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>
58 #include <linux/sunrpc/clnt.h>
60 #define NFSDDBG_FACILITY NFSDDBG_PROC
63 static time_t lease_time
= 90; /* default lease time */
64 static time_t user_lease_time
= 90;
65 static time_t boot_time
;
66 static u32 current_ownerid
= 1;
67 static u32 current_fileid
= 1;
68 static u32 current_delegid
= 1;
70 static stateid_t zerostateid
; /* bits all 0 */
71 static stateid_t onestateid
; /* bits all 1 */
72 static u64 current_sessionid
= 1;
74 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
75 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
77 /* forward declarations */
78 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
79 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
80 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
81 static void nfs4_set_recdir(char *recdir
);
85 /* Currently used for almost all code touching nfsv4 state: */
86 static DEFINE_MUTEX(client_mutex
);
89 * Currently used for the del_recall_lru and file hash table. In an
90 * effort to decrease the scope of the client_mutex, this spinlock may
91 * eventually cover more:
93 static DEFINE_SPINLOCK(recall_lock
);
95 static struct kmem_cache
*stateowner_slab
= NULL
;
96 static struct kmem_cache
*file_slab
= NULL
;
97 static struct kmem_cache
*stateid_slab
= NULL
;
98 static struct kmem_cache
*deleg_slab
= NULL
;
101 nfs4_lock_state(void)
103 mutex_lock(&client_mutex
);
107 nfs4_unlock_state(void)
109 mutex_unlock(&client_mutex
);
113 opaque_hashval(const void *ptr
, int nbytes
)
115 unsigned char *cptr
= (unsigned char *) ptr
;
125 static struct list_head del_recall_lru
;
128 put_nfs4_file(struct nfs4_file
*fi
)
130 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
131 list_del(&fi
->fi_hash
);
132 spin_unlock(&recall_lock
);
134 kmem_cache_free(file_slab
, fi
);
139 get_nfs4_file(struct nfs4_file
*fi
)
141 atomic_inc(&fi
->fi_ref
);
144 static int num_delegations
;
145 unsigned int max_delegations
;
148 * Open owner state (share locks)
151 /* hash tables for nfs4_stateowner */
152 #define OWNER_HASH_BITS 8
153 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
154 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
156 #define ownerid_hashval(id) \
157 ((id) & OWNER_HASH_MASK)
158 #define ownerstr_hashval(clientid, ownername) \
159 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
161 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
162 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
164 /* hash table for nfs4_file */
165 #define FILE_HASH_BITS 8
166 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
167 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
168 /* hash table for (open)nfs4_stateid */
169 #define STATEID_HASH_BITS 10
170 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
171 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
173 #define file_hashval(x) \
174 hash_ptr(x, FILE_HASH_BITS)
175 #define stateid_hashval(owner_id, file_id) \
176 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
178 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
179 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
181 static struct nfs4_delegation
*
182 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
184 struct nfs4_delegation
*dp
;
185 struct nfs4_file
*fp
= stp
->st_file
;
186 struct nfs4_cb_conn
*cb
= &stp
->st_stateowner
->so_client
->cl_cb_conn
;
188 dprintk("NFSD alloc_init_deleg\n");
189 if (fp
->fi_had_conflict
)
191 if (num_delegations
> max_delegations
)
193 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
197 INIT_LIST_HEAD(&dp
->dl_perfile
);
198 INIT_LIST_HEAD(&dp
->dl_perclnt
);
199 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
204 get_file(stp
->st_vfs_file
);
205 dp
->dl_vfs_file
= stp
->st_vfs_file
;
207 dp
->dl_ident
= cb
->cb_ident
;
208 dp
->dl_stateid
.si_boot
= get_seconds();
209 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
210 dp
->dl_stateid
.si_fileid
= 0;
211 dp
->dl_stateid
.si_generation
= 0;
212 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
214 atomic_set(&dp
->dl_count
, 1);
215 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
216 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
221 nfs4_put_delegation(struct nfs4_delegation
*dp
)
223 if (atomic_dec_and_test(&dp
->dl_count
)) {
224 dprintk("NFSD: freeing dp %p\n",dp
);
225 put_nfs4_file(dp
->dl_file
);
226 kmem_cache_free(deleg_slab
, dp
);
231 /* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
237 nfs4_close_delegation(struct nfs4_delegation
*dp
)
239 struct file
*filp
= dp
->dl_vfs_file
;
241 dprintk("NFSD: close_delegation dp %p\n",dp
);
242 dp
->dl_vfs_file
= NULL
;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
246 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
250 /* Called under the state lock. */
252 unhash_delegation(struct nfs4_delegation
*dp
)
254 list_del_init(&dp
->dl_perfile
);
255 list_del_init(&dp
->dl_perclnt
);
256 spin_lock(&recall_lock
);
257 list_del_init(&dp
->dl_recall_lru
);
258 spin_unlock(&recall_lock
);
259 nfs4_close_delegation(dp
);
260 nfs4_put_delegation(dp
);
267 /* Hash tables for nfs4_clientid state */
268 #define CLIENT_HASH_BITS 4
269 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
272 #define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
274 #define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
286 * client_lru holds client queue ordered by nfs4_client.cl_time
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
292 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
293 static int reclaim_str_hashtbl_size
= 0;
294 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
295 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
296 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
297 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
298 static struct list_head client_lru
;
299 static struct list_head close_lru
;
301 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
303 list_del(&stp
->st_hash
);
304 list_del(&stp
->st_perfile
);
305 list_del(&stp
->st_perstateowner
);
308 static void free_generic_stateid(struct nfs4_stateid
*stp
)
310 put_nfs4_file(stp
->st_file
);
311 kmem_cache_free(stateid_slab
, stp
);
314 static void release_lock_stateid(struct nfs4_stateid
*stp
)
316 unhash_generic_stateid(stp
);
317 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
318 free_generic_stateid(stp
);
321 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
323 struct nfs4_stateid
*stp
;
325 list_del(&sop
->so_idhash
);
326 list_del(&sop
->so_strhash
);
327 list_del(&sop
->so_perstateid
);
328 while (!list_empty(&sop
->so_stateids
)) {
329 stp
= list_first_entry(&sop
->so_stateids
,
330 struct nfs4_stateid
, st_perstateowner
);
331 release_lock_stateid(stp
);
335 static void release_lockowner(struct nfs4_stateowner
*sop
)
337 unhash_lockowner(sop
);
338 nfs4_put_stateowner(sop
);
342 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
344 struct nfs4_stateowner
*lock_sop
;
346 while (!list_empty(&open_stp
->st_lockowners
)) {
347 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
348 struct nfs4_stateowner
, so_perstateid
);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop
->so_is_open_owner
);
351 release_lockowner(lock_sop
);
355 static void release_open_stateid(struct nfs4_stateid
*stp
)
357 unhash_generic_stateid(stp
);
358 release_stateid_lockowners(stp
);
359 nfsd_close(stp
->st_vfs_file
);
360 free_generic_stateid(stp
);
363 static void unhash_openowner(struct nfs4_stateowner
*sop
)
365 struct nfs4_stateid
*stp
;
367 list_del(&sop
->so_idhash
);
368 list_del(&sop
->so_strhash
);
369 list_del(&sop
->so_perclient
);
370 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
371 while (!list_empty(&sop
->so_stateids
)) {
372 stp
= list_first_entry(&sop
->so_stateids
,
373 struct nfs4_stateid
, st_perstateowner
);
374 release_open_stateid(stp
);
378 static void release_openowner(struct nfs4_stateowner
*sop
)
380 unhash_openowner(sop
);
381 list_del(&sop
->so_close_lru
);
382 nfs4_put_stateowner(sop
);
385 static DEFINE_SPINLOCK(sessionid_lock
);
386 #define SESSION_HASH_SIZE 512
387 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
390 hash_sessionid(struct nfs4_sessionid
*sessionid
)
392 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
394 return sid
->sequence
% SESSION_HASH_SIZE
;
398 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
400 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
405 gen_sessionid(struct nfsd4_session
*ses
)
407 struct nfs4_client
*clp
= ses
->se_client
;
408 struct nfsd4_sessionid
*sid
;
410 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
411 sid
->clientid
= clp
->cl_clientid
;
412 sid
->sequence
= current_sessionid
++;
417 * Give the client the number of slots it requests bound by
418 * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
420 * If we run out of reserved DRC memory we should (up to a point) re-negotiate
421 * active sessions and reduce their slot usage to make rooom for new
422 * connections. For now we just fail the create session.
424 static int set_forechannel_maxreqs(struct nfsd4_channel_attrs
*fchan
)
428 if (fchan
->maxreqs
< 1)
430 else if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
431 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
433 mem
= fchan
->maxreqs
* NFSD_SLOT_CACHE_SIZE
;
435 spin_lock(&nfsd_drc_lock
);
436 if (mem
+ nfsd_drc_mem_used
> nfsd_drc_max_mem
)
437 mem
= ((nfsd_drc_max_mem
- nfsd_drc_mem_used
) /
438 NFSD_SLOT_CACHE_SIZE
) * NFSD_SLOT_CACHE_SIZE
;
439 nfsd_drc_mem_used
+= mem
;
440 spin_unlock(&nfsd_drc_lock
);
442 fchan
->maxreqs
= mem
/ NFSD_SLOT_CACHE_SIZE
;
443 if (fchan
->maxreqs
== 0)
444 return nfserr_resource
;
449 * fchan holds the client values on input, and the server values on output
451 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
452 struct nfsd4_channel_attrs
*session_fchan
,
453 struct nfsd4_channel_attrs
*fchan
)
456 __u32 maxcount
= svc_max_payload(rqstp
);
458 /* headerpadsz set to zero in encode routine */
460 /* Use the client's max request and max response size if possible */
461 if (fchan
->maxreq_sz
> maxcount
)
462 fchan
->maxreq_sz
= maxcount
;
463 session_fchan
->maxreq_sz
= fchan
->maxreq_sz
;
465 if (fchan
->maxresp_sz
> maxcount
)
466 fchan
->maxresp_sz
= maxcount
;
467 session_fchan
->maxresp_sz
= fchan
->maxresp_sz
;
469 session_fchan
->maxresp_cached
= NFSD_SLOT_CACHE_SIZE
;
470 fchan
->maxresp_cached
= session_fchan
->maxresp_cached
;
472 /* Use the client's maxops if possible */
473 if (fchan
->maxops
> NFSD_MAX_OPS_PER_COMPOUND
)
474 fchan
->maxops
= NFSD_MAX_OPS_PER_COMPOUND
;
475 session_fchan
->maxops
= fchan
->maxops
;
477 /* FIXME: Error means no more DRC pages so the server should
478 * recover pages from existing sessions. For now fail session
481 status
= set_forechannel_maxreqs(fchan
);
483 session_fchan
->maxreqs
= fchan
->maxreqs
;
488 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
489 struct nfsd4_create_session
*cses
)
491 struct nfsd4_session
*new, tmp
;
492 int idx
, status
= nfserr_resource
, slotsize
;
494 memset(&tmp
, 0, sizeof(tmp
));
496 /* FIXME: For now, we just accept the client back channel attributes. */
497 tmp
.se_bchannel
= cses
->back_channel
;
498 status
= init_forechannel_attrs(rqstp
, &tmp
.se_fchannel
,
499 &cses
->fore_channel
);
503 /* allocate struct nfsd4_session and slot table in one piece */
504 slotsize
= tmp
.se_fchannel
.maxreqs
* sizeof(struct nfsd4_slot
);
505 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
509 memcpy(new, &tmp
, sizeof(*new));
511 new->se_client
= clp
;
513 idx
= hash_sessionid(&new->se_sessionid
);
514 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
515 NFS4_MAX_SESSIONID_LEN
);
517 new->se_flags
= cses
->flags
;
518 kref_init(&new->se_ref
);
519 spin_lock(&sessionid_lock
);
520 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
521 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
522 spin_unlock(&sessionid_lock
);
529 /* caller must hold sessionid_lock */
530 static struct nfsd4_session
*
531 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
533 struct nfsd4_session
*elem
;
536 dump_sessionid(__func__
, sessionid
);
537 idx
= hash_sessionid(sessionid
);
538 dprintk("%s: idx is %d\n", __func__
, idx
);
539 /* Search in the appropriate list */
540 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
541 dump_sessionid("list traversal", &elem
->se_sessionid
);
542 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
543 NFS4_MAX_SESSIONID_LEN
)) {
548 dprintk("%s: session not found\n", __func__
);
552 /* caller must hold sessionid_lock */
554 unhash_session(struct nfsd4_session
*ses
)
556 list_del(&ses
->se_hash
);
557 list_del(&ses
->se_perclnt
);
561 release_session(struct nfsd4_session
*ses
)
563 spin_lock(&sessionid_lock
);
565 spin_unlock(&sessionid_lock
);
566 nfsd4_put_session(ses
);
569 static void nfsd4_release_respages(struct page
**respages
, short resused
);
572 free_session(struct kref
*kref
)
574 struct nfsd4_session
*ses
;
577 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
578 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++) {
579 struct nfsd4_cache_entry
*e
= &ses
->se_slots
[i
].sl_cache_entry
;
580 nfsd4_release_respages(e
->ce_respages
, e
->ce_resused
);
582 spin_lock(&nfsd_drc_lock
);
583 nfsd_drc_mem_used
-= ses
->se_fchannel
.maxreqs
* NFSD_SLOT_CACHE_SIZE
;
584 spin_unlock(&nfsd_drc_lock
);
589 renew_client(struct nfs4_client
*clp
)
592 * Move client to the end to the LRU list.
594 dprintk("renewing client (clientid %08x/%08x)\n",
595 clp
->cl_clientid
.cl_boot
,
596 clp
->cl_clientid
.cl_id
);
597 list_move_tail(&clp
->cl_lru
, &client_lru
);
598 clp
->cl_time
= get_seconds();
601 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
603 STALE_CLIENTID(clientid_t
*clid
)
605 if (clid
->cl_boot
== boot_time
)
607 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
608 clid
->cl_boot
, clid
->cl_id
, boot_time
);
613 * XXX Should we use a slab cache ?
614 * This type of memory management is somewhat inefficient, but we use it
615 * anyway since SETCLIENTID is not a common operation.
617 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
619 struct nfs4_client
*clp
;
621 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
624 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
625 if (clp
->cl_name
.data
== NULL
) {
629 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
630 clp
->cl_name
.len
= name
.len
;
635 shutdown_callback_client(struct nfs4_client
*clp
)
637 struct rpc_clnt
*clnt
= clp
->cl_cb_conn
.cb_client
;
641 * Callback threads take a reference on the client, so there
642 * should be no outstanding callbacks at this point.
644 clp
->cl_cb_conn
.cb_client
= NULL
;
645 rpc_shutdown_client(clnt
);
647 if (clp
->cl_cb_conn
.cb_cred
) {
648 put_rpccred(clp
->cl_cb_conn
.cb_cred
);
649 clp
->cl_cb_conn
.cb_cred
= NULL
;
654 free_client(struct nfs4_client
*clp
)
656 shutdown_callback_client(clp
);
657 if (clp
->cl_cred
.cr_group_info
)
658 put_group_info(clp
->cl_cred
.cr_group_info
);
659 kfree(clp
->cl_principal
);
660 kfree(clp
->cl_name
.data
);
665 put_nfs4_client(struct nfs4_client
*clp
)
667 if (atomic_dec_and_test(&clp
->cl_count
))
672 expire_client(struct nfs4_client
*clp
)
674 struct nfs4_stateowner
*sop
;
675 struct nfs4_delegation
*dp
;
676 struct list_head reaplist
;
678 dprintk("NFSD: expire_client cl_count %d\n",
679 atomic_read(&clp
->cl_count
));
681 INIT_LIST_HEAD(&reaplist
);
682 spin_lock(&recall_lock
);
683 while (!list_empty(&clp
->cl_delegations
)) {
684 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
685 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
687 list_del_init(&dp
->dl_perclnt
);
688 list_move(&dp
->dl_recall_lru
, &reaplist
);
690 spin_unlock(&recall_lock
);
691 while (!list_empty(&reaplist
)) {
692 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
693 list_del_init(&dp
->dl_recall_lru
);
694 unhash_delegation(dp
);
696 list_del(&clp
->cl_idhash
);
697 list_del(&clp
->cl_strhash
);
698 list_del(&clp
->cl_lru
);
699 while (!list_empty(&clp
->cl_openowners
)) {
700 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
701 release_openowner(sop
);
703 while (!list_empty(&clp
->cl_sessions
)) {
704 struct nfsd4_session
*ses
;
705 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
707 release_session(ses
);
709 put_nfs4_client(clp
);
712 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
714 struct nfs4_client
*clp
;
716 clp
= alloc_client(name
);
719 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
720 atomic_set(&clp
->cl_count
, 1);
721 atomic_set(&clp
->cl_cb_conn
.cb_set
, 0);
722 INIT_LIST_HEAD(&clp
->cl_idhash
);
723 INIT_LIST_HEAD(&clp
->cl_strhash
);
724 INIT_LIST_HEAD(&clp
->cl_openowners
);
725 INIT_LIST_HEAD(&clp
->cl_delegations
);
726 INIT_LIST_HEAD(&clp
->cl_sessions
);
727 INIT_LIST_HEAD(&clp
->cl_lru
);
731 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
733 memcpy(target
->cl_verifier
.data
, source
->data
,
734 sizeof(target
->cl_verifier
.data
));
737 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
739 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
740 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
743 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
745 target
->cr_uid
= source
->cr_uid
;
746 target
->cr_gid
= source
->cr_gid
;
747 target
->cr_group_info
= source
->cr_group_info
;
748 get_group_info(target
->cr_group_info
);
751 static int same_name(const char *n1
, const char *n2
)
753 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
757 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
759 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
763 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
765 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
768 /* XXX what about NGROUP */
770 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
772 return cr1
->cr_uid
== cr2
->cr_uid
;
775 static void gen_clid(struct nfs4_client
*clp
)
777 static u32 current_clientid
= 1;
779 clp
->cl_clientid
.cl_boot
= boot_time
;
780 clp
->cl_clientid
.cl_id
= current_clientid
++;
783 static void gen_confirm(struct nfs4_client
*clp
)
788 p
= (u32
*)clp
->cl_confirm
.data
;
789 *p
++ = get_seconds();
793 static int check_name(struct xdr_netobj name
)
797 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
798 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
805 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
807 unsigned int idhashval
;
809 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
810 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
811 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
812 list_add_tail(&clp
->cl_lru
, &client_lru
);
813 clp
->cl_time
= get_seconds();
817 move_to_confirmed(struct nfs4_client
*clp
)
819 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
820 unsigned int strhashval
;
822 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
823 list_del_init(&clp
->cl_strhash
);
824 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
825 strhashval
= clientstr_hashval(clp
->cl_recdir
);
826 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
830 static struct nfs4_client
*
831 find_confirmed_client(clientid_t
*clid
)
833 struct nfs4_client
*clp
;
834 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
836 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
837 if (same_clid(&clp
->cl_clientid
, clid
))
843 static struct nfs4_client
*
844 find_unconfirmed_client(clientid_t
*clid
)
846 struct nfs4_client
*clp
;
847 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
849 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
850 if (same_clid(&clp
->cl_clientid
, clid
))
857 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
858 * parameter. Matching is based on the fact the at least one of the
859 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
861 * FIXME: we need to unify the clientid namespaces for nfsv4.x
862 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
863 * and SET_CLIENTID{,_CONFIRM}
866 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
868 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
869 return use_exchange_id
== has_exchange_flags
;
872 static struct nfs4_client
*
873 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
874 bool use_exchange_id
)
876 struct nfs4_client
*clp
;
878 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
879 if (same_name(clp
->cl_recdir
, dname
) &&
880 match_clientid_establishment(clp
, use_exchange_id
))
886 static struct nfs4_client
*
887 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
888 bool use_exchange_id
)
890 struct nfs4_client
*clp
;
892 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
893 if (same_name(clp
->cl_recdir
, dname
) &&
894 match_clientid_establishment(clp
, use_exchange_id
))
901 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
903 struct nfs4_cb_conn
*cb
= &clp
->cl_cb_conn
;
905 /* Currently, we only support tcp for the callback channel */
906 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
909 cb
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
910 se
->se_callback_addr_len
,
911 (struct sockaddr
*) &cb
->cb_addr
,
912 sizeof(cb
->cb_addr
));
914 if (!cb
->cb_addrlen
|| cb
->cb_addr
.ss_family
!= AF_INET
)
917 cb
->cb_minorversion
= 0;
918 cb
->cb_prog
= se
->se_callback_prog
;
919 cb
->cb_ident
= se
->se_callback_ident
;
922 cb
->cb_addr
.ss_family
= AF_UNSPEC
;
924 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
925 "will not receive delegations\n",
926 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
932 nfsd4_set_statp(struct svc_rqst
*rqstp
, __be32
*statp
)
934 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
936 resp
->cstate
.statp
= statp
;
940 * Dereference the result pages.
943 nfsd4_release_respages(struct page
**respages
, short resused
)
947 dprintk("--> %s\n", __func__
);
948 for (i
= 0; i
< resused
; i
++) {
951 put_page(respages
[i
]);
957 nfsd4_copy_pages(struct page
**topages
, struct page
**frompages
, short count
)
961 for (i
= 0; i
< count
; i
++) {
962 topages
[i
] = frompages
[i
];
965 get_page(topages
[i
]);
970 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
971 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
972 * length of the XDR response is less than se_fmaxresp_cached
973 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
974 * of the reply (e.g. readdir).
976 * Store the base and length of the rq_req.head[0] page
977 * of the NFSv4.1 data, just past the rpc header.
980 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
982 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
983 struct svc_rqst
*rqstp
= resp
->rqstp
;
984 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
985 struct nfsd4_op
*op
= &args
->ops
[resp
->opcnt
];
986 struct kvec
*resv
= &rqstp
->rq_res
.head
[0];
988 dprintk("--> %s entry %p\n", __func__
, entry
);
990 /* Don't cache a failed OP_SEQUENCE. */
991 if (resp
->opcnt
== 1 && op
->opnum
== OP_SEQUENCE
&& resp
->cstate
.status
)
994 nfsd4_release_respages(entry
->ce_respages
, entry
->ce_resused
);
995 entry
->ce_opcnt
= resp
->opcnt
;
996 entry
->ce_status
= resp
->cstate
.status
;
999 * Don't need a page to cache just the sequence operation - the slot
1003 if (nfsd4_not_cached(resp
)) {
1004 entry
->ce_resused
= 0;
1005 entry
->ce_rpchdrlen
= 0;
1006 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__
,
1007 resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1010 entry
->ce_resused
= rqstp
->rq_resused
;
1011 if (entry
->ce_resused
> NFSD_PAGES_PER_SLOT
+ 1)
1012 entry
->ce_resused
= NFSD_PAGES_PER_SLOT
+ 1;
1013 nfsd4_copy_pages(entry
->ce_respages
, rqstp
->rq_respages
,
1015 entry
->ce_datav
.iov_base
= resp
->cstate
.statp
;
1016 entry
->ce_datav
.iov_len
= resv
->iov_len
- ((char *)resp
->cstate
.statp
-
1017 (char *)page_address(rqstp
->rq_respages
[0]));
1018 /* Current request rpc header length*/
1019 entry
->ce_rpchdrlen
= (char *)resp
->cstate
.statp
-
1020 (char *)page_address(rqstp
->rq_respages
[0]);
1024 * We keep the rpc header, but take the nfs reply from the replycache.
1027 nfsd41_copy_replay_data(struct nfsd4_compoundres
*resp
,
1028 struct nfsd4_cache_entry
*entry
)
1030 struct svc_rqst
*rqstp
= resp
->rqstp
;
1031 struct kvec
*resv
= &resp
->rqstp
->rq_res
.head
[0];
1034 /* Current request rpc header length*/
1035 len
= (char *)resp
->cstate
.statp
-
1036 (char *)page_address(rqstp
->rq_respages
[0]);
1037 if (entry
->ce_datav
.iov_len
+ len
> PAGE_SIZE
) {
1038 dprintk("%s v41 cached reply too large (%Zd).\n", __func__
,
1039 entry
->ce_datav
.iov_len
);
1042 /* copy the cached reply nfsd data past the current rpc header */
1043 memcpy((char *)resv
->iov_base
+ len
, entry
->ce_datav
.iov_base
,
1044 entry
->ce_datav
.iov_len
);
1045 resv
->iov_len
= len
+ entry
->ce_datav
.iov_len
;
1050 * Encode the replay sequence operation from the slot values.
1051 * If cachethis is FALSE encode the uncached rep error on the next
1052 * operation which sets resp->p and increments resp->opcnt for
1053 * nfs4svc_encode_compoundres.
1057 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1058 struct nfsd4_compoundres
*resp
)
1060 struct nfsd4_op
*op
;
1061 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1063 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1064 resp
->opcnt
, resp
->cstate
.slot
->sl_cache_entry
.ce_cachethis
);
1066 /* Encode the replayed sequence operation */
1067 op
= &args
->ops
[resp
->opcnt
- 1];
1068 nfsd4_encode_operation(resp
, op
);
1070 /* Return nfserr_retry_uncached_rep in next operation. */
1071 if (args
->opcnt
> 1 && slot
->sl_cache_entry
.ce_cachethis
== 0) {
1072 op
= &args
->ops
[resp
->opcnt
++];
1073 op
->status
= nfserr_retry_uncached_rep
;
1074 nfsd4_encode_operation(resp
, op
);
1080 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1081 * cached page. Replace any futher replay pages from the cache.
1084 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1085 struct nfsd4_sequence
*seq
)
1087 struct nfsd4_cache_entry
*entry
= &resp
->cstate
.slot
->sl_cache_entry
;
1090 dprintk("--> %s entry %p\n", __func__
, entry
);
1093 * If this is just the sequence operation, we did not keep
1094 * a page in the cache entry because we can just use the
1095 * slot info stored in struct nfsd4_sequence that was checked
1096 * against the slot in nfsd4_sequence().
1098 * This occurs when seq->cachethis is FALSE, or when the client
1099 * session inactivity timer fires and a solo sequence operation
1100 * is sent (lease renewal).
1102 seq
->maxslots
= resp
->cstate
.session
->se_fchannel
.maxreqs
;
1104 /* Either returns 0 or nfserr_retry_uncached */
1105 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1106 if (status
== nfserr_retry_uncached_rep
)
1109 if (!nfsd41_copy_replay_data(resp
, entry
)) {
1111 * Not enough room to use the replay rpc header, send the
1112 * cached header. Release all the allocated result pages.
1114 svc_free_res_pages(resp
->rqstp
);
1115 nfsd4_copy_pages(resp
->rqstp
->rq_respages
, entry
->ce_respages
,
1118 /* Release all but the first allocated result page */
1120 resp
->rqstp
->rq_resused
--;
1121 svc_free_res_pages(resp
->rqstp
);
1123 nfsd4_copy_pages(&resp
->rqstp
->rq_respages
[1],
1124 &entry
->ce_respages
[1],
1125 entry
->ce_resused
- 1);
1128 resp
->rqstp
->rq_resused
= entry
->ce_resused
;
1129 resp
->opcnt
= entry
->ce_opcnt
;
1130 resp
->cstate
.iovlen
= entry
->ce_datav
.iov_len
+ entry
->ce_rpchdrlen
;
1131 status
= entry
->ce_status
;
1137 * Set the exchange_id flags returned by the server.
1140 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1142 /* pNFS is not supported */
1143 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1145 /* Referrals are supported, Migration is not. */
1146 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1148 /* set the wire flags to return to client. */
1149 clid
->flags
= new->cl_exchange_flags
;
1153 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1154 struct nfsd4_compound_state
*cstate
,
1155 struct nfsd4_exchange_id
*exid
)
1157 struct nfs4_client
*unconf
, *conf
, *new;
1159 unsigned int strhashval
;
1160 char dname
[HEXDIR_LEN
];
1161 char addr_str
[INET6_ADDRSTRLEN
];
1162 nfs4_verifier verf
= exid
->verifier
;
1163 struct sockaddr
*sa
= svc_addr(rqstp
);
1165 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1166 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1167 "ip_addr=%s flags %x, spa_how %d\n",
1168 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1169 addr_str
, exid
->flags
, exid
->spa_how
);
1171 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1172 return nfserr_inval
;
1174 /* Currently only support SP4_NONE */
1175 switch (exid
->spa_how
) {
1179 return nfserr_encr_alg_unsupp
;
1181 BUG(); /* checked by xdr code */
1183 return nfserr_serverfault
; /* no excuse :-/ */
1186 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1191 strhashval
= clientstr_hashval(dname
);
1196 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1198 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1199 /* 18.35.4 case 8 */
1200 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1201 status
= nfserr_not_same
;
1204 /* Client reboot: destroy old state */
1205 expire_client(conf
);
1208 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1209 /* 18.35.4 case 9 */
1210 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1211 status
= nfserr_perm
;
1214 expire_client(conf
);
1218 * Set bit when the owner id and verifier map to an already
1219 * confirmed client id (18.35.3).
1221 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1224 * Falling into 18.35.4 case 2, possible router replay.
1225 * Leave confirmed record intact and return same result.
1227 copy_verf(conf
, &verf
);
1232 /* 18.35.4 case 7 */
1233 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1234 status
= nfserr_noent
;
1238 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1241 * Possible retry or client restart. Per 18.35.4 case 4,
1242 * a new unconfirmed record should be generated regardless
1243 * of whether any properties have changed.
1245 expire_client(unconf
);
1250 new = create_client(exid
->clname
, dname
);
1252 status
= nfserr_resource
;
1256 copy_verf(new, &verf
);
1257 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1258 rpc_copy_addr((struct sockaddr
*) &new->cl_addr
, sa
);
1261 add_to_unconfirmed(new, strhashval
);
1263 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1264 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1267 nfsd4_set_ex_flags(new, exid
);
1269 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1270 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1274 nfs4_unlock_state();
1276 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1281 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1283 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1286 /* The slot is in use, and no response has been sent. */
1288 if (seqid
== slot_seqid
)
1289 return nfserr_jukebox
;
1291 return nfserr_seq_misordered
;
1294 if (likely(seqid
== slot_seqid
+ 1))
1297 if (seqid
== slot_seqid
)
1298 return nfserr_replay_cache
;
1300 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1302 /* Misordered replay or misordered new request */
1303 return nfserr_seq_misordered
;
1307 * Cache the create session result into the create session single DRC
1308 * slot cache by saving the xdr structure. sl_seqid has been set.
1309 * Do this for solo or embedded create session operations.
1312 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1313 struct nfsd4_clid_slot
*slot
, int nfserr
)
1315 slot
->sl_status
= nfserr
;
1316 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1320 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1321 struct nfsd4_clid_slot
*slot
)
1323 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1324 return slot
->sl_status
;
1328 nfsd4_create_session(struct svc_rqst
*rqstp
,
1329 struct nfsd4_compound_state
*cstate
,
1330 struct nfsd4_create_session
*cr_ses
)
1332 struct sockaddr
*sa
= svc_addr(rqstp
);
1333 struct nfs4_client
*conf
, *unconf
;
1334 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1338 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1339 conf
= find_confirmed_client(&cr_ses
->clientid
);
1342 cs_slot
= &conf
->cl_cs_slot
;
1343 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1344 if (status
== nfserr_replay_cache
) {
1345 dprintk("Got a create_session replay! seqid= %d\n",
1347 /* Return the cached reply status */
1348 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1350 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1351 status
= nfserr_seq_misordered
;
1352 dprintk("Sequence misordered!\n");
1353 dprintk("Expected seqid= %d but got seqid= %d\n",
1354 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1357 cs_slot
->sl_seqid
++;
1358 } else if (unconf
) {
1359 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1360 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1361 status
= nfserr_clid_inuse
;
1365 cs_slot
= &unconf
->cl_cs_slot
;
1366 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1368 /* an unconfirmed replay returns misordered */
1369 status
= nfserr_seq_misordered
;
1373 cs_slot
->sl_seqid
++; /* from 0 to 1 */
1374 move_to_confirmed(unconf
);
1377 * We do not support RDMA or persistent sessions
1379 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1380 cr_ses
->flags
&= ~SESSION4_RDMA
;
1384 status
= nfserr_stale_clientid
;
1388 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1392 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1393 NFS4_MAX_SESSIONID_LEN
);
1394 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1397 /* cache solo and embedded create sessions under the state lock */
1398 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1400 nfs4_unlock_state();
1401 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1406 nfsd4_destroy_session(struct svc_rqst
*r
,
1407 struct nfsd4_compound_state
*cstate
,
1408 struct nfsd4_destroy_session
*sessionid
)
1410 struct nfsd4_session
*ses
;
1411 u32 status
= nfserr_badsession
;
1414 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1415 * - Should we return nfserr_back_chan_busy if waiting for
1416 * callbacks on to-be-destroyed session?
1417 * - Do we need to clear any callback info from previous session?
1420 dump_sessionid(__func__
, &sessionid
->sessionid
);
1421 spin_lock(&sessionid_lock
);
1422 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1424 spin_unlock(&sessionid_lock
);
1428 unhash_session(ses
);
1429 spin_unlock(&sessionid_lock
);
1431 /* wait for callbacks */
1432 shutdown_callback_client(ses
->se_client
);
1433 nfsd4_put_session(ses
);
1436 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1441 nfsd4_sequence(struct svc_rqst
*rqstp
,
1442 struct nfsd4_compound_state
*cstate
,
1443 struct nfsd4_sequence
*seq
)
1445 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1446 struct nfsd4_session
*session
;
1447 struct nfsd4_slot
*slot
;
1450 if (resp
->opcnt
!= 1)
1451 return nfserr_sequence_pos
;
1453 spin_lock(&sessionid_lock
);
1454 status
= nfserr_badsession
;
1455 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1459 status
= nfserr_badslot
;
1460 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1463 slot
= &session
->se_slots
[seq
->slotid
];
1464 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1466 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1467 if (status
== nfserr_replay_cache
) {
1468 cstate
->slot
= slot
;
1469 cstate
->session
= session
;
1470 /* Return the cached reply status and set cstate->status
1471 * for nfsd4_svc_encode_compoundres processing */
1472 status
= nfsd4_replay_cache_entry(resp
, seq
);
1473 cstate
->status
= nfserr_replay_cache
;
1479 /* Success! bump slot seqid */
1480 slot
->sl_inuse
= true;
1481 slot
->sl_seqid
= seq
->seqid
;
1482 slot
->sl_cache_entry
.ce_cachethis
= seq
->cachethis
;
1483 /* Always set the cache entry cachethis for solo sequence */
1484 if (nfsd4_is_solo_sequence(resp
))
1485 slot
->sl_cache_entry
.ce_cachethis
= 1;
1487 cstate
->slot
= slot
;
1488 cstate
->session
= session
;
1491 /* Renew the clientid on success and on replay.
1492 * Hold a session reference until done processing the compound:
1493 * nfsd4_put_session called only if the cstate slot is set.
1495 renew_client(session
->se_client
);
1496 nfsd4_get_session(session
);
1498 spin_unlock(&sessionid_lock
);
1499 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1504 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1505 struct nfsd4_setclientid
*setclid
)
1507 struct sockaddr
*sa
= svc_addr(rqstp
);
1508 struct xdr_netobj clname
= {
1509 .len
= setclid
->se_namelen
,
1510 .data
= setclid
->se_name
,
1512 nfs4_verifier clverifier
= setclid
->se_verf
;
1513 unsigned int strhashval
;
1514 struct nfs4_client
*conf
, *unconf
, *new;
1517 char dname
[HEXDIR_LEN
];
1519 if (!check_name(clname
))
1520 return nfserr_inval
;
1522 status
= nfs4_make_rec_clidname(dname
, &clname
);
1527 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1528 * We get here on a DRC miss.
1531 strhashval
= clientstr_hashval(dname
);
1534 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1536 /* RFC 3530 14.2.33 CASE 0: */
1537 status
= nfserr_clid_inuse
;
1538 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1539 char addr_str
[INET6_ADDRSTRLEN
];
1540 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1542 dprintk("NFSD: setclientid: string in use by client "
1543 "at %s\n", addr_str
);
1548 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1549 * has a description of SETCLIENTID request processing consisting
1550 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1552 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1553 status
= nfserr_resource
;
1556 * RFC 3530 14.2.33 CASE 4:
1557 * placed first, because it is the normal case
1560 expire_client(unconf
);
1561 new = create_client(clname
, dname
);
1565 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1567 * RFC 3530 14.2.33 CASE 1:
1568 * probable callback update
1571 /* Note this is removing unconfirmed {*x***},
1572 * which is stronger than RFC recommended {vxc**}.
1573 * This has the advantage that there is at most
1574 * one {*x***} in either list at any time.
1576 expire_client(unconf
);
1578 new = create_client(clname
, dname
);
1581 copy_clid(new, conf
);
1582 } else if (!unconf
) {
1584 * RFC 3530 14.2.33 CASE 2:
1585 * probable client reboot; state will be removed if
1588 new = create_client(clname
, dname
);
1594 * RFC 3530 14.2.33 CASE 3:
1595 * probable client reboot; state will be removed if
1598 expire_client(unconf
);
1599 new = create_client(clname
, dname
);
1604 copy_verf(new, &clverifier
);
1605 rpc_copy_addr((struct sockaddr
*) &new->cl_addr
, sa
);
1606 new->cl_flavor
= rqstp
->rq_flavor
;
1607 princ
= svc_gss_principal(rqstp
);
1609 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1610 if (new->cl_principal
== NULL
) {
1615 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
1617 gen_callback(new, setclid
);
1618 add_to_unconfirmed(new, strhashval
);
1619 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1620 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1621 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1624 nfs4_unlock_state();
1630 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1631 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1632 * bullets, labeled as CASE1 - CASE4 below.
1635 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1636 struct nfsd4_compound_state
*cstate
,
1637 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1639 struct sockaddr
*sa
= svc_addr(rqstp
);
1640 struct nfs4_client
*conf
, *unconf
;
1641 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1642 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1645 if (STALE_CLIENTID(clid
))
1646 return nfserr_stale_clientid
;
1648 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1649 * We get here on a DRC miss.
1654 conf
= find_confirmed_client(clid
);
1655 unconf
= find_unconfirmed_client(clid
);
1657 status
= nfserr_clid_inuse
;
1658 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
1660 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
1664 * section 14.2.34 of RFC 3530 has a description of
1665 * SETCLIENTID_CONFIRM request processing consisting
1666 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1668 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1670 * RFC 3530 14.2.34 CASE 1:
1673 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1674 status
= nfserr_clid_inuse
;
1676 /* XXX: We just turn off callbacks until we can handle
1677 * change request correctly. */
1678 atomic_set(&conf
->cl_cb_conn
.cb_set
, 0);
1679 expire_client(unconf
);
1683 } else if (conf
&& !unconf
) {
1685 * RFC 3530 14.2.34 CASE 2:
1686 * probable retransmitted request; play it safe and
1689 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1690 status
= nfserr_clid_inuse
;
1693 } else if (!conf
&& unconf
1694 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1696 * RFC 3530 14.2.34 CASE 3:
1697 * Normal case; new or rebooted client:
1699 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1700 status
= nfserr_clid_inuse
;
1703 clientstr_hashval(unconf
->cl_recdir
);
1704 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1707 nfsd4_remove_clid_dir(conf
);
1708 expire_client(conf
);
1710 move_to_confirmed(unconf
);
1712 nfsd4_probe_callback(conf
);
1715 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1716 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1719 * RFC 3530 14.2.34 CASE 4:
1720 * Client probably hasn't noticed that we rebooted yet.
1722 status
= nfserr_stale_clientid
;
1724 /* check that we have hit one of the cases...*/
1725 status
= nfserr_clid_inuse
;
1728 nfs4_unlock_state();
1732 /* OPEN Share state helper functions */
1733 static inline struct nfs4_file
*
1734 alloc_init_file(struct inode
*ino
)
1736 struct nfs4_file
*fp
;
1737 unsigned int hashval
= file_hashval(ino
);
1739 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1741 atomic_set(&fp
->fi_ref
, 1);
1742 INIT_LIST_HEAD(&fp
->fi_hash
);
1743 INIT_LIST_HEAD(&fp
->fi_stateids
);
1744 INIT_LIST_HEAD(&fp
->fi_delegations
);
1745 spin_lock(&recall_lock
);
1746 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1747 spin_unlock(&recall_lock
);
1748 fp
->fi_inode
= igrab(ino
);
1749 fp
->fi_id
= current_fileid
++;
1750 fp
->fi_had_conflict
= false;
1757 nfsd4_free_slab(struct kmem_cache
**slab
)
1761 kmem_cache_destroy(*slab
);
1766 nfsd4_free_slabs(void)
1768 nfsd4_free_slab(&stateowner_slab
);
1769 nfsd4_free_slab(&file_slab
);
1770 nfsd4_free_slab(&stateid_slab
);
1771 nfsd4_free_slab(&deleg_slab
);
1775 nfsd4_init_slabs(void)
1777 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1778 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1779 if (stateowner_slab
== NULL
)
1781 file_slab
= kmem_cache_create("nfsd4_files",
1782 sizeof(struct nfs4_file
), 0, 0, NULL
);
1783 if (file_slab
== NULL
)
1785 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1786 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1787 if (stateid_slab
== NULL
)
1789 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1790 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1791 if (deleg_slab
== NULL
)
1796 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1801 nfs4_free_stateowner(struct kref
*kref
)
1803 struct nfs4_stateowner
*sop
=
1804 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1805 kfree(sop
->so_owner
.data
);
1806 kmem_cache_free(stateowner_slab
, sop
);
1809 static inline struct nfs4_stateowner
*
1810 alloc_stateowner(struct xdr_netobj
*owner
)
1812 struct nfs4_stateowner
*sop
;
1814 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1815 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1816 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1817 sop
->so_owner
.len
= owner
->len
;
1818 kref_init(&sop
->so_ref
);
1821 kmem_cache_free(stateowner_slab
, sop
);
1826 static struct nfs4_stateowner
*
1827 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1828 struct nfs4_stateowner
*sop
;
1829 struct nfs4_replay
*rp
;
1830 unsigned int idhashval
;
1832 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1834 idhashval
= ownerid_hashval(current_ownerid
);
1835 INIT_LIST_HEAD(&sop
->so_idhash
);
1836 INIT_LIST_HEAD(&sop
->so_strhash
);
1837 INIT_LIST_HEAD(&sop
->so_perclient
);
1838 INIT_LIST_HEAD(&sop
->so_stateids
);
1839 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1840 INIT_LIST_HEAD(&sop
->so_close_lru
);
1842 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1843 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1844 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1845 sop
->so_is_open_owner
= 1;
1846 sop
->so_id
= current_ownerid
++;
1847 sop
->so_client
= clp
;
1848 sop
->so_seqid
= open
->op_seqid
;
1849 sop
->so_confirmed
= 0;
1850 rp
= &sop
->so_replay
;
1851 rp
->rp_status
= nfserr_serverfault
;
1853 rp
->rp_buf
= rp
->rp_ibuf
;
1858 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1859 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1860 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1862 INIT_LIST_HEAD(&stp
->st_hash
);
1863 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1864 INIT_LIST_HEAD(&stp
->st_lockowners
);
1865 INIT_LIST_HEAD(&stp
->st_perfile
);
1866 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1867 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1868 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1869 stp
->st_stateowner
= sop
;
1872 stp
->st_stateid
.si_boot
= get_seconds();
1873 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1874 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1875 stp
->st_stateid
.si_generation
= 0;
1876 stp
->st_access_bmap
= 0;
1877 stp
->st_deny_bmap
= 0;
1878 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
1879 &stp
->st_access_bmap
);
1880 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1881 stp
->st_openstp
= NULL
;
1885 move_to_close_lru(struct nfs4_stateowner
*sop
)
1887 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1889 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1890 sop
->so_time
= get_seconds();
1894 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1897 return (sop
->so_owner
.len
== owner
->len
) &&
1898 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1899 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1902 static struct nfs4_stateowner
*
1903 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1905 struct nfs4_stateowner
*so
= NULL
;
1907 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1908 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1914 /* search file_hashtbl[] for file */
1915 static struct nfs4_file
*
1916 find_file(struct inode
*ino
)
1918 unsigned int hashval
= file_hashval(ino
);
1919 struct nfs4_file
*fp
;
1921 spin_lock(&recall_lock
);
1922 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1923 if (fp
->fi_inode
== ino
) {
1925 spin_unlock(&recall_lock
);
1929 spin_unlock(&recall_lock
);
1933 static inline int access_valid(u32 x
, u32 minorversion
)
1935 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
1937 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
1939 x
&= ~NFS4_SHARE_ACCESS_MASK
;
1940 if (minorversion
&& x
) {
1941 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
1943 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
1945 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
1952 static inline int deny_valid(u32 x
)
1954 /* Note: unlike access bits, deny bits may be zero. */
1955 return x
<= NFS4_SHARE_DENY_BOTH
;
1959 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1960 * st_{access,deny}_bmap field of the stateid, in order to track not
1961 * only what share bits are currently in force, but also what
1962 * combinations of share bits previous opens have used. This allows us
1963 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1964 * return an error if the client attempt to downgrade to a combination
1965 * of share bits not explicable by closing some of its previous opens.
1967 * XXX: This enforcement is actually incomplete, since we don't keep
1968 * track of access/deny bit combinations; so, e.g., we allow:
1970 * OPEN allow read, deny write
1971 * OPEN allow both, deny none
1972 * DOWNGRADE allow read, deny none
1974 * which we should reject.
1977 set_access(unsigned int *access
, unsigned long bmap
) {
1981 for (i
= 1; i
< 4; i
++) {
1982 if (test_bit(i
, &bmap
))
1988 set_deny(unsigned int *deny
, unsigned long bmap
) {
1992 for (i
= 0; i
< 4; i
++) {
1993 if (test_bit(i
, &bmap
))
1999 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
2000 unsigned int access
, deny
;
2002 set_access(&access
, stp
->st_access_bmap
);
2003 set_deny(&deny
, stp
->st_deny_bmap
);
2004 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
2010 * Called to check deny when READ with all zero stateid or
2011 * WRITE with all zero or all one stateid
2014 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2016 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2017 struct nfs4_file
*fp
;
2018 struct nfs4_stateid
*stp
;
2021 dprintk("NFSD: nfs4_share_conflict\n");
2023 fp
= find_file(ino
);
2026 ret
= nfserr_locked
;
2027 /* Search for conflicting share reservations */
2028 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2029 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2030 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2040 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
2042 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
2043 drop_file_write_access(filp
);
2044 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
2049 * Spawn a thread to perform a recall on the delegation represented
2050 * by the lease (file_lock)
2052 * Called from break_lease() with lock_kernel() held.
2053 * Note: we assume break_lease will only call this *once* for any given
2057 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2059 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2061 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2065 /* We're assuming the state code never drops its reference
2066 * without first removing the lease. Since we're in this lease
2067 * callback (and since the lease code is serialized by the kernel
2068 * lock) we know the server hasn't removed the lease yet, we know
2069 * it's safe to take a reference: */
2070 atomic_inc(&dp
->dl_count
);
2071 atomic_inc(&dp
->dl_client
->cl_count
);
2073 spin_lock(&recall_lock
);
2074 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2075 spin_unlock(&recall_lock
);
2077 /* only place dl_time is set. protected by lock_kernel*/
2078 dp
->dl_time
= get_seconds();
2081 * We don't want the locks code to timeout the lease for us;
2082 * we'll remove it ourself if the delegation isn't returned
2085 fl
->fl_break_time
= 0;
2087 dp
->dl_file
->fi_had_conflict
= true;
2088 nfsd4_cb_recall(dp
);
2092 * The file_lock is being reapd.
2094 * Called by locks_free_lock() with lock_kernel() held.
2097 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2099 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2101 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2103 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2105 dp
->dl_flock
= NULL
;
2109 * Set the delegation file_lock back pointer.
2111 * Called from setlease() with lock_kernel() held.
2114 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2116 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2118 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2125 * Called from setlease() with lock_kernel() held
2128 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2130 struct nfs4_delegation
*onlistd
=
2131 (struct nfs4_delegation
*)onlist
->fl_owner
;
2132 struct nfs4_delegation
*tryd
=
2133 (struct nfs4_delegation
*)try->fl_owner
;
2135 if (onlist
->fl_lmops
!= try->fl_lmops
)
2138 return onlistd
->dl_client
== tryd
->dl_client
;
2143 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2146 return lease_modify(onlist
, arg
);
2151 static struct lock_manager_operations nfsd_lease_mng_ops
= {
2152 .fl_break
= nfsd_break_deleg_cb
,
2153 .fl_release_private
= nfsd_release_deleg_cb
,
2154 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2155 .fl_mylease
= nfsd_same_client_deleg_cb
,
2156 .fl_change
= nfsd_change_deleg_cb
,
2161 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2162 struct nfsd4_open
*open
)
2164 clientid_t
*clientid
= &open
->op_clientid
;
2165 struct nfs4_client
*clp
= NULL
;
2166 unsigned int strhashval
;
2167 struct nfs4_stateowner
*sop
= NULL
;
2169 if (!check_name(open
->op_owner
))
2170 return nfserr_inval
;
2172 if (STALE_CLIENTID(&open
->op_clientid
))
2173 return nfserr_stale_clientid
;
2175 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2176 sop
= find_openstateowner_str(strhashval
, open
);
2177 open
->op_stateowner
= sop
;
2179 /* Make sure the client's lease hasn't expired. */
2180 clp
= find_confirmed_client(clientid
);
2182 return nfserr_expired
;
2185 /* When sessions are used, skip open sequenceid processing */
2186 if (nfsd4_has_session(cstate
))
2188 if (!sop
->so_confirmed
) {
2189 /* Replace unconfirmed owners without checking for replay. */
2190 clp
= sop
->so_client
;
2191 release_openowner(sop
);
2192 open
->op_stateowner
= NULL
;
2195 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2196 if (sop
->so_replay
.rp_buflen
)
2197 return nfserr_replay_me
;
2198 /* The original OPEN failed so spectacularly
2199 * that we don't even have replay data saved!
2200 * Therefore, we have no choice but to continue
2201 * processing this OPEN; presumably, we'll
2202 * fail again for the same reason.
2204 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2207 if (open
->op_seqid
!= sop
->so_seqid
)
2208 return nfserr_bad_seqid
;
2210 if (open
->op_stateowner
== NULL
) {
2211 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2213 return nfserr_resource
;
2214 open
->op_stateowner
= sop
;
2216 list_del_init(&sop
->so_close_lru
);
2217 renew_client(sop
->so_client
);
2221 static inline __be32
2222 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2224 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2225 return nfserr_openmode
;
2230 static struct nfs4_delegation
*
2231 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2233 struct nfs4_delegation
*dp
;
2235 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2236 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2243 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2244 struct nfs4_delegation
**dp
)
2247 __be32 status
= nfserr_bad_stateid
;
2249 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2252 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
2253 RD_STATE
: WR_STATE
;
2254 status
= nfs4_check_delegmode(*dp
, flags
);
2258 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2262 open
->op_stateowner
->so_confirmed
= 1;
2267 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2269 struct nfs4_stateid
*local
;
2270 __be32 status
= nfserr_share_denied
;
2271 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2273 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2274 /* ignore lock owners */
2275 if (local
->st_stateowner
->so_is_open_owner
== 0)
2277 /* remember if we have seen this open owner */
2278 if (local
->st_stateowner
== sop
)
2280 /* check for conflicting share reservations */
2281 if (!test_share(local
, open
))
2289 static inline struct nfs4_stateid
*
2290 nfs4_alloc_stateid(void)
2292 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2296 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2297 struct nfs4_delegation
*dp
,
2298 struct svc_fh
*cur_fh
, int flags
)
2300 struct nfs4_stateid
*stp
;
2302 stp
= nfs4_alloc_stateid();
2304 return nfserr_resource
;
2307 get_file(dp
->dl_vfs_file
);
2308 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2311 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2314 if (status
== nfserr_dropit
)
2315 status
= nfserr_jukebox
;
2316 kmem_cache_free(stateid_slab
, stp
);
2324 static inline __be32
2325 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2326 struct nfsd4_open
*open
)
2328 struct iattr iattr
= {
2329 .ia_valid
= ATTR_SIZE
,
2332 if (!open
->op_truncate
)
2334 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2335 return nfserr_inval
;
2336 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2340 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2342 struct file
*filp
= stp
->st_vfs_file
;
2343 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2344 unsigned int share_access
, new_writer
;
2347 set_access(&share_access
, stp
->st_access_bmap
);
2348 new_writer
= (~share_access
) & open
->op_share_access
2349 & NFS4_SHARE_ACCESS_WRITE
;
2352 int err
= get_write_access(inode
);
2354 return nfserrno(err
);
2355 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2357 return nfserrno(err
);
2358 file_take_write(filp
);
2360 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2363 put_write_access(inode
);
2366 /* remember the open */
2367 filp
->f_mode
|= open
->op_share_access
;
2368 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2369 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2376 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2378 open
->op_stateowner
->so_confirmed
= 1;
2379 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2383 * Attempt to hand out a delegation.
2386 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2388 struct nfs4_delegation
*dp
;
2389 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2390 struct nfs4_cb_conn
*cb
= &sop
->so_client
->cl_cb_conn
;
2391 struct file_lock fl
, *flp
= &fl
;
2392 int status
, flag
= 0;
2394 flag
= NFS4_OPEN_DELEGATE_NONE
;
2395 open
->op_recall
= 0;
2396 switch (open
->op_claim_type
) {
2397 case NFS4_OPEN_CLAIM_PREVIOUS
:
2398 if (!atomic_read(&cb
->cb_set
))
2399 open
->op_recall
= 1;
2400 flag
= open
->op_delegate_type
;
2401 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2404 case NFS4_OPEN_CLAIM_NULL
:
2405 /* Let's not give out any delegations till everyone's
2406 * had the chance to reclaim theirs.... */
2407 if (locks_in_grace())
2409 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
2411 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2412 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2414 flag
= NFS4_OPEN_DELEGATE_READ
;
2420 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2422 flag
= NFS4_OPEN_DELEGATE_NONE
;
2425 locks_init_lock(&fl
);
2426 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2427 fl
.fl_flags
= FL_LEASE
;
2428 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2429 fl
.fl_end
= OFFSET_MAX
;
2430 fl
.fl_owner
= (fl_owner_t
)dp
;
2431 fl
.fl_file
= stp
->st_vfs_file
;
2432 fl
.fl_pid
= current
->tgid
;
2434 /* vfs_setlease checks to see if delegation should be handed out.
2435 * the lock_manager callbacks fl_mylease and fl_change are used
2437 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2438 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2439 unhash_delegation(dp
);
2440 flag
= NFS4_OPEN_DELEGATE_NONE
;
2444 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2446 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2447 dp
->dl_stateid
.si_boot
,
2448 dp
->dl_stateid
.si_stateownerid
,
2449 dp
->dl_stateid
.si_fileid
,
2450 dp
->dl_stateid
.si_generation
);
2452 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2453 && flag
== NFS4_OPEN_DELEGATE_NONE
2454 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2455 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2456 open
->op_delegate_type
= flag
;
2460 * called with nfs4_lock_state() held.
2463 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2465 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2466 struct nfs4_file
*fp
= NULL
;
2467 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2468 struct nfs4_stateid
*stp
= NULL
;
2469 struct nfs4_delegation
*dp
= NULL
;
2472 status
= nfserr_inval
;
2473 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2474 || !deny_valid(open
->op_share_deny
))
2477 * Lookup file; if found, lookup stateid and check open request,
2478 * and check for delegations in the process of being recalled.
2479 * If not found, create the nfs4_file struct
2481 fp
= find_file(ino
);
2483 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2485 status
= nfs4_check_deleg(fp
, open
, &dp
);
2489 status
= nfserr_bad_stateid
;
2490 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2492 status
= nfserr_resource
;
2493 fp
= alloc_init_file(ino
);
2499 * OPEN the file, or upgrade an existing OPEN.
2500 * If truncate fails, the OPEN fails.
2503 /* Stateid was found, this is an OPEN upgrade */
2504 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2507 update_stateid(&stp
->st_stateid
);
2509 /* Stateid was not found, this is a new OPEN */
2511 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2512 flags
|= NFSD_MAY_READ
;
2513 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2514 flags
|= NFSD_MAY_WRITE
;
2515 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2518 init_stateid(stp
, fp
, open
);
2519 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2521 release_open_stateid(stp
);
2524 if (nfsd4_has_session(&resp
->cstate
))
2525 update_stateid(&stp
->st_stateid
);
2527 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2529 if (nfsd4_has_session(&resp
->cstate
))
2530 open
->op_stateowner
->so_confirmed
= 1;
2533 * Attempt to hand out a delegation. No error return, because the
2534 * OPEN succeeds even if we fail.
2536 nfs4_open_delegation(current_fh
, open
, stp
);
2540 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2541 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
2542 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
2546 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2547 nfs4_set_claim_prev(open
);
2549 * To finish the open response, we just need to set the rflags.
2551 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2552 if (!open
->op_stateowner
->so_confirmed
&&
2553 !nfsd4_has_session(&resp
->cstate
))
2554 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2560 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2563 struct nfs4_client
*clp
;
2567 dprintk("process_renew(%08x/%08x): starting\n",
2568 clid
->cl_boot
, clid
->cl_id
);
2569 status
= nfserr_stale_clientid
;
2570 if (STALE_CLIENTID(clid
))
2572 clp
= find_confirmed_client(clid
);
2573 status
= nfserr_expired
;
2575 /* We assume the client took too long to RENEW. */
2576 dprintk("nfsd4_renew: clientid not found!\n");
2580 status
= nfserr_cb_path_down
;
2581 if (!list_empty(&clp
->cl_delegations
)
2582 && !atomic_read(&clp
->cl_cb_conn
.cb_set
))
2586 nfs4_unlock_state();
2590 struct lock_manager nfsd4_manager
= {
2594 nfsd4_end_grace(void)
2596 dprintk("NFSD: end of grace period\n");
2597 nfsd4_recdir_purge_old();
2598 locks_end_grace(&nfsd4_manager
);
2602 nfs4_laundromat(void)
2604 struct nfs4_client
*clp
;
2605 struct nfs4_stateowner
*sop
;
2606 struct nfs4_delegation
*dp
;
2607 struct list_head
*pos
, *next
, reaplist
;
2608 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
2609 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
2610 time_t u
, test_val
= NFSD_LEASE_TIME
;
2614 dprintk("NFSD: laundromat service - starting\n");
2615 if (locks_in_grace())
2617 list_for_each_safe(pos
, next
, &client_lru
) {
2618 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2619 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2620 t
= clp
->cl_time
- cutoff
;
2621 if (clientid_val
> t
)
2625 dprintk("NFSD: purging unused client (clientid %08x)\n",
2626 clp
->cl_clientid
.cl_id
);
2627 nfsd4_remove_clid_dir(clp
);
2630 INIT_LIST_HEAD(&reaplist
);
2631 spin_lock(&recall_lock
);
2632 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2633 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2634 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2635 u
= dp
->dl_time
- cutoff
;
2640 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2642 list_move(&dp
->dl_recall_lru
, &reaplist
);
2644 spin_unlock(&recall_lock
);
2645 list_for_each_safe(pos
, next
, &reaplist
) {
2646 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2647 list_del_init(&dp
->dl_recall_lru
);
2648 unhash_delegation(dp
);
2650 test_val
= NFSD_LEASE_TIME
;
2651 list_for_each_safe(pos
, next
, &close_lru
) {
2652 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2653 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2654 u
= sop
->so_time
- cutoff
;
2659 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2661 release_openowner(sop
);
2663 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2664 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2665 nfs4_unlock_state();
2666 return clientid_val
;
2669 static struct workqueue_struct
*laundry_wq
;
2670 static void laundromat_main(struct work_struct
*);
2671 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2674 laundromat_main(struct work_struct
*not_used
)
2678 t
= nfs4_laundromat();
2679 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2680 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2683 static struct nfs4_stateowner
*
2684 search_close_lru(u32 st_id
, int flags
)
2686 struct nfs4_stateowner
*local
= NULL
;
2688 if (flags
& CLOSE_STATE
) {
2689 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2690 if (local
->so_id
== st_id
)
2698 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2700 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2704 STALE_STATEID(stateid_t
*stateid
)
2706 if (time_after((unsigned long)boot_time
,
2707 (unsigned long)stateid
->si_boot
)) {
2708 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2709 stateid
->si_boot
, stateid
->si_stateownerid
,
2710 stateid
->si_fileid
, stateid
->si_generation
);
2717 EXPIRED_STATEID(stateid_t
*stateid
)
2719 if (time_before((unsigned long)boot_time
,
2720 ((unsigned long)stateid
->si_boot
)) &&
2721 time_before((unsigned long)(stateid
->si_boot
+ lease_time
), get_seconds())) {
2722 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2723 stateid
->si_boot
, stateid
->si_stateownerid
,
2724 stateid
->si_fileid
, stateid
->si_generation
);
2731 stateid_error_map(stateid_t
*stateid
)
2733 if (STALE_STATEID(stateid
))
2734 return nfserr_stale_stateid
;
2735 if (EXPIRED_STATEID(stateid
))
2736 return nfserr_expired
;
2738 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2739 stateid
->si_boot
, stateid
->si_stateownerid
,
2740 stateid
->si_fileid
, stateid
->si_generation
);
2741 return nfserr_bad_stateid
;
2745 access_permit_read(unsigned long access_bmap
)
2747 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2748 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2749 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2753 access_permit_write(unsigned long access_bmap
)
2755 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2756 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2760 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2762 __be32 status
= nfserr_openmode
;
2764 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2766 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2773 static inline __be32
2774 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2776 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2778 else if (locks_in_grace()) {
2779 /* Answer in remaining cases depends on existance of
2780 * conflicting state; so we must wait out the grace period. */
2781 return nfserr_grace
;
2782 } else if (flags
& WR_STATE
)
2783 return nfs4_share_conflict(current_fh
,
2784 NFS4_SHARE_DENY_WRITE
);
2785 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2786 return nfs4_share_conflict(current_fh
,
2787 NFS4_SHARE_DENY_READ
);
2791 * Allow READ/WRITE during grace period on recovered state only for files
2792 * that are not able to provide mandatory locking.
2795 grace_disallows_io(struct inode
*inode
)
2797 return locks_in_grace() && mandatory_lock(inode
);
2800 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
2803 * When sessions are used the stateid generation number is ignored
2806 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
2809 /* If the client sends us a stateid from the future, it's buggy: */
2810 if (in
->si_generation
> ref
->si_generation
)
2811 return nfserr_bad_stateid
;
2813 * The following, however, can happen. For example, if the
2814 * client sends an open and some IO at the same time, the open
2815 * may bump si_generation while the IO is still in flight.
2816 * Thanks to hard links and renames, the client never knows what
2817 * file an open will affect. So it could avoid that situation
2818 * only by serializing all opens and IO from the same open
2819 * owner. To recover from the old_stateid error, the client
2820 * will just have to retry the IO:
2822 if (in
->si_generation
< ref
->si_generation
)
2823 return nfserr_old_stateid
;
2828 static int is_delegation_stateid(stateid_t
*stateid
)
2830 return stateid
->si_fileid
== 0;
2834 * Checks for stateid operations
2837 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
2838 stateid_t
*stateid
, int flags
, struct file
**filpp
)
2840 struct nfs4_stateid
*stp
= NULL
;
2841 struct nfs4_delegation
*dp
= NULL
;
2842 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2843 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2849 if (grace_disallows_io(ino
))
2850 return nfserr_grace
;
2852 if (nfsd4_has_session(cstate
))
2853 flags
|= HAS_SESSION
;
2855 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2856 return check_special_stateids(current_fh
, stateid
, flags
);
2858 status
= nfserr_stale_stateid
;
2859 if (STALE_STATEID(stateid
))
2862 status
= nfserr_bad_stateid
;
2863 if (is_delegation_stateid(stateid
)) {
2864 dp
= find_delegation_stateid(ino
, stateid
);
2866 status
= stateid_error_map(stateid
);
2869 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
2873 status
= nfs4_check_delegmode(dp
, flags
);
2876 renew_client(dp
->dl_client
);
2878 *filpp
= dp
->dl_vfs_file
;
2879 } else { /* open or lock stateid */
2880 stp
= find_stateid(stateid
, flags
);
2882 status
= stateid_error_map(stateid
);
2885 if (nfs4_check_fh(current_fh
, stp
))
2887 if (!stp
->st_stateowner
->so_confirmed
)
2889 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
2893 status
= nfs4_check_openmode(stp
, flags
);
2896 renew_client(stp
->st_stateowner
->so_client
);
2898 *filpp
= stp
->st_vfs_file
;
2908 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2909 RD_STATE
: WR_STATE
;
2913 * Checks for sequence id mutating operations.
2916 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
2917 stateid_t
*stateid
, int flags
,
2918 struct nfs4_stateowner
**sopp
,
2919 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2921 struct nfs4_stateid
*stp
;
2922 struct nfs4_stateowner
*sop
;
2923 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2926 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2927 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2928 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2929 stateid
->si_generation
);
2934 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2935 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2936 return nfserr_bad_stateid
;
2939 if (STALE_STATEID(stateid
))
2940 return nfserr_stale_stateid
;
2942 if (nfsd4_has_session(cstate
))
2943 flags
|= HAS_SESSION
;
2946 * We return BAD_STATEID if filehandle doesn't match stateid,
2947 * the confirmed flag is incorrecly set, or the generation
2948 * number is incorrect.
2950 stp
= find_stateid(stateid
, flags
);
2953 * Also, we should make sure this isn't just the result of
2956 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2958 return stateid_error_map(stateid
);
2964 *sopp
= sop
= stp
->st_stateowner
;
2967 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
2968 struct nfs4_client
*clp
= sop
->so_client
;
2972 lkflg
= setlkflg(lock
->lk_type
);
2974 if (lock
->lk_is_new
) {
2975 if (!sop
->so_is_open_owner
)
2976 return nfserr_bad_stateid
;
2977 if (!(flags
& HAS_SESSION
) &&
2978 !same_clid(&clp
->cl_clientid
, lockclid
))
2979 return nfserr_bad_stateid
;
2980 /* stp is the open stateid */
2981 status
= nfs4_check_openmode(stp
, lkflg
);
2985 /* stp is the lock stateid */
2986 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
2992 if (nfs4_check_fh(current_fh
, stp
)) {
2993 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2994 return nfserr_bad_stateid
;
2998 * We now validate the seqid and stateid generation numbers.
2999 * For the moment, we ignore the possibility of
3000 * generation number wraparound.
3002 if (!(flags
& HAS_SESSION
) && seqid
!= sop
->so_seqid
)
3005 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3006 dprintk("NFSD: preprocess_seqid_op: expected"
3007 " unconfirmed stateowner!\n");
3008 return nfserr_bad_stateid
;
3010 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3011 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3012 " confirmed yet!\n");
3013 return nfserr_bad_stateid
;
3015 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, flags
);
3018 renew_client(sop
->so_client
);
3022 if (seqid
== sop
->so_seqid
- 1) {
3023 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3024 /* indicate replay to calling function */
3025 return nfserr_replay_me
;
3027 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3028 sop
->so_seqid
, seqid
);
3030 return nfserr_bad_seqid
;
3034 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3035 struct nfsd4_open_confirm
*oc
)
3038 struct nfs4_stateowner
*sop
;
3039 struct nfs4_stateid
*stp
;
3041 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3042 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3043 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3045 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3051 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3052 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3053 CONFIRM
| OPEN_STATE
,
3054 &oc
->oc_stateowner
, &stp
, NULL
)))
3057 sop
= oc
->oc_stateowner
;
3058 sop
->so_confirmed
= 1;
3059 update_stateid(&stp
->st_stateid
);
3060 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3061 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3062 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
3063 stp
->st_stateid
.si_boot
,
3064 stp
->st_stateid
.si_stateownerid
,
3065 stp
->st_stateid
.si_fileid
,
3066 stp
->st_stateid
.si_generation
);
3068 nfsd4_create_clid_dir(sop
->so_client
);
3070 if (oc
->oc_stateowner
) {
3071 nfs4_get_stateowner(oc
->oc_stateowner
);
3072 cstate
->replay_owner
= oc
->oc_stateowner
;
3074 nfs4_unlock_state();
3080 * unset all bits in union bitmap (bmap) that
3081 * do not exist in share (from successful OPEN_DOWNGRADE)
3084 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3087 for (i
= 1; i
< 4; i
++) {
3088 if ((i
& access
) != i
)
3089 __clear_bit(i
, bmap
);
3094 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3097 for (i
= 0; i
< 4; i
++) {
3098 if ((i
& deny
) != i
)
3099 __clear_bit(i
, bmap
);
3104 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3105 struct nfsd4_compound_state
*cstate
,
3106 struct nfsd4_open_downgrade
*od
)
3109 struct nfs4_stateid
*stp
;
3110 unsigned int share_access
;
3112 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3113 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3114 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3116 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3117 || !deny_valid(od
->od_share_deny
))
3118 return nfserr_inval
;
3121 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3125 &od
->od_stateowner
, &stp
, NULL
)))
3128 status
= nfserr_inval
;
3129 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3130 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3131 stp
->st_access_bmap
, od
->od_share_access
);
3134 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3135 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3136 stp
->st_deny_bmap
, od
->od_share_deny
);
3139 set_access(&share_access
, stp
->st_access_bmap
);
3140 nfs4_file_downgrade(stp
->st_vfs_file
,
3141 share_access
& ~od
->od_share_access
);
3143 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3144 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3146 update_stateid(&stp
->st_stateid
);
3147 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3150 if (od
->od_stateowner
) {
3151 nfs4_get_stateowner(od
->od_stateowner
);
3152 cstate
->replay_owner
= od
->od_stateowner
;
3154 nfs4_unlock_state();
3159 * nfs4_unlock_state() called after encode
3162 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3163 struct nfsd4_close
*close
)
3166 struct nfs4_stateid
*stp
;
3168 dprintk("NFSD: nfsd4_close on file %.*s\n",
3169 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3170 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3173 /* check close_lru for replay */
3174 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3177 OPEN_STATE
| CLOSE_STATE
,
3178 &close
->cl_stateowner
, &stp
, NULL
)))
3181 update_stateid(&stp
->st_stateid
);
3182 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3184 /* release_stateid() calls nfsd_close() if needed */
3185 release_open_stateid(stp
);
3187 /* place unused nfs4_stateowners on so_close_lru list to be
3188 * released by the laundromat service after the lease period
3189 * to enable us to handle CLOSE replay
3191 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3192 move_to_close_lru(close
->cl_stateowner
);
3194 if (close
->cl_stateowner
) {
3195 nfs4_get_stateowner(close
->cl_stateowner
);
3196 cstate
->replay_owner
= close
->cl_stateowner
;
3198 nfs4_unlock_state();
3203 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3204 struct nfsd4_delegreturn
*dr
)
3206 struct nfs4_delegation
*dp
;
3207 stateid_t
*stateid
= &dr
->dr_stateid
;
3208 struct inode
*inode
;
3212 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3214 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3216 if (nfsd4_has_session(cstate
))
3217 flags
|= HAS_SESSION
;
3219 status
= nfserr_bad_stateid
;
3220 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3222 status
= nfserr_stale_stateid
;
3223 if (STALE_STATEID(stateid
))
3225 status
= nfserr_bad_stateid
;
3226 if (!is_delegation_stateid(stateid
))
3228 dp
= find_delegation_stateid(inode
, stateid
);
3230 status
= stateid_error_map(stateid
);
3233 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3236 renew_client(dp
->dl_client
);
3238 unhash_delegation(dp
);
3240 nfs4_unlock_state();
3247 * Lock owner state (byte-range locks)
3249 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3250 #define LOCK_HASH_BITS 8
3251 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3252 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3255 end_offset(u64 start
, u64 len
)
3260 return end
>= start
? end
: NFS4_MAX_UINT64
;
3263 /* last octet in a range */
3265 last_byte_offset(u64 start
, u64 len
)
3271 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3274 #define lockownerid_hashval(id) \
3275 ((id) & LOCK_HASH_MASK)
3277 static inline unsigned int
3278 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3279 struct xdr_netobj
*ownername
)
3281 return (file_hashval(inode
) + cl_id
3282 + opaque_hashval(ownername
->data
, ownername
->len
))
3286 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3287 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3288 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3290 static struct nfs4_stateid
*
3291 find_stateid(stateid_t
*stid
, int flags
)
3293 struct nfs4_stateid
*local
;
3294 u32 st_id
= stid
->si_stateownerid
;
3295 u32 f_id
= stid
->si_fileid
;
3296 unsigned int hashval
;
3298 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3299 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3300 hashval
= stateid_hashval(st_id
, f_id
);
3301 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3302 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3303 (local
->st_stateid
.si_fileid
== f_id
))
3308 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3309 hashval
= stateid_hashval(st_id
, f_id
);
3310 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3311 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3312 (local
->st_stateid
.si_fileid
== f_id
))
3319 static struct nfs4_delegation
*
3320 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3322 struct nfs4_file
*fp
;
3323 struct nfs4_delegation
*dl
;
3325 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3326 stid
->si_boot
, stid
->si_stateownerid
,
3327 stid
->si_fileid
, stid
->si_generation
);
3329 fp
= find_file(ino
);
3332 dl
= find_delegation_file(fp
, stid
);
3338 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3339 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3340 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3341 * locking, this prevents us from being completely protocol-compliant. The
3342 * real solution to this problem is to start using unsigned file offsets in
3343 * the VFS, but this is a very deep change!
3346 nfs4_transform_lock_offset(struct file_lock
*lock
)
3348 if (lock
->fl_start
< 0)
3349 lock
->fl_start
= OFFSET_MAX
;
3350 if (lock
->fl_end
< 0)
3351 lock
->fl_end
= OFFSET_MAX
;
3354 /* Hack!: For now, we're defining this just so we can use a pointer to it
3355 * as a unique cookie to identify our (NFSv4's) posix locks. */
3356 static struct lock_manager_operations nfsd_posix_mng_ops
= {
3360 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3362 struct nfs4_stateowner
*sop
;
3365 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3366 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3367 hval
= lockownerid_hashval(sop
->so_id
);
3368 kref_get(&sop
->so_ref
);
3370 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3372 deny
->ld_sop
= NULL
;
3373 deny
->ld_clientid
.cl_boot
= 0;
3374 deny
->ld_clientid
.cl_id
= 0;
3376 deny
->ld_start
= fl
->fl_start
;
3377 deny
->ld_length
= NFS4_MAX_UINT64
;
3378 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3379 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3380 deny
->ld_type
= NFS4_READ_LT
;
3381 if (fl
->fl_type
!= F_RDLCK
)
3382 deny
->ld_type
= NFS4_WRITE_LT
;
3385 static struct nfs4_stateowner
*
3386 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3387 struct xdr_netobj
*owner
)
3389 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3390 struct nfs4_stateowner
*op
;
3392 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3393 if (same_owner_str(op
, owner
, clid
))
3400 * Alloc a lock owner structure.
3401 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3404 * strhashval = lock_ownerstr_hashval
3407 static struct nfs4_stateowner
*
3408 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3409 struct nfs4_stateowner
*sop
;
3410 struct nfs4_replay
*rp
;
3411 unsigned int idhashval
;
3413 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3415 idhashval
= lockownerid_hashval(current_ownerid
);
3416 INIT_LIST_HEAD(&sop
->so_idhash
);
3417 INIT_LIST_HEAD(&sop
->so_strhash
);
3418 INIT_LIST_HEAD(&sop
->so_perclient
);
3419 INIT_LIST_HEAD(&sop
->so_stateids
);
3420 INIT_LIST_HEAD(&sop
->so_perstateid
);
3421 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3423 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3424 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3425 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3426 sop
->so_is_open_owner
= 0;
3427 sop
->so_id
= current_ownerid
++;
3428 sop
->so_client
= clp
;
3429 /* It is the openowner seqid that will be incremented in encode in the
3430 * case of new lockowners; so increment the lock seqid manually: */
3431 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3432 sop
->so_confirmed
= 1;
3433 rp
= &sop
->so_replay
;
3434 rp
->rp_status
= nfserr_serverfault
;
3436 rp
->rp_buf
= rp
->rp_ibuf
;
3440 static struct nfs4_stateid
*
3441 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3443 struct nfs4_stateid
*stp
;
3444 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3446 stp
= nfs4_alloc_stateid();
3449 INIT_LIST_HEAD(&stp
->st_hash
);
3450 INIT_LIST_HEAD(&stp
->st_perfile
);
3451 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3452 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3453 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3454 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3455 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3456 stp
->st_stateowner
= sop
;
3459 stp
->st_stateid
.si_boot
= get_seconds();
3460 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3461 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3462 stp
->st_stateid
.si_generation
= 0;
3463 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3464 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3465 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3466 stp
->st_openstp
= open_stp
;
3473 check_lock_length(u64 offset
, u64 length
)
3475 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3476 LOFF_OVERFLOW(offset
, length
)));
3483 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3484 struct nfsd4_lock
*lock
)
3486 struct nfs4_stateowner
*open_sop
= NULL
;
3487 struct nfs4_stateowner
*lock_sop
= NULL
;
3488 struct nfs4_stateid
*lock_stp
;
3490 struct file_lock file_lock
;
3491 struct file_lock conflock
;
3493 unsigned int strhashval
;
3497 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3498 (long long) lock
->lk_offset
,
3499 (long long) lock
->lk_length
);
3501 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3502 return nfserr_inval
;
3504 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3505 S_IFREG
, NFSD_MAY_LOCK
))) {
3506 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3512 if (lock
->lk_is_new
) {
3514 * Client indicates that this is a new lockowner.
3515 * Use open owner and open stateid to create lock owner and
3518 struct nfs4_stateid
*open_stp
= NULL
;
3519 struct nfs4_file
*fp
;
3521 status
= nfserr_stale_clientid
;
3522 if (!nfsd4_has_session(cstate
) &&
3523 STALE_CLIENTID(&lock
->lk_new_clientid
))
3526 /* validate and update open stateid and open seqid */
3527 status
= nfs4_preprocess_seqid_op(cstate
,
3528 lock
->lk_new_open_seqid
,
3529 &lock
->lk_new_open_stateid
,
3531 &lock
->lk_replay_owner
, &open_stp
,
3535 open_sop
= lock
->lk_replay_owner
;
3536 /* create lockowner and lock stateid */
3537 fp
= open_stp
->st_file
;
3538 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3539 open_sop
->so_client
->cl_clientid
.cl_id
,
3540 &lock
->v
.new.owner
);
3541 /* XXX: Do we need to check for duplicate stateowners on
3542 * the same file, or should they just be allowed (and
3543 * create new stateids)? */
3544 status
= nfserr_resource
;
3545 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3546 open_sop
->so_client
, open_stp
, lock
);
3547 if (lock_sop
== NULL
)
3549 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3550 if (lock_stp
== NULL
)
3553 /* lock (lock owner + lock stateid) already exists */
3554 status
= nfs4_preprocess_seqid_op(cstate
,
3555 lock
->lk_old_lock_seqid
,
3556 &lock
->lk_old_lock_stateid
,
3558 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3561 lock_sop
= lock
->lk_replay_owner
;
3563 /* lock->lk_replay_owner and lock_stp have been created or found */
3564 filp
= lock_stp
->st_vfs_file
;
3566 status
= nfserr_grace
;
3567 if (locks_in_grace() && !lock
->lk_reclaim
)
3569 status
= nfserr_no_grace
;
3570 if (!locks_in_grace() && lock
->lk_reclaim
)
3573 locks_init_lock(&file_lock
);
3574 switch (lock
->lk_type
) {
3577 file_lock
.fl_type
= F_RDLCK
;
3581 case NFS4_WRITEW_LT
:
3582 file_lock
.fl_type
= F_WRLCK
;
3586 status
= nfserr_inval
;
3589 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3590 file_lock
.fl_pid
= current
->tgid
;
3591 file_lock
.fl_file
= filp
;
3592 file_lock
.fl_flags
= FL_POSIX
;
3593 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3595 file_lock
.fl_start
= lock
->lk_offset
;
3596 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3597 nfs4_transform_lock_offset(&file_lock
);
3600 * Try to lock the file in the VFS.
3601 * Note: locks.c uses the BKL to protect the inode's lock list.
3604 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3606 case 0: /* success! */
3607 update_stateid(&lock_stp
->st_stateid
);
3608 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3612 case (EAGAIN
): /* conflock holds conflicting lock */
3613 status
= nfserr_denied
;
3614 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3615 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3618 status
= nfserr_deadlock
;
3621 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3622 status
= nfserr_resource
;
3626 if (status
&& lock
->lk_is_new
&& lock_sop
)
3627 release_lockowner(lock_sop
);
3628 if (lock
->lk_replay_owner
) {
3629 nfs4_get_stateowner(lock
->lk_replay_owner
);
3630 cstate
->replay_owner
= lock
->lk_replay_owner
;
3632 nfs4_unlock_state();
3637 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3638 * so we do a temporary open here just to get an open file to pass to
3639 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3642 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3647 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3650 err
= vfs_test_lock(file
, lock
);
3659 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3660 struct nfsd4_lockt
*lockt
)
3662 struct inode
*inode
;
3663 struct file_lock file_lock
;
3667 if (locks_in_grace())
3668 return nfserr_grace
;
3670 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3671 return nfserr_inval
;
3673 lockt
->lt_stateowner
= NULL
;
3676 status
= nfserr_stale_clientid
;
3677 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3680 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3681 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3682 if (status
== nfserr_symlink
)
3683 status
= nfserr_inval
;
3687 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3688 locks_init_lock(&file_lock
);
3689 switch (lockt
->lt_type
) {
3692 file_lock
.fl_type
= F_RDLCK
;
3695 case NFS4_WRITEW_LT
:
3696 file_lock
.fl_type
= F_WRLCK
;
3699 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3700 status
= nfserr_inval
;
3704 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3705 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3706 if (lockt
->lt_stateowner
)
3707 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3708 file_lock
.fl_pid
= current
->tgid
;
3709 file_lock
.fl_flags
= FL_POSIX
;
3711 file_lock
.fl_start
= lockt
->lt_offset
;
3712 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3714 nfs4_transform_lock_offset(&file_lock
);
3717 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3719 status
= nfserrno(error
);
3722 if (file_lock
.fl_type
!= F_UNLCK
) {
3723 status
= nfserr_denied
;
3724 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3727 nfs4_unlock_state();
3732 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3733 struct nfsd4_locku
*locku
)
3735 struct nfs4_stateid
*stp
;
3736 struct file
*filp
= NULL
;
3737 struct file_lock file_lock
;
3741 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3742 (long long) locku
->lu_offset
,
3743 (long long) locku
->lu_length
);
3745 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3746 return nfserr_inval
;
3750 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3754 &locku
->lu_stateowner
, &stp
, NULL
)))
3757 filp
= stp
->st_vfs_file
;
3759 locks_init_lock(&file_lock
);
3760 file_lock
.fl_type
= F_UNLCK
;
3761 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3762 file_lock
.fl_pid
= current
->tgid
;
3763 file_lock
.fl_file
= filp
;
3764 file_lock
.fl_flags
= FL_POSIX
;
3765 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3766 file_lock
.fl_start
= locku
->lu_offset
;
3768 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3769 nfs4_transform_lock_offset(&file_lock
);
3772 * Try to unlock the file in the VFS.
3774 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3776 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3780 * OK, unlock succeeded; the only thing left to do is update the stateid.
3782 update_stateid(&stp
->st_stateid
);
3783 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3786 if (locku
->lu_stateowner
) {
3787 nfs4_get_stateowner(locku
->lu_stateowner
);
3788 cstate
->replay_owner
= locku
->lu_stateowner
;
3790 nfs4_unlock_state();
3794 status
= nfserrno(err
);
3800 * 1: locks held by lockowner
3801 * 0: no locks held by lockowner
3804 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3806 struct file_lock
**flpp
;
3807 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3811 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3812 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3823 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3824 struct nfsd4_compound_state
*cstate
,
3825 struct nfsd4_release_lockowner
*rlockowner
)
3827 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3828 struct nfs4_stateowner
*sop
;
3829 struct nfs4_stateid
*stp
;
3830 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3831 struct list_head matches
;
3835 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3836 clid
->cl_boot
, clid
->cl_id
);
3838 /* XXX check for lease expiration */
3840 status
= nfserr_stale_clientid
;
3841 if (STALE_CLIENTID(clid
))
3846 status
= nfserr_locks_held
;
3847 /* XXX: we're doing a linear search through all the lockowners.
3848 * Yipes! For now we'll just hope clients aren't really using
3849 * release_lockowner much, but eventually we have to fix these
3850 * data structures. */
3851 INIT_LIST_HEAD(&matches
);
3852 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3853 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3854 if (!same_owner_str(sop
, owner
, clid
))
3856 list_for_each_entry(stp
, &sop
->so_stateids
,
3858 if (check_for_locks(stp
->st_vfs_file
, sop
))
3860 /* Note: so_perclient unused for lockowners,
3861 * so it's OK to fool with here. */
3862 list_add(&sop
->so_perclient
, &matches
);
3866 /* Clients probably won't expect us to return with some (but not all)
3867 * of the lockowner state released; so don't release any until all
3868 * have been checked. */
3870 while (!list_empty(&matches
)) {
3871 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3873 /* unhash_stateowner deletes so_perclient only
3874 * for openowners. */
3875 list_del(&sop
->so_perclient
);
3876 release_lockowner(sop
);
3879 nfs4_unlock_state();
3883 static inline struct nfs4_client_reclaim
*
3886 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3890 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3892 unsigned int strhashval
= clientstr_hashval(name
);
3893 struct nfs4_client
*clp
;
3895 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3900 * failure => all reset bets are off, nfserr_no_grace...
3903 nfs4_client_to_reclaim(const char *name
)
3905 unsigned int strhashval
;
3906 struct nfs4_client_reclaim
*crp
= NULL
;
3908 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3909 crp
= alloc_reclaim();
3912 strhashval
= clientstr_hashval(name
);
3913 INIT_LIST_HEAD(&crp
->cr_strhash
);
3914 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3915 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3916 reclaim_str_hashtbl_size
++;
3921 nfs4_release_reclaim(void)
3923 struct nfs4_client_reclaim
*crp
= NULL
;
3926 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3927 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3928 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3929 struct nfs4_client_reclaim
, cr_strhash
);
3930 list_del(&crp
->cr_strhash
);
3932 reclaim_str_hashtbl_size
--;
3935 BUG_ON(reclaim_str_hashtbl_size
);
3939 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3940 static struct nfs4_client_reclaim
*
3941 nfs4_find_reclaim_client(clientid_t
*clid
)
3943 unsigned int strhashval
;
3944 struct nfs4_client
*clp
;
3945 struct nfs4_client_reclaim
*crp
= NULL
;
3948 /* find clientid in conf_id_hashtbl */
3949 clp
= find_confirmed_client(clid
);
3953 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3954 clp
->cl_name
.len
, clp
->cl_name
.data
,
3957 /* find clp->cl_name in reclaim_str_hashtbl */
3958 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3959 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3960 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3968 * Called from OPEN. Look for clientid in reclaim list.
3971 nfs4_check_open_reclaim(clientid_t
*clid
)
3973 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3976 /* initialization to perform at module load time: */
3979 nfs4_state_init(void)
3983 status
= nfsd4_init_slabs();
3986 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3987 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
3988 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
3989 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
3990 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
3991 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
3993 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
3994 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
3995 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3996 INIT_LIST_HEAD(&file_hashtbl
[i
]);
3998 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
3999 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4000 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
4002 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
4003 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4004 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
4006 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4007 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4008 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4010 memset(&onestateid
, ~0, sizeof(stateid_t
));
4011 INIT_LIST_HEAD(&close_lru
);
4012 INIT_LIST_HEAD(&client_lru
);
4013 INIT_LIST_HEAD(&del_recall_lru
);
4014 reclaim_str_hashtbl_size
= 0;
4019 nfsd4_load_reboot_recovery_data(void)
4024 nfsd4_init_recdir(user_recovery_dirname
);
4025 status
= nfsd4_recdir_load();
4026 nfs4_unlock_state();
4028 printk("NFSD: Failure reading reboot recovery data\n");
4032 get_nfs4_grace_period(void)
4034 return max(user_lease_time
, lease_time
) * HZ
;
4038 * Since the lifetime of a delegation isn't limited to that of an open, a
4039 * client may quite reasonably hang on to a delegation as long as it has
4040 * the inode cached. This becomes an obvious problem the first time a
4041 * client's inode cache approaches the size of the server's total memory.
4043 * For now we avoid this problem by imposing a hard limit on the number
4044 * of delegations, which varies according to the server's memory size.
4047 set_max_delegations(void)
4050 * Allow at most 4 delegations per megabyte of RAM. Quick
4051 * estimates suggest that in the worst case (where every delegation
4052 * is for a different inode), a delegation could take about 1.5K,
4053 * giving a worst case usage of about 6% of memory.
4055 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4058 /* initialization to perform when the nfsd service is started: */
4061 __nfs4_state_start(void)
4063 unsigned long grace_time
;
4065 boot_time
= get_seconds();
4066 grace_time
= get_nfs4_grace_period();
4067 lease_time
= user_lease_time
;
4068 locks_start_grace(&nfsd4_manager
);
4069 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4071 laundry_wq
= create_singlethread_workqueue("nfsd4");
4072 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
4073 set_max_delegations();
4077 nfs4_state_start(void)
4081 nfsd4_load_reboot_recovery_data();
4082 __nfs4_state_start();
4088 nfs4_lease_time(void)
4094 __nfs4_state_shutdown(void)
4097 struct nfs4_client
*clp
= NULL
;
4098 struct nfs4_delegation
*dp
= NULL
;
4099 struct list_head
*pos
, *next
, reaplist
;
4101 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4102 while (!list_empty(&conf_id_hashtbl
[i
])) {
4103 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4106 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4107 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4111 INIT_LIST_HEAD(&reaplist
);
4112 spin_lock(&recall_lock
);
4113 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4114 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4115 list_move(&dp
->dl_recall_lru
, &reaplist
);
4117 spin_unlock(&recall_lock
);
4118 list_for_each_safe(pos
, next
, &reaplist
) {
4119 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4120 list_del_init(&dp
->dl_recall_lru
);
4121 unhash_delegation(dp
);
4124 nfsd4_shutdown_recdir();
4129 nfs4_state_shutdown(void)
4131 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4132 destroy_workqueue(laundry_wq
);
4133 locks_end_grace(&nfsd4_manager
);
4135 nfs4_release_reclaim();
4136 __nfs4_state_shutdown();
4137 nfs4_unlock_state();
4141 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4142 * accessed when nfsd is starting.
4145 nfs4_set_recdir(char *recdir
)
4147 strcpy(user_recovery_dirname
, recdir
);
4151 * Change the NFSv4 recovery directory to recdir.
4154 nfs4_reset_recoverydir(char *recdir
)
4159 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4163 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4164 nfs4_set_recdir(recdir
);
4172 nfs4_recoverydir(void)
4174 return user_recovery_dirname
;
4178 * Called when leasetime is changed.
4180 * The only way the protocol gives us to handle on-the-fly lease changes is to
4181 * simulate a reboot. Instead of doing that, we just wait till the next time
4182 * we start to register any changes in lease time. If the administrator
4183 * really wants to change the lease time *now*, they can go ahead and bring
4184 * nfsd down and then back up again after changing the lease time.
4186 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4187 * when nfsd is starting
4190 nfs4_reset_lease(time_t leasetime
)
4192 user_lease_time
= leasetime
;