exchange_flags
;
734 static struct nfs4_client
*
735 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
736 bool use_exchange_id
)
738 struct nfs4_client
*clp
;
740 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
741 if (same_name(clp
->cl_recdir
, dname
) &&
742 match_clientid_establishment(clp
, use_exchange_id
))
748 static struct nfs4_client
*
749 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
750 bool use_exchange_id
)
752 struct nfs4_client
*clp
;
754 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
755 if (same_name(clp
->cl_recdir
, dname
) &&
756 match_clientid_establishment(clp
, use_exchange_id
))
762 /* a helper function for parse_callback */
764 parse_octet(unsigned int *lenp
, char **addrp
)
766 unsigned int len
= *lenp
;
778 if ((c
< '0') || (c
> '9')) {
784 n
= (n
* 10) + (c
- '0');
795 /* parse and set the setclientid ipv4 callback address */
797 parse_ipv4(unsigned int addr_len
, char *addr_val
, unsigned int *cbaddrp
, unsigned short *cbportp
)
802 u32 addrlen
= addr_len
;
803 char *addr
= addr_val
;
808 for(i
= 4; i
> 0 ; i
--) {
809 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
812 cbaddr
|= (temp
<< shift
);
820 for(i
= 2; i
> 0 ; i
--) {
821 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
824 cbport
|= (temp
<< shift
);
833 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
835 struct nfs4_callback
*cb
= &clp
->cl_callback
;
837 /* Currently, we only support tcp for the callback channel */
838 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
841 if ( !(parse_ipv4(se
->se_callback_addr_len
, se
->se_callback_addr_val
,
842 &cb
->cb_addr
, &cb
->cb_port
)))
844 cb
->cb_prog
= se
->se_callback_prog
;
845 cb
->cb_ident
= se
->se_callback_ident
;
848 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
849 "will not receive delegations\n",
850 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
856 nfsd4_set_statp(struct svc_rqst
*rqstp
, __be32
*statp
)
858 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
860 resp
->cstate
.statp
= statp
;
864 * Dereference the result pages.
867 nfsd4_release_respages(struct page
**respages
, short resused
)
871 dprintk("--> %s\n", __func__
);
872 for (i
= 0; i
< resused
; i
++) {
875 put_page(respages
[i
]);
881 nfsd4_copy_pages(struct page
**topages
, struct page
**frompages
, short count
)
885 for (i
= 0; i
< count
; i
++) {
886 topages
[i
] = frompages
[i
];
889 get_page(topages
[i
]);
894 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
895 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
896 * length of the XDR response is less than se_fmaxresp_cached
897 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
898 * of the reply (e.g. readdir).
900 * Store the base and length of the rq_req.head[0] page
901 * of the NFSv4.1 data, just past the rpc header.
904 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
906 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
907 struct svc_rqst
*rqstp
= resp
->rqstp
;
908 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
909 struct nfsd4_op
*op
= &args
->ops
[resp
->opcnt
];
910 struct kvec
*resv
= &rqstp
->rq_res
.head
[0];
912 dprintk("--> %s entry %p\n", __func__
, entry
);
914 /* Don't cache a failed OP_SEQUENCE. */
915 if (resp
->opcnt
== 1 && op
->opnum
== OP_SEQUENCE
&& resp
->cstate
.status
)
917 nfsd4_release_respages(entry
->ce_respages
, entry
->ce_resused
);
918 entry
->ce_resused
= rqstp
->rq_resused
;
919 if (entry
->ce_resused
> NFSD_PAGES_PER_SLOT
+ 1)
920 entry
->ce_resused
= NFSD_PAGES_PER_SLOT
+ 1;
921 nfsd4_copy_pages(entry
->ce_respages
, rqstp
->rq_respages
,
923 entry
->ce_status
= resp
->cstate
.status
;
924 entry
->ce_datav
.iov_base
= resp
->cstate
.statp
;
925 entry
->ce_datav
.iov_len
= resv
->iov_len
- ((char *)resp
->cstate
.statp
-
926 (char *)page_address(rqstp
->rq_respages
[0]));
927 entry
->ce_opcnt
= resp
->opcnt
;
928 /* Current request rpc header length*/
929 entry
->ce_rpchdrlen
= (char *)resp
->cstate
.statp
-
930 (char *)page_address(rqstp
->rq_respages
[0]);
934 * We keep the rpc header, but take the nfs reply from the replycache.
937 nfsd41_copy_replay_data(struct nfsd4_compoundres
*resp
,
938 struct nfsd4_cache_entry
*entry
)
940 struct svc_rqst
*rqstp
= resp
->rqstp
;
941 struct kvec
*resv
= &resp
->rqstp
->rq_res
.head
[0];
944 /* Current request rpc header length*/
945 len
= (char *)resp
->cstate
.statp
-
946 (char *)page_address(rqstp
->rq_respages
[0]);
947 if (entry
->ce_datav
.iov_len
+ len
> PAGE_SIZE
) {
948 dprintk("%s v41 cached reply too large (%Zd).\n", __func__
,
949 entry
->ce_datav
.iov_len
);
952 /* copy the cached reply nfsd data past the current rpc header */
953 memcpy((char *)resv
->iov_base
+ len
, entry
->ce_datav
.iov_base
,
954 entry
->ce_datav
.iov_len
);
955 resv
->iov_len
= len
+ entry
->ce_datav
.iov_len
;
960 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
961 * cached page. Replace any futher replay pages from the cache.
964 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
)
966 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
969 dprintk("--> %s entry %p\n", __func__
, entry
);
972 if (!nfsd41_copy_replay_data(resp
, entry
)) {
974 * Not enough room to use the replay rpc header, send the
975 * cached header. Release all the allocated result pages.
977 svc_free_res_pages(resp
->rqstp
);
978 nfsd4_copy_pages(resp
->rqstp
->rq_respages
, entry
->ce_respages
,
981 /* Release all but the first allocated result page */
983 resp
->rqstp
->rq_resused
--;
984 svc_free_res_pages(resp
->rqstp
);
986 nfsd4_copy_pages(&resp
->rqstp
->rq_respages
[1],
987 &entry
->ce_respages
[1],
988 entry
->ce_resused
- 1);
991 resp
->rqstp
->rq_resused
= entry
->ce_resused
;
992 resp
->opcnt
= entry
->ce_opcnt
;
993 resp
->cstate
.iovlen
= entry
->ce_datav
.iov_len
+ entry
->ce_rpchdrlen
;
994 status
= entry
->ce_status
;
1000 * Set the exchange_id flags returned by the server.
1003 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1005 /* pNFS is not supported */
1006 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1008 /* Referrals are supported, Migration is not. */
1009 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1011 /* set the wire flags to return to client. */
1012 clid
->flags
= new->cl_exchange_flags
;
1016 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1017 struct nfsd4_compound_state
*cstate
,
1018 struct nfsd4_exchange_id
*exid
)
1020 struct nfs4_client
*unconf
, *conf
, *new;
1022 unsigned int strhashval
;
1023 char dname
[HEXDIR_LEN
];
1024 nfs4_verifier verf
= exid
->verifier
;
1025 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1027 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1028 " ip_addr=%u flags %x, spa_how %d\n",
1029 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1030 ip_addr
, exid
->flags
, exid
->spa_how
);
1032 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1033 return nfserr_inval
;
1035 /* Currently only support SP4_NONE */
1036 switch (exid
->spa_how
) {
1040 return nfserr_encr_alg_unsupp
;
1042 BUG(); /* checked by xdr code */
1044 return nfserr_serverfault
; /* no excuse :-/ */
1047 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1052 strhashval
= clientstr_hashval(dname
);
1057 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1059 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1060 /* 18.35.4 case 8 */
1061 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1062 status
= nfserr_not_same
;
1065 /* Client reboot: destroy old state */
1066 expire_client(conf
);
1069 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1070 /* 18.35.4 case 9 */
1071 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1072 status
= nfserr_perm
;
1075 expire_client(conf
);
1078 if (ip_addr
!= conf
->cl_addr
&&
1079 !(exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
)) {
1080 /* Client collision. 18.35.4 case 3 */
1081 status
= nfserr_clid_inuse
;
1085 * Set bit when the owner id and verifier map to an already
1086 * confirmed client id (18.35.3).
1088 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1091 * Falling into 18.35.4 case 2, possible router replay.
1092 * Leave confirmed record intact and return same result.
1094 copy_verf(conf
, &verf
);
1098 /* 18.35.4 case 7 */
1099 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1100 status
= nfserr_noent
;
1105 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1108 * Possible retry or client restart. Per 18.35.4 case 4,
1109 * a new unconfirmed record should be generated regardless
1110 * of whether any properties have changed.
1112 expire_client(unconf
);
1117 new = create_client(exid
->clname
, dname
);
1119 status
= nfserr_resource
;
1123 copy_verf(new, &verf
);
1124 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1125 new->cl_addr
= ip_addr
;
1128 add_to_unconfirmed(new, strhashval
);
1130 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1131 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1133 new->cl_seqid
= exid
->seqid
= 1;
1134 nfsd4_set_ex_flags(new, exid
);
1136 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1137 new->cl_seqid
, new->cl_exchange_flags
);
1141 nfs4_unlock_state();
1143 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1148 check_slot_seqid(u32 seqid
, struct nfsd4_slot
*slot
)
1150 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__
, seqid
,
1153 /* The slot is in use, and no response has been sent. */
1154 if (slot
->sl_inuse
) {
1155 if (seqid
== slot
->sl_seqid
)
1156 return nfserr_jukebox
;
1158 return nfserr_seq_misordered
;
1161 if (likely(seqid
== slot
->sl_seqid
+ 1))
1164 if (seqid
== slot
->sl_seqid
)
1165 return nfserr_replay_cache
;
1167 if (seqid
== 1 && (slot
->sl_seqid
+ 1) == 0)
1169 /* Misordered replay or misordered new request */
1170 return nfserr_seq_misordered
;
1174 nfsd4_create_session(struct svc_rqst
*rqstp
,
1175 struct nfsd4_compound_state
*cstate
,
1176 struct nfsd4_create_session
*cr_ses
)
1178 return -1; /* stub */
1182 nfsd4_destroy_session(struct svc_rqst
*r
,
1183 struct nfsd4_compound_state
*cstate
,
1184 struct nfsd4_destroy_session
*sessionid
)
1186 return -1; /* stub */
1190 nfsd4_sequence(struct svc_rqst
*rqstp
,
1191 struct nfsd4_compound_state
*cstate
,
1192 struct nfsd4_sequence
*seq
)
1194 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1195 struct nfsd4_session
*session
;
1196 struct nfsd4_slot
*slot
;
1199 if (resp
->opcnt
!= 1)
1200 return nfserr_sequence_pos
;
1202 spin_lock(&sessionid_lock
);
1203 status
= nfserr_badsession
;
1204 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1208 status
= nfserr_badslot
;
1209 if (seq
->slotid
>= session
->se_fnumslots
)
1212 slot
= &session
->se_slots
[seq
->slotid
];
1213 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1215 status
= check_slot_seqid(seq
->seqid
, slot
);
1216 if (status
== nfserr_replay_cache
) {
1217 cstate
->slot
= slot
;
1218 cstate
->session
= session
;
1219 /* Return the cached reply status and set cstate->status
1220 * for nfsd4_svc_encode_compoundres processing*/
1221 status
= nfsd4_replay_cache_entry(resp
);
1222 cstate
->status
= nfserr_replay_cache
;
1228 /* Success! bump slot seqid */
1229 slot
->sl_inuse
= true;
1230 slot
->sl_seqid
= seq
->seqid
;
1232 cstate
->slot
= slot
;
1233 cstate
->session
= session
;
1236 /* Renew the clientid on success and on replay.
1237 * Hold a session reference until done processing the compound:
1238 * nfsd4_put_session called only if the cstate slot is set.
1240 renew_client(session
->se_client
);
1241 nfsd4_get_session(session
);
1243 spin_unlock(&sessionid_lock
);
1244 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1249 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1250 struct nfsd4_setclientid
*setclid
)
1252 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1253 struct xdr_netobj clname
= {
1254 .len
= setclid
->se_namelen
,
1255 .data
= setclid
->se_name
,
1257 nfs4_verifier clverifier
= setclid
->se_verf
;
1258 unsigned int strhashval
;
1259 struct nfs4_client
*conf
, *unconf
, *new;
1262 char dname
[HEXDIR_LEN
];
1264 if (!check_name(clname
))
1265 return nfserr_inval
;
1267 status
= nfs4_make_rec_clidname(dname
, &clname
);
1272 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1273 * We get here on a DRC miss.
1276 strhashval
= clientstr_hashval(dname
);
1279 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1281 /* RFC 3530 14.2.33 CASE 0: */
1282 status
= nfserr_clid_inuse
;
1283 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1284 dprintk("NFSD: setclientid: string in use by client"
1285 " at %pI4\n", &conf
->cl_addr
);
1290 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1291 * has a description of SETCLIENTID request processing consisting
1292 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1294 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1295 status
= nfserr_resource
;
1298 * RFC 3530 14.2.33 CASE 4:
1299 * placed first, because it is the normal case
1302 expire_client(unconf
);
1303 new = create_client(clname
, dname
);
1307 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1309 * RFC 3530 14.2.33 CASE 1:
1310 * probable callback update
1313 /* Note this is removing unconfirmed {*x***},
1314 * which is stronger than RFC recommended {vxc**}.
1315 * This has the advantage that there is at most
1316 * one {*x***} in either list at any time.
1318 expire_client(unconf
);
1320 new = create_client(clname
, dname
);
1323 copy_clid(new, conf
);
1324 } else if (!unconf
) {
1326 * RFC 3530 14.2.33 CASE 2:
1327 * probable client reboot; state will be removed if
1330 new = create_client(clname
, dname
);
1336 * RFC 3530 14.2.33 CASE 3:
1337 * probable client reboot; state will be removed if
1340 expire_client(unconf
);
1341 new = create_client(clname
, dname
);
1346 copy_verf(new, &clverifier
);
1347 new->cl_addr
= sin
->sin_addr
.s_addr
;
1348 new->cl_flavor
= rqstp
->rq_flavor
;
1349 princ
= svc_gss_principal(rqstp
);
1351 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1352 if (new->cl_principal
== NULL
) {
1357 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1359 gen_callback(new, setclid
);
1360 add_to_unconfirmed(new, strhashval
);
1361 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1362 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1363 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1366 nfs4_unlock_state();
1372 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1373 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1374 * bullets, labeled as CASE1 - CASE4 below.
1377 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1378 struct nfsd4_compound_state
*cstate
,
1379 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1381 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1382 struct nfs4_client
*conf
, *unconf
;
1383 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1384 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1387 if (STALE_CLIENTID(clid
))
1388 return nfserr_stale_clientid
;
1390 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1391 * We get here on a DRC miss.
1396 conf
= find_confirmed_client(clid
);
1397 unconf
= find_unconfirmed_client(clid
);
1399 status
= nfserr_clid_inuse
;
1400 if (conf
&& conf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1402 if (unconf
&& unconf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1406 * section 14.2.34 of RFC 3530 has a description of
1407 * SETCLIENTID_CONFIRM request processing consisting
1408 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1410 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1412 * RFC 3530 14.2.34 CASE 1:
1415 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1416 status
= nfserr_clid_inuse
;
1418 /* XXX: We just turn off callbacks until we can handle
1419 * change request correctly. */
1420 atomic_set(&conf
->cl_callback
.cb_set
, 0);
1422 nfsd4_remove_clid_dir(unconf
);
1423 expire_client(unconf
);
1427 } else if (conf
&& !unconf
) {
1429 * RFC 3530 14.2.34 CASE 2:
1430 * probable retransmitted request; play it safe and
1433 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1434 status
= nfserr_clid_inuse
;
1437 } else if (!conf
&& unconf
1438 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1440 * RFC 3530 14.2.34 CASE 3:
1441 * Normal case; new or rebooted client:
1443 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1444 status
= nfserr_clid_inuse
;
1447 clientstr_hashval(unconf
->cl_recdir
);
1448 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1451 nfsd4_remove_clid_dir(conf
);
1452 expire_client(conf
);
1454 move_to_confirmed(unconf
);
1456 nfsd4_probe_callback(conf
);
1459 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1460 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1463 * RFC 3530 14.2.34 CASE 4:
1464 * Client probably hasn't noticed that we rebooted yet.
1466 status
= nfserr_stale_clientid
;
1468 /* check that we have hit one of the cases...*/
1469 status
= nfserr_clid_inuse
;
1472 nfs4_unlock_state();
1476 /* OPEN Share state helper functions */
1477 static inline struct nfs4_file
*
1478 alloc_init_file(struct inode
*ino
)
1480 struct nfs4_file
*fp
;
1481 unsigned int hashval
= file_hashval(ino
);
1483 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1485 atomic_set(&fp
->fi_ref
, 1);
1486 INIT_LIST_HEAD(&fp
->fi_hash
);
1487 INIT_LIST_HEAD(&fp
->fi_stateids
);
1488 INIT_LIST_HEAD(&fp
->fi_delegations
);
1489 spin_lock(&recall_lock
);
1490 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1491 spin_unlock(&recall_lock
);
1492 fp
->fi_inode
= igrab(ino
);
1493 fp
->fi_id
= current_fileid
++;
1494 fp
->fi_had_conflict
= false;
1501 nfsd4_free_slab(struct kmem_cache
**slab
)
1505 kmem_cache_destroy(*slab
);
1510 nfsd4_free_slabs(void)
1512 nfsd4_free_slab(&stateowner_slab
);
1513 nfsd4_free_slab(&file_slab
);
1514 nfsd4_free_slab(&stateid_slab
);
1515 nfsd4_free_slab(&deleg_slab
);
1519 nfsd4_init_slabs(void)
1521 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1522 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1523 if (stateowner_slab
== NULL
)
1525 file_slab
= kmem_cache_create("nfsd4_files",
1526 sizeof(struct nfs4_file
), 0, 0, NULL
);
1527 if (file_slab
== NULL
)
1529 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1530 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1531 if (stateid_slab
== NULL
)
1533 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1534 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1535 if (deleg_slab
== NULL
)
1540 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1545 nfs4_free_stateowner(struct kref
*kref
)
1547 struct nfs4_stateowner
*sop
=
1548 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1549 kfree(sop
->so_owner
.data
);
1550 kmem_cache_free(stateowner_slab
, sop
);
1553 static inline struct nfs4_stateowner
*
1554 alloc_stateowner(struct xdr_netobj
*owner
)
1556 struct nfs4_stateowner
*sop
;
1558 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1559 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1560 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1561 sop
->so_owner
.len
= owner
->len
;
1562 kref_init(&sop
->so_ref
);
1565 kmem_cache_free(stateowner_slab
, sop
);
1570 static struct nfs4_stateowner
*
1571 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1572 struct nfs4_stateowner
*sop
;
1573 struct nfs4_replay
*rp
;
1574 unsigned int idhashval
;
1576 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1578 idhashval
= ownerid_hashval(current_ownerid
);
1579 INIT_LIST_HEAD(&sop
->so_idhash
);
1580 INIT_LIST_HEAD(&sop
->so_strhash
);
1581 INIT_LIST_HEAD(&sop
->so_perclient
);
1582 INIT_LIST_HEAD(&sop
->so_stateids
);
1583 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1584 INIT_LIST_HEAD(&sop
->so_close_lru
);
1586 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1587 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1588 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1589 sop
->so_is_open_owner
= 1;
1590 sop
->so_id
= current_ownerid
++;
1591 sop
->so_client
= clp
;
1592 sop
->so_seqid
= open
->op_seqid
;
1593 sop
->so_confirmed
= 0;
1594 rp
= &sop
->so_replay
;
1595 rp
->rp_status
= nfserr_serverfault
;
1597 rp
->rp_buf
= rp
->rp_ibuf
;
1602 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1603 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1604 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1606 INIT_LIST_HEAD(&stp
->st_hash
);
1607 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1608 INIT_LIST_HEAD(&stp
->st_lockowners
);
1609 INIT_LIST_HEAD(&stp
->st_perfile
);
1610 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1611 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1612 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1613 stp
->st_stateowner
= sop
;
1616 stp
->st_stateid
.si_boot
= boot_time
;
1617 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1618 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1619 stp
->st_stateid
.si_generation
= 0;
1620 stp
->st_access_bmap
= 0;
1621 stp
->st_deny_bmap
= 0;
1622 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
1623 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1624 stp
->st_openstp
= NULL
;
1628 move_to_close_lru(struct nfs4_stateowner
*sop
)
1630 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1632 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1633 sop
->so_time
= get_seconds();
1637 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1640 return (sop
->so_owner
.len
== owner
->len
) &&
1641 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1642 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1645 static struct nfs4_stateowner
*
1646 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1648 struct nfs4_stateowner
*so
= NULL
;
1650 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1651 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1657 /* search file_hashtbl[] for file */
1658 static struct nfs4_file
*
1659 find_file(struct inode
*ino
)
1661 unsigned int hashval
= file_hashval(ino
);
1662 struct nfs4_file
*fp
;
1664 spin_lock(&recall_lock
);
1665 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1666 if (fp
->fi_inode
== ino
) {
1668 spin_unlock(&recall_lock
);
1672 spin_unlock(&recall_lock
);
1676 static inline int access_valid(u32 x
)
1678 if (x
< NFS4_SHARE_ACCESS_READ
)
1680 if (x
> NFS4_SHARE_ACCESS_BOTH
)
1685 static inline int deny_valid(u32 x
)
1687 /* Note: unlike access bits, deny bits may be zero. */
1688 return x
<= NFS4_SHARE_DENY_BOTH
;
1692 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1693 * st_{access,deny}_bmap field of the stateid, in order to track not
1694 * only what share bits are currently in force, but also what
1695 * combinations of share bits previous opens have used. This allows us
1696 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1697 * return an error if the client attempt to downgrade to a combination
1698 * of share bits not explicable by closing some of its previous opens.
1700 * XXX: This enforcement is actually incomplete, since we don't keep
1701 * track of access/deny bit combinations; so, e.g., we allow:
1703 * OPEN allow read, deny write
1704 * OPEN allow both, deny none
1705 * DOWNGRADE allow read, deny none
1707 * which we should reject.
1710 set_access(unsigned int *access
, unsigned long bmap
) {
1714 for (i
= 1; i
< 4; i
++) {
1715 if (test_bit(i
, &bmap
))
1721 set_deny(unsigned int *deny
, unsigned long bmap
) {
1725 for (i
= 0; i
< 4; i
++) {
1726 if (test_bit(i
, &bmap
))
1732 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
1733 unsigned int access
, deny
;
1735 set_access(&access
, stp
->st_access_bmap
);
1736 set_deny(&deny
, stp
->st_deny_bmap
);
1737 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
1743 * Called to check deny when READ with all zero stateid or
1744 * WRITE with all zero or all one stateid
1747 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
1749 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
1750 struct nfs4_file
*fp
;
1751 struct nfs4_stateid
*stp
;
1754 dprintk("NFSD: nfs4_share_conflict\n");
1756 fp
= find_file(ino
);
1759 ret
= nfserr_locked
;
1760 /* Search for conflicting share reservations */
1761 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
1762 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
1763 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
1773 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
1775 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
1776 drop_file_write_access(filp
);
1777 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
1782 * Recall a delegation
1785 do_recall(void *__dp
)
1787 struct nfs4_delegation
*dp
= __dp
;
1789 dp
->dl_file
->fi_had_conflict
= true;
1790 nfsd4_cb_recall(dp
);
1795 * Spawn a thread to perform a recall on the delegation represented
1796 * by the lease (file_lock)
1798 * Called from break_lease() with lock_kernel() held.
1799 * Note: we assume break_lease will only call this *once* for any given
1803 void nfsd_break_deleg_cb(struct file_lock
*fl
)
1805 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
1806 struct task_struct
*t
;
1808 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
1812 /* We're assuming the state code never drops its reference
1813 * without first removing the lease. Since we're in this lease
1814 * callback (and since the lease code is serialized by the kernel
1815 * lock) we know the server hasn't removed the lease yet, we know
1816 * it's safe to take a reference: */
1817 atomic_inc(&dp
->dl_count
);
1818 atomic_inc(&dp
->dl_client
->cl_count
);
1820 spin_lock(&recall_lock
);
1821 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
1822 spin_unlock(&recall_lock
);
1824 /* only place dl_time is set. protected by lock_kernel*/
1825 dp
->dl_time
= get_seconds();
1828 * We don't want the locks code to timeout the lease for us;
1829 * we'll remove it ourself if the delegation isn't returned
1832 fl
->fl_break_time
= 0;
1834 t
= kthread_run(do_recall
, dp
, "%s", "nfs4_cb_recall");
1836 struct nfs4_client
*clp
= dp
->dl_client
;
1838 printk(KERN_INFO
"NFSD: Callback thread failed for "
1839 "for client (clientid %08x/%08x)\n",
1840 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1841 put_nfs4_client(dp
->dl_client
);
1842 nfs4_put_delegation(dp
);
1847 * The file_lock is being reapd.
1849 * Called by locks_free_lock() with lock_kernel() held.
1852 void nfsd_release_deleg_cb(struct file_lock
*fl
)
1854 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
1856 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
1858 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
1860 dp
->dl_flock
= NULL
;
1864 * Set the delegation file_lock back pointer.
1866 * Called from setlease() with lock_kernel() held.
1869 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
1871 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
1873 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
1880 * Called from setlease() with lock_kernel() held
1883 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
1885 struct nfs4_delegation
*onlistd
=
1886 (struct nfs4_delegation
*)onlist
->fl_owner
;
1887 struct nfs4_delegation
*tryd
=
1888 (struct nfs4_delegation
*)try->fl_owner
;
1890 if (onlist
->fl_lmops
!= try->fl_lmops
)
1893 return onlistd
->dl_client
== tryd
->dl_client
;
1898 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
1901 return lease_modify(onlist
, arg
);
1906 static struct lock_manager_operations nfsd_lease_mng_ops
= {
1907 .fl_break
= nfsd_break_deleg_cb
,
1908 .fl_release_private
= nfsd_release_deleg_cb
,
1909 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
1910 .fl_mylease
= nfsd_same_client_deleg_cb
,
1911 .fl_change
= nfsd_change_deleg_cb
,
1916 nfsd4_process_open1(struct nfsd4_open
*open
)
1918 clientid_t
*clientid
= &open
->op_clientid
;
1919 struct nfs4_client
*clp
= NULL
;
1920 unsigned int strhashval
;
1921 struct nfs4_stateowner
*sop
= NULL
;
1923 if (!check_name(open
->op_owner
))
1924 return nfserr_inval
;
1926 if (STALE_CLIENTID(&open
->op_clientid
))
1927 return nfserr_stale_clientid
;
1929 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
1930 sop
= find_openstateowner_str(strhashval
, open
);
1931 open
->op_stateowner
= sop
;
1933 /* Make sure the client's lease hasn't expired. */
1934 clp
= find_confirmed_client(clientid
);
1936 return nfserr_expired
;
1939 if (!sop
->so_confirmed
) {
1940 /* Replace unconfirmed owners without checking for replay. */
1941 clp
= sop
->so_client
;
1942 release_openowner(sop
);
1943 open
->op_stateowner
= NULL
;
1946 if (open
->op_seqid
== sop
->so_seqid
- 1) {
1947 if (sop
->so_replay
.rp_buflen
)
1948 return nfserr_replay_me
;
1949 /* The original OPEN failed so spectacularly
1950 * that we don't even have replay data saved!
1951 * Therefore, we have no choice but to continue
1952 * processing this OPEN; presumably, we'll
1953 * fail again for the same reason.
1955 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1958 if (open
->op_seqid
!= sop
->so_seqid
)
1959 return nfserr_bad_seqid
;
1961 if (open
->op_stateowner
== NULL
) {
1962 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
1964 return nfserr_resource
;
1965 open
->op_stateowner
= sop
;
1967 list_del_init(&sop
->so_close_lru
);
1968 renew_client(sop
->so_client
);
1972 static inline __be32
1973 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
1975 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
1976 return nfserr_openmode
;
1981 static struct nfs4_delegation
*
1982 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
1984 struct nfs4_delegation
*dp
;
1986 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
1987 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
1994 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
1995 struct nfs4_delegation
**dp
)
1998 __be32 status
= nfserr_bad_stateid
;
2000 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2003 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
2004 RD_STATE
: WR_STATE
;
2005 status
= nfs4_check_delegmode(*dp
, flags
);
2009 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2013 open
->op_stateowner
->so_confirmed
= 1;
2018 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2020 struct nfs4_stateid
*local
;
2021 __be32 status
= nfserr_share_denied
;
2022 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2024 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2025 /* ignore lock owners */
2026 if (local
->st_stateowner
->so_is_open_owner
== 0)
2028 /* remember if we have seen this open owner */
2029 if (local
->st_stateowner
== sop
)
2031 /* check for conflicting share reservations */
2032 if (!test_share(local
, open
))
2040 static inline struct nfs4_stateid
*
2041 nfs4_alloc_stateid(void)
2043 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2047 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2048 struct nfs4_delegation
*dp
,
2049 struct svc_fh
*cur_fh
, int flags
)
2051 struct nfs4_stateid
*stp
;
2053 stp
= nfs4_alloc_stateid();
2055 return nfserr_resource
;
2058 get_file(dp
->dl_vfs_file
);
2059 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2062 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2065 if (status
== nfserr_dropit
)
2066 status
= nfserr_jukebox
;
2067 kmem_cache_free(stateid_slab
, stp
);
2075 static inline __be32
2076 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2077 struct nfsd4_open
*open
)
2079 struct iattr iattr
= {
2080 .ia_valid
= ATTR_SIZE
,
2083 if (!open
->op_truncate
)
2085 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2086 return nfserr_inval
;
2087 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2091 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2093 struct file
*filp
= stp
->st_vfs_file
;
2094 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2095 unsigned int share_access
, new_writer
;
2098 set_access(&share_access
, stp
->st_access_bmap
);
2099 new_writer
= (~share_access
) & open
->op_share_access
2100 & NFS4_SHARE_ACCESS_WRITE
;
2103 int err
= get_write_access(inode
);
2105 return nfserrno(err
);
2106 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2108 return nfserrno(err
);
2109 file_take_write(filp
);
2111 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2114 put_write_access(inode
);
2117 /* remember the open */
2118 filp
->f_mode
|= open
->op_share_access
;
2119 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2120 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2127 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2129 open
->op_stateowner
->so_confirmed
= 1;
2130 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2134 * Attempt to hand out a delegation.
2137 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2139 struct nfs4_delegation
*dp
;
2140 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2141 struct nfs4_callback
*cb
= &sop
->so_client
->cl_callback
;
2142 struct file_lock fl
, *flp
= &fl
;
2143 int status
, flag
= 0;
2145 flag
= NFS4_OPEN_DELEGATE_NONE
;
2146 open
->op_recall
= 0;
2147 switch (open
->op_claim_type
) {
2148 case NFS4_OPEN_CLAIM_PREVIOUS
:
2149 if (!atomic_read(&cb
->cb_set
))
2150 open
->op_recall
= 1;
2151 flag
= open
->op_delegate_type
;
2152 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2155 case NFS4_OPEN_CLAIM_NULL
:
2156 /* Let's not give out any delegations till everyone's
2157 * had the chance to reclaim theirs.... */
2158 if (locks_in_grace())
2160 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
2162 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2163 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2165 flag
= NFS4_OPEN_DELEGATE_READ
;
2171 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2173 flag
= NFS4_OPEN_DELEGATE_NONE
;
2176 locks_init_lock(&fl
);
2177 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2178 fl
.fl_flags
= FL_LEASE
;
2179 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2180 fl
.fl_end
= OFFSET_MAX
;
2181 fl
.fl_owner
= (fl_owner_t
)dp
;
2182 fl
.fl_file
= stp
->st_vfs_file
;
2183 fl
.fl_pid
= current
->tgid
;
2185 /* vfs_setlease checks to see if delegation should be handed out.
2186 * the lock_manager callbacks fl_mylease and fl_change are used
2188 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2189 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2190 unhash_delegation(dp
);
2191 flag
= NFS4_OPEN_DELEGATE_NONE
;
2195 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2197 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2198 dp
->dl_stateid
.si_boot
,
2199 dp
->dl_stateid
.si_stateownerid
,
2200 dp
->dl_stateid
.si_fileid
,
2201 dp
->dl_stateid
.si_generation
);
2203 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2204 && flag
== NFS4_OPEN_DELEGATE_NONE
2205 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2206 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2207 open
->op_delegate_type
= flag
;
2211 * called with nfs4_lock_state() held.
2214 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2216 struct nfs4_file
*fp
= NULL
;
2217 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2218 struct nfs4_stateid
*stp
= NULL
;
2219 struct nfs4_delegation
*dp
= NULL
;
2222 status
= nfserr_inval
;
2223 if (!access_valid(open
->op_share_access
)
2224 || !deny_valid(open
->op_share_deny
))
2227 * Lookup file; if found, lookup stateid and check open request,
2228 * and check for delegations in the process of being recalled.
2229 * If not found, create the nfs4_file struct
2231 fp
= find_file(ino
);
2233 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2235 status
= nfs4_check_deleg(fp
, open
, &dp
);
2239 status
= nfserr_bad_stateid
;
2240 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2242 status
= nfserr_resource
;
2243 fp
= alloc_init_file(ino
);
2249 * OPEN the file, or upgrade an existing OPEN.
2250 * If truncate fails, the OPEN fails.
2253 /* Stateid was found, this is an OPEN upgrade */
2254 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2257 update_stateid(&stp
->st_stateid
);
2259 /* Stateid was not found, this is a new OPEN */
2261 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2262 flags
|= NFSD_MAY_READ
;
2263 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2264 flags
|= NFSD_MAY_WRITE
;
2265 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2268 init_stateid(stp
, fp
, open
);
2269 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2271 release_open_stateid(stp
);
2275 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2278 * Attempt to hand out a delegation. No error return, because the
2279 * OPEN succeeds even if we fail.
2281 nfs4_open_delegation(current_fh
, open
, stp
);
2285 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2286 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
2287 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
2291 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2292 nfs4_set_claim_prev(open
);
2294 * To finish the open response, we just need to set the rflags.
2296 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2297 if (!open
->op_stateowner
->so_confirmed
)
2298 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2304 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2307 struct nfs4_client
*clp
;
2311 dprintk("process_renew(%08x/%08x): starting\n",
2312 clid
->cl_boot
, clid
->cl_id
);
2313 status
= nfserr_stale_clientid
;
2314 if (STALE_CLIENTID(clid
))
2316 clp
= find_confirmed_client(clid
);
2317 status
= nfserr_expired
;
2319 /* We assume the client took too long to RENEW. */
2320 dprintk("nfsd4_renew: clientid not found!\n");
2324 status
= nfserr_cb_path_down
;
2325 if (!list_empty(&clp
->cl_delegations
)
2326 && !atomic_read(&clp
->cl_callback
.cb_set
))
2330 nfs4_unlock_state();
2334 struct lock_manager nfsd4_manager
= {
2338 nfsd4_end_grace(void)
2340 dprintk("NFSD: end of grace period\n");
2341 nfsd4_recdir_purge_old();
2342 locks_end_grace(&nfsd4_manager
);
2346 nfs4_laundromat(void)
2348 struct nfs4_client
*clp
;
2349 struct nfs4_stateowner
*sop
;
2350 struct nfs4_delegation
*dp
;
2351 struct list_head
*pos
, *next
, reaplist
;
2352 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
2353 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
2354 time_t u
, test_val
= NFSD_LEASE_TIME
;
2358 dprintk("NFSD: laundromat service - starting\n");
2359 if (locks_in_grace())
2361 list_for_each_safe(pos
, next
, &client_lru
) {
2362 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2363 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2364 t
= clp
->cl_time
- cutoff
;
2365 if (clientid_val
> t
)
2369 dprintk("NFSD: purging unused client (clientid %08x)\n",
2370 clp
->cl_clientid
.cl_id
);
2371 nfsd4_remove_clid_dir(clp
);
2374 INIT_LIST_HEAD(&reaplist
);
2375 spin_lock(&recall_lock
);
2376 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2377 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2378 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2379 u
= dp
->dl_time
- cutoff
;
2384 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2386 list_move(&dp
->dl_recall_lru
, &reaplist
);
2388 spin_unlock(&recall_lock
);
2389 list_for_each_safe(pos
, next
, &reaplist
) {
2390 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2391 list_del_init(&dp
->dl_recall_lru
);
2392 unhash_delegation(dp
);
2394 test_val
= NFSD_LEASE_TIME
;
2395 list_for_each_safe(pos
, next
, &close_lru
) {
2396 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2397 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2398 u
= sop
->so_time
- cutoff
;
2403 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2405 release_openowner(sop
);
2407 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2408 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2409 nfs4_unlock_state();
2410 return clientid_val
;
2413 static struct workqueue_struct
*laundry_wq
;
2414 static void laundromat_main(struct work_struct
*);
2415 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2418 laundromat_main(struct work_struct
*not_used
)
2422 t
= nfs4_laundromat();
2423 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2424 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2427 static struct nfs4_stateowner
*
2428 search_close_lru(u32 st_id
, int flags
)
2430 struct nfs4_stateowner
*local
= NULL
;
2432 if (flags
& CLOSE_STATE
) {
2433 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2434 if (local
->so_id
== st_id
)
2442 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2444 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2448 STALE_STATEID(stateid_t
*stateid
)
2450 if (stateid
->si_boot
== boot_time
)
2452 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2453 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2454 stateid
->si_generation
);
2459 access_permit_read(unsigned long access_bmap
)
2461 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2462 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2463 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2467 access_permit_write(unsigned long access_bmap
)
2469 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2470 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2474 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2476 __be32 status
= nfserr_openmode
;
2478 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2480 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2487 static inline __be32
2488 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2490 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2492 else if (locks_in_grace()) {
2493 /* Answer in remaining cases depends on existance of
2494 * conflicting state; so we must wait out the grace period. */
2495 return nfserr_grace
;
2496 } else if (flags
& WR_STATE
)
2497 return nfs4_share_conflict(current_fh
,
2498 NFS4_SHARE_DENY_WRITE
);
2499 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2500 return nfs4_share_conflict(current_fh
,
2501 NFS4_SHARE_DENY_READ
);
2505 * Allow READ/WRITE during grace period on recovered state only for files
2506 * that are not able to provide mandatory locking.
2509 grace_disallows_io(struct inode
*inode
)
2511 return locks_in_grace() && mandatory_lock(inode
);
2514 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
)
2516 /* If the client sends us a stateid from the future, it's buggy: */
2517 if (in
->si_generation
> ref
->si_generation
)
2518 return nfserr_bad_stateid
;
2520 * The following, however, can happen. For example, if the
2521 * client sends an open and some IO at the same time, the open
2522 * may bump si_generation while the IO is still in flight.
2523 * Thanks to hard links and renames, the client never knows what
2524 * file an open will affect. So it could avoid that situation
2525 * only by serializing all opens and IO from the same open
2526 * owner. To recover from the old_stateid error, the client
2527 * will just have to retry the IO:
2529 if (in
->si_generation
< ref
->si_generation
)
2530 return nfserr_old_stateid
;
2534 static int is_delegation_stateid(stateid_t
*stateid
)
2536 return stateid
->si_fileid
== 0;
2540 * Checks for stateid operations
2543 nfs4_preprocess_stateid_op(struct svc_fh
*current_fh
, stateid_t
*stateid
, int flags
, struct file
**filpp
)
2545 struct nfs4_stateid
*stp
= NULL
;
2546 struct nfs4_delegation
*dp
= NULL
;
2547 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2553 if (grace_disallows_io(ino
))
2554 return nfserr_grace
;
2556 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2557 return check_special_stateids(current_fh
, stateid
, flags
);
2559 status
= nfserr_stale_stateid
;
2560 if (STALE_STATEID(stateid
))
2563 status
= nfserr_bad_stateid
;
2564 if (is_delegation_stateid(stateid
)) {
2565 dp
= find_delegation_stateid(ino
, stateid
);
2568 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
);
2571 status
= nfs4_check_delegmode(dp
, flags
);
2574 renew_client(dp
->dl_client
);
2576 *filpp
= dp
->dl_vfs_file
;
2577 } else { /* open or lock stateid */
2578 stp
= find_stateid(stateid
, flags
);
2581 if (nfs4_check_fh(current_fh
, stp
))
2583 if (!stp
->st_stateowner
->so_confirmed
)
2585 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2588 status
= nfs4_check_openmode(stp
, flags
);
2591 renew_client(stp
->st_stateowner
->so_client
);
2593 *filpp
= stp
->st_vfs_file
;
2603 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2604 RD_STATE
: WR_STATE
;
2608 * Checks for sequence id mutating operations.
2611 nfs4_preprocess_seqid_op(struct svc_fh
*current_fh
, u32 seqid
, stateid_t
*stateid
, int flags
, struct nfs4_stateowner
**sopp
, struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2613 struct nfs4_stateid
*stp
;
2614 struct nfs4_stateowner
*sop
;
2617 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2618 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2619 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2620 stateid
->si_generation
);
2625 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2626 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2627 return nfserr_bad_stateid
;
2630 if (STALE_STATEID(stateid
))
2631 return nfserr_stale_stateid
;
2633 * We return BAD_STATEID if filehandle doesn't match stateid,
2634 * the confirmed flag is incorrecly set, or the generation
2635 * number is incorrect.
2637 stp
= find_stateid(stateid
, flags
);
2640 * Also, we should make sure this isn't just the result of
2643 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2645 return nfserr_bad_stateid
;
2651 *sopp
= sop
= stp
->st_stateowner
;
2654 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
2655 struct nfs4_client
*clp
= sop
->so_client
;
2659 lkflg
= setlkflg(lock
->lk_type
);
2661 if (lock
->lk_is_new
) {
2662 if (!sop
->so_is_open_owner
)
2663 return nfserr_bad_stateid
;
2664 if (!same_clid(&clp
->cl_clientid
, lockclid
))
2665 return nfserr_bad_stateid
;
2666 /* stp is the open stateid */
2667 status
= nfs4_check_openmode(stp
, lkflg
);
2671 /* stp is the lock stateid */
2672 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
2678 if (nfs4_check_fh(current_fh
, stp
)) {
2679 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2680 return nfserr_bad_stateid
;
2684 * We now validate the seqid and stateid generation numbers.
2685 * For the moment, we ignore the possibility of
2686 * generation number wraparound.
2688 if (seqid
!= sop
->so_seqid
)
2691 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
2692 dprintk("NFSD: preprocess_seqid_op: expected"
2693 " unconfirmed stateowner!\n");
2694 return nfserr_bad_stateid
;
2696 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
2697 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2698 " confirmed yet!\n");
2699 return nfserr_bad_stateid
;
2701 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2704 renew_client(sop
->so_client
);
2708 if (seqid
== sop
->so_seqid
- 1) {
2709 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2710 /* indicate replay to calling function */
2711 return nfserr_replay_me
;
2713 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2714 sop
->so_seqid
, seqid
);
2716 return nfserr_bad_seqid
;
2720 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2721 struct nfsd4_open_confirm
*oc
)
2724 struct nfs4_stateowner
*sop
;
2725 struct nfs4_stateid
*stp
;
2727 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2728 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2729 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2731 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
2737 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2738 oc
->oc_seqid
, &oc
->oc_req_stateid
,
2739 CONFIRM
| OPEN_STATE
,
2740 &oc
->oc_stateowner
, &stp
, NULL
)))
2743 sop
= oc
->oc_stateowner
;
2744 sop
->so_confirmed
= 1;
2745 update_stateid(&stp
->st_stateid
);
2746 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2747 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2748 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
2749 stp
->st_stateid
.si_boot
,
2750 stp
->st_stateid
.si_stateownerid
,
2751 stp
->st_stateid
.si_fileid
,
2752 stp
->st_stateid
.si_generation
);
2754 nfsd4_create_clid_dir(sop
->so_client
);
2756 if (oc
->oc_stateowner
) {
2757 nfs4_get_stateowner(oc
->oc_stateowner
);
2758 cstate
->replay_owner
= oc
->oc_stateowner
;
2760 nfs4_unlock_state();
2766 * unset all bits in union bitmap (bmap) that
2767 * do not exist in share (from successful OPEN_DOWNGRADE)
2770 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
2773 for (i
= 1; i
< 4; i
++) {
2774 if ((i
& access
) != i
)
2775 __clear_bit(i
, bmap
);
2780 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
2783 for (i
= 0; i
< 4; i
++) {
2784 if ((i
& deny
) != i
)
2785 __clear_bit(i
, bmap
);
2790 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
2791 struct nfsd4_compound_state
*cstate
,
2792 struct nfsd4_open_downgrade
*od
)
2795 struct nfs4_stateid
*stp
;
2796 unsigned int share_access
;
2798 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2799 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2800 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2802 if (!access_valid(od
->od_share_access
)
2803 || !deny_valid(od
->od_share_deny
))
2804 return nfserr_inval
;
2807 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2811 &od
->od_stateowner
, &stp
, NULL
)))
2814 status
= nfserr_inval
;
2815 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
2816 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2817 stp
->st_access_bmap
, od
->od_share_access
);
2820 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
2821 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2822 stp
->st_deny_bmap
, od
->od_share_deny
);
2825 set_access(&share_access
, stp
->st_access_bmap
);
2826 nfs4_file_downgrade(stp
->st_vfs_file
,
2827 share_access
& ~od
->od_share_access
);
2829 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
2830 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
2832 update_stateid(&stp
->st_stateid
);
2833 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2836 if (od
->od_stateowner
) {
2837 nfs4_get_stateowner(od
->od_stateowner
);
2838 cstate
->replay_owner
= od
->od_stateowner
;
2840 nfs4_unlock_state();
2845 * nfs4_unlock_state() called after encode
2848 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2849 struct nfsd4_close
*close
)
2852 struct nfs4_stateid
*stp
;
2854 dprintk("NFSD: nfsd4_close on file %.*s\n",
2855 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2856 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2859 /* check close_lru for replay */
2860 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2863 OPEN_STATE
| CLOSE_STATE
,
2864 &close
->cl_stateowner
, &stp
, NULL
)))
2867 update_stateid(&stp
->st_stateid
);
2868 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2870 /* release_stateid() calls nfsd_close() if needed */
2871 release_open_stateid(stp
);
2873 /* place unused nfs4_stateowners on so_close_lru list to be
2874 * released by the laundromat service after the lease period
2875 * to enable us to handle CLOSE replay
2877 if (list_empty(&close
->cl_stateowner
->so_stateids
))
2878 move_to_close_lru(close
->cl_stateowner
);
2880 if (close
->cl_stateowner
) {
2881 nfs4_get_stateowner(close
->cl_stateowner
);
2882 cstate
->replay_owner
= close
->cl_stateowner
;
2884 nfs4_unlock_state();
2889 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2890 struct nfsd4_delegreturn
*dr
)
2892 struct nfs4_delegation
*dp
;
2893 stateid_t
*stateid
= &dr
->dr_stateid
;
2894 struct inode
*inode
;
2897 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
2899 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
2902 status
= nfserr_bad_stateid
;
2903 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2905 status
= nfserr_stale_stateid
;
2906 if (STALE_STATEID(stateid
))
2908 status
= nfserr_bad_stateid
;
2909 if (!is_delegation_stateid(stateid
))
2911 dp
= find_delegation_stateid(inode
, stateid
);
2914 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
);
2917 renew_client(dp
->dl_client
);
2919 unhash_delegation(dp
);
2921 nfs4_unlock_state();
2928 * Lock owner state (byte-range locks)
2930 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2931 #define LOCK_HASH_BITS 8
2932 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2933 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2936 end_offset(u64 start
, u64 len
)
2941 return end
>= start
? end
: NFS4_MAX_UINT64
;
2944 /* last octet in a range */
2946 last_byte_offset(u64 start
, u64 len
)
2952 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
2955 #define lockownerid_hashval(id) \
2956 ((id) & LOCK_HASH_MASK)
2958 static inline unsigned int
2959 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
2960 struct xdr_netobj
*ownername
)
2962 return (file_hashval(inode
) + cl_id
2963 + opaque_hashval(ownername
->data
, ownername
->len
))
2967 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
2968 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
2969 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
2971 static struct nfs4_stateid
*
2972 find_stateid(stateid_t
*stid
, int flags
)
2974 struct nfs4_stateid
*local
;
2975 u32 st_id
= stid
->si_stateownerid
;
2976 u32 f_id
= stid
->si_fileid
;
2977 unsigned int hashval
;
2979 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
2980 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
2981 hashval
= stateid_hashval(st_id
, f_id
);
2982 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
2983 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
2984 (local
->st_stateid
.si_fileid
== f_id
))
2989 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
2990 hashval
= stateid_hashval(st_id
, f_id
);
2991 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
2992 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
2993 (local
->st_stateid
.si_fileid
== f_id
))
3000 static struct nfs4_delegation
*
3001 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3003 struct nfs4_file
*fp
;
3004 struct nfs4_delegation
*dl
;
3006 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3007 stid
->si_boot
, stid
->si_stateownerid
,
3008 stid
->si_fileid
, stid
->si_generation
);
3010 fp
= find_file(ino
);
3013 dl
= find_delegation_file(fp
, stid
);
3019 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3020 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3021 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3022 * locking, this prevents us from being completely protocol-compliant. The
3023 * real solution to this problem is to start using unsigned file offsets in
3024 * the VFS, but this is a very deep change!
3027 nfs4_transform_lock_offset(struct file_lock
*lock
)
3029 if (lock
->fl_start
< 0)
3030 lock
->fl_start
= OFFSET_MAX
;
3031 if (lock
->fl_end
< 0)
3032 lock
->fl_end
= OFFSET_MAX
;
3035 /* Hack!: For now, we're defining this just so we can use a pointer to it
3036 * as a unique cookie to identify our (NFSv4's) posix locks. */
3037 static struct lock_manager_operations nfsd_posix_mng_ops
= {
3041 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3043 struct nfs4_stateowner
*sop
;
3046 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3047 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3048 hval
= lockownerid_hashval(sop
->so_id
);
3049 kref_get(&sop
->so_ref
);
3051 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3053 deny
->ld_sop
= NULL
;
3054 deny
->ld_clientid
.cl_boot
= 0;
3055 deny
->ld_clientid
.cl_id
= 0;
3057 deny
->ld_start
= fl
->fl_start
;
3058 deny
->ld_length
= NFS4_MAX_UINT64
;
3059 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3060 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3061 deny
->ld_type
= NFS4_READ_LT
;
3062 if (fl
->fl_type
!= F_RDLCK
)
3063 deny
->ld_type
= NFS4_WRITE_LT
;
3066 static struct nfs4_stateowner
*
3067 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3068 struct xdr_netobj
*owner
)
3070 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3071 struct nfs4_stateowner
*op
;
3073 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3074 if (same_owner_str(op
, owner
, clid
))
3081 * Alloc a lock owner structure.
3082 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3085 * strhashval = lock_ownerstr_hashval
3088 static struct nfs4_stateowner
*
3089 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3090 struct nfs4_stateowner
*sop
;
3091 struct nfs4_replay
*rp
;
3092 unsigned int idhashval
;
3094 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3096 idhashval
= lockownerid_hashval(current_ownerid
);
3097 INIT_LIST_HEAD(&sop
->so_idhash
);
3098 INIT_LIST_HEAD(&sop
->so_strhash
);
3099 INIT_LIST_HEAD(&sop
->so_perclient
);
3100 INIT_LIST_HEAD(&sop
->so_stateids
);
3101 INIT_LIST_HEAD(&sop
->so_perstateid
);
3102 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3104 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3105 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3106 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3107 sop
->so_is_open_owner
= 0;
3108 sop
->so_id
= current_ownerid
++;
3109 sop
->so_client
= clp
;
3110 /* It is the openowner seqid that will be incremented in encode in the
3111 * case of new lockowners; so increment the lock seqid manually: */
3112 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3113 sop
->so_confirmed
= 1;
3114 rp
= &sop
->so_replay
;
3115 rp
->rp_status
= nfserr_serverfault
;
3117 rp
->rp_buf
= rp
->rp_ibuf
;
3121 static struct nfs4_stateid
*
3122 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3124 struct nfs4_stateid
*stp
;
3125 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3127 stp
= nfs4_alloc_stateid();
3130 INIT_LIST_HEAD(&stp
->st_hash
);
3131 INIT_LIST_HEAD(&stp
->st_perfile
);
3132 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3133 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3134 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3135 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3136 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3137 stp
->st_stateowner
= sop
;
3140 stp
->st_stateid
.si_boot
= boot_time
;
3141 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3142 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3143 stp
->st_stateid
.si_generation
= 0;
3144 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3145 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3146 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3147 stp
->st_openstp
= open_stp
;
3154 check_lock_length(u64 offset
, u64 length
)
3156 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3157 LOFF_OVERFLOW(offset
, length
)));
3164 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3165 struct nfsd4_lock
*lock
)
3167 struct nfs4_stateowner
*open_sop
= NULL
;
3168 struct nfs4_stateowner
*lock_sop
= NULL
;
3169 struct nfs4_stateid
*lock_stp
;
3171 struct file_lock file_lock
;
3172 struct file_lock conflock
;
3174 unsigned int strhashval
;
3178 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3179 (long long) lock
->lk_offset
,
3180 (long long) lock
->lk_length
);
3182 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3183 return nfserr_inval
;
3185 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3186 S_IFREG
, NFSD_MAY_LOCK
))) {
3187 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3193 if (lock
->lk_is_new
) {
3195 * Client indicates that this is a new lockowner.
3196 * Use open owner and open stateid to create lock owner and
3199 struct nfs4_stateid
*open_stp
= NULL
;
3200 struct nfs4_file
*fp
;
3202 status
= nfserr_stale_clientid
;
3203 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
3206 /* validate and update open stateid and open seqid */
3207 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3208 lock
->lk_new_open_seqid
,
3209 &lock
->lk_new_open_stateid
,
3211 &lock
->lk_replay_owner
, &open_stp
,
3215 open_sop
= lock
->lk_replay_owner
;
3216 /* create lockowner and lock stateid */
3217 fp
= open_stp
->st_file
;
3218 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3219 open_sop
->so_client
->cl_clientid
.cl_id
,
3220 &lock
->v
.new.owner
);
3221 /* XXX: Do we need to check for duplicate stateowners on
3222 * the same file, or should they just be allowed (and
3223 * create new stateids)? */
3224 status
= nfserr_resource
;
3225 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3226 open_sop
->so_client
, open_stp
, lock
);
3227 if (lock_sop
== NULL
)
3229 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3230 if (lock_stp
== NULL
)
3233 /* lock (lock owner + lock stateid) already exists */
3234 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3235 lock
->lk_old_lock_seqid
,
3236 &lock
->lk_old_lock_stateid
,
3238 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3241 lock_sop
= lock
->lk_replay_owner
;
3243 /* lock->lk_replay_owner and lock_stp have been created or found */
3244 filp
= lock_stp
->st_vfs_file
;
3246 status
= nfserr_grace
;
3247 if (locks_in_grace() && !lock
->lk_reclaim
)
3249 status
= nfserr_no_grace
;
3250 if (!locks_in_grace() && lock
->lk_reclaim
)
3253 locks_init_lock(&file_lock
);
3254 switch (lock
->lk_type
) {
3257 file_lock
.fl_type
= F_RDLCK
;
3261 case NFS4_WRITEW_LT
:
3262 file_lock
.fl_type
= F_WRLCK
;
3266 status
= nfserr_inval
;
3269 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3270 file_lock
.fl_pid
= current
->tgid
;
3271 file_lock
.fl_file
= filp
;
3272 file_lock
.fl_flags
= FL_POSIX
;
3273 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3275 file_lock
.fl_start
= lock
->lk_offset
;
3276 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3277 nfs4_transform_lock_offset(&file_lock
);
3280 * Try to lock the file in the VFS.
3281 * Note: locks.c uses the BKL to protect the inode's lock list.
3284 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3286 case 0: /* success! */
3287 update_stateid(&lock_stp
->st_stateid
);
3288 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3292 case (EAGAIN
): /* conflock holds conflicting lock */
3293 status
= nfserr_denied
;
3294 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3295 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3298 status
= nfserr_deadlock
;
3301 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3302 status
= nfserr_resource
;
3306 if (status
&& lock
->lk_is_new
&& lock_sop
)
3307 release_lockowner(lock_sop
);
3308 if (lock
->lk_replay_owner
) {
3309 nfs4_get_stateowner(lock
->lk_replay_owner
);
3310 cstate
->replay_owner
= lock
->lk_replay_owner
;
3312 nfs4_unlock_state();
3317 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3318 * so we do a temporary open here just to get an open file to pass to
3319 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3322 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3327 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3330 err
= vfs_test_lock(file
, lock
);
3339 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3340 struct nfsd4_lockt
*lockt
)
3342 struct inode
*inode
;
3343 struct file_lock file_lock
;
3347 if (locks_in_grace())
3348 return nfserr_grace
;
3350 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3351 return nfserr_inval
;
3353 lockt
->lt_stateowner
= NULL
;
3356 status
= nfserr_stale_clientid
;
3357 if (STALE_CLIENTID(&lockt
->lt_clientid
))
3360 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3361 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3362 if (status
== nfserr_symlink
)
3363 status
= nfserr_inval
;
3367 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3368 locks_init_lock(&file_lock
);
3369 switch (lockt
->lt_type
) {
3372 file_lock
.fl_type
= F_RDLCK
;
3375 case NFS4_WRITEW_LT
:
3376 file_lock
.fl_type
= F_WRLCK
;
3379 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3380 status
= nfserr_inval
;
3384 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3385 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3386 if (lockt
->lt_stateowner
)
3387 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3388 file_lock
.fl_pid
= current
->tgid
;
3389 file_lock
.fl_flags
= FL_POSIX
;
3391 file_lock
.fl_start
= lockt
->lt_offset
;
3392 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3394 nfs4_transform_lock_offset(&file_lock
);
3397 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3399 status
= nfserrno(error
);
3402 if (file_lock
.fl_type
!= F_UNLCK
) {
3403 status
= nfserr_denied
;
3404 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3407 nfs4_unlock_state();
3412 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3413 struct nfsd4_locku
*locku
)
3415 struct nfs4_stateid
*stp
;
3416 struct file
*filp
= NULL
;
3417 struct file_lock file_lock
;
3421 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3422 (long long) locku
->lu_offset
,
3423 (long long) locku
->lu_length
);
3425 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3426 return nfserr_inval
;
3430 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3434 &locku
->lu_stateowner
, &stp
, NULL
)))
3437 filp
= stp
->st_vfs_file
;
3439 locks_init_lock(&file_lock
);
3440 file_lock
.fl_type
= F_UNLCK
;
3441 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3442 file_lock
.fl_pid
= current
->tgid
;
3443 file_lock
.fl_file
= filp
;
3444 file_lock
.fl_flags
= FL_POSIX
;
3445 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3446 file_lock
.fl_start
= locku
->lu_offset
;
3448 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3449 nfs4_transform_lock_offset(&file_lock
);
3452 * Try to unlock the file in the VFS.
3454 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3456 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3460 * OK, unlock succeeded; the only thing left to do is update the stateid.
3462 update_stateid(&stp
->st_stateid
);
3463 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3466 if (locku
->lu_stateowner
) {
3467 nfs4_get_stateowner(locku
->lu_stateowner
);
3468 cstate
->replay_owner
= locku
->lu_stateowner
;
3470 nfs4_unlock_state();
3474 status
= nfserrno(err
);
3480 * 1: locks held by lockowner
3481 * 0: no locks held by lockowner
3484 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3486 struct file_lock
**flpp
;
3487 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3491 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3492 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3503 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3504 struct nfsd4_compound_state
*cstate
,
3505 struct nfsd4_release_lockowner
*rlockowner
)
3507 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3508 struct nfs4_stateowner
*sop
;
3509 struct nfs4_stateid
*stp
;
3510 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3511 struct list_head matches
;
3515 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3516 clid
->cl_boot
, clid
->cl_id
);
3518 /* XXX check for lease expiration */
3520 status
= nfserr_stale_clientid
;
3521 if (STALE_CLIENTID(clid
))
3526 status
= nfserr_locks_held
;
3527 /* XXX: we're doing a linear search through all the lockowners.
3528 * Yipes! For now we'll just hope clients aren't really using
3529 * release_lockowner much, but eventually we have to fix these
3530 * data structures. */
3531 INIT_LIST_HEAD(&matches
);
3532 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3533 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3534 if (!same_owner_str(sop
, owner
, clid
))
3536 list_for_each_entry(stp
, &sop
->so_stateids
,
3538 if (check_for_locks(stp
->st_vfs_file
, sop
))
3540 /* Note: so_perclient unused for lockowners,
3541 * so it's OK to fool with here. */
3542 list_add(&sop
->so_perclient
, &matches
);
3546 /* Clients probably won't expect us to return with some (but not all)
3547 * of the lockowner state released; so don't release any until all
3548 * have been checked. */
3550 while (!list_empty(&matches
)) {
3551 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3553 /* unhash_stateowner deletes so_perclient only
3554 * for openowners. */
3555 list_del(&sop
->so_perclient
);
3556 release_lockowner(sop
);
3559 nfs4_unlock_state();
3563 static inline struct nfs4_client_reclaim
*
3566 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3570 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3572 unsigned int strhashval
= clientstr_hashval(name
);
3573 struct nfs4_client
*clp
;
3575 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3580 * failure => all reset bets are off, nfserr_no_grace...
3583 nfs4_client_to_reclaim(const char *name
)
3585 unsigned int strhashval
;
3586 struct nfs4_client_reclaim
*crp
= NULL
;
3588 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3589 crp
= alloc_reclaim();
3592 strhashval
= clientstr_hashval(name
);
3593 INIT_LIST_HEAD(&crp
->cr_strhash
);
3594 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3595 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3596 reclaim_str_hashtbl_size
++;
3601 nfs4_release_reclaim(void)
3603 struct nfs4_client_reclaim
*crp
= NULL
;
3606 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3607 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3608 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3609 struct nfs4_client_reclaim
, cr_strhash
);
3610 list_del(&crp
->cr_strhash
);
3612 reclaim_str_hashtbl_size
--;
3615 BUG_ON(reclaim_str_hashtbl_size
);
3619 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3620 static struct nfs4_client_reclaim
*
3621 nfs4_find_reclaim_client(clientid_t
*clid
)
3623 unsigned int strhashval
;
3624 struct nfs4_client
*clp
;
3625 struct nfs4_client_reclaim
*crp
= NULL
;
3628 /* find clientid in conf_id_hashtbl */
3629 clp
= find_confirmed_client(clid
);
3633 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3634 clp
->cl_name
.len
, clp
->cl_name
.data
,
3637 /* find clp->cl_name in reclaim_str_hashtbl */
3638 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3639 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3640 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3648 * Called from OPEN. Look for clientid in reclaim list.
3651 nfs4_check_open_reclaim(clientid_t
*clid
)
3653 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3656 /* initialization to perform at module load time: */
3659 nfs4_state_init(void)
3663 status
= nfsd4_init_slabs();
3666 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3667 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
3668 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
3669 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
3670 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
3672 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
3673 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
3674 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3675 INIT_LIST_HEAD(&file_hashtbl
[i
]);
3677 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
3678 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
3679 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
3681 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
3682 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
3683 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
3685 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3686 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
3687 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
3689 memset(&onestateid
, ~0, sizeof(stateid_t
));
3690 INIT_LIST_HEAD(&close_lru
);
3691 INIT_LIST_HEAD(&client_lru
);
3692 INIT_LIST_HEAD(&del_recall_lru
);
3693 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++)
3694 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
3695 reclaim_str_hashtbl_size
= 0;
3700 nfsd4_load_reboot_recovery_data(void)
3705 nfsd4_init_recdir(user_recovery_dirname
);
3706 status
= nfsd4_recdir_load();
3707 nfs4_unlock_state();
3709 printk("NFSD: Failure reading reboot recovery data\n");
3713 get_nfs4_grace_period(void)
3715 return max(user_lease_time
, lease_time
) * HZ
;
3719 * Since the lifetime of a delegation isn't limited to that of an open, a
3720 * client may quite reasonably hang on to a delegation as long as it has
3721 * the inode cached. This becomes an obvious problem the first time a
3722 * client's inode cache approaches the size of the server's total memory.
3724 * For now we avoid this problem by imposing a hard limit on the number
3725 * of delegations, which varies according to the server's memory size.
3728 set_max_delegations(void)
3731 * Allow at most 4 delegations per megabyte of RAM. Quick
3732 * estimates suggest that in the worst case (where every delegation
3733 * is for a different inode), a delegation could take about 1.5K,
3734 * giving a worst case usage of about 6% of memory.
3736 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
3739 /* initialization to perform when the nfsd service is started: */
3742 __nfs4_state_start(void)
3744 unsigned long grace_time
;
3746 boot_time
= get_seconds();
3747 grace_time
= get_nfs4_grace_period();
3748 lease_time
= user_lease_time
;
3749 locks_start_grace(&nfsd4_manager
);
3750 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
3752 laundry_wq
= create_singlethread_workqueue("nfsd4");
3753 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
3754 set_max_delegations();
3758 nfs4_state_start(void)
3762 nfsd4_load_reboot_recovery_data();
3763 __nfs4_state_start();
3769 nfs4_lease_time(void)
3775 __nfs4_state_shutdown(void)
3778 struct nfs4_client
*clp
= NULL
;
3779 struct nfs4_delegation
*dp
= NULL
;
3780 struct list_head
*pos
, *next
, reaplist
;
3782 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3783 while (!list_empty(&conf_id_hashtbl
[i
])) {
3784 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
3787 while (!list_empty(&unconf_str_hashtbl
[i
])) {
3788 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
3792 INIT_LIST_HEAD(&reaplist
);
3793 spin_lock(&recall_lock
);
3794 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3795 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3796 list_move(&dp
->dl_recall_lru
, &reaplist
);
3798 spin_unlock(&recall_lock
);
3799 list_for_each_safe(pos
, next
, &reaplist
) {
3800 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3801 list_del_init(&dp
->dl_recall_lru
);
3802 unhash_delegation(dp
);
3805 nfsd4_shutdown_recdir();
3810 nfs4_state_shutdown(void)
3812 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
3813 destroy_workqueue(laundry_wq
);
3814 locks_end_grace(&nfsd4_manager
);
3816 nfs4_release_reclaim();
3817 __nfs4_state_shutdown();
3818 nfs4_unlock_state();
3822 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3823 * accessed when nfsd is starting.
3826 nfs4_set_recdir(char *recdir
)
3828 strcpy(user_recovery_dirname
, recdir
);
3832 * Change the NFSv4 recovery directory to recdir.
3835 nfs4_reset_recoverydir(char *recdir
)
3840 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
3844 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
3845 nfs4_set_recdir(recdir
);
3853 nfs4_recoverydir(void)
3855 return user_recovery_dirname
;
3859 * Called when leasetime is changed.
3861 * The only way the protocol gives us to handle on-the-fly lease changes is to
3862 * simulate a reboot. Instead of doing that, we just wait till the next time
3863 * we start to register any changes in lease time. If the administrator
3864 * really wants to change the lease time *now*, they can go ahead and bring
3865 * nfsd down and then back up again after changing the lease time.
3867 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3868 * when nfsd is starting
3871 nfs4_reset_lease(time_t leasetime
)
3873 user_lease_time
= leasetime
;