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 */
71 static u64 current_sessionid
= 1;
73 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
74 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
76 /* forward declarations */
77 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
78 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
79 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
80 static void nfs4_set_recdir(char *recdir
);
84 /* Currently used for almost all code touching nfsv4 state: */
85 static DEFINE_MUTEX(client_mutex
);
88 * Currently used for the del_recall_lru and file hash table. In an
89 * effort to decrease the scope of the client_mutex, this spinlock may
90 * eventually cover more:
92 static DEFINE_SPINLOCK(recall_lock
);
94 static struct kmem_cache
*stateowner_slab
= NULL
;
95 static struct kmem_cache
*file_slab
= NULL
;
96 static struct kmem_cache
*stateid_slab
= NULL
;
97 static struct kmem_cache
*deleg_slab
= NULL
;
100 nfs4_lock_state(void)
102 mutex_lock(&client_mutex
);
106 nfs4_unlock_state(void)
108 mutex_unlock(&client_mutex
);
112 opaque_hashval(const void *ptr
, int nbytes
)
114 unsigned char *cptr
= (unsigned char *) ptr
;
124 static struct list_head del_recall_lru
;
127 put_nfs4_file(struct nfs4_file
*fi
)
129 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
130 list_del(&fi
->fi_hash
);
131 spin_unlock(&recall_lock
);
133 kmem_cache_free(file_slab
, fi
);
138 get_nfs4_file(struct nfs4_file
*fi
)
140 atomic_inc(&fi
->fi_ref
);
143 static int num_delegations
;
144 unsigned int max_delegations
;
147 * Open owner state (share locks)
150 /* hash tables for nfs4_stateowner */
151 #define OWNER_HASH_BITS 8
152 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
153 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155 #define ownerid_hashval(id) \
156 ((id) & OWNER_HASH_MASK)
157 #define ownerstr_hashval(clientid, ownername) \
158 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
161 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
163 /* hash table for nfs4_file */
164 #define FILE_HASH_BITS 8
165 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
166 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
167 /* hash table for (open)nfs4_stateid */
168 #define STATEID_HASH_BITS 10
169 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
170 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172 #define file_hashval(x) \
173 hash_ptr(x, FILE_HASH_BITS)
174 #define stateid_hashval(owner_id, file_id) \
175 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
178 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
180 static struct nfs4_delegation
*
181 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
183 struct nfs4_delegation
*dp
;
184 struct nfs4_file
*fp
= stp
->st_file
;
185 struct nfs4_callback
*cb
= &stp
->st_stateowner
->so_client
->cl_callback
;
187 dprintk("NFSD alloc_init_deleg\n");
188 if (fp
->fi_had_conflict
)
190 if (num_delegations
> max_delegations
)
192 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
196 INIT_LIST_HEAD(&dp
->dl_perfile
);
197 INIT_LIST_HEAD(&dp
->dl_perclnt
);
198 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
203 get_file(stp
->st_vfs_file
);
204 dp
->dl_vfs_file
= stp
->st_vfs_file
;
206 dp
->dl_recall
.cbr_dp
= NULL
;
207 dp
->dl_recall
.cbr_ident
= cb
->cb_ident
;
208 dp
->dl_recall
.cbr_trunc
= 0;
209 dp
->dl_stateid
.si_boot
= boot_time
;
210 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
211 dp
->dl_stateid
.si_fileid
= 0;
212 dp
->dl_stateid
.si_generation
= 0;
213 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
215 atomic_set(&dp
->dl_count
, 1);
216 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
217 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
222 nfs4_put_delegation(struct nfs4_delegation
*dp
)
224 if (atomic_dec_and_test(&dp
->dl_count
)) {
225 dprintk("NFSD: freeing dp %p\n",dp
);
226 put_nfs4_file(dp
->dl_file
);
227 kmem_cache_free(deleg_slab
, dp
);
232 /* Remove the associated file_lock first, then remove the delegation.
233 * lease_modify() is called to remove the FS_LEASE file_lock from
234 * the i_flock list, eventually calling nfsd's lock_manager
235 * fl_release_callback.
238 nfs4_close_delegation(struct nfs4_delegation
*dp
)
240 struct file
*filp
= dp
->dl_vfs_file
;
242 dprintk("NFSD: close_delegation dp %p\n",dp
);
243 dp
->dl_vfs_file
= NULL
;
244 /* The following nfsd_close may not actually close the file,
245 * but we want to remove the lease in any case. */
247 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
251 /* Called under the state lock. */
253 unhash_delegation(struct nfs4_delegation
*dp
)
255 list_del_init(&dp
->dl_perfile
);
256 list_del_init(&dp
->dl_perclnt
);
257 spin_lock(&recall_lock
);
258 list_del_init(&dp
->dl_recall_lru
);
259 spin_unlock(&recall_lock
);
260 nfs4_close_delegation(dp
);
261 nfs4_put_delegation(dp
);
268 /* Hash tables for nfs4_clientid state */
269 #define CLIENT_HASH_BITS 4
270 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
271 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
273 #define clientid_hashval(id) \
274 ((id) & CLIENT_HASH_MASK)
275 #define clientstr_hashval(name) \
276 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
278 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
279 * used in reboot/reset lease grace period processing
281 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
282 * setclientid_confirmed info.
284 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
287 * client_lru holds client queue ordered by nfs4_client.cl_time
290 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
291 * for last close replay.
293 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
294 static int reclaim_str_hashtbl_size
= 0;
295 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
296 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
297 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
298 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
299 static struct list_head client_lru
;
300 static struct list_head close_lru
;
302 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
304 list_del(&stp
->st_hash
);
305 list_del(&stp
->st_perfile
);
306 list_del(&stp
->st_perstateowner
);
309 static void free_generic_stateid(struct nfs4_stateid
*stp
)
311 put_nfs4_file(stp
->st_file
);
312 kmem_cache_free(stateid_slab
, stp
);
315 static void release_lock_stateid(struct nfs4_stateid
*stp
)
317 unhash_generic_stateid(stp
);
318 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
319 free_generic_stateid(stp
);
322 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
324 struct nfs4_stateid
*stp
;
326 list_del(&sop
->so_idhash
);
327 list_del(&sop
->so_strhash
);
328 list_del(&sop
->so_perstateid
);
329 while (!list_empty(&sop
->so_stateids
)) {
330 stp
= list_first_entry(&sop
->so_stateids
,
331 struct nfs4_stateid
, st_perstateowner
);
332 release_lock_stateid(stp
);
336 static void release_lockowner(struct nfs4_stateowner
*sop
)
338 unhash_lockowner(sop
);
339 nfs4_put_stateowner(sop
);
343 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
345 struct nfs4_stateowner
*lock_sop
;
347 while (!list_empty(&open_stp
->st_lockowners
)) {
348 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
349 struct nfs4_stateowner
, so_perstateid
);
350 /* list_del(&open_stp->st_lockowners); */
351 BUG_ON(lock_sop
->so_is_open_owner
);
352 release_lockowner(lock_sop
);
356 static void release_open_stateid(struct nfs4_stateid
*stp
)
358 unhash_generic_stateid(stp
);
359 release_stateid_lockowners(stp
);
360 nfsd_close(stp
->st_vfs_file
);
361 free_generic_stateid(stp
);
364 static void unhash_openowner(struct nfs4_stateowner
*sop
)
366 struct nfs4_stateid
*stp
;
368 list_del(&sop
->so_idhash
);
369 list_del(&sop
->so_strhash
);
370 list_del(&sop
->so_perclient
);
371 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
372 while (!list_empty(&sop
->so_stateids
)) {
373 stp
= list_first_entry(&sop
->so_stateids
,
374 struct nfs4_stateid
, st_perstateowner
);
375 release_open_stateid(stp
);
379 static void release_openowner(struct nfs4_stateowner
*sop
)
381 unhash_openowner(sop
);
382 list_del(&sop
->so_close_lru
);
383 nfs4_put_stateowner(sop
);
386 static DEFINE_SPINLOCK(sessionid_lock
);
387 #define SESSION_HASH_SIZE 512
388 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
391 hash_sessionid(struct nfs4_sessionid
*sessionid
)
393 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
395 return sid
->sequence
% SESSION_HASH_SIZE
;
399 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
401 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
402 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
406 gen_sessionid(struct nfsd4_session
*ses
)
408 struct nfs4_client
*clp
= ses
->se_client
;
409 struct nfsd4_sessionid
*sid
;
411 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
412 sid
->clientid
= clp
->cl_clientid
;
413 sid
->sequence
= current_sessionid
++;
418 * Give the client the number of slots it requests bound by
419 * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
421 * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
422 * should (up to a point) re-negotiate active sessions and reduce their
423 * slot usage to make rooom for new connections. For now we just fail the
426 static int set_forechannel_maxreqs(struct nfsd4_channel_attrs
*fchan
)
428 int status
= 0, np
= fchan
->maxreqs
* NFSD_PAGES_PER_SLOT
;
430 spin_lock(&nfsd_serv
->sv_lock
);
431 if (np
+ nfsd_serv
->sv_drc_pages_used
> nfsd_serv
->sv_drc_max_pages
)
432 np
= nfsd_serv
->sv_drc_max_pages
- nfsd_serv
->sv_drc_pages_used
;
433 nfsd_serv
->sv_drc_pages_used
+= np
;
434 spin_unlock(&nfsd_serv
->sv_lock
);
437 status
= nfserr_resource
;
440 fchan
->maxreqs
= np
/ NFSD_PAGES_PER_SLOT
;
446 * fchan holds the client values on input, and the server values on output
448 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
449 struct nfsd4_session
*session
,
450 struct nfsd4_channel_attrs
*fchan
)
453 __u32 maxcount
= svc_max_payload(rqstp
);
455 /* headerpadsz set to zero in encode routine */
457 /* Use the client's max request and max response size if possible */
458 if (fchan
->maxreq_sz
> maxcount
)
459 fchan
->maxreq_sz
= maxcount
;
460 session
->se_fmaxreq_sz
= fchan
->maxreq_sz
;
462 if (fchan
->maxresp_sz
> maxcount
)
463 fchan
->maxresp_sz
= maxcount
;
464 session
->se_fmaxresp_sz
= fchan
->maxresp_sz
;
466 /* Set the max response cached size our default which is
467 * a multiple of PAGE_SIZE and small */
468 session
->se_fmaxresp_cached
= NFSD_PAGES_PER_SLOT
* PAGE_SIZE
;
469 fchan
->maxresp_cached
= session
->se_fmaxresp_cached
;
471 /* Use the client's maxops if possible */
472 if (fchan
->maxops
> NFSD_MAX_OPS_PER_COMPOUND
)
473 fchan
->maxops
= NFSD_MAX_OPS_PER_COMPOUND
;
474 session
->se_fmaxops
= fchan
->maxops
;
476 /* try to use the client requested number of slots */
477 if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
478 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
480 /* FIXME: Error means no more DRC pages so the server should
481 * recover pages from existing sessions. For now fail session
484 status
= set_forechannel_maxreqs(fchan
);
486 session
->se_fnumslots
= fchan
->maxreqs
;
491 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
492 struct nfsd4_create_session
*cses
)
494 struct nfsd4_session
*new, tmp
;
495 int idx
, status
= nfserr_resource
, slotsize
;
497 memset(&tmp
, 0, sizeof(tmp
));
499 /* FIXME: For now, we just accept the client back channel attributes. */
500 status
= init_forechannel_attrs(rqstp
, &tmp
, &cses
->fore_channel
);
504 /* allocate struct nfsd4_session and slot table in one piece */
505 slotsize
= tmp
.se_fnumslots
* sizeof(struct nfsd4_slot
);
506 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
510 memcpy(new, &tmp
, sizeof(*new));
512 new->se_client
= clp
;
514 idx
= hash_sessionid(&new->se_sessionid
);
515 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
516 NFS4_MAX_SESSIONID_LEN
);
518 new->se_flags
= cses
->flags
;
519 kref_init(&new->se_ref
);
520 spin_lock(&sessionid_lock
);
521 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
522 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
523 spin_unlock(&sessionid_lock
);
530 /* caller must hold sessionid_lock */
531 static struct nfsd4_session
*
532 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
534 struct nfsd4_session
*elem
;
537 dump_sessionid(__func__
, sessionid
);
538 idx
= hash_sessionid(sessionid
);
539 dprintk("%s: idx is %d\n", __func__
, idx
);
540 /* Search in the appropriate list */
541 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
542 dump_sessionid("list traversal", &elem
->se_sessionid
);
543 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
544 NFS4_MAX_SESSIONID_LEN
)) {
549 dprintk("%s: session not found\n", __func__
);
553 /* caller must hold sessionid_lock */
555 unhash_session(struct nfsd4_session
*ses
)
557 list_del(&ses
->se_hash
);
558 list_del(&ses
->se_perclnt
);
562 release_session(struct nfsd4_session
*ses
)
564 spin_lock(&sessionid_lock
);
566 spin_unlock(&sessionid_lock
);
567 nfsd4_put_session(ses
);
570 static void nfsd4_release_respages(struct page
**respages
, short resused
);
573 free_session(struct kref
*kref
)
575 struct nfsd4_session
*ses
;
578 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
579 for (i
= 0; i
< ses
->se_fnumslots
; i
++) {
580 struct nfsd4_cache_entry
*e
= &ses
->se_slots
[i
].sl_cache_entry
;
581 nfsd4_release_respages(e
->ce_respages
, e
->ce_resused
);
583 kfree(ses
->se_slots
);
588 renew_client(struct nfs4_client
*clp
)
591 * Move client to the end to the LRU list.
593 dprintk("renewing client (clientid %08x/%08x)\n",
594 clp
->cl_clientid
.cl_boot
,
595 clp
->cl_clientid
.cl_id
);
596 list_move_tail(&clp
->cl_lru
, &client_lru
);
597 clp
->cl_time
= get_seconds();
600 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
602 STALE_CLIENTID(clientid_t
*clid
)
604 if (clid
->cl_boot
== boot_time
)
606 dprintk("NFSD stale clientid (%08x/%08x)\n",
607 clid
->cl_boot
, clid
->cl_id
);
612 * XXX Should we use a slab cache ?
613 * This type of memory management is somewhat inefficient, but we use it
614 * anyway since SETCLIENTID is not a common operation.
616 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
618 struct nfs4_client
*clp
;
620 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
623 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
624 if (clp
->cl_name
.data
== NULL
) {
628 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
629 clp
->cl_name
.len
= name
.len
;
634 shutdown_callback_client(struct nfs4_client
*clp
)
636 struct rpc_clnt
*clnt
= clp
->cl_callback
.cb_client
;
640 * Callback threads take a reference on the client, so there
641 * should be no outstanding callbacks at this point.
643 clp
->cl_callback
.cb_client
= NULL
;
644 rpc_shutdown_client(clnt
);
649 free_client(struct nfs4_client
*clp
)
651 shutdown_callback_client(clp
);
652 nfsd4_release_respages(clp
->cl_slot
.sl_cache_entry
.ce_respages
,
653 clp
->cl_slot
.sl_cache_entry
.ce_resused
);
654 if (clp
->cl_cred
.cr_group_info
)
655 put_group_info(clp
->cl_cred
.cr_group_info
);
656 kfree(clp
->cl_principal
);
657 kfree(clp
->cl_name
.data
);
662 put_nfs4_client(struct nfs4_client
*clp
)
664 if (atomic_dec_and_test(&clp
->cl_count
))
669 expire_client(struct nfs4_client
*clp
)
671 struct nfs4_stateowner
*sop
;
672 struct nfs4_delegation
*dp
;
673 struct list_head reaplist
;
675 dprintk("NFSD: expire_client cl_count %d\n",
676 atomic_read(&clp
->cl_count
));
678 INIT_LIST_HEAD(&reaplist
);
679 spin_lock(&recall_lock
);
680 while (!list_empty(&clp
->cl_delegations
)) {
681 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
682 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
684 list_del_init(&dp
->dl_perclnt
);
685 list_move(&dp
->dl_recall_lru
, &reaplist
);
687 spin_unlock(&recall_lock
);
688 while (!list_empty(&reaplist
)) {
689 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
690 list_del_init(&dp
->dl_recall_lru
);
691 unhash_delegation(dp
);
693 list_del(&clp
->cl_idhash
);
694 list_del(&clp
->cl_strhash
);
695 list_del(&clp
->cl_lru
);
696 while (!list_empty(&clp
->cl_openowners
)) {
697 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
698 release_openowner(sop
);
700 while (!list_empty(&clp
->cl_sessions
)) {
701 struct nfsd4_session
*ses
;
702 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
704 release_session(ses
);
706 put_nfs4_client(clp
);
709 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
711 struct nfs4_client
*clp
;
713 clp
= alloc_client(name
);
716 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
717 atomic_set(&clp
->cl_count
, 1);
718 atomic_set(&clp
->cl_callback
.cb_set
, 0);
719 INIT_LIST_HEAD(&clp
->cl_idhash
);
720 INIT_LIST_HEAD(&clp
->cl_strhash
);
721 INIT_LIST_HEAD(&clp
->cl_openowners
);
722 INIT_LIST_HEAD(&clp
->cl_delegations
);
723 INIT_LIST_HEAD(&clp
->cl_sessions
);
724 INIT_LIST_HEAD(&clp
->cl_lru
);
728 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
730 memcpy(target
->cl_verifier
.data
, source
->data
,
731 sizeof(target
->cl_verifier
.data
));
734 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
736 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
737 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
740 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
742 target
->cr_uid
= source
->cr_uid
;
743 target
->cr_gid
= source
->cr_gid
;
744 target
->cr_group_info
= source
->cr_group_info
;
745 get_group_info(target
->cr_group_info
);
748 static int same_name(const char *n1
, const char *n2
)
750 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
754 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
756 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
760 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
762 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
765 /* XXX what about NGROUP */
767 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
769 return cr1
->cr_uid
== cr2
->cr_uid
;
772 static void gen_clid(struct nfs4_client
*clp
)
774 static u32 current_clientid
= 1;
776 clp
->cl_clientid
.cl_boot
= boot_time
;
777 clp
->cl_clientid
.cl_id
= current_clientid
++;
780 static void gen_confirm(struct nfs4_client
*clp
)
785 p
= (u32
*)clp
->cl_confirm
.data
;
786 *p
++ = get_seconds();
790 static int check_name(struct xdr_netobj name
)
794 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
795 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
802 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
804 unsigned int idhashval
;
806 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
807 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
808 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
809 list_add_tail(&clp
->cl_lru
, &client_lru
);
810 clp
->cl_time
= get_seconds();
814 move_to_confirmed(struct nfs4_client
*clp
)
816 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
817 unsigned int strhashval
;
819 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
820 list_del_init(&clp
->cl_strhash
);
821 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
822 strhashval
= clientstr_hashval(clp
->cl_recdir
);
823 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
827 static struct nfs4_client
*
828 find_confirmed_client(clientid_t
*clid
)
830 struct nfs4_client
*clp
;
831 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
833 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
834 if (same_clid(&clp
->cl_clientid
, clid
))
840 static struct nfs4_client
*
841 find_unconfirmed_client(clientid_t
*clid
)
843 struct nfs4_client
*clp
;
844 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
846 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
847 if (same_clid(&clp
->cl_clientid
, clid
))
854 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
855 * parameter. Matching is based on the fact the at least one of the
856 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
858 * FIXME: we need to unify the clientid namespaces for nfsv4.x
859 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
860 * and SET_CLIENTID{,_CONFIRM}
863 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
865 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
866 return use_exchange_id
== has_exchange_flags
;
869 static struct nfs4_client
*
870 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
871 bool use_exchange_id
)
873 struct nfs4_client
*clp
;
875 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
876 if (same_name(clp
->cl_recdir
, dname
) &&
877 match_clientid_establishment(clp
, use_exchange_id
))
883 static struct nfs4_client
*
884 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
885 bool use_exchange_id
)
887 struct nfs4_client
*clp
;
889 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
890 if (same_name(clp
->cl_recdir
, dname
) &&
891 match_clientid_establishment(clp
, use_exchange_id
))
897 /* a helper function for parse_callback */
899 parse_octet(unsigned int *lenp
, char **addrp
)
901 unsigned int len
= *lenp
;
913 if ((c
< '0') || (c
> '9')) {
919 n
= (n
* 10) + (c
- '0');
930 /* parse and set the setclientid ipv4 callback address */
932 parse_ipv4(unsigned int addr_len
, char *addr_val
, unsigned int *cbaddrp
, unsigned short *cbportp
)
937 u32 addrlen
= addr_len
;
938 char *addr
= addr_val
;
943 for(i
= 4; i
> 0 ; i
--) {
944 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
947 cbaddr
|= (temp
<< shift
);
955 for(i
= 2; i
> 0 ; i
--) {
956 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
959 cbport
|= (temp
<< shift
);
968 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
970 struct nfs4_callback
*cb
= &clp
->cl_callback
;
972 /* Currently, we only support tcp for the callback channel */
973 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
976 if ( !(parse_ipv4(se
->se_callback_addr_len
, se
->se_callback_addr_val
,
977 &cb
->cb_addr
, &cb
->cb_port
)))
979 cb
->cb_prog
= se
->se_callback_prog
;
980 cb
->cb_ident
= se
->se_callback_ident
;
983 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
984 "will not receive delegations\n",
985 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
991 nfsd4_set_statp(struct svc_rqst
*rqstp
, __be32
*statp
)
993 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
995 resp
->cstate
.statp
= statp
;
999 * Dereference the result pages.
1002 nfsd4_release_respages(struct page
**respages
, short resused
)
1006 dprintk("--> %s\n", __func__
);
1007 for (i
= 0; i
< resused
; i
++) {
1010 put_page(respages
[i
]);
1016 nfsd4_copy_pages(struct page
**topages
, struct page
**frompages
, short count
)
1020 for (i
= 0; i
< count
; i
++) {
1021 topages
[i
] = frompages
[i
];
1024 get_page(topages
[i
]);
1029 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1030 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1031 * length of the XDR response is less than se_fmaxresp_cached
1032 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1033 * of the reply (e.g. readdir).
1035 * Store the base and length of the rq_req.head[0] page
1036 * of the NFSv4.1 data, just past the rpc header.
1039 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1041 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1042 struct svc_rqst
*rqstp
= resp
->rqstp
;
1043 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1044 struct nfsd4_op
*op
= &args
->ops
[resp
->opcnt
];
1045 struct kvec
*resv
= &rqstp
->rq_res
.head
[0];
1047 dprintk("--> %s entry %p\n", __func__
, entry
);
1049 /* Don't cache a failed OP_SEQUENCE. */
1050 if (resp
->opcnt
== 1 && op
->opnum
== OP_SEQUENCE
&& resp
->cstate
.status
)
1053 nfsd4_release_respages(entry
->ce_respages
, entry
->ce_resused
);
1054 entry
->ce_opcnt
= resp
->opcnt
;
1055 entry
->ce_status
= resp
->cstate
.status
;
1058 * Don't need a page to cache just the sequence operation - the slot
1062 if (nfsd4_not_cached(resp
)) {
1063 entry
->ce_resused
= 0;
1064 entry
->ce_rpchdrlen
= 0;
1065 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__
,
1066 resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1069 entry
->ce_resused
= rqstp
->rq_resused
;
1070 if (entry
->ce_resused
> NFSD_PAGES_PER_SLOT
+ 1)
1071 entry
->ce_resused
= NFSD_PAGES_PER_SLOT
+ 1;
1072 nfsd4_copy_pages(entry
->ce_respages
, rqstp
->rq_respages
,
1074 entry
->ce_datav
.iov_base
= resp
->cstate
.statp
;
1075 entry
->ce_datav
.iov_len
= resv
->iov_len
- ((char *)resp
->cstate
.statp
-
1076 (char *)page_address(rqstp
->rq_respages
[0]));
1077 /* Current request rpc header length*/
1078 entry
->ce_rpchdrlen
= (char *)resp
->cstate
.statp
-
1079 (char *)page_address(rqstp
->rq_respages
[0]);
1083 * We keep the rpc header, but take the nfs reply from the replycache.
1086 nfsd41_copy_replay_data(struct nfsd4_compoundres
*resp
,
1087 struct nfsd4_cache_entry
*entry
)
1089 struct svc_rqst
*rqstp
= resp
->rqstp
;
1090 struct kvec
*resv
= &resp
->rqstp
->rq_res
.head
[0];
1093 /* Current request rpc header length*/
1094 len
= (char *)resp
->cstate
.statp
-
1095 (char *)page_address(rqstp
->rq_respages
[0]);
1096 if (entry
->ce_datav
.iov_len
+ len
> PAGE_SIZE
) {
1097 dprintk("%s v41 cached reply too large (%Zd).\n", __func__
,
1098 entry
->ce_datav
.iov_len
);
1101 /* copy the cached reply nfsd data past the current rpc header */
1102 memcpy((char *)resv
->iov_base
+ len
, entry
->ce_datav
.iov_base
,
1103 entry
->ce_datav
.iov_len
);
1104 resv
->iov_len
= len
+ entry
->ce_datav
.iov_len
;
1109 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1110 * cached page. Replace any futher replay pages from the cache.
1113 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1114 struct nfsd4_sequence
*seq
)
1116 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1119 dprintk("--> %s entry %p\n", __func__
, entry
);
1122 * If this is just the sequence operation, we did not keep
1123 * a page in the cache entry because we can just use the
1124 * slot info stored in struct nfsd4_sequence that was checked
1125 * against the slot in nfsd4_sequence().
1127 * This occurs when seq->cachethis is FALSE, or when the client
1128 * session inactivity timer fires and a solo sequence operation
1129 * is sent (lease renewal).
1131 if (seq
&& nfsd4_not_cached(resp
)) {
1132 seq
->maxslots
= resp
->cstate
.session
->se_fnumslots
;
1136 if (!nfsd41_copy_replay_data(resp
, entry
)) {
1138 * Not enough room to use the replay rpc header, send the
1139 * cached header. Release all the allocated result pages.
1141 svc_free_res_pages(resp
->rqstp
);
1142 nfsd4_copy_pages(resp
->rqstp
->rq_respages
, entry
->ce_respages
,
1145 /* Release all but the first allocated result page */
1147 resp
->rqstp
->rq_resused
--;
1148 svc_free_res_pages(resp
->rqstp
);
1150 nfsd4_copy_pages(&resp
->rqstp
->rq_respages
[1],
1151 &entry
->ce_respages
[1],
1152 entry
->ce_resused
- 1);
1155 resp
->rqstp
->rq_resused
= entry
->ce_resused
;
1156 resp
->opcnt
= entry
->ce_opcnt
;
1157 resp
->cstate
.iovlen
= entry
->ce_datav
.iov_len
+ entry
->ce_rpchdrlen
;
1158 status
= entry
->ce_status
;
1164 * Set the exchange_id flags returned by the server.
1167 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1169 /* pNFS is not supported */
1170 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1172 /* Referrals are supported, Migration is not. */
1173 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1175 /* set the wire flags to return to client. */
1176 clid
->flags
= new->cl_exchange_flags
;
1180 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1181 struct nfsd4_compound_state
*cstate
,
1182 struct nfsd4_exchange_id
*exid
)
1184 struct nfs4_client
*unconf
, *conf
, *new;
1186 unsigned int strhashval
;
1187 char dname
[HEXDIR_LEN
];
1188 nfs4_verifier verf
= exid
->verifier
;
1189 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1191 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1192 " ip_addr=%u flags %x, spa_how %d\n",
1193 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1194 ip_addr
, exid
->flags
, exid
->spa_how
);
1196 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1197 return nfserr_inval
;
1199 /* Currently only support SP4_NONE */
1200 switch (exid
->spa_how
) {
1204 return nfserr_encr_alg_unsupp
;
1206 BUG(); /* checked by xdr code */
1208 return nfserr_serverfault
; /* no excuse :-/ */
1211 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1216 strhashval
= clientstr_hashval(dname
);
1221 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1223 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1224 /* 18.35.4 case 8 */
1225 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1226 status
= nfserr_not_same
;
1229 /* Client reboot: destroy old state */
1230 expire_client(conf
);
1233 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1234 /* 18.35.4 case 9 */
1235 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1236 status
= nfserr_perm
;
1239 expire_client(conf
);
1242 if (ip_addr
!= conf
->cl_addr
&&
1243 !(exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
)) {
1244 /* Client collision. 18.35.4 case 3 */
1245 status
= nfserr_clid_inuse
;
1249 * Set bit when the owner id and verifier map to an already
1250 * confirmed client id (18.35.3).
1252 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1255 * Falling into 18.35.4 case 2, possible router replay.
1256 * Leave confirmed record intact and return same result.
1258 copy_verf(conf
, &verf
);
1262 /* 18.35.4 case 7 */
1263 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1264 status
= nfserr_noent
;
1269 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1272 * Possible retry or client restart. Per 18.35.4 case 4,
1273 * a new unconfirmed record should be generated regardless
1274 * of whether any properties have changed.
1276 expire_client(unconf
);
1281 new = create_client(exid
->clname
, dname
);
1283 status
= nfserr_resource
;
1287 copy_verf(new, &verf
);
1288 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1289 new->cl_addr
= ip_addr
;
1292 add_to_unconfirmed(new, strhashval
);
1294 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1295 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1297 new->cl_slot
.sl_seqid
= 0;
1299 nfsd4_set_ex_flags(new, exid
);
1301 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1302 new->cl_slot
.sl_seqid
, new->cl_exchange_flags
);
1306 nfs4_unlock_state();
1308 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1313 check_slot_seqid(u32 seqid
, struct nfsd4_slot
*slot
)
1315 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__
, seqid
,
1318 /* The slot is in use, and no response has been sent. */
1319 if (slot
->sl_inuse
) {
1320 if (seqid
== slot
->sl_seqid
)
1321 return nfserr_jukebox
;
1323 return nfserr_seq_misordered
;
1326 if (likely(seqid
== slot
->sl_seqid
+ 1))
1329 if (seqid
== slot
->sl_seqid
)
1330 return nfserr_replay_cache
;
1332 if (seqid
== 1 && (slot
->sl_seqid
+ 1) == 0)
1334 /* Misordered replay or misordered new request */
1335 return nfserr_seq_misordered
;
1339 nfsd4_create_session(struct svc_rqst
*rqstp
,
1340 struct nfsd4_compound_state
*cstate
,
1341 struct nfsd4_create_session
*cr_ses
)
1343 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1344 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1345 struct nfs4_client
*conf
, *unconf
;
1346 struct nfsd4_slot
*slot
= NULL
;
1350 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1351 conf
= find_confirmed_client(&cr_ses
->clientid
);
1354 slot
= &conf
->cl_slot
;
1355 status
= check_slot_seqid(cr_ses
->seqid
, slot
);
1356 if (status
== nfserr_replay_cache
) {
1357 dprintk("Got a create_session replay! seqid= %d\n",
1359 cstate
->slot
= slot
;
1360 cstate
->status
= status
;
1361 /* Return the cached reply status */
1362 status
= nfsd4_replay_cache_entry(resp
, NULL
);
1364 } else if (cr_ses
->seqid
!= conf
->cl_slot
.sl_seqid
+ 1) {
1365 status
= nfserr_seq_misordered
;
1366 dprintk("Sequence misordered!\n");
1367 dprintk("Expected seqid= %d but got seqid= %d\n",
1368 slot
->sl_seqid
, cr_ses
->seqid
);
1371 conf
->cl_slot
.sl_seqid
++;
1372 } else if (unconf
) {
1373 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1374 (ip_addr
!= unconf
->cl_addr
)) {
1375 status
= nfserr_clid_inuse
;
1379 slot
= &unconf
->cl_slot
;
1380 status
= check_slot_seqid(cr_ses
->seqid
, slot
);
1382 /* an unconfirmed replay returns misordered */
1383 status
= nfserr_seq_misordered
;
1387 slot
->sl_seqid
++; /* from 0 to 1 */
1388 move_to_confirmed(unconf
);
1391 * We do not support RDMA or persistent sessions
1393 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1394 cr_ses
->flags
&= ~SESSION4_RDMA
;
1398 status
= nfserr_stale_clientid
;
1402 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1406 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1407 NFS4_MAX_SESSIONID_LEN
);
1408 cr_ses
->seqid
= slot
->sl_seqid
;
1410 slot
->sl_inuse
= true;
1411 cstate
->slot
= slot
;
1412 /* Ensure a page is used for the cache */
1413 slot
->sl_cache_entry
.ce_cachethis
= 1;
1415 nfs4_unlock_state();
1416 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1421 nfsd4_destroy_session(struct svc_rqst
*r
,
1422 struct nfsd4_compound_state
*cstate
,
1423 struct nfsd4_destroy_session
*sessionid
)
1425 return -1; /* stub */
1429 nfsd4_sequence(struct svc_rqst
*rqstp
,
1430 struct nfsd4_compound_state
*cstate
,
1431 struct nfsd4_sequence
*seq
)
1433 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1434 struct nfsd4_session
*session
;
1435 struct nfsd4_slot
*slot
;
1438 if (resp
->opcnt
!= 1)
1439 return nfserr_sequence_pos
;
1441 spin_lock(&sessionid_lock
);
1442 status
= nfserr_badsession
;
1443 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1447 status
= nfserr_badslot
;
1448 if (seq
->slotid
>= session
->se_fnumslots
)
1451 slot
= &session
->se_slots
[seq
->slotid
];
1452 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1454 status
= check_slot_seqid(seq
->seqid
, slot
);
1455 if (status
== nfserr_replay_cache
) {
1456 cstate
->slot
= slot
;
1457 cstate
->session
= session
;
1458 /* Return the cached reply status and set cstate->status
1459 * for nfsd4_svc_encode_compoundres processing */
1460 status
= nfsd4_replay_cache_entry(resp
, seq
);
1461 cstate
->status
= nfserr_replay_cache
;
1467 /* Success! bump slot seqid */
1468 slot
->sl_inuse
= true;
1469 slot
->sl_seqid
= seq
->seqid
;
1470 slot
->sl_cache_entry
.ce_cachethis
= seq
->cachethis
;
1471 /* Always set the cache entry cachethis for solo sequence */
1472 if (nfsd4_is_solo_sequence(resp
))
1473 slot
->sl_cache_entry
.ce_cachethis
= 1;
1475 cstate
->slot
= slot
;
1476 cstate
->session
= session
;
1479 /* Renew the clientid on success and on replay.
1480 * Hold a session reference until done processing the compound:
1481 * nfsd4_put_session called only if the cstate slot is set.
1483 renew_client(session
->se_client
);
1484 nfsd4_get_session(session
);
1486 spin_unlock(&sessionid_lock
);
1487 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1492 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1493 struct nfsd4_setclientid
*setclid
)
1495 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1496 struct xdr_netobj clname
= {
1497 .len
= setclid
->se_namelen
,
1498 .data
= setclid
->se_name
,
1500 nfs4_verifier clverifier
= setclid
->se_verf
;
1501 unsigned int strhashval
;
1502 struct nfs4_client
*conf
, *unconf
, *new;
1505 char dname
[HEXDIR_LEN
];
1507 if (!check_name(clname
))
1508 return nfserr_inval
;
1510 status
= nfs4_make_rec_clidname(dname
, &clname
);
1515 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1516 * We get here on a DRC miss.
1519 strhashval
= clientstr_hashval(dname
);
1522 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1524 /* RFC 3530 14.2.33 CASE 0: */
1525 status
= nfserr_clid_inuse
;
1526 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1527 dprintk("NFSD: setclientid: string in use by client"
1528 " at %pI4\n", &conf
->cl_addr
);
1533 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1534 * has a description of SETCLIENTID request processing consisting
1535 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1537 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1538 status
= nfserr_resource
;
1541 * RFC 3530 14.2.33 CASE 4:
1542 * placed first, because it is the normal case
1545 expire_client(unconf
);
1546 new = create_client(clname
, dname
);
1550 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1552 * RFC 3530 14.2.33 CASE 1:
1553 * probable callback update
1556 /* Note this is removing unconfirmed {*x***},
1557 * which is stronger than RFC recommended {vxc**}.
1558 * This has the advantage that there is at most
1559 * one {*x***} in either list at any time.
1561 expire_client(unconf
);
1563 new = create_client(clname
, dname
);
1566 copy_clid(new, conf
);
1567 } else if (!unconf
) {
1569 * RFC 3530 14.2.33 CASE 2:
1570 * probable client reboot; state will be removed if
1573 new = create_client(clname
, dname
);
1579 * RFC 3530 14.2.33 CASE 3:
1580 * probable client reboot; state will be removed if
1583 expire_client(unconf
);
1584 new = create_client(clname
, dname
);
1589 copy_verf(new, &clverifier
);
1590 new->cl_addr
= sin
->sin_addr
.s_addr
;
1591 new->cl_flavor
= rqstp
->rq_flavor
;
1592 princ
= svc_gss_principal(rqstp
);
1594 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1595 if (new->cl_principal
== NULL
) {
1600 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1602 gen_callback(new, setclid
);
1603 add_to_unconfirmed(new, strhashval
);
1604 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1605 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1606 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1609 nfs4_unlock_state();
1615 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1616 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1617 * bullets, labeled as CASE1 - CASE4 below.
1620 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1621 struct nfsd4_compound_state
*cstate
,
1622 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1624 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1625 struct nfs4_client
*conf
, *unconf
;
1626 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1627 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1630 if (STALE_CLIENTID(clid
))
1631 return nfserr_stale_clientid
;
1633 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1634 * We get here on a DRC miss.
1639 conf
= find_confirmed_client(clid
);
1640 unconf
= find_unconfirmed_client(clid
);
1642 status
= nfserr_clid_inuse
;
1643 if (conf
&& conf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1645 if (unconf
&& unconf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1649 * section 14.2.34 of RFC 3530 has a description of
1650 * SETCLIENTID_CONFIRM request processing consisting
1651 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1653 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1655 * RFC 3530 14.2.34 CASE 1:
1658 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1659 status
= nfserr_clid_inuse
;
1661 /* XXX: We just turn off callbacks until we can handle
1662 * change request correctly. */
1663 atomic_set(&conf
->cl_callback
.cb_set
, 0);
1665 nfsd4_remove_clid_dir(unconf
);
1666 expire_client(unconf
);
1670 } else if (conf
&& !unconf
) {
1672 * RFC 3530 14.2.34 CASE 2:
1673 * probable retransmitted request; play it safe and
1676 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1677 status
= nfserr_clid_inuse
;
1680 } else if (!conf
&& unconf
1681 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1683 * RFC 3530 14.2.34 CASE 3:
1684 * Normal case; new or rebooted client:
1686 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1687 status
= nfserr_clid_inuse
;
1690 clientstr_hashval(unconf
->cl_recdir
);
1691 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1694 nfsd4_remove_clid_dir(conf
);
1695 expire_client(conf
);
1697 move_to_confirmed(unconf
);
1699 nfsd4_probe_callback(conf
);
1702 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1703 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1706 * RFC 3530 14.2.34 CASE 4:
1707 * Client probably hasn't noticed that we rebooted yet.
1709 status
= nfserr_stale_clientid
;
1711 /* check that we have hit one of the cases...*/
1712 status
= nfserr_clid_inuse
;
1715 nfs4_unlock_state();
1719 /* OPEN Share state helper functions */
1720 static inline struct nfs4_file
*
1721 alloc_init_file(struct inode
*ino
)
1723 struct nfs4_file
*fp
;
1724 unsigned int hashval
= file_hashval(ino
);
1726 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1728 atomic_set(&fp
->fi_ref
, 1);
1729 INIT_LIST_HEAD(&fp
->fi_hash
);
1730 INIT_LIST_HEAD(&fp
->fi_stateids
);
1731 INIT_LIST_HEAD(&fp
->fi_delegations
);
1732 spin_lock(&recall_lock
);
1733 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1734 spin_unlock(&recall_lock
);
1735 fp
->fi_inode
= igrab(ino
);
1736 fp
->fi_id
= current_fileid
++;
1737 fp
->fi_had_conflict
= false;
1744 nfsd4_free_slab(struct kmem_cache
**slab
)
1748 kmem_cache_destroy(*slab
);
1753 nfsd4_free_slabs(void)
1755 nfsd4_free_slab(&stateowner_slab
);
1756 nfsd4_free_slab(&file_slab
);
1757 nfsd4_free_slab(&stateid_slab
);
1758 nfsd4_free_slab(&deleg_slab
);
1762 nfsd4_init_slabs(void)
1764 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1765 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1766 if (stateowner_slab
== NULL
)
1768 file_slab
= kmem_cache_create("nfsd4_files",
1769 sizeof(struct nfs4_file
), 0, 0, NULL
);
1770 if (file_slab
== NULL
)
1772 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1773 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1774 if (stateid_slab
== NULL
)
1776 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1777 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1778 if (deleg_slab
== NULL
)
1783 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1788 nfs4_free_stateowner(struct kref
*kref
)
1790 struct nfs4_stateowner
*sop
=
1791 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1792 kfree(sop
->so_owner
.data
);
1793 kmem_cache_free(stateowner_slab
, sop
);
1796 static inline struct nfs4_stateowner
*
1797 alloc_stateowner(struct xdr_netobj
*owner
)
1799 struct nfs4_stateowner
*sop
;
1801 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1802 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1803 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1804 sop
->so_owner
.len
= owner
->len
;
1805 kref_init(&sop
->so_ref
);
1808 kmem_cache_free(stateowner_slab
, sop
);
1813 static struct nfs4_stateowner
*
1814 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1815 struct nfs4_stateowner
*sop
;
1816 struct nfs4_replay
*rp
;
1817 unsigned int idhashval
;
1819 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1821 idhashval
= ownerid_hashval(current_ownerid
);
1822 INIT_LIST_HEAD(&sop
->so_idhash
);
1823 INIT_LIST_HEAD(&sop
->so_strhash
);
1824 INIT_LIST_HEAD(&sop
->so_perclient
);
1825 INIT_LIST_HEAD(&sop
->so_stateids
);
1826 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1827 INIT_LIST_HEAD(&sop
->so_close_lru
);
1829 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1830 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1831 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1832 sop
->so_is_open_owner
= 1;
1833 sop
->so_id
= current_ownerid
++;
1834 sop
->so_client
= clp
;
1835 sop
->so_seqid
= open
->op_seqid
;
1836 sop
->so_confirmed
= 0;
1837 rp
= &sop
->so_replay
;
1838 rp
->rp_status
= nfserr_serverfault
;
1840 rp
->rp_buf
= rp
->rp_ibuf
;
1845 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1846 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1847 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1849 INIT_LIST_HEAD(&stp
->st_hash
);
1850 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1851 INIT_LIST_HEAD(&stp
->st_lockowners
);
1852 INIT_LIST_HEAD(&stp
->st_perfile
);
1853 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1854 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1855 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1856 stp
->st_stateowner
= sop
;
1859 stp
->st_stateid
.si_boot
= boot_time
;
1860 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1861 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1862 stp
->st_stateid
.si_generation
= 0;
1863 stp
->st_access_bmap
= 0;
1864 stp
->st_deny_bmap
= 0;
1865 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
1866 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1867 stp
->st_openstp
= NULL
;
1871 move_to_close_lru(struct nfs4_stateowner
*sop
)
1873 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1875 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1876 sop
->so_time
= get_seconds();
1880 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1883 return (sop
->so_owner
.len
== owner
->len
) &&
1884 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1885 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1888 static struct nfs4_stateowner
*
1889 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1891 struct nfs4_stateowner
*so
= NULL
;
1893 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1894 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1900 /* search file_hashtbl[] for file */
1901 static struct nfs4_file
*
1902 find_file(struct inode
*ino
)
1904 unsigned int hashval
= file_hashval(ino
);
1905 struct nfs4_file
*fp
;
1907 spin_lock(&recall_lock
);
1908 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1909 if (fp
->fi_inode
== ino
) {
1911 spin_unlock(&recall_lock
);
1915 spin_unlock(&recall_lock
);
1919 static inline int access_valid(u32 x
)
1921 if (x
< NFS4_SHARE_ACCESS_READ
)
1923 if (x
> NFS4_SHARE_ACCESS_BOTH
)
1928 static inline int deny_valid(u32 x
)
1930 /* Note: unlike access bits, deny bits may be zero. */
1931 return x
<= NFS4_SHARE_DENY_BOTH
;
1935 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1936 * st_{access,deny}_bmap field of the stateid, in order to track not
1937 * only what share bits are currently in force, but also what
1938 * combinations of share bits previous opens have used. This allows us
1939 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1940 * return an error if the client attempt to downgrade to a combination
1941 * of share bits not explicable by closing some of its previous opens.
1943 * XXX: This enforcement is actually incomplete, since we don't keep
1944 * track of access/deny bit combinations; so, e.g., we allow:
1946 * OPEN allow read, deny write
1947 * OPEN allow both, deny none
1948 * DOWNGRADE allow read, deny none
1950 * which we should reject.
1953 set_access(unsigned int *access
, unsigned long bmap
) {
1957 for (i
= 1; i
< 4; i
++) {
1958 if (test_bit(i
, &bmap
))
1964 set_deny(unsigned int *deny
, unsigned long bmap
) {
1968 for (i
= 0; i
< 4; i
++) {
1969 if (test_bit(i
, &bmap
))
1975 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
1976 unsigned int access
, deny
;
1978 set_access(&access
, stp
->st_access_bmap
);
1979 set_deny(&deny
, stp
->st_deny_bmap
);
1980 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
1986 * Called to check deny when READ with all zero stateid or
1987 * WRITE with all zero or all one stateid
1990 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
1992 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
1993 struct nfs4_file
*fp
;
1994 struct nfs4_stateid
*stp
;
1997 dprintk("NFSD: nfs4_share_conflict\n");
1999 fp
= find_file(ino
);
2002 ret
= nfserr_locked
;
2003 /* Search for conflicting share reservations */
2004 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2005 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2006 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2016 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
2018 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
2019 drop_file_write_access(filp
);
2020 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
2025 * Recall a delegation
2028 do_recall(void *__dp
)
2030 struct nfs4_delegation
*dp
= __dp
;
2032 dp
->dl_file
->fi_had_conflict
= true;
2033 nfsd4_cb_recall(dp
);
2038 * Spawn a thread to perform a recall on the delegation represented
2039 * by the lease (file_lock)
2041 * Called from break_lease() with lock_kernel() held.
2042 * Note: we assume break_lease will only call this *once* for any given
2046 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2048 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2049 struct task_struct
*t
;
2051 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2055 /* We're assuming the state code never drops its reference
2056 * without first removing the lease. Since we're in this lease
2057 * callback (and since the lease code is serialized by the kernel
2058 * lock) we know the server hasn't removed the lease yet, we know
2059 * it's safe to take a reference: */
2060 atomic_inc(&dp
->dl_count
);
2061 atomic_inc(&dp
->dl_client
->cl_count
);
2063 spin_lock(&recall_lock
);
2064 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2065 spin_unlock(&recall_lock
);
2067 /* only place dl_time is set. protected by lock_kernel*/
2068 dp
->dl_time
= get_seconds();
2071 * We don't want the locks code to timeout the lease for us;
2072 * we'll remove it ourself if the delegation isn't returned
2075 fl
->fl_break_time
= 0;
2077 t
= kthread_run(do_recall
, dp
, "%s", "nfs4_cb_recall");
2079 struct nfs4_client
*clp
= dp
->dl_client
;
2081 printk(KERN_INFO
"NFSD: Callback thread failed for "
2082 "for client (clientid %08x/%08x)\n",
2083 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
2084 put_nfs4_client(dp
->dl_client
);
2085 nfs4_put_delegation(dp
);
2090 * The file_lock is being reapd.
2092 * Called by locks_free_lock() with lock_kernel() held.
2095 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2097 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2099 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2101 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2103 dp
->dl_flock
= NULL
;
2107 * Set the delegation file_lock back pointer.
2109 * Called from setlease() with lock_kernel() held.
2112 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2114 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2116 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2123 * Called from setlease() with lock_kernel() held
2126 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2128 struct nfs4_delegation
*onlistd
=
2129 (struct nfs4_delegation
*)onlist
->fl_owner
;
2130 struct nfs4_delegation
*tryd
=
2131 (struct nfs4_delegation
*)try->fl_owner
;
2133 if (onlist
->fl_lmops
!= try->fl_lmops
)
2136 return onlistd
->dl_client
== tryd
->dl_client
;
2141 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2144 return lease_modify(onlist
, arg
);
2149 static struct lock_manager_operations nfsd_lease_mng_ops
= {
2150 .fl_break
= nfsd_break_deleg_cb
,
2151 .fl_release_private
= nfsd_release_deleg_cb
,
2152 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2153 .fl_mylease
= nfsd_same_client_deleg_cb
,
2154 .fl_change
= nfsd_change_deleg_cb
,
2159 nfsd4_process_open1(struct nfsd4_open
*open
)
2161 clientid_t
*clientid
= &open
->op_clientid
;
2162 struct nfs4_client
*clp
= NULL
;
2163 unsigned int strhashval
;
2164 struct nfs4_stateowner
*sop
= NULL
;
2166 if (!check_name(open
->op_owner
))
2167 return nfserr_inval
;
2169 if (STALE_CLIENTID(&open
->op_clientid
))
2170 return nfserr_stale_clientid
;
2172 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2173 sop
= find_openstateowner_str(strhashval
, open
);
2174 open
->op_stateowner
= sop
;
2176 /* Make sure the client's lease hasn't expired. */
2177 clp
= find_confirmed_client(clientid
);
2179 return nfserr_expired
;
2182 if (!sop
->so_confirmed
) {
2183 /* Replace unconfirmed owners without checking for replay. */
2184 clp
= sop
->so_client
;
2185 release_openowner(sop
);
2186 open
->op_stateowner
= NULL
;
2189 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2190 if (sop
->so_replay
.rp_buflen
)
2191 return nfserr_replay_me
;
2192 /* The original OPEN failed so spectacularly
2193 * that we don't even have replay data saved!
2194 * Therefore, we have no choice but to continue
2195 * processing this OPEN; presumably, we'll
2196 * fail again for the same reason.
2198 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2201 if (open
->op_seqid
!= sop
->so_seqid
)
2202 return nfserr_bad_seqid
;
2204 if (open
->op_stateowner
== NULL
) {
2205 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2207 return nfserr_resource
;
2208 open
->op_stateowner
= sop
;
2210 list_del_init(&sop
->so_close_lru
);
2211 renew_client(sop
->so_client
);
2215 static inline __be32
2216 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2218 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2219 return nfserr_openmode
;
2224 static struct nfs4_delegation
*
2225 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2227 struct nfs4_delegation
*dp
;
2229 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2230 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2237 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2238 struct nfs4_delegation
**dp
)
2241 __be32 status
= nfserr_bad_stateid
;
2243 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2246 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
2247 RD_STATE
: WR_STATE
;
2248 status
= nfs4_check_delegmode(*dp
, flags
);
2252 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2256 open
->op_stateowner
->so_confirmed
= 1;
2261 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2263 struct nfs4_stateid
*local
;
2264 __be32 status
= nfserr_share_denied
;
2265 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2267 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2268 /* ignore lock owners */
2269 if (local
->st_stateowner
->so_is_open_owner
== 0)
2271 /* remember if we have seen this open owner */
2272 if (local
->st_stateowner
== sop
)
2274 /* check for conflicting share reservations */
2275 if (!test_share(local
, open
))
2283 static inline struct nfs4_stateid
*
2284 nfs4_alloc_stateid(void)
2286 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2290 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2291 struct nfs4_delegation
*dp
,
2292 struct svc_fh
*cur_fh
, int flags
)
2294 struct nfs4_stateid
*stp
;
2296 stp
= nfs4_alloc_stateid();
2298 return nfserr_resource
;
2301 get_file(dp
->dl_vfs_file
);
2302 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2305 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2308 if (status
== nfserr_dropit
)
2309 status
= nfserr_jukebox
;
2310 kmem_cache_free(stateid_slab
, stp
);
2318 static inline __be32
2319 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2320 struct nfsd4_open
*open
)
2322 struct iattr iattr
= {
2323 .ia_valid
= ATTR_SIZE
,
2326 if (!open
->op_truncate
)
2328 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2329 return nfserr_inval
;
2330 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2334 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2336 struct file
*filp
= stp
->st_vfs_file
;
2337 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2338 unsigned int share_access
, new_writer
;
2341 set_access(&share_access
, stp
->st_access_bmap
);
2342 new_writer
= (~share_access
) & open
->op_share_access
2343 & NFS4_SHARE_ACCESS_WRITE
;
2346 int err
= get_write_access(inode
);
2348 return nfserrno(err
);
2349 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2351 return nfserrno(err
);
2352 file_take_write(filp
);
2354 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2357 put_write_access(inode
);
2360 /* remember the open */
2361 filp
->f_mode
|= open
->op_share_access
;
2362 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2363 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2370 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2372 open
->op_stateowner
->so_confirmed
= 1;
2373 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2377 * Attempt to hand out a delegation.
2380 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2382 struct nfs4_delegation
*dp
;
2383 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2384 struct nfs4_callback
*cb
= &sop
->so_client
->cl_callback
;
2385 struct file_lock fl
, *flp
= &fl
;
2386 int status
, flag
= 0;
2388 flag
= NFS4_OPEN_DELEGATE_NONE
;
2389 open
->op_recall
= 0;
2390 switch (open
->op_claim_type
) {
2391 case NFS4_OPEN_CLAIM_PREVIOUS
:
2392 if (!atomic_read(&cb
->cb_set
))
2393 open
->op_recall
= 1;
2394 flag
= open
->op_delegate_type
;
2395 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2398 case NFS4_OPEN_CLAIM_NULL
:
2399 /* Let's not give out any delegations till everyone's
2400 * had the chance to reclaim theirs.... */
2401 if (locks_in_grace())
2403 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
2405 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2406 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2408 flag
= NFS4_OPEN_DELEGATE_READ
;
2414 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2416 flag
= NFS4_OPEN_DELEGATE_NONE
;
2419 locks_init_lock(&fl
);
2420 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2421 fl
.fl_flags
= FL_LEASE
;
2422 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2423 fl
.fl_end
= OFFSET_MAX
;
2424 fl
.fl_owner
= (fl_owner_t
)dp
;
2425 fl
.fl_file
= stp
->st_vfs_file
;
2426 fl
.fl_pid
= current
->tgid
;
2428 /* vfs_setlease checks to see if delegation should be handed out.
2429 * the lock_manager callbacks fl_mylease and fl_change are used
2431 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2432 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2433 unhash_delegation(dp
);
2434 flag
= NFS4_OPEN_DELEGATE_NONE
;
2438 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2440 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2441 dp
->dl_stateid
.si_boot
,
2442 dp
->dl_stateid
.si_stateownerid
,
2443 dp
->dl_stateid
.si_fileid
,
2444 dp
->dl_stateid
.si_generation
);
2446 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2447 && flag
== NFS4_OPEN_DELEGATE_NONE
2448 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2449 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2450 open
->op_delegate_type
= flag
;
2454 * called with nfs4_lock_state() held.
2457 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2459 struct nfs4_file
*fp
= NULL
;
2460 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2461 struct nfs4_stateid
*stp
= NULL
;
2462 struct nfs4_delegation
*dp
= NULL
;
2465 status
= nfserr_inval
;
2466 if (!access_valid(open
->op_share_access
)
2467 || !deny_valid(open
->op_share_deny
))
2470 * Lookup file; if found, lookup stateid and check open request,
2471 * and check for delegations in the process of being recalled.
2472 * If not found, create the nfs4_file struct
2474 fp
= find_file(ino
);
2476 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2478 status
= nfs4_check_deleg(fp
, open
, &dp
);
2482 status
= nfserr_bad_stateid
;
2483 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2485 status
= nfserr_resource
;
2486 fp
= alloc_init_file(ino
);
2492 * OPEN the file, or upgrade an existing OPEN.
2493 * If truncate fails, the OPEN fails.
2496 /* Stateid was found, this is an OPEN upgrade */
2497 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2500 update_stateid(&stp
->st_stateid
);
2502 /* Stateid was not found, this is a new OPEN */
2504 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2505 flags
|= NFSD_MAY_READ
;
2506 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2507 flags
|= NFSD_MAY_WRITE
;
2508 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2511 init_stateid(stp
, fp
, open
);
2512 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2514 release_open_stateid(stp
);
2518 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2521 * Attempt to hand out a delegation. No error return, because the
2522 * OPEN succeeds even if we fail.
2524 nfs4_open_delegation(current_fh
, open
, stp
);
2528 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2529 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
2530 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
2534 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2535 nfs4_set_claim_prev(open
);
2537 * To finish the open response, we just need to set the rflags.
2539 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2540 if (!open
->op_stateowner
->so_confirmed
)
2541 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2547 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2550 struct nfs4_client
*clp
;
2554 dprintk("process_renew(%08x/%08x): starting\n",
2555 clid
->cl_boot
, clid
->cl_id
);
2556 status
= nfserr_stale_clientid
;
2557 if (STALE_CLIENTID(clid
))
2559 clp
= find_confirmed_client(clid
);
2560 status
= nfserr_expired
;
2562 /* We assume the client took too long to RENEW. */
2563 dprintk("nfsd4_renew: clientid not found!\n");
2567 status
= nfserr_cb_path_down
;
2568 if (!list_empty(&clp
->cl_delegations
)
2569 && !atomic_read(&clp
->cl_callback
.cb_set
))
2573 nfs4_unlock_state();
2577 struct lock_manager nfsd4_manager
= {
2581 nfsd4_end_grace(void)
2583 dprintk("NFSD: end of grace period\n");
2584 nfsd4_recdir_purge_old();
2585 locks_end_grace(&nfsd4_manager
);
2589 nfs4_laundromat(void)
2591 struct nfs4_client
*clp
;
2592 struct nfs4_stateowner
*sop
;
2593 struct nfs4_delegation
*dp
;
2594 struct list_head
*pos
, *next
, reaplist
;
2595 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
2596 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
2597 time_t u
, test_val
= NFSD_LEASE_TIME
;
2601 dprintk("NFSD: laundromat service - starting\n");
2602 if (locks_in_grace())
2604 list_for_each_safe(pos
, next
, &client_lru
) {
2605 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2606 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2607 t
= clp
->cl_time
- cutoff
;
2608 if (clientid_val
> t
)
2612 dprintk("NFSD: purging unused client (clientid %08x)\n",
2613 clp
->cl_clientid
.cl_id
);
2614 nfsd4_remove_clid_dir(clp
);
2617 INIT_LIST_HEAD(&reaplist
);
2618 spin_lock(&recall_lock
);
2619 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2620 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2621 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2622 u
= dp
->dl_time
- cutoff
;
2627 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2629 list_move(&dp
->dl_recall_lru
, &reaplist
);
2631 spin_unlock(&recall_lock
);
2632 list_for_each_safe(pos
, next
, &reaplist
) {
2633 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2634 list_del_init(&dp
->dl_recall_lru
);
2635 unhash_delegation(dp
);
2637 test_val
= NFSD_LEASE_TIME
;
2638 list_for_each_safe(pos
, next
, &close_lru
) {
2639 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2640 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2641 u
= sop
->so_time
- cutoff
;
2646 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2648 release_openowner(sop
);
2650 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2651 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2652 nfs4_unlock_state();
2653 return clientid_val
;
2656 static struct workqueue_struct
*laundry_wq
;
2657 static void laundromat_main(struct work_struct
*);
2658 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2661 laundromat_main(struct work_struct
*not_used
)
2665 t
= nfs4_laundromat();
2666 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2667 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2670 static struct nfs4_stateowner
*
2671 search_close_lru(u32 st_id
, int flags
)
2673 struct nfs4_stateowner
*local
= NULL
;
2675 if (flags
& CLOSE_STATE
) {
2676 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2677 if (local
->so_id
== st_id
)
2685 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2687 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2691 STALE_STATEID(stateid_t
*stateid
)
2693 if (stateid
->si_boot
== boot_time
)
2695 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2696 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2697 stateid
->si_generation
);
2702 access_permit_read(unsigned long access_bmap
)
2704 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2705 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2706 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2710 access_permit_write(unsigned long access_bmap
)
2712 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2713 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2717 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2719 __be32 status
= nfserr_openmode
;
2721 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2723 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2730 static inline __be32
2731 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2733 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2735 else if (locks_in_grace()) {
2736 /* Answer in remaining cases depends on existance of
2737 * conflicting state; so we must wait out the grace period. */
2738 return nfserr_grace
;
2739 } else if (flags
& WR_STATE
)
2740 return nfs4_share_conflict(current_fh
,
2741 NFS4_SHARE_DENY_WRITE
);
2742 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2743 return nfs4_share_conflict(current_fh
,
2744 NFS4_SHARE_DENY_READ
);
2748 * Allow READ/WRITE during grace period on recovered state only for files
2749 * that are not able to provide mandatory locking.
2752 grace_disallows_io(struct inode
*inode
)
2754 return locks_in_grace() && mandatory_lock(inode
);
2757 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
)
2759 /* If the client sends us a stateid from the future, it's buggy: */
2760 if (in
->si_generation
> ref
->si_generation
)
2761 return nfserr_bad_stateid
;
2763 * The following, however, can happen. For example, if the
2764 * client sends an open and some IO at the same time, the open
2765 * may bump si_generation while the IO is still in flight.
2766 * Thanks to hard links and renames, the client never knows what
2767 * file an open will affect. So it could avoid that situation
2768 * only by serializing all opens and IO from the same open
2769 * owner. To recover from the old_stateid error, the client
2770 * will just have to retry the IO:
2772 if (in
->si_generation
< ref
->si_generation
)
2773 return nfserr_old_stateid
;
2777 static int is_delegation_stateid(stateid_t
*stateid
)
2779 return stateid
->si_fileid
== 0;
2783 * Checks for stateid operations
2786 nfs4_preprocess_stateid_op(struct svc_fh
*current_fh
, stateid_t
*stateid
, int flags
, struct file
**filpp
)
2788 struct nfs4_stateid
*stp
= NULL
;
2789 struct nfs4_delegation
*dp
= NULL
;
2790 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2796 if (grace_disallows_io(ino
))
2797 return nfserr_grace
;
2799 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2800 return check_special_stateids(current_fh
, stateid
, flags
);
2802 status
= nfserr_stale_stateid
;
2803 if (STALE_STATEID(stateid
))
2806 status
= nfserr_bad_stateid
;
2807 if (is_delegation_stateid(stateid
)) {
2808 dp
= find_delegation_stateid(ino
, stateid
);
2811 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
);
2814 status
= nfs4_check_delegmode(dp
, flags
);
2817 renew_client(dp
->dl_client
);
2819 *filpp
= dp
->dl_vfs_file
;
2820 } else { /* open or lock stateid */
2821 stp
= find_stateid(stateid
, flags
);
2824 if (nfs4_check_fh(current_fh
, stp
))
2826 if (!stp
->st_stateowner
->so_confirmed
)
2828 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2831 status
= nfs4_check_openmode(stp
, flags
);
2834 renew_client(stp
->st_stateowner
->so_client
);
2836 *filpp
= stp
->st_vfs_file
;
2846 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2847 RD_STATE
: WR_STATE
;
2851 * Checks for sequence id mutating operations.
2854 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
)
2856 struct nfs4_stateid
*stp
;
2857 struct nfs4_stateowner
*sop
;
2860 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2861 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2862 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2863 stateid
->si_generation
);
2868 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2869 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2870 return nfserr_bad_stateid
;
2873 if (STALE_STATEID(stateid
))
2874 return nfserr_stale_stateid
;
2876 * We return BAD_STATEID if filehandle doesn't match stateid,
2877 * the confirmed flag is incorrecly set, or the generation
2878 * number is incorrect.
2880 stp
= find_stateid(stateid
, flags
);
2883 * Also, we should make sure this isn't just the result of
2886 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2888 return nfserr_bad_stateid
;
2894 *sopp
= sop
= stp
->st_stateowner
;
2897 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
2898 struct nfs4_client
*clp
= sop
->so_client
;
2902 lkflg
= setlkflg(lock
->lk_type
);
2904 if (lock
->lk_is_new
) {
2905 if (!sop
->so_is_open_owner
)
2906 return nfserr_bad_stateid
;
2907 if (!same_clid(&clp
->cl_clientid
, lockclid
))
2908 return nfserr_bad_stateid
;
2909 /* stp is the open stateid */
2910 status
= nfs4_check_openmode(stp
, lkflg
);
2914 /* stp is the lock stateid */
2915 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
2921 if (nfs4_check_fh(current_fh
, stp
)) {
2922 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2923 return nfserr_bad_stateid
;
2927 * We now validate the seqid and stateid generation numbers.
2928 * For the moment, we ignore the possibility of
2929 * generation number wraparound.
2931 if (seqid
!= sop
->so_seqid
)
2934 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
2935 dprintk("NFSD: preprocess_seqid_op: expected"
2936 " unconfirmed stateowner!\n");
2937 return nfserr_bad_stateid
;
2939 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
2940 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2941 " confirmed yet!\n");
2942 return nfserr_bad_stateid
;
2944 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2947 renew_client(sop
->so_client
);
2951 if (seqid
== sop
->so_seqid
- 1) {
2952 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2953 /* indicate replay to calling function */
2954 return nfserr_replay_me
;
2956 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2957 sop
->so_seqid
, seqid
);
2959 return nfserr_bad_seqid
;
2963 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2964 struct nfsd4_open_confirm
*oc
)
2967 struct nfs4_stateowner
*sop
;
2968 struct nfs4_stateid
*stp
;
2970 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2971 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2972 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2974 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
2980 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2981 oc
->oc_seqid
, &oc
->oc_req_stateid
,
2982 CONFIRM
| OPEN_STATE
,
2983 &oc
->oc_stateowner
, &stp
, NULL
)))
2986 sop
= oc
->oc_stateowner
;
2987 sop
->so_confirmed
= 1;
2988 update_stateid(&stp
->st_stateid
);
2989 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2990 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2991 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
2992 stp
->st_stateid
.si_boot
,
2993 stp
->st_stateid
.si_stateownerid
,
2994 stp
->st_stateid
.si_fileid
,
2995 stp
->st_stateid
.si_generation
);
2997 nfsd4_create_clid_dir(sop
->so_client
);
2999 if (oc
->oc_stateowner
) {
3000 nfs4_get_stateowner(oc
->oc_stateowner
);
3001 cstate
->replay_owner
= oc
->oc_stateowner
;
3003 nfs4_unlock_state();
3009 * unset all bits in union bitmap (bmap) that
3010 * do not exist in share (from successful OPEN_DOWNGRADE)
3013 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3016 for (i
= 1; i
< 4; i
++) {
3017 if ((i
& access
) != i
)
3018 __clear_bit(i
, bmap
);
3023 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3026 for (i
= 0; i
< 4; i
++) {
3027 if ((i
& deny
) != i
)
3028 __clear_bit(i
, bmap
);
3033 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3034 struct nfsd4_compound_state
*cstate
,
3035 struct nfsd4_open_downgrade
*od
)
3038 struct nfs4_stateid
*stp
;
3039 unsigned int share_access
;
3041 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3042 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3043 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3045 if (!access_valid(od
->od_share_access
)
3046 || !deny_valid(od
->od_share_deny
))
3047 return nfserr_inval
;
3050 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3054 &od
->od_stateowner
, &stp
, NULL
)))
3057 status
= nfserr_inval
;
3058 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3059 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3060 stp
->st_access_bmap
, od
->od_share_access
);
3063 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3064 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3065 stp
->st_deny_bmap
, od
->od_share_deny
);
3068 set_access(&share_access
, stp
->st_access_bmap
);
3069 nfs4_file_downgrade(stp
->st_vfs_file
,
3070 share_access
& ~od
->od_share_access
);
3072 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3073 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3075 update_stateid(&stp
->st_stateid
);
3076 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3079 if (od
->od_stateowner
) {
3080 nfs4_get_stateowner(od
->od_stateowner
);
3081 cstate
->replay_owner
= od
->od_stateowner
;
3083 nfs4_unlock_state();
3088 * nfs4_unlock_state() called after encode
3091 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3092 struct nfsd4_close
*close
)
3095 struct nfs4_stateid
*stp
;
3097 dprintk("NFSD: nfsd4_close on file %.*s\n",
3098 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3099 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3102 /* check close_lru for replay */
3103 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3106 OPEN_STATE
| CLOSE_STATE
,
3107 &close
->cl_stateowner
, &stp
, NULL
)))
3110 update_stateid(&stp
->st_stateid
);
3111 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3113 /* release_stateid() calls nfsd_close() if needed */
3114 release_open_stateid(stp
);
3116 /* place unused nfs4_stateowners on so_close_lru list to be
3117 * released by the laundromat service after the lease period
3118 * to enable us to handle CLOSE replay
3120 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3121 move_to_close_lru(close
->cl_stateowner
);
3123 if (close
->cl_stateowner
) {
3124 nfs4_get_stateowner(close
->cl_stateowner
);
3125 cstate
->replay_owner
= close
->cl_stateowner
;
3127 nfs4_unlock_state();
3132 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3133 struct nfsd4_delegreturn
*dr
)
3135 struct nfs4_delegation
*dp
;
3136 stateid_t
*stateid
= &dr
->dr_stateid
;
3137 struct inode
*inode
;
3140 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3142 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3145 status
= nfserr_bad_stateid
;
3146 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3148 status
= nfserr_stale_stateid
;
3149 if (STALE_STATEID(stateid
))
3151 status
= nfserr_bad_stateid
;
3152 if (!is_delegation_stateid(stateid
))
3154 dp
= find_delegation_stateid(inode
, stateid
);
3157 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
);
3160 renew_client(dp
->dl_client
);
3162 unhash_delegation(dp
);
3164 nfs4_unlock_state();
3171 * Lock owner state (byte-range locks)
3173 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3174 #define LOCK_HASH_BITS 8
3175 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3176 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3179 end_offset(u64 start
, u64 len
)
3184 return end
>= start
? end
: NFS4_MAX_UINT64
;
3187 /* last octet in a range */
3189 last_byte_offset(u64 start
, u64 len
)
3195 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3198 #define lockownerid_hashval(id) \
3199 ((id) & LOCK_HASH_MASK)
3201 static inline unsigned int
3202 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3203 struct xdr_netobj
*ownername
)
3205 return (file_hashval(inode
) + cl_id
3206 + opaque_hashval(ownername
->data
, ownername
->len
))
3210 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3211 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3212 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3214 static struct nfs4_stateid
*
3215 find_stateid(stateid_t
*stid
, int flags
)
3217 struct nfs4_stateid
*local
;
3218 u32 st_id
= stid
->si_stateownerid
;
3219 u32 f_id
= stid
->si_fileid
;
3220 unsigned int hashval
;
3222 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3223 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3224 hashval
= stateid_hashval(st_id
, f_id
);
3225 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3226 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3227 (local
->st_stateid
.si_fileid
== f_id
))
3232 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3233 hashval
= stateid_hashval(st_id
, f_id
);
3234 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3235 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3236 (local
->st_stateid
.si_fileid
== f_id
))
3243 static struct nfs4_delegation
*
3244 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3246 struct nfs4_file
*fp
;
3247 struct nfs4_delegation
*dl
;
3249 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3250 stid
->si_boot
, stid
->si_stateownerid
,
3251 stid
->si_fileid
, stid
->si_generation
);
3253 fp
= find_file(ino
);
3256 dl
= find_delegation_file(fp
, stid
);
3262 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3263 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3264 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3265 * locking, this prevents us from being completely protocol-compliant. The
3266 * real solution to this problem is to start using unsigned file offsets in
3267 * the VFS, but this is a very deep change!
3270 nfs4_transform_lock_offset(struct file_lock
*lock
)
3272 if (lock
->fl_start
< 0)
3273 lock
->fl_start
= OFFSET_MAX
;
3274 if (lock
->fl_end
< 0)
3275 lock
->fl_end
= OFFSET_MAX
;
3278 /* Hack!: For now, we're defining this just so we can use a pointer to it
3279 * as a unique cookie to identify our (NFSv4's) posix locks. */
3280 static struct lock_manager_operations nfsd_posix_mng_ops
= {
3284 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3286 struct nfs4_stateowner
*sop
;
3289 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3290 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3291 hval
= lockownerid_hashval(sop
->so_id
);
3292 kref_get(&sop
->so_ref
);
3294 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3296 deny
->ld_sop
= NULL
;
3297 deny
->ld_clientid
.cl_boot
= 0;
3298 deny
->ld_clientid
.cl_id
= 0;
3300 deny
->ld_start
= fl
->fl_start
;
3301 deny
->ld_length
= NFS4_MAX_UINT64
;
3302 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3303 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3304 deny
->ld_type
= NFS4_READ_LT
;
3305 if (fl
->fl_type
!= F_RDLCK
)
3306 deny
->ld_type
= NFS4_WRITE_LT
;
3309 static struct nfs4_stateowner
*
3310 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3311 struct xdr_netobj
*owner
)
3313 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3314 struct nfs4_stateowner
*op
;
3316 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3317 if (same_owner_str(op
, owner
, clid
))
3324 * Alloc a lock owner structure.
3325 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3328 * strhashval = lock_ownerstr_hashval
3331 static struct nfs4_stateowner
*
3332 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3333 struct nfs4_stateowner
*sop
;
3334 struct nfs4_replay
*rp
;
3335 unsigned int idhashval
;
3337 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3339 idhashval
= lockownerid_hashval(current_ownerid
);
3340 INIT_LIST_HEAD(&sop
->so_idhash
);
3341 INIT_LIST_HEAD(&sop
->so_strhash
);
3342 INIT_LIST_HEAD(&sop
->so_perclient
);
3343 INIT_LIST_HEAD(&sop
->so_stateids
);
3344 INIT_LIST_HEAD(&sop
->so_perstateid
);
3345 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3347 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3348 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3349 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3350 sop
->so_is_open_owner
= 0;
3351 sop
->so_id
= current_ownerid
++;
3352 sop
->so_client
= clp
;
3353 /* It is the openowner seqid that will be incremented in encode in the
3354 * case of new lockowners; so increment the lock seqid manually: */
3355 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3356 sop
->so_confirmed
= 1;
3357 rp
= &sop
->so_replay
;
3358 rp
->rp_status
= nfserr_serverfault
;
3360 rp
->rp_buf
= rp
->rp_ibuf
;
3364 static struct nfs4_stateid
*
3365 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3367 struct nfs4_stateid
*stp
;
3368 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3370 stp
= nfs4_alloc_stateid();
3373 INIT_LIST_HEAD(&stp
->st_hash
);
3374 INIT_LIST_HEAD(&stp
->st_perfile
);
3375 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3376 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3377 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3378 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3379 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3380 stp
->st_stateowner
= sop
;
3383 stp
->st_stateid
.si_boot
= boot_time
;
3384 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3385 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3386 stp
->st_stateid
.si_generation
= 0;
3387 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3388 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3389 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3390 stp
->st_openstp
= open_stp
;
3397 check_lock_length(u64 offset
, u64 length
)
3399 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3400 LOFF_OVERFLOW(offset
, length
)));
3407 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3408 struct nfsd4_lock
*lock
)
3410 struct nfs4_stateowner
*open_sop
= NULL
;
3411 struct nfs4_stateowner
*lock_sop
= NULL
;
3412 struct nfs4_stateid
*lock_stp
;
3414 struct file_lock file_lock
;
3415 struct file_lock conflock
;
3417 unsigned int strhashval
;
3421 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3422 (long long) lock
->lk_offset
,
3423 (long long) lock
->lk_length
);
3425 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3426 return nfserr_inval
;
3428 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3429 S_IFREG
, NFSD_MAY_LOCK
))) {
3430 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3436 if (lock
->lk_is_new
) {
3438 * Client indicates that this is a new lockowner.
3439 * Use open owner and open stateid to create lock owner and
3442 struct nfs4_stateid
*open_stp
= NULL
;
3443 struct nfs4_file
*fp
;
3445 status
= nfserr_stale_clientid
;
3446 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
3449 /* validate and update open stateid and open seqid */
3450 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3451 lock
->lk_new_open_seqid
,
3452 &lock
->lk_new_open_stateid
,
3454 &lock
->lk_replay_owner
, &open_stp
,
3458 open_sop
= lock
->lk_replay_owner
;
3459 /* create lockowner and lock stateid */
3460 fp
= open_stp
->st_file
;
3461 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3462 open_sop
->so_client
->cl_clientid
.cl_id
,
3463 &lock
->v
.new.owner
);
3464 /* XXX: Do we need to check for duplicate stateowners on
3465 * the same file, or should they just be allowed (and
3466 * create new stateids)? */
3467 status
= nfserr_resource
;
3468 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3469 open_sop
->so_client
, open_stp
, lock
);
3470 if (lock_sop
== NULL
)
3472 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3473 if (lock_stp
== NULL
)
3476 /* lock (lock owner + lock stateid) already exists */
3477 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3478 lock
->lk_old_lock_seqid
,
3479 &lock
->lk_old_lock_stateid
,
3481 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3484 lock_sop
= lock
->lk_replay_owner
;
3486 /* lock->lk_replay_owner and lock_stp have been created or found */
3487 filp
= lock_stp
->st_vfs_file
;
3489 status
= nfserr_grace
;
3490 if (locks_in_grace() && !lock
->lk_reclaim
)
3492 status
= nfserr_no_grace
;
3493 if (!locks_in_grace() && lock
->lk_reclaim
)
3496 locks_init_lock(&file_lock
);
3497 switch (lock
->lk_type
) {
3500 file_lock
.fl_type
= F_RDLCK
;
3504 case NFS4_WRITEW_LT
:
3505 file_lock
.fl_type
= F_WRLCK
;
3509 status
= nfserr_inval
;
3512 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3513 file_lock
.fl_pid
= current
->tgid
;
3514 file_lock
.fl_file
= filp
;
3515 file_lock
.fl_flags
= FL_POSIX
;
3516 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3518 file_lock
.fl_start
= lock
->lk_offset
;
3519 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3520 nfs4_transform_lock_offset(&file_lock
);
3523 * Try to lock the file in the VFS.
3524 * Note: locks.c uses the BKL to protect the inode's lock list.
3527 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3529 case 0: /* success! */
3530 update_stateid(&lock_stp
->st_stateid
);
3531 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3535 case (EAGAIN
): /* conflock holds conflicting lock */
3536 status
= nfserr_denied
;
3537 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3538 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3541 status
= nfserr_deadlock
;
3544 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3545 status
= nfserr_resource
;
3549 if (status
&& lock
->lk_is_new
&& lock_sop
)
3550 release_lockowner(lock_sop
);
3551 if (lock
->lk_replay_owner
) {
3552 nfs4_get_stateowner(lock
->lk_replay_owner
);
3553 cstate
->replay_owner
= lock
->lk_replay_owner
;
3555 nfs4_unlock_state();
3560 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3561 * so we do a temporary open here just to get an open file to pass to
3562 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3565 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3570 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3573 err
= vfs_test_lock(file
, lock
);
3582 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3583 struct nfsd4_lockt
*lockt
)
3585 struct inode
*inode
;
3586 struct file_lock file_lock
;
3590 if (locks_in_grace())
3591 return nfserr_grace
;
3593 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3594 return nfserr_inval
;
3596 lockt
->lt_stateowner
= NULL
;
3599 status
= nfserr_stale_clientid
;
3600 if (STALE_CLIENTID(&lockt
->lt_clientid
))
3603 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3604 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3605 if (status
== nfserr_symlink
)
3606 status
= nfserr_inval
;
3610 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3611 locks_init_lock(&file_lock
);
3612 switch (lockt
->lt_type
) {
3615 file_lock
.fl_type
= F_RDLCK
;
3618 case NFS4_WRITEW_LT
:
3619 file_lock
.fl_type
= F_WRLCK
;
3622 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3623 status
= nfserr_inval
;
3627 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3628 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3629 if (lockt
->lt_stateowner
)
3630 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3631 file_lock
.fl_pid
= current
->tgid
;
3632 file_lock
.fl_flags
= FL_POSIX
;
3634 file_lock
.fl_start
= lockt
->lt_offset
;
3635 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3637 nfs4_transform_lock_offset(&file_lock
);
3640 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3642 status
= nfserrno(error
);
3645 if (file_lock
.fl_type
!= F_UNLCK
) {
3646 status
= nfserr_denied
;
3647 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3650 nfs4_unlock_state();
3655 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3656 struct nfsd4_locku
*locku
)
3658 struct nfs4_stateid
*stp
;
3659 struct file
*filp
= NULL
;
3660 struct file_lock file_lock
;
3664 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3665 (long long) locku
->lu_offset
,
3666 (long long) locku
->lu_length
);
3668 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3669 return nfserr_inval
;
3673 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
3677 &locku
->lu_stateowner
, &stp
, NULL
)))
3680 filp
= stp
->st_vfs_file
;
3682 locks_init_lock(&file_lock
);
3683 file_lock
.fl_type
= F_UNLCK
;
3684 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3685 file_lock
.fl_pid
= current
->tgid
;
3686 file_lock
.fl_file
= filp
;
3687 file_lock
.fl_flags
= FL_POSIX
;
3688 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3689 file_lock
.fl_start
= locku
->lu_offset
;
3691 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3692 nfs4_transform_lock_offset(&file_lock
);
3695 * Try to unlock the file in the VFS.
3697 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3699 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3703 * OK, unlock succeeded; the only thing left to do is update the stateid.
3705 update_stateid(&stp
->st_stateid
);
3706 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3709 if (locku
->lu_stateowner
) {
3710 nfs4_get_stateowner(locku
->lu_stateowner
);
3711 cstate
->replay_owner
= locku
->lu_stateowner
;
3713 nfs4_unlock_state();
3717 status
= nfserrno(err
);
3723 * 1: locks held by lockowner
3724 * 0: no locks held by lockowner
3727 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3729 struct file_lock
**flpp
;
3730 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3734 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3735 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3746 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3747 struct nfsd4_compound_state
*cstate
,
3748 struct nfsd4_release_lockowner
*rlockowner
)
3750 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3751 struct nfs4_stateowner
*sop
;
3752 struct nfs4_stateid
*stp
;
3753 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3754 struct list_head matches
;
3758 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3759 clid
->cl_boot
, clid
->cl_id
);
3761 /* XXX check for lease expiration */
3763 status
= nfserr_stale_clientid
;
3764 if (STALE_CLIENTID(clid
))
3769 status
= nfserr_locks_held
;
3770 /* XXX: we're doing a linear search through all the lockowners.
3771 * Yipes! For now we'll just hope clients aren't really using
3772 * release_lockowner much, but eventually we have to fix these
3773 * data structures. */
3774 INIT_LIST_HEAD(&matches
);
3775 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3776 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3777 if (!same_owner_str(sop
, owner
, clid
))
3779 list_for_each_entry(stp
, &sop
->so_stateids
,
3781 if (check_for_locks(stp
->st_vfs_file
, sop
))
3783 /* Note: so_perclient unused for lockowners,
3784 * so it's OK to fool with here. */
3785 list_add(&sop
->so_perclient
, &matches
);
3789 /* Clients probably won't expect us to return with some (but not all)
3790 * of the lockowner state released; so don't release any until all
3791 * have been checked. */
3793 while (!list_empty(&matches
)) {
3794 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3796 /* unhash_stateowner deletes so_perclient only
3797 * for openowners. */
3798 list_del(&sop
->so_perclient
);
3799 release_lockowner(sop
);
3802 nfs4_unlock_state();
3806 static inline struct nfs4_client_reclaim
*
3809 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3813 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3815 unsigned int strhashval
= clientstr_hashval(name
);
3816 struct nfs4_client
*clp
;
3818 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3823 * failure => all reset bets are off, nfserr_no_grace...
3826 nfs4_client_to_reclaim(const char *name
)
3828 unsigned int strhashval
;
3829 struct nfs4_client_reclaim
*crp
= NULL
;
3831 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3832 crp
= alloc_reclaim();
3835 strhashval
= clientstr_hashval(name
);
3836 INIT_LIST_HEAD(&crp
->cr_strhash
);
3837 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3838 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3839 reclaim_str_hashtbl_size
++;
3844 nfs4_release_reclaim(void)
3846 struct nfs4_client_reclaim
*crp
= NULL
;
3849 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3850 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3851 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3852 struct nfs4_client_reclaim
, cr_strhash
);
3853 list_del(&crp
->cr_strhash
);
3855 reclaim_str_hashtbl_size
--;
3858 BUG_ON(reclaim_str_hashtbl_size
);
3862 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3863 static struct nfs4_client_reclaim
*
3864 nfs4_find_reclaim_client(clientid_t
*clid
)
3866 unsigned int strhashval
;
3867 struct nfs4_client
*clp
;
3868 struct nfs4_client_reclaim
*crp
= NULL
;
3871 /* find clientid in conf_id_hashtbl */
3872 clp
= find_confirmed_client(clid
);
3876 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3877 clp
->cl_name
.len
, clp
->cl_name
.data
,
3880 /* find clp->cl_name in reclaim_str_hashtbl */
3881 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3882 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3883 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3891 * Called from OPEN. Look for clientid in reclaim list.
3894 nfs4_check_open_reclaim(clientid_t
*clid
)
3896 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3899 /* initialization to perform at module load time: */
3902 nfs4_state_init(void)
3906 status
= nfsd4_init_slabs();
3909 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3910 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
3911 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
3912 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
3913 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
3915 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
3916 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
3917 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3918 INIT_LIST_HEAD(&file_hashtbl
[i
]);
3920 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
3921 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
3922 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
3924 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
3925 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
3926 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
3928 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3929 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
3930 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
3932 memset(&onestateid
, ~0, sizeof(stateid_t
));
3933 INIT_LIST_HEAD(&close_lru
);
3934 INIT_LIST_HEAD(&client_lru
);
3935 INIT_LIST_HEAD(&del_recall_lru
);
3936 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++)
3937 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
3938 reclaim_str_hashtbl_size
= 0;
3943 nfsd4_load_reboot_recovery_data(void)
3948 nfsd4_init_recdir(user_recovery_dirname
);
3949 status
= nfsd4_recdir_load();
3950 nfs4_unlock_state();
3952 printk("NFSD: Failure reading reboot recovery data\n");
3956 get_nfs4_grace_period(void)
3958 return max(user_lease_time
, lease_time
) * HZ
;
3962 * Since the lifetime of a delegation isn't limited to that of an open, a
3963 * client may quite reasonably hang on to a delegation as long as it has
3964 * the inode cached. This becomes an obvious problem the first time a
3965 * client's inode cache approaches the size of the server's total memory.
3967 * For now we avoid this problem by imposing a hard limit on the number
3968 * of delegations, which varies according to the server's memory size.
3971 set_max_delegations(void)
3974 * Allow at most 4 delegations per megabyte of RAM. Quick
3975 * estimates suggest that in the worst case (where every delegation
3976 * is for a different inode), a delegation could take about 1.5K,
3977 * giving a worst case usage of about 6% of memory.
3979 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
3982 /* initialization to perform when the nfsd service is started: */
3985 __nfs4_state_start(void)
3987 unsigned long grace_time
;
3989 boot_time
= get_seconds();
3990 grace_time
= get_nfs4_grace_period();
3991 lease_time
= user_lease_time
;
3992 locks_start_grace(&nfsd4_manager
);
3993 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
3995 laundry_wq
= create_singlethread_workqueue("nfsd4");
3996 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
3997 set_max_delegations();
4001 nfs4_state_start(void)
4005 nfsd4_load_reboot_recovery_data();
4006 __nfs4_state_start();
4012 nfs4_lease_time(void)
4018 __nfs4_state_shutdown(void)
4021 struct nfs4_client
*clp
= NULL
;
4022 struct nfs4_delegation
*dp
= NULL
;
4023 struct list_head
*pos
, *next
, reaplist
;
4025 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4026 while (!list_empty(&conf_id_hashtbl
[i
])) {
4027 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4030 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4031 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4035 INIT_LIST_HEAD(&reaplist
);
4036 spin_lock(&recall_lock
);
4037 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4038 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4039 list_move(&dp
->dl_recall_lru
, &reaplist
);
4041 spin_unlock(&recall_lock
);
4042 list_for_each_safe(pos
, next
, &reaplist
) {
4043 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4044 list_del_init(&dp
->dl_recall_lru
);
4045 unhash_delegation(dp
);
4048 nfsd4_shutdown_recdir();
4053 nfs4_state_shutdown(void)
4055 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4056 destroy_workqueue(laundry_wq
);
4057 locks_end_grace(&nfsd4_manager
);
4059 nfs4_release_reclaim();
4060 __nfs4_state_shutdown();
4061 nfs4_unlock_state();
4065 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4066 * accessed when nfsd is starting.
4069 nfs4_set_recdir(char *recdir
)
4071 strcpy(user_recovery_dirname
, recdir
);
4075 * Change the NFSv4 recovery directory to recdir.
4078 nfs4_reset_recoverydir(char *recdir
)
4083 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4087 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4088 nfs4_set_recdir(recdir
);
4096 nfs4_recoverydir(void)
4098 return user_recovery_dirname
;
4102 * Called when leasetime is changed.
4104 * The only way the protocol gives us to handle on-the-fly lease changes is to
4105 * simulate a reboot. Instead of doing that, we just wait till the next time
4106 * we start to register any changes in lease time. If the administrator
4107 * really wants to change the lease time *now*, they can go ahead and bring
4108 * nfsd down and then back up again after changing the lease time.
4110 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4111 * when nfsd is starting
4114 nfs4_reset_lease(time_t leasetime
)
4116 user_lease_time
= leasetime
;