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_cb_conn
*cb
= &stp
->st_stateowner
->so_client
->cl_cb_conn
;
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_ident
= cb
->cb_ident
;
207 dp
->dl_stateid
.si_boot
= get_seconds();
208 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
209 dp
->dl_stateid
.si_fileid
= 0;
210 dp
->dl_stateid
.si_generation
= 0;
211 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
213 atomic_set(&dp
->dl_count
, 1);
214 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
215 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
220 nfs4_put_delegation(struct nfs4_delegation
*dp
)
222 if (atomic_dec_and_test(&dp
->dl_count
)) {
223 dprintk("NFSD: freeing dp %p\n",dp
);
224 put_nfs4_file(dp
->dl_file
);
225 kmem_cache_free(deleg_slab
, dp
);
230 /* Remove the associated file_lock first, then remove the delegation.
231 * lease_modify() is called to remove the FS_LEASE file_lock from
232 * the i_flock list, eventually calling nfsd's lock_manager
233 * fl_release_callback.
236 nfs4_close_delegation(struct nfs4_delegation
*dp
)
238 struct file
*filp
= dp
->dl_vfs_file
;
240 dprintk("NFSD: close_delegation dp %p\n",dp
);
241 dp
->dl_vfs_file
= NULL
;
242 /* The following nfsd_close may not actually close the file,
243 * but we want to remove the lease in any case. */
245 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
249 /* Called under the state lock. */
251 unhash_delegation(struct nfs4_delegation
*dp
)
253 list_del_init(&dp
->dl_perfile
);
254 list_del_init(&dp
->dl_perclnt
);
255 spin_lock(&recall_lock
);
256 list_del_init(&dp
->dl_recall_lru
);
257 spin_unlock(&recall_lock
);
258 nfs4_close_delegation(dp
);
259 nfs4_put_delegation(dp
);
266 /* Hash tables for nfs4_clientid state */
267 #define CLIENT_HASH_BITS 4
268 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
269 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271 #define clientid_hashval(id) \
272 ((id) & CLIENT_HASH_MASK)
273 #define clientstr_hashval(name) \
274 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
276 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
277 * used in reboot/reset lease grace period processing
279 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
280 * setclientid_confirmed info.
282 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
285 * client_lru holds client queue ordered by nfs4_client.cl_time
288 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
289 * for last close replay.
291 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
292 static int reclaim_str_hashtbl_size
= 0;
293 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
294 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
295 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
296 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
297 static struct list_head client_lru
;
298 static struct list_head close_lru
;
300 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
302 list_del(&stp
->st_hash
);
303 list_del(&stp
->st_perfile
);
304 list_del(&stp
->st_perstateowner
);
307 static void free_generic_stateid(struct nfs4_stateid
*stp
)
309 put_nfs4_file(stp
->st_file
);
310 kmem_cache_free(stateid_slab
, stp
);
313 static void release_lock_stateid(struct nfs4_stateid
*stp
)
315 unhash_generic_stateid(stp
);
316 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
317 free_generic_stateid(stp
);
320 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
322 struct nfs4_stateid
*stp
;
324 list_del(&sop
->so_idhash
);
325 list_del(&sop
->so_strhash
);
326 list_del(&sop
->so_perstateid
);
327 while (!list_empty(&sop
->so_stateids
)) {
328 stp
= list_first_entry(&sop
->so_stateids
,
329 struct nfs4_stateid
, st_perstateowner
);
330 release_lock_stateid(stp
);
334 static void release_lockowner(struct nfs4_stateowner
*sop
)
336 unhash_lockowner(sop
);
337 nfs4_put_stateowner(sop
);
341 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
343 struct nfs4_stateowner
*lock_sop
;
345 while (!list_empty(&open_stp
->st_lockowners
)) {
346 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
347 struct nfs4_stateowner
, so_perstateid
);
348 /* list_del(&open_stp->st_lockowners); */
349 BUG_ON(lock_sop
->so_is_open_owner
);
350 release_lockowner(lock_sop
);
354 static void release_open_stateid(struct nfs4_stateid
*stp
)
356 unhash_generic_stateid(stp
);
357 release_stateid_lockowners(stp
);
358 nfsd_close(stp
->st_vfs_file
);
359 free_generic_stateid(stp
);
362 static void unhash_openowner(struct nfs4_stateowner
*sop
)
364 struct nfs4_stateid
*stp
;
366 list_del(&sop
->so_idhash
);
367 list_del(&sop
->so_strhash
);
368 list_del(&sop
->so_perclient
);
369 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
370 while (!list_empty(&sop
->so_stateids
)) {
371 stp
= list_first_entry(&sop
->so_stateids
,
372 struct nfs4_stateid
, st_perstateowner
);
373 release_open_stateid(stp
);
377 static void release_openowner(struct nfs4_stateowner
*sop
)
379 unhash_openowner(sop
);
380 list_del(&sop
->so_close_lru
);
381 nfs4_put_stateowner(sop
);
384 static DEFINE_SPINLOCK(sessionid_lock
);
385 #define SESSION_HASH_SIZE 512
386 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
389 hash_sessionid(struct nfs4_sessionid
*sessionid
)
391 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
393 return sid
->sequence
% SESSION_HASH_SIZE
;
397 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
399 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
400 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
404 gen_sessionid(struct nfsd4_session
*ses
)
406 struct nfs4_client
*clp
= ses
->se_client
;
407 struct nfsd4_sessionid
*sid
;
409 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
410 sid
->clientid
= clp
->cl_clientid
;
411 sid
->sequence
= current_sessionid
++;
416 * Give the client the number of slots it requests bound by
417 * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
419 * If we run out of reserved DRC memory we should (up to a point) re-negotiate
420 * active sessions and reduce their slot usage to make rooom for new
421 * connections. For now we just fail the create session.
423 static int set_forechannel_maxreqs(struct nfsd4_channel_attrs
*fchan
)
427 if (fchan
->maxreqs
< 1)
429 else if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
430 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
432 mem
= fchan
->maxreqs
* NFSD_SLOT_CACHE_SIZE
;
434 spin_lock(&nfsd_drc_lock
);
435 if (mem
+ nfsd_drc_mem_used
> nfsd_drc_max_mem
)
436 mem
= ((nfsd_drc_max_mem
- nfsd_drc_mem_used
) /
437 NFSD_SLOT_CACHE_SIZE
) * NFSD_SLOT_CACHE_SIZE
;
438 nfsd_drc_mem_used
+= mem
;
439 spin_unlock(&nfsd_drc_lock
);
441 fchan
->maxreqs
= mem
/ NFSD_SLOT_CACHE_SIZE
;
442 if (fchan
->maxreqs
== 0)
443 return nfserr_resource
;
448 * fchan holds the client values on input, and the server values on output
450 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
451 struct nfsd4_channel_attrs
*session_fchan
,
452 struct nfsd4_channel_attrs
*fchan
)
455 __u32 maxcount
= svc_max_payload(rqstp
);
457 /* headerpadsz set to zero in encode routine */
459 /* Use the client's max request and max response size if possible */
460 if (fchan
->maxreq_sz
> maxcount
)
461 fchan
->maxreq_sz
= maxcount
;
462 session_fchan
->maxreq_sz
= fchan
->maxreq_sz
;
464 if (fchan
->maxresp_sz
> maxcount
)
465 fchan
->maxresp_sz
= maxcount
;
466 session_fchan
->maxresp_sz
= fchan
->maxresp_sz
;
468 session_fchan
->maxresp_cached
= NFSD_SLOT_CACHE_SIZE
;
469 fchan
->maxresp_cached
= session_fchan
->maxresp_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_fchan
->maxops
= fchan
->maxops
;
476 /* FIXME: Error means no more DRC pages so the server should
477 * recover pages from existing sessions. For now fail session
480 status
= set_forechannel_maxreqs(fchan
);
482 session_fchan
->maxreqs
= fchan
->maxreqs
;
487 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
488 struct nfsd4_create_session
*cses
)
490 struct nfsd4_session
*new, tmp
;
491 int idx
, status
= nfserr_resource
, slotsize
;
493 memset(&tmp
, 0, sizeof(tmp
));
495 /* FIXME: For now, we just accept the client back channel attributes. */
496 tmp
.se_bchannel
= cses
->back_channel
;
497 status
= init_forechannel_attrs(rqstp
, &tmp
.se_fchannel
,
498 &cses
->fore_channel
);
502 /* allocate struct nfsd4_session and slot table in one piece */
503 slotsize
= tmp
.se_fchannel
.maxreqs
* sizeof(struct nfsd4_slot
);
504 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
508 memcpy(new, &tmp
, sizeof(*new));
510 new->se_client
= clp
;
512 idx
= hash_sessionid(&new->se_sessionid
);
513 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
514 NFS4_MAX_SESSIONID_LEN
);
516 new->se_flags
= cses
->flags
;
517 kref_init(&new->se_ref
);
518 spin_lock(&sessionid_lock
);
519 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
520 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
521 spin_unlock(&sessionid_lock
);
528 /* caller must hold sessionid_lock */
529 static struct nfsd4_session
*
530 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
532 struct nfsd4_session
*elem
;
535 dump_sessionid(__func__
, sessionid
);
536 idx
= hash_sessionid(sessionid
);
537 dprintk("%s: idx is %d\n", __func__
, idx
);
538 /* Search in the appropriate list */
539 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
540 dump_sessionid("list traversal", &elem
->se_sessionid
);
541 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
542 NFS4_MAX_SESSIONID_LEN
)) {
547 dprintk("%s: session not found\n", __func__
);
551 /* caller must hold sessionid_lock */
553 unhash_session(struct nfsd4_session
*ses
)
555 list_del(&ses
->se_hash
);
556 list_del(&ses
->se_perclnt
);
560 release_session(struct nfsd4_session
*ses
)
562 spin_lock(&sessionid_lock
);
564 spin_unlock(&sessionid_lock
);
565 nfsd4_put_session(ses
);
568 static void nfsd4_release_respages(struct page
**respages
, short resused
);
571 free_session(struct kref
*kref
)
573 struct nfsd4_session
*ses
;
576 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
577 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++) {
578 struct nfsd4_cache_entry
*e
= &ses
->se_slots
[i
].sl_cache_entry
;
579 nfsd4_release_respages(e
->ce_respages
, e
->ce_resused
);
581 spin_lock(&nfsd_drc_lock
);
582 nfsd_drc_mem_used
-= ses
->se_fchannel
.maxreqs
* NFSD_SLOT_CACHE_SIZE
;
583 spin_unlock(&nfsd_drc_lock
);
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) boot_time %08lx\n",
607 clid
->cl_boot
, clid
->cl_id
, boot_time
);
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_cb_conn
.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_cb_conn
.cb_client
= NULL
;
644 rpc_shutdown_client(clnt
);
646 if (clp
->cl_cb_conn
.cb_cred
) {
647 put_rpccred(clp
->cl_cb_conn
.cb_cred
);
648 clp
->cl_cb_conn
.cb_cred
= NULL
;
653 free_client(struct nfs4_client
*clp
)
655 shutdown_callback_client(clp
);
656 if (clp
->cl_cred
.cr_group_info
)
657 put_group_info(clp
->cl_cred
.cr_group_info
);
658 kfree(clp
->cl_principal
);
659 kfree(clp
->cl_name
.data
);
664 put_nfs4_client(struct nfs4_client
*clp
)
666 if (atomic_dec_and_test(&clp
->cl_count
))
671 expire_client(struct nfs4_client
*clp
)
673 struct nfs4_stateowner
*sop
;
674 struct nfs4_delegation
*dp
;
675 struct list_head reaplist
;
677 dprintk("NFSD: expire_client cl_count %d\n",
678 atomic_read(&clp
->cl_count
));
680 INIT_LIST_HEAD(&reaplist
);
681 spin_lock(&recall_lock
);
682 while (!list_empty(&clp
->cl_delegations
)) {
683 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
684 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
686 list_del_init(&dp
->dl_perclnt
);
687 list_move(&dp
->dl_recall_lru
, &reaplist
);
689 spin_unlock(&recall_lock
);
690 while (!list_empty(&reaplist
)) {
691 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
692 list_del_init(&dp
->dl_recall_lru
);
693 unhash_delegation(dp
);
695 list_del(&clp
->cl_idhash
);
696 list_del(&clp
->cl_strhash
);
697 list_del(&clp
->cl_lru
);
698 while (!list_empty(&clp
->cl_openowners
)) {
699 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
700 release_openowner(sop
);
702 while (!list_empty(&clp
->cl_sessions
)) {
703 struct nfsd4_session
*ses
;
704 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
706 release_session(ses
);
708 put_nfs4_client(clp
);
711 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
713 struct nfs4_client
*clp
;
715 clp
= alloc_client(name
);
718 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
719 atomic_set(&clp
->cl_count
, 1);
720 atomic_set(&clp
->cl_cb_conn
.cb_set
, 0);
721 INIT_LIST_HEAD(&clp
->cl_idhash
);
722 INIT_LIST_HEAD(&clp
->cl_strhash
);
723 INIT_LIST_HEAD(&clp
->cl_openowners
);
724 INIT_LIST_HEAD(&clp
->cl_delegations
);
725 INIT_LIST_HEAD(&clp
->cl_sessions
);
726 INIT_LIST_HEAD(&clp
->cl_lru
);
730 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
732 memcpy(target
->cl_verifier
.data
, source
->data
,
733 sizeof(target
->cl_verifier
.data
));
736 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
738 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
739 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
742 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
744 target
->cr_uid
= source
->cr_uid
;
745 target
->cr_gid
= source
->cr_gid
;
746 target
->cr_group_info
= source
->cr_group_info
;
747 get_group_info(target
->cr_group_info
);
750 static int same_name(const char *n1
, const char *n2
)
752 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
756 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
758 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
762 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
764 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
767 /* XXX what about NGROUP */
769 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
771 return cr1
->cr_uid
== cr2
->cr_uid
;
774 static void gen_clid(struct nfs4_client
*clp
)
776 static u32 current_clientid
= 1;
778 clp
->cl_clientid
.cl_boot
= boot_time
;
779 clp
->cl_clientid
.cl_id
= current_clientid
++;
782 static void gen_confirm(struct nfs4_client
*clp
)
787 p
= (u32
*)clp
->cl_confirm
.data
;
788 *p
++ = get_seconds();
792 static int check_name(struct xdr_netobj name
)
796 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
797 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
804 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
806 unsigned int idhashval
;
808 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
809 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
810 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
811 list_add_tail(&clp
->cl_lru
, &client_lru
);
812 clp
->cl_time
= get_seconds();
816 move_to_confirmed(struct nfs4_client
*clp
)
818 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
819 unsigned int strhashval
;
821 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
822 list_del_init(&clp
->cl_strhash
);
823 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
824 strhashval
= clientstr_hashval(clp
->cl_recdir
);
825 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
829 static struct nfs4_client
*
830 find_confirmed_client(clientid_t
*clid
)
832 struct nfs4_client
*clp
;
833 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
835 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
836 if (same_clid(&clp
->cl_clientid
, clid
))
842 static struct nfs4_client
*
843 find_unconfirmed_client(clientid_t
*clid
)
845 struct nfs4_client
*clp
;
846 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
848 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
849 if (same_clid(&clp
->cl_clientid
, clid
))
856 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
857 * parameter. Matching is based on the fact the at least one of the
858 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
860 * FIXME: we need to unify the clientid namespaces for nfsv4.x
861 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
862 * and SET_CLIENTID{,_CONFIRM}
865 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
867 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
868 return use_exchange_id
== has_exchange_flags
;
871 static struct nfs4_client
*
872 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
873 bool use_exchange_id
)
875 struct nfs4_client
*clp
;
877 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
878 if (same_name(clp
->cl_recdir
, dname
) &&
879 match_clientid_establishment(clp
, use_exchange_id
))
885 static struct nfs4_client
*
886 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
887 bool use_exchange_id
)
889 struct nfs4_client
*clp
;
891 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
892 if (same_name(clp
->cl_recdir
, dname
) &&
893 match_clientid_establishment(clp
, use_exchange_id
))
899 /* a helper function for parse_callback */
901 parse_octet(unsigned int *lenp
, char **addrp
)
903 unsigned int len
= *lenp
;
915 if ((c
< '0') || (c
> '9')) {
921 n
= (n
* 10) + (c
- '0');
932 /* parse and set the setclientid ipv4 callback address */
934 parse_ipv4(unsigned int addr_len
, char *addr_val
, unsigned int *cbaddrp
, unsigned short *cbportp
)
939 u32 addrlen
= addr_len
;
940 char *addr
= addr_val
;
945 for(i
= 4; i
> 0 ; i
--) {
946 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
949 cbaddr
|= (temp
<< shift
);
957 for(i
= 2; i
> 0 ; i
--) {
958 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
961 cbport
|= (temp
<< shift
);
970 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
972 struct nfs4_cb_conn
*cb
= &clp
->cl_cb_conn
;
974 /* Currently, we only support tcp for the callback channel */
975 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
978 if ( !(parse_ipv4(se
->se_callback_addr_len
, se
->se_callback_addr_val
,
979 &cb
->cb_addr
, &cb
->cb_port
)))
981 cb
->cb_minorversion
= 0;
982 cb
->cb_prog
= se
->se_callback_prog
;
983 cb
->cb_ident
= se
->se_callback_ident
;
986 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
987 "will not receive delegations\n",
988 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
994 nfsd4_set_statp(struct svc_rqst
*rqstp
, __be32
*statp
)
996 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
998 resp
->cstate
.statp
= statp
;
1002 * Dereference the result pages.
1005 nfsd4_release_respages(struct page
**respages
, short resused
)
1009 dprintk("--> %s\n", __func__
);
1010 for (i
= 0; i
< resused
; i
++) {
1013 put_page(respages
[i
]);
1019 nfsd4_copy_pages(struct page
**topages
, struct page
**frompages
, short count
)
1023 for (i
= 0; i
< count
; i
++) {
1024 topages
[i
] = frompages
[i
];
1027 get_page(topages
[i
]);
1032 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1033 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1034 * length of the XDR response is less than se_fmaxresp_cached
1035 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1036 * of the reply (e.g. readdir).
1038 * Store the base and length of the rq_req.head[0] page
1039 * of the NFSv4.1 data, just past the rpc header.
1042 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1044 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1045 struct svc_rqst
*rqstp
= resp
->rqstp
;
1046 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1047 struct nfsd4_op
*op
= &args
->ops
[resp
->opcnt
];
1048 struct kvec
*resv
= &rqstp
->rq_res
.head
[0];
1050 dprintk("--> %s entry %p\n", __func__
, entry
);
1052 /* Don't cache a failed OP_SEQUENCE. */
1053 if (resp
->opcnt
== 1 && op
->opnum
== OP_SEQUENCE
&& resp
->cstate
.status
)
1056 nfsd4_release_respages(entry
->ce_respages
, entry
->ce_resused
);
1057 entry
->ce_opcnt
= resp
->opcnt
;
1058 entry
->ce_status
= resp
->cstate
.status
;
1061 * Don't need a page to cache just the sequence operation - the slot
1065 if (nfsd4_not_cached(resp
)) {
1066 entry
->ce_resused
= 0;
1067 entry
->ce_rpchdrlen
= 0;
1068 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__
,
1069 resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1072 entry
->ce_resused
= rqstp
->rq_resused
;
1073 if (entry
->ce_resused
> NFSD_PAGES_PER_SLOT
+ 1)
1074 entry
->ce_resused
= NFSD_PAGES_PER_SLOT
+ 1;
1075 nfsd4_copy_pages(entry
->ce_respages
, rqstp
->rq_respages
,
1077 entry
->ce_datav
.iov_base
= resp
->cstate
.statp
;
1078 entry
->ce_datav
.iov_len
= resv
->iov_len
- ((char *)resp
->cstate
.statp
-
1079 (char *)page_address(rqstp
->rq_respages
[0]));
1080 /* Current request rpc header length*/
1081 entry
->ce_rpchdrlen
= (char *)resp
->cstate
.statp
-
1082 (char *)page_address(rqstp
->rq_respages
[0]);
1086 * We keep the rpc header, but take the nfs reply from the replycache.
1089 nfsd41_copy_replay_data(struct nfsd4_compoundres
*resp
,
1090 struct nfsd4_cache_entry
*entry
)
1092 struct svc_rqst
*rqstp
= resp
->rqstp
;
1093 struct kvec
*resv
= &resp
->rqstp
->rq_res
.head
[0];
1096 /* Current request rpc header length*/
1097 len
= (char *)resp
->cstate
.statp
-
1098 (char *)page_address(rqstp
->rq_respages
[0]);
1099 if (entry
->ce_datav
.iov_len
+ len
> PAGE_SIZE
) {
1100 dprintk("%s v41 cached reply too large (%Zd).\n", __func__
,
1101 entry
->ce_datav
.iov_len
);
1104 /* copy the cached reply nfsd data past the current rpc header */
1105 memcpy((char *)resv
->iov_base
+ len
, entry
->ce_datav
.iov_base
,
1106 entry
->ce_datav
.iov_len
);
1107 resv
->iov_len
= len
+ entry
->ce_datav
.iov_len
;
1112 * Encode the replay sequence operation from the slot values.
1113 * If cachethis is FALSE encode the uncached rep error on the next
1114 * operation which sets resp->p and increments resp->opcnt for
1115 * nfs4svc_encode_compoundres.
1119 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1120 struct nfsd4_compoundres
*resp
)
1122 struct nfsd4_op
*op
;
1123 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1125 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1126 resp
->opcnt
, resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1128 /* Encode the replayed sequence operation */
1129 op
= &args
->ops
[resp
->opcnt
- 1];
1130 nfsd4_encode_operation(resp
, op
);
1132 /* Return nfserr_retry_uncached_rep in next operation. */
1133 if (args
->opcnt
> 1 && slot
->sl_cache_entry
.ce_cachethis
== 0) {
1134 op
= &args
->ops
[resp
->opcnt
++];
1135 op
->status
= nfserr_retry_uncached_rep
;
1136 nfsd4_encode_operation(resp
, op
);
1142 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1143 * cached page. Replace any futher replay pages from the cache.
1146 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1147 struct nfsd4_sequence
*seq
)
1149 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1152 dprintk("--> %s entry %p\n", __func__
, entry
);
1155 * If this is just the sequence operation, we did not keep
1156 * a page in the cache entry because we can just use the
1157 * slot info stored in struct nfsd4_sequence that was checked
1158 * against the slot in nfsd4_sequence().
1160 * This occurs when seq->cachethis is FALSE, or when the client
1161 * session inactivity timer fires and a solo sequence operation
1162 * is sent (lease renewal).
1164 seq
->maxslots
= resp
->cstate
.session
->se_fchannel
.maxreqs
;
1166 /* Either returns 0 or nfserr_retry_uncached */
1167 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1168 if (status
== nfserr_retry_uncached_rep
)
1171 if (!nfsd41_copy_replay_data(resp
, entry
)) {
1173 * Not enough room to use the replay rpc header, send the
1174 * cached header. Release all the allocated result pages.
1176 svc_free_res_pages(resp
->rqstp
);
1177 nfsd4_copy_pages(resp
->rqstp
->rq_respages
, entry
->ce_respages
,
1180 /* Release all but the first allocated result page */
1182 resp
->rqstp
->rq_resused
--;
1183 svc_free_res_pages(resp
->rqstp
);
1185 nfsd4_copy_pages(&resp
->rqstp
->rq_respages
[1],
1186 &entry
->ce_respages
[1],
1187 entry
->ce_resused
- 1);
1190 resp
->rqstp
->rq_resused
= entry
->ce_resused
;
1191 resp
->opcnt
= entry
->ce_opcnt
;
1192 resp
->cstate
.iovlen
= entry
->ce_datav
.iov_len
+ entry
->ce_rpchdrlen
;
1193 status
= entry
->ce_status
;
1199 * Set the exchange_id flags returned by the server.
1202 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1204 /* pNFS is not supported */
1205 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1207 /* Referrals are supported, Migration is not. */
1208 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1210 /* set the wire flags to return to client. */
1211 clid
->flags
= new->cl_exchange_flags
;
1215 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1216 struct nfsd4_compound_state
*cstate
,
1217 struct nfsd4_exchange_id
*exid
)
1219 struct nfs4_client
*unconf
, *conf
, *new;
1221 unsigned int strhashval
;
1222 char dname
[HEXDIR_LEN
];
1223 nfs4_verifier verf
= exid
->verifier
;
1224 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1226 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1227 " ip_addr=%u flags %x, spa_how %d\n",
1228 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1229 ip_addr
, exid
->flags
, exid
->spa_how
);
1231 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1232 return nfserr_inval
;
1234 /* Currently only support SP4_NONE */
1235 switch (exid
->spa_how
) {
1239 return nfserr_encr_alg_unsupp
;
1241 BUG(); /* checked by xdr code */
1243 return nfserr_serverfault
; /* no excuse :-/ */
1246 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1251 strhashval
= clientstr_hashval(dname
);
1256 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1258 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1259 /* 18.35.4 case 8 */
1260 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1261 status
= nfserr_not_same
;
1264 /* Client reboot: destroy old state */
1265 expire_client(conf
);
1268 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1269 /* 18.35.4 case 9 */
1270 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1271 status
= nfserr_perm
;
1274 expire_client(conf
);
1278 * Set bit when the owner id and verifier map to an already
1279 * confirmed client id (18.35.3).
1281 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1284 * Falling into 18.35.4 case 2, possible router replay.
1285 * Leave confirmed record intact and return same result.
1287 copy_verf(conf
, &verf
);
1292 /* 18.35.4 case 7 */
1293 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1294 status
= nfserr_noent
;
1298 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1301 * Possible retry or client restart. Per 18.35.4 case 4,
1302 * a new unconfirmed record should be generated regardless
1303 * of whether any properties have changed.
1305 expire_client(unconf
);
1310 new = create_client(exid
->clname
, dname
);
1312 status
= nfserr_resource
;
1316 copy_verf(new, &verf
);
1317 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1318 new->cl_addr
= ip_addr
;
1321 add_to_unconfirmed(new, strhashval
);
1323 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1324 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1327 nfsd4_set_ex_flags(new, exid
);
1329 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1330 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1334 nfs4_unlock_state();
1336 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1341 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1343 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1346 /* The slot is in use, and no response has been sent. */
1348 if (seqid
== slot_seqid
)
1349 return nfserr_jukebox
;
1351 return nfserr_seq_misordered
;
1354 if (likely(seqid
== slot_seqid
+ 1))
1357 if (seqid
== slot_seqid
)
1358 return nfserr_replay_cache
;
1360 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1362 /* Misordered replay or misordered new request */
1363 return nfserr_seq_misordered
;
1367 * Cache the create session result into the create session single DRC
1368 * slot cache by saving the xdr structure. sl_seqid has been set.
1369 * Do this for solo or embedded create session operations.
1372 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1373 struct nfsd4_clid_slot
*slot
, int nfserr
)
1375 slot
->sl_status
= nfserr
;
1376 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1380 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1381 struct nfsd4_clid_slot
*slot
)
1383 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1384 return slot
->sl_status
;
1388 nfsd4_create_session(struct svc_rqst
*rqstp
,
1389 struct nfsd4_compound_state
*cstate
,
1390 struct nfsd4_create_session
*cr_ses
)
1392 u32 ip_addr
= svc_addr_in(rqstp
)->sin_addr
.s_addr
;
1393 struct nfs4_client
*conf
, *unconf
;
1394 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1398 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1399 conf
= find_confirmed_client(&cr_ses
->clientid
);
1402 cs_slot
= &conf
->cl_cs_slot
;
1403 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1404 if (status
== nfserr_replay_cache
) {
1405 dprintk("Got a create_session replay! seqid= %d\n",
1407 /* Return the cached reply status */
1408 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1410 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1411 status
= nfserr_seq_misordered
;
1412 dprintk("Sequence misordered!\n");
1413 dprintk("Expected seqid= %d but got seqid= %d\n",
1414 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1417 cs_slot
->sl_seqid
++;
1418 } else if (unconf
) {
1419 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1420 (ip_addr
!= unconf
->cl_addr
)) {
1421 status
= nfserr_clid_inuse
;
1425 cs_slot
= &unconf
->cl_cs_slot
;
1426 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1428 /* an unconfirmed replay returns misordered */
1429 status
= nfserr_seq_misordered
;
1433 cs_slot
->sl_seqid
++; /* from 0 to 1 */
1434 move_to_confirmed(unconf
);
1437 * We do not support RDMA or persistent sessions
1439 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1440 cr_ses
->flags
&= ~SESSION4_RDMA
;
1444 status
= nfserr_stale_clientid
;
1448 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1452 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1453 NFS4_MAX_SESSIONID_LEN
);
1454 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1457 /* cache solo and embedded create sessions under the state lock */
1458 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1460 nfs4_unlock_state();
1461 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1466 nfsd4_destroy_session(struct svc_rqst
*r
,
1467 struct nfsd4_compound_state
*cstate
,
1468 struct nfsd4_destroy_session
*sessionid
)
1470 struct nfsd4_session
*ses
;
1471 u32 status
= nfserr_badsession
;
1474 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1475 * - Should we return nfserr_back_chan_busy if waiting for
1476 * callbacks on to-be-destroyed session?
1477 * - Do we need to clear any callback info from previous session?
1480 dump_sessionid(__func__
, &sessionid
->sessionid
);
1481 spin_lock(&sessionid_lock
);
1482 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1484 spin_unlock(&sessionid_lock
);
1488 unhash_session(ses
);
1489 spin_unlock(&sessionid_lock
);
1491 /* wait for callbacks */
1492 shutdown_callback_client(ses
->se_client
);
1493 nfsd4_put_session(ses
);
1496 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1501 nfsd4_sequence(struct svc_rqst
*rqstp
,
1502 struct nfsd4_compound_state
*cstate
,
1503 struct nfsd4_sequence
*seq
)
1505 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1506 struct nfsd4_session
*session
;
1507 struct nfsd4_slot
*slot
;
1510 if (resp
->opcnt
!= 1)
1511 return nfserr_sequence_pos
;
1513 spin_lock(&sessionid_lock
);
1514 status
= nfserr_badsession
;
1515 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1519 status
= nfserr_badslot
;
1520 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1523 slot
= &session
->se_slots
[seq
->slotid
];
1524 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1526 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1527 if (status
== nfserr_replay_cache
) {
1528 cstate
->slot
= slot
;
1529 cstate
->session
= session
;
1530 /* Return the cached reply status and set cstate->status
1531 * for nfsd4_svc_encode_compoundres processing */
1532 status
= nfsd4_replay_cache_entry(resp
, seq
);
1533 cstate
->status
= nfserr_replay_cache
;
1539 /* Success! bump slot seqid */
1540 slot
->sl_inuse
= true;
1541 slot
->sl_seqid
= seq
->seqid
;
1542 slot
->sl_cache_entry
.ce_cachethis
= seq
->cachethis
;
1543 /* Always set the cache entry cachethis for solo sequence */
1544 if (nfsd4_is_solo_sequence(resp
))
1545 slot
->sl_cache_entry
.ce_cachethis
= 1;
1547 cstate
->slot
= slot
;
1548 cstate
->session
= session
;
1551 /* Renew the clientid on success and on replay.
1552 * Hold a session reference until done processing the compound:
1553 * nfsd4_put_session called only if the cstate slot is set.
1555 renew_client(session
->se_client
);
1556 nfsd4_get_session(session
);
1558 spin_unlock(&sessionid_lock
);
1559 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1564 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1565 struct nfsd4_setclientid
*setclid
)
1567 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1568 struct xdr_netobj clname
= {
1569 .len
= setclid
->se_namelen
,
1570 .data
= setclid
->se_name
,
1572 nfs4_verifier clverifier
= setclid
->se_verf
;
1573 unsigned int strhashval
;
1574 struct nfs4_client
*conf
, *unconf
, *new;
1577 char dname
[HEXDIR_LEN
];
1579 if (!check_name(clname
))
1580 return nfserr_inval
;
1582 status
= nfs4_make_rec_clidname(dname
, &clname
);
1587 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1588 * We get here on a DRC miss.
1591 strhashval
= clientstr_hashval(dname
);
1594 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1596 /* RFC 3530 14.2.33 CASE 0: */
1597 status
= nfserr_clid_inuse
;
1598 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1599 dprintk("NFSD: setclientid: string in use by client"
1600 " at %pI4\n", &conf
->cl_addr
);
1605 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1606 * has a description of SETCLIENTID request processing consisting
1607 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1609 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1610 status
= nfserr_resource
;
1613 * RFC 3530 14.2.33 CASE 4:
1614 * placed first, because it is the normal case
1617 expire_client(unconf
);
1618 new = create_client(clname
, dname
);
1622 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1624 * RFC 3530 14.2.33 CASE 1:
1625 * probable callback update
1628 /* Note this is removing unconfirmed {*x***},
1629 * which is stronger than RFC recommended {vxc**}.
1630 * This has the advantage that there is at most
1631 * one {*x***} in either list at any time.
1633 expire_client(unconf
);
1635 new = create_client(clname
, dname
);
1638 copy_clid(new, conf
);
1639 } else if (!unconf
) {
1641 * RFC 3530 14.2.33 CASE 2:
1642 * probable client reboot; state will be removed if
1645 new = create_client(clname
, dname
);
1651 * RFC 3530 14.2.33 CASE 3:
1652 * probable client reboot; state will be removed if
1655 expire_client(unconf
);
1656 new = create_client(clname
, dname
);
1661 copy_verf(new, &clverifier
);
1662 new->cl_addr
= sin
->sin_addr
.s_addr
;
1663 new->cl_flavor
= rqstp
->rq_flavor
;
1664 princ
= svc_gss_principal(rqstp
);
1666 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1667 if (new->cl_principal
== NULL
) {
1672 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1674 gen_callback(new, setclid
);
1675 add_to_unconfirmed(new, strhashval
);
1676 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1677 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1678 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1681 nfs4_unlock_state();
1687 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1688 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1689 * bullets, labeled as CASE1 - CASE4 below.
1692 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1693 struct nfsd4_compound_state
*cstate
,
1694 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1696 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
1697 struct nfs4_client
*conf
, *unconf
;
1698 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1699 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1702 if (STALE_CLIENTID(clid
))
1703 return nfserr_stale_clientid
;
1705 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1706 * We get here on a DRC miss.
1711 conf
= find_confirmed_client(clid
);
1712 unconf
= find_unconfirmed_client(clid
);
1714 status
= nfserr_clid_inuse
;
1715 if (conf
&& conf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1717 if (unconf
&& unconf
->cl_addr
!= sin
->sin_addr
.s_addr
)
1721 * section 14.2.34 of RFC 3530 has a description of
1722 * SETCLIENTID_CONFIRM request processing consisting
1723 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1725 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1727 * RFC 3530 14.2.34 CASE 1:
1730 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1731 status
= nfserr_clid_inuse
;
1733 /* XXX: We just turn off callbacks until we can handle
1734 * change request correctly. */
1735 atomic_set(&conf
->cl_cb_conn
.cb_set
, 0);
1736 expire_client(unconf
);
1740 } else if (conf
&& !unconf
) {
1742 * RFC 3530 14.2.34 CASE 2:
1743 * probable retransmitted request; play it safe and
1746 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1747 status
= nfserr_clid_inuse
;
1750 } else if (!conf
&& unconf
1751 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1753 * RFC 3530 14.2.34 CASE 3:
1754 * Normal case; new or rebooted client:
1756 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1757 status
= nfserr_clid_inuse
;
1760 clientstr_hashval(unconf
->cl_recdir
);
1761 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1764 nfsd4_remove_clid_dir(conf
);
1765 expire_client(conf
);
1767 move_to_confirmed(unconf
);
1769 nfsd4_probe_callback(conf
);
1772 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1773 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1776 * RFC 3530 14.2.34 CASE 4:
1777 * Client probably hasn't noticed that we rebooted yet.
1779 status
= nfserr_stale_clientid
;
1781 /* check that we have hit one of the cases...*/
1782 status
= nfserr_clid_inuse
;
1785 nfs4_unlock_state();
1789 /* OPEN Share state helper functions */
1790 static inline struct nfs4_file
*
1791 alloc_init_file(struct inode
*ino
)
1793 struct nfs4_file
*fp
;
1794 unsigned int hashval
= file_hashval(ino
);
1796 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1798 atomic_set(&fp
->fi_ref
, 1);
1799 INIT_LIST_HEAD(&fp
->fi_hash
);
1800 INIT_LIST_HEAD(&fp
->fi_stateids
);
1801 INIT_LIST_HEAD(&fp
->fi_delegations
);
1802 spin_lock(&recall_lock
);
1803 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1804 spin_unlock(&recall_lock
);
1805 fp
->fi_inode
= igrab(ino
);
1806 fp
->fi_id
= current_fileid
++;
1807 fp
->fi_had_conflict
= false;
1814 nfsd4_free_slab(struct kmem_cache
**slab
)
1818 kmem_cache_destroy(*slab
);
1823 nfsd4_free_slabs(void)
1825 nfsd4_free_slab(&stateowner_slab
);
1826 nfsd4_free_slab(&file_slab
);
1827 nfsd4_free_slab(&stateid_slab
);
1828 nfsd4_free_slab(&deleg_slab
);
1832 nfsd4_init_slabs(void)
1834 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1835 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1836 if (stateowner_slab
== NULL
)
1838 file_slab
= kmem_cache_create("nfsd4_files",
1839 sizeof(struct nfs4_file
), 0, 0, NULL
);
1840 if (file_slab
== NULL
)
1842 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1843 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1844 if (stateid_slab
== NULL
)
1846 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1847 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1848 if (deleg_slab
== NULL
)
1853 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1858 nfs4_free_stateowner(struct kref
*kref
)
1860 struct nfs4_stateowner
*sop
=
1861 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1862 kfree(sop
->so_owner
.data
);
1863 kmem_cache_free(stateowner_slab
, sop
);
1866 static inline struct nfs4_stateowner
*
1867 alloc_stateowner(struct xdr_netobj
*owner
)
1869 struct nfs4_stateowner
*sop
;
1871 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1872 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1873 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1874 sop
->so_owner
.len
= owner
->len
;
1875 kref_init(&sop
->so_ref
);
1878 kmem_cache_free(stateowner_slab
, sop
);
1883 static struct nfs4_stateowner
*
1884 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1885 struct nfs4_stateowner
*sop
;
1886 struct nfs4_replay
*rp
;
1887 unsigned int idhashval
;
1889 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1891 idhashval
= ownerid_hashval(current_ownerid
);
1892 INIT_LIST_HEAD(&sop
->so_idhash
);
1893 INIT_LIST_HEAD(&sop
->so_strhash
);
1894 INIT_LIST_HEAD(&sop
->so_perclient
);
1895 INIT_LIST_HEAD(&sop
->so_stateids
);
1896 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1897 INIT_LIST_HEAD(&sop
->so_close_lru
);
1899 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1900 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1901 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1902 sop
->so_is_open_owner
= 1;
1903 sop
->so_id
= current_ownerid
++;
1904 sop
->so_client
= clp
;
1905 sop
->so_seqid
= open
->op_seqid
;
1906 sop
->so_confirmed
= 0;
1907 rp
= &sop
->so_replay
;
1908 rp
->rp_status
= nfserr_serverfault
;
1910 rp
->rp_buf
= rp
->rp_ibuf
;
1915 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1916 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1917 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1919 INIT_LIST_HEAD(&stp
->st_hash
);
1920 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1921 INIT_LIST_HEAD(&stp
->st_lockowners
);
1922 INIT_LIST_HEAD(&stp
->st_perfile
);
1923 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1924 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1925 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1926 stp
->st_stateowner
= sop
;
1929 stp
->st_stateid
.si_boot
= get_seconds();
1930 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1931 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1932 stp
->st_stateid
.si_generation
= 0;
1933 stp
->st_access_bmap
= 0;
1934 stp
->st_deny_bmap
= 0;
1935 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
1936 &stp
->st_access_bmap
);
1937 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1938 stp
->st_openstp
= NULL
;
1942 move_to_close_lru(struct nfs4_stateowner
*sop
)
1944 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1946 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1947 sop
->so_time
= get_seconds();
1951 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1954 return (sop
->so_owner
.len
== owner
->len
) &&
1955 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1956 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1959 static struct nfs4_stateowner
*
1960 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1962 struct nfs4_stateowner
*so
= NULL
;
1964 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1965 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1971 /* search file_hashtbl[] for file */
1972 static struct nfs4_file
*
1973 find_file(struct inode
*ino
)
1975 unsigned int hashval
= file_hashval(ino
);
1976 struct nfs4_file
*fp
;
1978 spin_lock(&recall_lock
);
1979 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1980 if (fp
->fi_inode
== ino
) {
1982 spin_unlock(&recall_lock
);
1986 spin_unlock(&recall_lock
);
1990 static inline int access_valid(u32 x
, u32 minorversion
)
1992 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
1994 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
1996 x
&= ~NFS4_SHARE_ACCESS_MASK
;
1997 if (minorversion
&& x
) {
1998 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
2000 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
2002 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
2009 static inline int deny_valid(u32 x
)
2011 /* Note: unlike access bits, deny bits may be zero. */
2012 return x
<= NFS4_SHARE_DENY_BOTH
;
2016 * We store the NONE, READ, WRITE, and BOTH bits separately in the
2017 * st_{access,deny}_bmap field of the stateid, in order to track not
2018 * only what share bits are currently in force, but also what
2019 * combinations of share bits previous opens have used. This allows us
2020 * to enforce the recommendation of rfc 3530 14.2.19 that the server
2021 * return an error if the client attempt to downgrade to a combination
2022 * of share bits not explicable by closing some of its previous opens.
2024 * XXX: This enforcement is actually incomplete, since we don't keep
2025 * track of access/deny bit combinations; so, e.g., we allow:
2027 * OPEN allow read, deny write
2028 * OPEN allow both, deny none
2029 * DOWNGRADE allow read, deny none
2031 * which we should reject.
2034 set_access(unsigned int *access
, unsigned long bmap
) {
2038 for (i
= 1; i
< 4; i
++) {
2039 if (test_bit(i
, &bmap
))
2045 set_deny(unsigned int *deny
, unsigned long bmap
) {
2049 for (i
= 0; i
< 4; i
++) {
2050 if (test_bit(i
, &bmap
))
2056 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
2057 unsigned int access
, deny
;
2059 set_access(&access
, stp
->st_access_bmap
);
2060 set_deny(&deny
, stp
->st_deny_bmap
);
2061 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
2067 * Called to check deny when READ with all zero stateid or
2068 * WRITE with all zero or all one stateid
2071 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2073 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2074 struct nfs4_file
*fp
;
2075 struct nfs4_stateid
*stp
;
2078 dprintk("NFSD: nfs4_share_conflict\n");
2080 fp
= find_file(ino
);
2083 ret
= nfserr_locked
;
2084 /* Search for conflicting share reservations */
2085 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2086 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2087 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2097 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
2099 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
2100 drop_file_write_access(filp
);
2101 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
2106 * Spawn a thread to perform a recall on the delegation represented
2107 * by the lease (file_lock)
2109 * Called from break_lease() with lock_kernel() held.
2110 * Note: we assume break_lease will only call this *once* for any given
2114 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2116 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2118 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2122 /* We're assuming the state code never drops its reference
2123 * without first removing the lease. Since we're in this lease
2124 * callback (and since the lease code is serialized by the kernel
2125 * lock) we know the server hasn't removed the lease yet, we know
2126 * it's safe to take a reference: */
2127 atomic_inc(&dp
->dl_count
);
2128 atomic_inc(&dp
->dl_client
->cl_count
);
2130 spin_lock(&recall_lock
);
2131 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2132 spin_unlock(&recall_lock
);
2134 /* only place dl_time is set. protected by lock_kernel*/
2135 dp
->dl_time
= get_seconds();
2138 * We don't want the locks code to timeout the lease for us;
2139 * we'll remove it ourself if the delegation isn't returned
2142 fl
->fl_break_time
= 0;
2144 dp
->dl_file
->fi_had_conflict
= true;
2145 nfsd4_cb_recall(dp
);
2149 * The file_lock is being reapd.
2151 * Called by locks_free_lock() with lock_kernel() held.
2154 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2156 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2158 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2160 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2162 dp
->dl_flock
= NULL
;
2166 * Set the delegation file_lock back pointer.
2168 * Called from setlease() with lock_kernel() held.
2171 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2173 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2175 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2182 * Called from setlease() with lock_kernel() held
2185 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2187 struct nfs4_delegation
*onlistd
=
2188 (struct nfs4_delegation
*)onlist
->fl_owner
;
2189 struct nfs4_delegation
*tryd
=
2190 (struct nfs4_delegation
*)try->fl_owner
;
2192 if (onlist
->fl_lmops
!= try->fl_lmops
)
2195 return onlistd
->dl_client
== tryd
->dl_client
;
2200 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2203 return lease_modify(onlist
, arg
);
2208 static struct lock_manager_operations nfsd_lease_mng_ops
= {
2209 .fl_break
= nfsd_break_deleg_cb
,
2210 .fl_release_private
= nfsd_release_deleg_cb
,
2211 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2212 .fl_mylease
= nfsd_same_client_deleg_cb
,
2213 .fl_change
= nfsd_change_deleg_cb
,
2218 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2219 struct nfsd4_open
*open
)
2221 clientid_t
*clientid
= &open
->op_clientid
;
2222 struct nfs4_client
*clp
= NULL
;
2223 unsigned int strhashval
;
2224 struct nfs4_stateowner
*sop
= NULL
;
2226 if (!check_name(open
->op_owner
))
2227 return nfserr_inval
;
2229 if (STALE_CLIENTID(&open
->op_clientid
))
2230 return nfserr_stale_clientid
;
2232 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2233 sop
= find_openstateowner_str(strhashval
, open
);
2234 open
->op_stateowner
= sop
;
2236 /* Make sure the client's lease hasn't expired. */
2237 clp
= find_confirmed_client(clientid
);
2239 return nfserr_expired
;
2242 /* When sessions are used, skip open sequenceid processing */
2243 if (nfsd4_has_session(cstate
))
2245 if (!sop
->so_confirmed
) {
2246 /* Replace unconfirmed owners without checking for replay. */
2247 clp
= sop
->so_client
;
2248 release_openowner(sop
);
2249 open
->op_stateowner
= NULL
;
2252 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2253 if (sop
->so_replay
.rp_buflen
)
2254 return nfserr_replay_me
;
2255 /* The original OPEN failed so spectacularly
2256 * that we don't even have replay data saved!
2257 * Therefore, we have no choice but to continue
2258 * processing this OPEN; presumably, we'll
2259 * fail again for the same reason.
2261 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2264 if (open
->op_seqid
!= sop
->so_seqid
)
2265 return nfserr_bad_seqid
;
2267 if (open
->op_stateowner
== NULL
) {
2268 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2270 return nfserr_resource
;
2271 open
->op_stateowner
= sop
;
2273 list_del_init(&sop
->so_close_lru
);
2274 renew_client(sop
->so_client
);
2278 static inline __be32
2279 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2281 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2282 return nfserr_openmode
;
2287 static struct nfs4_delegation
*
2288 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2290 struct nfs4_delegation
*dp
;
2292 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2293 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2300 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2301 struct nfs4_delegation
**dp
)
2304 __be32 status
= nfserr_bad_stateid
;
2306 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2309 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
2310 RD_STATE
: WR_STATE
;
2311 status
= nfs4_check_delegmode(*dp
, flags
);
2315 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2319 open
->op_stateowner
->so_confirmed
= 1;
2324 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2326 struct nfs4_stateid
*local
;
2327 __be32 status
= nfserr_share_denied
;
2328 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2330 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2331 /* ignore lock owners */
2332 if (local
->st_stateowner
->so_is_open_owner
== 0)
2334 /* remember if we have seen this open owner */
2335 if (local
->st_stateowner
== sop
)
2337 /* check for conflicting share reservations */
2338 if (!test_share(local
, open
))
2346 static inline struct nfs4_stateid
*
2347 nfs4_alloc_stateid(void)
2349 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2353 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2354 struct nfs4_delegation
*dp
,
2355 struct svc_fh
*cur_fh
, int flags
)
2357 struct nfs4_stateid
*stp
;
2359 stp
= nfs4_alloc_stateid();
2361 return nfserr_resource
;
2364 get_file(dp
->dl_vfs_file
);
2365 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2368 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2371 if (status
== nfserr_dropit
)
2372 status
= nfserr_jukebox
;
2373 kmem_cache_free(stateid_slab
, stp
);
2381 static inline __be32
2382 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2383 struct nfsd4_open
*open
)
2385 struct iattr iattr
= {
2386 .ia_valid
= ATTR_SIZE
,
2389 if (!open
->op_truncate
)
2391 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2392 return nfserr_inval
;
2393 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2397 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2399 struct file
*filp
= stp
->st_vfs_file
;
2400 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2401 unsigned int share_access
, new_writer
;
2404 set_access(&share_access
, stp
->st_access_bmap
);
2405 new_writer
= (~share_access
) & open
->op_share_access
2406 & NFS4_SHARE_ACCESS_WRITE
;
2409 int err
= get_write_access(inode
);
2411 return nfserrno(err
);
2412 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2414 return nfserrno(err
);
2415 file_take_write(filp
);
2417 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2420 put_write_access(inode
);
2423 /* remember the open */
2424 filp
->f_mode
|= open
->op_share_access
;
2425 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2426 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2433 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2435 open
->op_stateowner
->so_confirmed
= 1;
2436 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2440 * Attempt to hand out a delegation.
2443 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2445 struct nfs4_delegation
*dp
;
2446 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2447 struct nfs4_cb_conn
*cb
= &sop
->so_client
->cl_cb_conn
;
2448 struct file_lock fl
, *flp
= &fl
;
2449 int status
, flag
= 0;
2451 flag
= NFS4_OPEN_DELEGATE_NONE
;
2452 open
->op_recall
= 0;
2453 switch (open
->op_claim_type
) {
2454 case NFS4_OPEN_CLAIM_PREVIOUS
:
2455 if (!atomic_read(&cb
->cb_set
))
2456 open
->op_recall
= 1;
2457 flag
= open
->op_delegate_type
;
2458 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2461 case NFS4_OPEN_CLAIM_NULL
:
2462 /* Let's not give out any delegations till everyone's
2463 * had the chance to reclaim theirs.... */
2464 if (locks_in_grace())
2466 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
2468 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2469 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2471 flag
= NFS4_OPEN_DELEGATE_READ
;
2477 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2479 flag
= NFS4_OPEN_DELEGATE_NONE
;
2482 locks_init_lock(&fl
);
2483 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2484 fl
.fl_flags
= FL_LEASE
;
2485 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2486 fl
.fl_end
= OFFSET_MAX
;
2487 fl
.fl_owner
= (fl_owner_t
)dp
;
2488 fl
.fl_file
= stp
->st_vfs_file
;
2489 fl
.fl_pid
= current
->tgid
;
2491 /* vfs_setlease checks to see if delegation should be handed out.
2492 * the lock_manager callbacks fl_mylease and fl_change are used
2494 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2495 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2496 unhash_delegation(dp
);
2497 flag
= NFS4_OPEN_DELEGATE_NONE
;
2501 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2503 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2504 dp
->dl_stateid
.si_boot
,
2505 dp
->dl_stateid
.si_stateownerid
,
2506 dp
->dl_stateid
.si_fileid
,
2507 dp
->dl_stateid
.si_generation
);
2509 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2510 && flag
== NFS4_OPEN_DELEGATE_NONE
2511 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2512 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2513 open
->op_delegate_type
= flag
;
2517 * called with nfs4_lock_state() held.
2520 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2522 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2523 struct nfs4_file
*fp
= NULL
;
2524 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2525 struct nfs4_stateid
*stp
= NULL
;
2526 struct nfs4_delegation
*dp
= NULL
;
2529 status
= nfserr_inval
;
2530 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2531 || !deny_valid(open
->op_share_deny
))
2534 * Lookup file; if found, lookup stateid and check open request,
2535 * and check for delegations in the process of being recalled.
2536 * If not found, create the nfs4_file struct
2538 fp
= find_file(ino
);
2540 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2542 status
= nfs4_check_deleg(fp
, open
, &dp
);
2546 status
= nfserr_bad_stateid
;
2547 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2549 status
= nfserr_resource
;
2550 fp
= alloc_init_file(ino
);
2556 * OPEN the file, or upgrade an existing OPEN.
2557 * If truncate fails, the OPEN fails.
2560 /* Stateid was found, this is an OPEN upgrade */
2561 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2564 update_stateid(&stp
->st_stateid
);
2566 /* Stateid was not found, this is a new OPEN */
2568 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2569 flags
|= NFSD_MAY_READ
;
2570 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2571 flags
|= NFSD_MAY_WRITE
;
2572 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2575 init_stateid(stp
, fp
, open
);
2576 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2578 release_open_stateid(stp
);
2581 if (nfsd4_has_session(&resp
->cstate
))
2582 update_stateid(&stp
->st_stateid
);
2584 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2586 if (nfsd4_has_session(&resp
->cstate
))
2587 open
->op_stateowner
->so_confirmed
= 1;
2590 * Attempt to hand out a delegation. No error return, because the
2591 * OPEN succeeds even if we fail.
2593 nfs4_open_delegation(current_fh
, open
, stp
);
2597 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2598 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
2599 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
2603 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2604 nfs4_set_claim_prev(open
);
2606 * To finish the open response, we just need to set the rflags.
2608 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2609 if (!open
->op_stateowner
->so_confirmed
&&
2610 !nfsd4_has_session(&resp
->cstate
))
2611 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2617 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2620 struct nfs4_client
*clp
;
2624 dprintk("process_renew(%08x/%08x): starting\n",
2625 clid
->cl_boot
, clid
->cl_id
);
2626 status
= nfserr_stale_clientid
;
2627 if (STALE_CLIENTID(clid
))
2629 clp
= find_confirmed_client(clid
);
2630 status
= nfserr_expired
;
2632 /* We assume the client took too long to RENEW. */
2633 dprintk("nfsd4_renew: clientid not found!\n");
2637 status
= nfserr_cb_path_down
;
2638 if (!list_empty(&clp
->cl_delegations
)
2639 && !atomic_read(&clp
->cl_cb_conn
.cb_set
))
2643 nfs4_unlock_state();
2647 struct lock_manager nfsd4_manager
= {
2651 nfsd4_end_grace(void)
2653 dprintk("NFSD: end of grace period\n");
2654 nfsd4_recdir_purge_old();
2655 locks_end_grace(&nfsd4_manager
);
2659 nfs4_laundromat(void)
2661 struct nfs4_client
*clp
;
2662 struct nfs4_stateowner
*sop
;
2663 struct nfs4_delegation
*dp
;
2664 struct list_head
*pos
, *next
, reaplist
;
2665 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
2666 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
2667 time_t u
, test_val
= NFSD_LEASE_TIME
;
2671 dprintk("NFSD: laundromat service - starting\n");
2672 if (locks_in_grace())
2674 list_for_each_safe(pos
, next
, &client_lru
) {
2675 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2676 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2677 t
= clp
->cl_time
- cutoff
;
2678 if (clientid_val
> t
)
2682 dprintk("NFSD: purging unused client (clientid %08x)\n",
2683 clp
->cl_clientid
.cl_id
);
2684 nfsd4_remove_clid_dir(clp
);
2687 INIT_LIST_HEAD(&reaplist
);
2688 spin_lock(&recall_lock
);
2689 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2690 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2691 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2692 u
= dp
->dl_time
- cutoff
;
2697 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2699 list_move(&dp
->dl_recall_lru
, &reaplist
);
2701 spin_unlock(&recall_lock
);
2702 list_for_each_safe(pos
, next
, &reaplist
) {
2703 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2704 list_del_init(&dp
->dl_recall_lru
);
2705 unhash_delegation(dp
);
2707 test_val
= NFSD_LEASE_TIME
;
2708 list_for_each_safe(pos
, next
, &close_lru
) {
2709 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2710 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2711 u
= sop
->so_time
- cutoff
;
2716 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2718 release_openowner(sop
);
2720 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2721 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2722 nfs4_unlock_state();
2723 return clientid_val
;
2726 static struct workqueue_struct
*laundry_wq
;
2727 static void laundromat_main(struct work_struct
*);
2728 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2731 laundromat_main(struct work_struct
*not_used
)
2735 t
= nfs4_laundromat();
2736 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2737 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2740 static struct nfs4_stateowner
*
2741 search_close_lru(u32 st_id
, int flags
)
2743 struct nfs4_stateowner
*local
= NULL
;
2745 if (flags
& CLOSE_STATE
) {
2746 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2747 if (local
->so_id
== st_id
)
2755 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2757 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2761 STALE_STATEID(stateid_t
*stateid
)
2763 if (time_after((unsigned long)boot_time
,
2764 (unsigned long)stateid
->si_boot
)) {
2765 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2766 stateid
->si_boot
, stateid
->si_stateownerid
,
2767 stateid
->si_fileid
, stateid
->si_generation
);
2774 EXPIRED_STATEID(stateid_t
*stateid
)
2776 if (time_before((unsigned long)boot_time
,
2777 ((unsigned long)stateid
->si_boot
)) &&
2778 time_before((unsigned long)(stateid
->si_boot
+ lease_time
), get_seconds())) {
2779 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2780 stateid
->si_boot
, stateid
->si_stateownerid
,
2781 stateid
->si_fileid
, stateid
->si_generation
);
2788 stateid_error_map(stateid_t
*stateid
)
2790 if (STALE_STATEID(stateid
))
2791 return nfserr_stale_stateid
;
2792 if (EXPIRED_STATEID(stateid
))
2793 return nfserr_expired
;
2795 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2796 stateid
->si_boot
, stateid
->si_stateownerid
,
2797 stateid
->si_fileid
, stateid
->si_generation
);
2798 return nfserr_bad_stateid
;
2802 access_permit_read(unsigned long access_bmap
)
2804 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2805 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2806 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2810 access_permit_write(unsigned long access_bmap
)
2812 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2813 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2817 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2819 __be32 status
= nfserr_openmode
;
2821 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2823 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2830 static inline __be32
2831 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2833 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2835 else if (locks_in_grace()) {
2836 /* Answer in remaining cases depends on existance of
2837 * conflicting state; so we must wait out the grace period. */
2838 return nfserr_grace
;
2839 } else if (flags
& WR_STATE
)
2840 return nfs4_share_conflict(current_fh
,
2841 NFS4_SHARE_DENY_WRITE
);
2842 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2843 return nfs4_share_conflict(current_fh
,
2844 NFS4_SHARE_DENY_READ
);
2848 * Allow READ/WRITE during grace period on recovered state only for files
2849 * that are not able to provide mandatory locking.
2852 grace_disallows_io(struct inode
*inode
)
2854 return locks_in_grace() && mandatory_lock(inode
);
2857 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
2860 * When sessions are used the stateid generation number is ignored
2863 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
2866 /* If the client sends us a stateid from the future, it's buggy: */
2867 if (in
->si_generation
> ref
->si_generation
)
2868 return nfserr_bad_stateid
;
2870 * The following, however, can happen. For example, if the
2871 * client sends an open and some IO at the same time, the open
2872 * may bump si_generation while the IO is still in flight.
2873 * Thanks to hard links and renames, the client never knows what
2874 * file an open will affect. So it could avoid that situation
2875 * only by serializing all opens and IO from the same open
2876 * owner. To recover from the old_stateid error, the client
2877 * will just have to retry the IO:
2879 if (in
->si_generation
< ref
->si_generation
)
2880 return nfserr_old_stateid
;
2885 static int is_delegation_stateid(stateid_t
*stateid
)
2887 return stateid
->si_fileid
== 0;
2891 * Checks for stateid operations
2894 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
2895 stateid_t
*stateid
, int flags
, struct file
**filpp
)
2897 struct nfs4_stateid
*stp
= NULL
;
2898 struct nfs4_delegation
*dp
= NULL
;
2899 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2900 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2906 if (grace_disallows_io(ino
))
2907 return nfserr_grace
;
2909 if (nfsd4_has_session(cstate
))
2910 flags
|= HAS_SESSION
;
2912 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2913 return check_special_stateids(current_fh
, stateid
, flags
);
2915 status
= nfserr_stale_stateid
;
2916 if (STALE_STATEID(stateid
))
2919 status
= nfserr_bad_stateid
;
2920 if (is_delegation_stateid(stateid
)) {
2921 dp
= find_delegation_stateid(ino
, stateid
);
2923 status
= stateid_error_map(stateid
);
2926 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
2930 status
= nfs4_check_delegmode(dp
, flags
);
2933 renew_client(dp
->dl_client
);
2935 *filpp
= dp
->dl_vfs_file
;
2936 } else { /* open or lock stateid */
2937 stp
= find_stateid(stateid
, flags
);
2939 status
= stateid_error_map(stateid
);
2942 if (nfs4_check_fh(current_fh
, stp
))
2944 if (!stp
->st_stateowner
->so_confirmed
)
2946 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
2950 status
= nfs4_check_openmode(stp
, flags
);
2953 renew_client(stp
->st_stateowner
->so_client
);
2955 *filpp
= stp
->st_vfs_file
;
2965 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2966 RD_STATE
: WR_STATE
;
2970 * Checks for sequence id mutating operations.
2973 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
2974 stateid_t
*stateid
, int flags
,
2975 struct nfs4_stateowner
**sopp
,
2976 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2978 struct nfs4_stateid
*stp
;
2979 struct nfs4_stateowner
*sop
;
2980 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2983 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2984 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2985 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2986 stateid
->si_generation
);
2991 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2992 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2993 return nfserr_bad_stateid
;
2996 if (STALE_STATEID(stateid
))
2997 return nfserr_stale_stateid
;
2999 if (nfsd4_has_session(cstate
))
3000 flags
|= HAS_SESSION
;
3003 * We return BAD_STATEID if filehandle doesn't match stateid,
3004 * the confirmed flag is incorrecly set, or the generation
3005 * number is incorrect.
3007 stp
= find_stateid(stateid
, flags
);
3010 * Also, we should make sure this isn't just the result of
3013 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
3015 return stateid_error_map(stateid
);
3021 *sopp
= sop
= stp
->st_stateowner
;
3024 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
3025 struct nfs4_client
*clp
= sop
->so_client
;
3029 lkflg
= setlkflg(lock
->lk_type
);
3031 if (lock
->lk_is_new
) {
3032 if (!sop
->so_is_open_owner
)
3033 return nfserr_bad_stateid
;
3034 if (!(flags
& HAS_SESSION
) &&
3035 !same_clid(&clp
->cl_clientid
, lockclid
))
3036 return nfserr_bad_stateid
;
3037 /* stp is the open stateid */
3038 status
= nfs4_check_openmode(stp
, lkflg
);
3042 /* stp is the lock stateid */
3043 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
3049 if (nfs4_check_fh(current_fh
, stp
)) {
3050 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3051 return nfserr_bad_stateid
;
3055 * We now validate the seqid and stateid generation numbers.
3056 * For the moment, we ignore the possibility of
3057 * generation number wraparound.
3059 if (!(flags
& HAS_SESSION
) && seqid
!= sop
->so_seqid
)
3062 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3063 dprintk("NFSD: preprocess_seqid_op: expected"
3064 " unconfirmed stateowner!\n");
3065 return nfserr_bad_stateid
;
3067 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3068 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3069 " confirmed yet!\n");
3070 return nfserr_bad_stateid
;
3072 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, flags
);
3075 renew_client(sop
->so_client
);
3079 if (seqid
== sop
->so_seqid
- 1) {
3080 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3081 /* indicate replay to calling function */
3082 return nfserr_replay_me
;
3084 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3085 sop
->so_seqid
, seqid
);
3087 return nfserr_bad_seqid
;
3091 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3092 struct nfsd4_open_confirm
*oc
)
3095 struct nfs4_stateowner
*sop
;
3096 struct nfs4_stateid
*stp
;
3098 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3099 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3100 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3102 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3108 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3109 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3110 CONFIRM
| OPEN_STATE
,
3111 &oc
->oc_stateowner
, &stp
, NULL
)))
3114 sop
= oc
->oc_stateowner
;
3115 sop
->so_confirmed
= 1;
3116 update_stateid(&stp
->st_stateid
);
3117 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3118 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3119 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
3120 stp
->st_stateid
.si_boot
,
3121 stp
->st_stateid
.si_stateownerid
,
3122 stp
->st_stateid
.si_fileid
,
3123 stp
->st_stateid
.si_generation
);
3125 nfsd4_create_clid_dir(sop
->so_client
);
3127 if (oc
->oc_stateowner
) {
3128 nfs4_get_stateowner(oc
->oc_stateowner
);
3129 cstate
->replay_owner
= oc
->oc_stateowner
;
3131 nfs4_unlock_state();
3137 * unset all bits in union bitmap (bmap) that
3138 * do not exist in share (from successful OPEN_DOWNGRADE)
3141 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3144 for (i
= 1; i
< 4; i
++) {
3145 if ((i
& access
) != i
)
3146 __clear_bit(i
, bmap
);
3151 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3154 for (i
= 0; i
< 4; i
++) {
3155 if ((i
& deny
) != i
)
3156 __clear_bit(i
, bmap
);
3161 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3162 struct nfsd4_compound_state
*cstate
,
3163 struct nfsd4_open_downgrade
*od
)
3166 struct nfs4_stateid
*stp
;
3167 unsigned int share_access
;
3169 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3170 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3171 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3173 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3174 || !deny_valid(od
->od_share_deny
))
3175 return nfserr_inval
;
3178 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3182 &od
->od_stateowner
, &stp
, NULL
)))
3185 status
= nfserr_inval
;
3186 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3187 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3188 stp
->st_access_bmap
, od
->od_share_access
);
3191 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3192 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3193 stp
->st_deny_bmap
, od
->od_share_deny
);
3196 set_access(&share_access
, stp
->st_access_bmap
);
3197 nfs4_file_downgrade(stp
->st_vfs_file
,
3198 share_access
& ~od
->od_share_access
);
3200 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3201 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3203 update_stateid(&stp
->st_stateid
);
3204 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3207 if (od
->od_stateowner
) {
3208 nfs4_get_stateowner(od
->od_stateowner
);
3209 cstate
->replay_owner
= od
->od_stateowner
;
3211 nfs4_unlock_state();
3216 * nfs4_unlock_state() called after encode
3219 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3220 struct nfsd4_close
*close
)
3223 struct nfs4_stateid
*stp
;
3225 dprintk("NFSD: nfsd4_close on file %.*s\n",
3226 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3227 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3230 /* check close_lru for replay */
3231 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3234 OPEN_STATE
| CLOSE_STATE
,
3235 &close
->cl_stateowner
, &stp
, NULL
)))
3238 update_stateid(&stp
->st_stateid
);
3239 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3241 /* release_stateid() calls nfsd_close() if needed */
3242 release_open_stateid(stp
);
3244 /* place unused nfs4_stateowners on so_close_lru list to be
3245 * released by the laundromat service after the lease period
3246 * to enable us to handle CLOSE replay
3248 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3249 move_to_close_lru(close
->cl_stateowner
);
3251 if (close
->cl_stateowner
) {
3252 nfs4_get_stateowner(close
->cl_stateowner
);
3253 cstate
->replay_owner
= close
->cl_stateowner
;
3255 nfs4_unlock_state();
3260 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3261 struct nfsd4_delegreturn
*dr
)
3263 struct nfs4_delegation
*dp
;
3264 stateid_t
*stateid
= &dr
->dr_stateid
;
3265 struct inode
*inode
;
3269 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3271 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3273 if (nfsd4_has_session(cstate
))
3274 flags
|= HAS_SESSION
;
3276 status
= nfserr_bad_stateid
;
3277 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3279 status
= nfserr_stale_stateid
;
3280 if (STALE_STATEID(stateid
))
3282 status
= nfserr_bad_stateid
;
3283 if (!is_delegation_stateid(stateid
))
3285 dp
= find_delegation_stateid(inode
, stateid
);
3287 status
= stateid_error_map(stateid
);
3290 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3293 renew_client(dp
->dl_client
);
3295 unhash_delegation(dp
);
3297 nfs4_unlock_state();
3304 * Lock owner state (byte-range locks)
3306 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3307 #define LOCK_HASH_BITS 8
3308 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3309 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3312 end_offset(u64 start
, u64 len
)
3317 return end
>= start
? end
: NFS4_MAX_UINT64
;
3320 /* last octet in a range */
3322 last_byte_offset(u64 start
, u64 len
)
3328 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3331 #define lockownerid_hashval(id) \
3332 ((id) & LOCK_HASH_MASK)
3334 static inline unsigned int
3335 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3336 struct xdr_netobj
*ownername
)
3338 return (file_hashval(inode
) + cl_id
3339 + opaque_hashval(ownername
->data
, ownername
->len
))
3343 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3344 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3345 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3347 static struct nfs4_stateid
*
3348 find_stateid(stateid_t
*stid
, int flags
)
3350 struct nfs4_stateid
*local
;
3351 u32 st_id
= stid
->si_stateownerid
;
3352 u32 f_id
= stid
->si_fileid
;
3353 unsigned int hashval
;
3355 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3356 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3357 hashval
= stateid_hashval(st_id
, f_id
);
3358 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3359 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3360 (local
->st_stateid
.si_fileid
== f_id
))
3365 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3366 hashval
= stateid_hashval(st_id
, f_id
);
3367 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3368 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3369 (local
->st_stateid
.si_fileid
== f_id
))
3376 static struct nfs4_delegation
*
3377 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3379 struct nfs4_file
*fp
;
3380 struct nfs4_delegation
*dl
;
3382 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3383 stid
->si_boot
, stid
->si_stateownerid
,
3384 stid
->si_fileid
, stid
->si_generation
);
3386 fp
= find_file(ino
);
3389 dl
= find_delegation_file(fp
, stid
);
3395 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3396 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3397 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3398 * locking, this prevents us from being completely protocol-compliant. The
3399 * real solution to this problem is to start using unsigned file offsets in
3400 * the VFS, but this is a very deep change!
3403 nfs4_transform_lock_offset(struct file_lock
*lock
)
3405 if (lock
->fl_start
< 0)
3406 lock
->fl_start
= OFFSET_MAX
;
3407 if (lock
->fl_end
< 0)
3408 lock
->fl_end
= OFFSET_MAX
;
3411 /* Hack!: For now, we're defining this just so we can use a pointer to it
3412 * as a unique cookie to identify our (NFSv4's) posix locks. */
3413 static struct lock_manager_operations nfsd_posix_mng_ops
= {
3417 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3419 struct nfs4_stateowner
*sop
;
3422 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3423 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3424 hval
= lockownerid_hashval(sop
->so_id
);
3425 kref_get(&sop
->so_ref
);
3427 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3429 deny
->ld_sop
= NULL
;
3430 deny
->ld_clientid
.cl_boot
= 0;
3431 deny
->ld_clientid
.cl_id
= 0;
3433 deny
->ld_start
= fl
->fl_start
;
3434 deny
->ld_length
= NFS4_MAX_UINT64
;
3435 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3436 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3437 deny
->ld_type
= NFS4_READ_LT
;
3438 if (fl
->fl_type
!= F_RDLCK
)
3439 deny
->ld_type
= NFS4_WRITE_LT
;
3442 static struct nfs4_stateowner
*
3443 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3444 struct xdr_netobj
*owner
)
3446 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3447 struct nfs4_stateowner
*op
;
3449 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3450 if (same_owner_str(op
, owner
, clid
))
3457 * Alloc a lock owner structure.
3458 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3461 * strhashval = lock_ownerstr_hashval
3464 static struct nfs4_stateowner
*
3465 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3466 struct nfs4_stateowner
*sop
;
3467 struct nfs4_replay
*rp
;
3468 unsigned int idhashval
;
3470 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3472 idhashval
= lockownerid_hashval(current_ownerid
);
3473 INIT_LIST_HEAD(&sop
->so_idhash
);
3474 INIT_LIST_HEAD(&sop
->so_strhash
);
3475 INIT_LIST_HEAD(&sop
->so_perclient
);
3476 INIT_LIST_HEAD(&sop
->so_stateids
);
3477 INIT_LIST_HEAD(&sop
->so_perstateid
);
3478 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3480 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3481 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3482 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3483 sop
->so_is_open_owner
= 0;
3484 sop
->so_id
= current_ownerid
++;
3485 sop
->so_client
= clp
;
3486 /* It is the openowner seqid that will be incremented in encode in the
3487 * case of new lockowners; so increment the lock seqid manually: */
3488 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3489 sop
->so_confirmed
= 1;
3490 rp
= &sop
->so_replay
;
3491 rp
->rp_status
= nfserr_serverfault
;
3493 rp
->rp_buf
= rp
->rp_ibuf
;
3497 static struct nfs4_stateid
*
3498 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3500 struct nfs4_stateid
*stp
;
3501 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3503 stp
= nfs4_alloc_stateid();
3506 INIT_LIST_HEAD(&stp
->st_hash
);
3507 INIT_LIST_HEAD(&stp
->st_perfile
);
3508 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3509 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3510 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3511 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3512 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3513 stp
->st_stateowner
= sop
;
3516 stp
->st_stateid
.si_boot
= get_seconds();
3517 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3518 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3519 stp
->st_stateid
.si_generation
= 0;
3520 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3521 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3522 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3523 stp
->st_openstp
= open_stp
;
3530 check_lock_length(u64 offset
, u64 length
)
3532 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3533 LOFF_OVERFLOW(offset
, length
)));
3540 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3541 struct nfsd4_lock
*lock
)
3543 struct nfs4_stateowner
*open_sop
= NULL
;
3544 struct nfs4_stateowner
*lock_sop
= NULL
;
3545 struct nfs4_stateid
*lock_stp
;
3547 struct file_lock file_lock
;
3548 struct file_lock conflock
;
3550 unsigned int strhashval
;
3554 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3555 (long long) lock
->lk_offset
,
3556 (long long) lock
->lk_length
);
3558 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3559 return nfserr_inval
;
3561 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3562 S_IFREG
, NFSD_MAY_LOCK
))) {
3563 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3569 if (lock
->lk_is_new
) {
3571 * Client indicates that this is a new lockowner.
3572 * Use open owner and open stateid to create lock owner and
3575 struct nfs4_stateid
*open_stp
= NULL
;
3576 struct nfs4_file
*fp
;
3578 status
= nfserr_stale_clientid
;
3579 if (!nfsd4_has_session(cstate
) &&
3580 STALE_CLIENTID(&lock
->lk_new_clientid
))
3583 /* validate and update open stateid and open seqid */
3584 status
= nfs4_preprocess_seqid_op(cstate
,
3585 lock
->lk_new_open_seqid
,
3586 &lock
->lk_new_open_stateid
,
3588 &lock
->lk_replay_owner
, &open_stp
,
3592 open_sop
= lock
->lk_replay_owner
;
3593 /* create lockowner and lock stateid */
3594 fp
= open_stp
->st_file
;
3595 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3596 open_sop
->so_client
->cl_clientid
.cl_id
,
3597 &lock
->v
.new.owner
);
3598 /* XXX: Do we need to check for duplicate stateowners on
3599 * the same file, or should they just be allowed (and
3600 * create new stateids)? */
3601 status
= nfserr_resource
;
3602 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3603 open_sop
->so_client
, open_stp
, lock
);
3604 if (lock_sop
== NULL
)
3606 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3607 if (lock_stp
== NULL
)
3610 /* lock (lock owner + lock stateid) already exists */
3611 status
= nfs4_preprocess_seqid_op(cstate
,
3612 lock
->lk_old_lock_seqid
,
3613 &lock
->lk_old_lock_stateid
,
3615 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3618 lock_sop
= lock
->lk_replay_owner
;
3620 /* lock->lk_replay_owner and lock_stp have been created or found */
3621 filp
= lock_stp
->st_vfs_file
;
3623 status
= nfserr_grace
;
3624 if (locks_in_grace() && !lock
->lk_reclaim
)
3626 status
= nfserr_no_grace
;
3627 if (!locks_in_grace() && lock
->lk_reclaim
)
3630 locks_init_lock(&file_lock
);
3631 switch (lock
->lk_type
) {
3634 file_lock
.fl_type
= F_RDLCK
;
3638 case NFS4_WRITEW_LT
:
3639 file_lock
.fl_type
= F_WRLCK
;
3643 status
= nfserr_inval
;
3646 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3647 file_lock
.fl_pid
= current
->tgid
;
3648 file_lock
.fl_file
= filp
;
3649 file_lock
.fl_flags
= FL_POSIX
;
3650 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3652 file_lock
.fl_start
= lock
->lk_offset
;
3653 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3654 nfs4_transform_lock_offset(&file_lock
);
3657 * Try to lock the file in the VFS.
3658 * Note: locks.c uses the BKL to protect the inode's lock list.
3661 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3663 case 0: /* success! */
3664 update_stateid(&lock_stp
->st_stateid
);
3665 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3669 case (EAGAIN
): /* conflock holds conflicting lock */
3670 status
= nfserr_denied
;
3671 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3672 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3675 status
= nfserr_deadlock
;
3678 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3679 status
= nfserr_resource
;
3683 if (status
&& lock
->lk_is_new
&& lock_sop
)
3684 release_lockowner(lock_sop
);
3685 if (lock
->lk_replay_owner
) {
3686 nfs4_get_stateowner(lock
->lk_replay_owner
);
3687 cstate
->replay_owner
= lock
->lk_replay_owner
;
3689 nfs4_unlock_state();
3694 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3695 * so we do a temporary open here just to get an open file to pass to
3696 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3699 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3704 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3707 err
= vfs_test_lock(file
, lock
);
3716 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3717 struct nfsd4_lockt
*lockt
)
3719 struct inode
*inode
;
3720 struct file_lock file_lock
;
3724 if (locks_in_grace())
3725 return nfserr_grace
;
3727 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3728 return nfserr_inval
;
3730 lockt
->lt_stateowner
= NULL
;
3733 status
= nfserr_stale_clientid
;
3734 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3737 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3738 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3739 if (status
== nfserr_symlink
)
3740 status
= nfserr_inval
;
3744 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3745 locks_init_lock(&file_lock
);
3746 switch (lockt
->lt_type
) {
3749 file_lock
.fl_type
= F_RDLCK
;
3752 case NFS4_WRITEW_LT
:
3753 file_lock
.fl_type
= F_WRLCK
;
3756 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3757 status
= nfserr_inval
;
3761 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3762 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3763 if (lockt
->lt_stateowner
)
3764 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3765 file_lock
.fl_pid
= current
->tgid
;
3766 file_lock
.fl_flags
= FL_POSIX
;
3768 file_lock
.fl_start
= lockt
->lt_offset
;
3769 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3771 nfs4_transform_lock_offset(&file_lock
);
3774 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3776 status
= nfserrno(error
);
3779 if (file_lock
.fl_type
!= F_UNLCK
) {
3780 status
= nfserr_denied
;
3781 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3784 nfs4_unlock_state();
3789 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3790 struct nfsd4_locku
*locku
)
3792 struct nfs4_stateid
*stp
;
3793 struct file
*filp
= NULL
;
3794 struct file_lock file_lock
;
3798 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3799 (long long) locku
->lu_offset
,
3800 (long long) locku
->lu_length
);
3802 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3803 return nfserr_inval
;
3807 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3811 &locku
->lu_stateowner
, &stp
, NULL
)))
3814 filp
= stp
->st_vfs_file
;
3816 locks_init_lock(&file_lock
);
3817 file_lock
.fl_type
= F_UNLCK
;
3818 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3819 file_lock
.fl_pid
= current
->tgid
;
3820 file_lock
.fl_file
= filp
;
3821 file_lock
.fl_flags
= FL_POSIX
;
3822 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3823 file_lock
.fl_start
= locku
->lu_offset
;
3825 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3826 nfs4_transform_lock_offset(&file_lock
);
3829 * Try to unlock the file in the VFS.
3831 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3833 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3837 * OK, unlock succeeded; the only thing left to do is update the stateid.
3839 update_stateid(&stp
->st_stateid
);
3840 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3843 if (locku
->lu_stateowner
) {
3844 nfs4_get_stateowner(locku
->lu_stateowner
);
3845 cstate
->replay_owner
= locku
->lu_stateowner
;
3847 nfs4_unlock_state();
3851 status
= nfserrno(err
);
3857 * 1: locks held by lockowner
3858 * 0: no locks held by lockowner
3861 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3863 struct file_lock
**flpp
;
3864 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3868 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3869 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3880 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3881 struct nfsd4_compound_state
*cstate
,
3882 struct nfsd4_release_lockowner
*rlockowner
)
3884 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3885 struct nfs4_stateowner
*sop
;
3886 struct nfs4_stateid
*stp
;
3887 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3888 struct list_head matches
;
3892 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3893 clid
->cl_boot
, clid
->cl_id
);
3895 /* XXX check for lease expiration */
3897 status
= nfserr_stale_clientid
;
3898 if (STALE_CLIENTID(clid
))
3903 status
= nfserr_locks_held
;
3904 /* XXX: we're doing a linear search through all the lockowners.
3905 * Yipes! For now we'll just hope clients aren't really using
3906 * release_lockowner much, but eventually we have to fix these
3907 * data structures. */
3908 INIT_LIST_HEAD(&matches
);
3909 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3910 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3911 if (!same_owner_str(sop
, owner
, clid
))
3913 list_for_each_entry(stp
, &sop
->so_stateids
,
3915 if (check_for_locks(stp
->st_vfs_file
, sop
))
3917 /* Note: so_perclient unused for lockowners,
3918 * so it's OK to fool with here. */
3919 list_add(&sop
->so_perclient
, &matches
);
3923 /* Clients probably won't expect us to return with some (but not all)
3924 * of the lockowner state released; so don't release any until all
3925 * have been checked. */
3927 while (!list_empty(&matches
)) {
3928 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3930 /* unhash_stateowner deletes so_perclient only
3931 * for openowners. */
3932 list_del(&sop
->so_perclient
);
3933 release_lockowner(sop
);
3936 nfs4_unlock_state();
3940 static inline struct nfs4_client_reclaim
*
3943 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3947 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3949 unsigned int strhashval
= clientstr_hashval(name
);
3950 struct nfs4_client
*clp
;
3952 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3957 * failure => all reset bets are off, nfserr_no_grace...
3960 nfs4_client_to_reclaim(const char *name
)
3962 unsigned int strhashval
;
3963 struct nfs4_client_reclaim
*crp
= NULL
;
3965 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3966 crp
= alloc_reclaim();
3969 strhashval
= clientstr_hashval(name
);
3970 INIT_LIST_HEAD(&crp
->cr_strhash
);
3971 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3972 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3973 reclaim_str_hashtbl_size
++;
3978 nfs4_release_reclaim(void)
3980 struct nfs4_client_reclaim
*crp
= NULL
;
3983 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3984 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3985 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3986 struct nfs4_client_reclaim
, cr_strhash
);
3987 list_del(&crp
->cr_strhash
);
3989 reclaim_str_hashtbl_size
--;
3992 BUG_ON(reclaim_str_hashtbl_size
);
3996 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3997 static struct nfs4_client_reclaim
*
3998 nfs4_find_reclaim_client(clientid_t
*clid
)
4000 unsigned int strhashval
;
4001 struct nfs4_client
*clp
;
4002 struct nfs4_client_reclaim
*crp
= NULL
;
4005 /* find clientid in conf_id_hashtbl */
4006 clp
= find_confirmed_client(clid
);
4010 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4011 clp
->cl_name
.len
, clp
->cl_name
.data
,
4014 /* find clp->cl_name in reclaim_str_hashtbl */
4015 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4016 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4017 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4025 * Called from OPEN. Look for clientid in reclaim list.
4028 nfs4_check_open_reclaim(clientid_t
*clid
)
4030 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4033 /* initialization to perform at module load time: */
4036 nfs4_state_init(void)
4040 status
= nfsd4_init_slabs();
4043 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4044 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4045 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4046 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4047 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4048 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4050 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4051 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4052 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4053 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4055 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4056 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4057 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
4059 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
4060 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4061 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
4063 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4064 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4065 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4067 memset(&onestateid
, ~0, sizeof(stateid_t
));
4068 INIT_LIST_HEAD(&close_lru
);
4069 INIT_LIST_HEAD(&client_lru
);
4070 INIT_LIST_HEAD(&del_recall_lru
);
4071 reclaim_str_hashtbl_size
= 0;
4076 nfsd4_load_reboot_recovery_data(void)
4081 nfsd4_init_recdir(user_recovery_dirname
);
4082 status
= nfsd4_recdir_load();
4083 nfs4_unlock_state();
4085 printk("NFSD: Failure reading reboot recovery data\n");
4089 get_nfs4_grace_period(void)
4091 return max(user_lease_time
, lease_time
) * HZ
;
4095 * Since the lifetime of a delegation isn't limited to that of an open, a
4096 * client may quite reasonably hang on to a delegation as long as it has
4097 * the inode cached. This becomes an obvious problem the first time a
4098 * client's inode cache approaches the size of the server's total memory.
4100 * For now we avoid this problem by imposing a hard limit on the number
4101 * of delegations, which varies according to the server's memory size.
4104 set_max_delegations(void)
4107 * Allow at most 4 delegations per megabyte of RAM. Quick
4108 * estimates suggest that in the worst case (where every delegation
4109 * is for a different inode), a delegation could take about 1.5K,
4110 * giving a worst case usage of about 6% of memory.
4112 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4115 /* initialization to perform when the nfsd service is started: */
4118 __nfs4_state_start(void)
4120 unsigned long grace_time
;
4122 boot_time
= get_seconds();
4123 grace_time
= get_nfs4_grace_period();
4124 lease_time
= user_lease_time
;
4125 locks_start_grace(&nfsd4_manager
);
4126 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4128 laundry_wq
= create_singlethread_workqueue("nfsd4");
4129 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
4130 set_max_delegations();
4134 nfs4_state_start(void)
4138 nfsd4_load_reboot_recovery_data();
4139 __nfs4_state_start();
4145 nfs4_lease_time(void)
4151 __nfs4_state_shutdown(void)
4154 struct nfs4_client
*clp
= NULL
;
4155 struct nfs4_delegation
*dp
= NULL
;
4156 struct list_head
*pos
, *next
, reaplist
;
4158 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4159 while (!list_empty(&conf_id_hashtbl
[i
])) {
4160 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4163 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4164 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4168 INIT_LIST_HEAD(&reaplist
);
4169 spin_lock(&recall_lock
);
4170 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4171 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4172 list_move(&dp
->dl_recall_lru
, &reaplist
);
4174 spin_unlock(&recall_lock
);
4175 list_for_each_safe(pos
, next
, &reaplist
) {
4176 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4177 list_del_init(&dp
->dl_recall_lru
);
4178 unhash_delegation(dp
);
4181 nfsd4_shutdown_recdir();
4186 nfs4_state_shutdown(void)
4188 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4189 destroy_workqueue(laundry_wq
);
4190 locks_end_grace(&nfsd4_manager
);
4192 nfs4_release_reclaim();
4193 __nfs4_state_shutdown();
4194 nfs4_unlock_state();
4198 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4199 * accessed when nfsd is starting.
4202 nfs4_set_recdir(char *recdir
)
4204 strcpy(user_recovery_dirname
, recdir
);
4208 * Change the NFSv4 recovery directory to recdir.
4211 nfs4_reset_recoverydir(char *recdir
)
4216 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4220 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4221 nfs4_set_recdir(recdir
);
4229 nfs4_recoverydir(void)
4231 return user_recovery_dirname
;
4235 * Called when leasetime is changed.
4237 * The only way the protocol gives us to handle on-the-fly lease changes is to
4238 * simulate a reboot. Instead of doing that, we just wait till the next time
4239 * we start to register any changes in lease time. If the administrator
4240 * really wants to change the lease time *now*, they can go ahead and bring
4241 * nfsd down and then back up again after changing the lease time.
4243 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4244 * when nfsd is starting
4247 nfs4_reset_lease(time_t leasetime
)
4249 user_lease_time
= leasetime
;