2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
36 #include <linux/smp_lock.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease
= 90; /* default lease time */
49 time_t nfsd4_grace
= 90;
50 static time_t boot_time
;
51 static u32 current_ownerid
= 1;
52 static u32 current_fileid
= 1;
53 static u32 current_delegid
= 1;
55 static stateid_t zerostateid
; /* bits all 0 */
56 static stateid_t onestateid
; /* bits all 1 */
57 static u64 current_sessionid
= 1;
59 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
62 /* forward declarations */
63 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
64 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
65 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
66 static void nfs4_set_recdir(char *recdir
);
70 /* Currently used for almost all code touching nfsv4 state: */
71 static DEFINE_MUTEX(client_mutex
);
74 * Currently used for the del_recall_lru and file hash table. In an
75 * effort to decrease the scope of the client_mutex, this spinlock may
76 * eventually cover more:
78 static DEFINE_SPINLOCK(recall_lock
);
80 static struct kmem_cache
*stateowner_slab
= NULL
;
81 static struct kmem_cache
*file_slab
= NULL
;
82 static struct kmem_cache
*stateid_slab
= NULL
;
83 static struct kmem_cache
*deleg_slab
= NULL
;
88 mutex_lock(&client_mutex
);
92 nfs4_unlock_state(void)
94 mutex_unlock(&client_mutex
);
98 opaque_hashval(const void *ptr
, int nbytes
)
100 unsigned char *cptr
= (unsigned char *) ptr
;
110 static struct list_head del_recall_lru
;
113 put_nfs4_file(struct nfs4_file
*fi
)
115 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
116 list_del(&fi
->fi_hash
);
117 spin_unlock(&recall_lock
);
119 kmem_cache_free(file_slab
, fi
);
124 get_nfs4_file(struct nfs4_file
*fi
)
126 atomic_inc(&fi
->fi_ref
);
129 static int num_delegations
;
130 unsigned int max_delegations
;
133 * Open owner state (share locks)
136 /* hash tables for nfs4_stateowner */
137 #define OWNER_HASH_BITS 8
138 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
139 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
141 #define ownerid_hashval(id) \
142 ((id) & OWNER_HASH_MASK)
143 #define ownerstr_hashval(clientid, ownername) \
144 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
146 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
147 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
149 /* hash table for nfs4_file */
150 #define FILE_HASH_BITS 8
151 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
153 /* hash table for (open)nfs4_stateid */
154 #define STATEID_HASH_BITS 10
155 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
156 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
158 #define file_hashval(x) \
159 hash_ptr(x, FILE_HASH_BITS)
160 #define stateid_hashval(owner_id, file_id) \
161 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
163 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
164 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
166 static struct nfs4_delegation
*
167 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
169 struct nfs4_delegation
*dp
;
170 struct nfs4_file
*fp
= stp
->st_file
;
171 struct nfs4_cb_conn
*cb
= &stp
->st_stateowner
->so_client
->cl_cb_conn
;
173 dprintk("NFSD alloc_init_deleg\n");
174 if (fp
->fi_had_conflict
)
176 if (num_delegations
> max_delegations
)
178 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
182 INIT_LIST_HEAD(&dp
->dl_perfile
);
183 INIT_LIST_HEAD(&dp
->dl_perclnt
);
184 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
189 get_file(stp
->st_vfs_file
);
190 dp
->dl_vfs_file
= stp
->st_vfs_file
;
192 dp
->dl_ident
= cb
->cb_ident
;
193 dp
->dl_stateid
.si_boot
= boot_time
;
194 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
195 dp
->dl_stateid
.si_fileid
= 0;
196 dp
->dl_stateid
.si_generation
= 0;
197 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
199 atomic_set(&dp
->dl_count
, 1);
200 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
201 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
202 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
207 nfs4_put_delegation(struct nfs4_delegation
*dp
)
209 if (atomic_dec_and_test(&dp
->dl_count
)) {
210 dprintk("NFSD: freeing dp %p\n",dp
);
211 put_nfs4_file(dp
->dl_file
);
212 kmem_cache_free(deleg_slab
, dp
);
217 /* Remove the associated file_lock first, then remove the delegation.
218 * lease_modify() is called to remove the FS_LEASE file_lock from
219 * the i_flock list, eventually calling nfsd's lock_manager
220 * fl_release_callback.
223 nfs4_close_delegation(struct nfs4_delegation
*dp
)
225 struct file
*filp
= dp
->dl_vfs_file
;
227 dprintk("NFSD: close_delegation dp %p\n",dp
);
228 dp
->dl_vfs_file
= NULL
;
229 /* The following nfsd_close may not actually close the file,
230 * but we want to remove the lease in any case. */
232 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
236 /* Called under the state lock. */
238 unhash_delegation(struct nfs4_delegation
*dp
)
240 list_del_init(&dp
->dl_perfile
);
241 list_del_init(&dp
->dl_perclnt
);
242 spin_lock(&recall_lock
);
243 list_del_init(&dp
->dl_recall_lru
);
244 spin_unlock(&recall_lock
);
245 nfs4_close_delegation(dp
);
246 nfs4_put_delegation(dp
);
253 /* client_lock protects the client lru list and session hash table */
254 static DEFINE_SPINLOCK(client_lock
);
256 /* Hash tables for nfs4_clientid state */
257 #define CLIENT_HASH_BITS 4
258 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
261 #define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
263 #define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
275 * client_lru holds client queue ordered by nfs4_client.cl_time
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
281 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
282 static int reclaim_str_hashtbl_size
= 0;
283 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
284 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
285 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
286 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
287 static struct list_head client_lru
;
288 static struct list_head close_lru
;
290 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
292 list_del(&stp
->st_hash
);
293 list_del(&stp
->st_perfile
);
294 list_del(&stp
->st_perstateowner
);
297 static void free_generic_stateid(struct nfs4_stateid
*stp
)
299 put_nfs4_file(stp
->st_file
);
300 kmem_cache_free(stateid_slab
, stp
);
303 static void release_lock_stateid(struct nfs4_stateid
*stp
)
305 unhash_generic_stateid(stp
);
306 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
307 free_generic_stateid(stp
);
310 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
312 struct nfs4_stateid
*stp
;
314 list_del(&sop
->so_idhash
);
315 list_del(&sop
->so_strhash
);
316 list_del(&sop
->so_perstateid
);
317 while (!list_empty(&sop
->so_stateids
)) {
318 stp
= list_first_entry(&sop
->so_stateids
,
319 struct nfs4_stateid
, st_perstateowner
);
320 release_lock_stateid(stp
);
324 static void release_lockowner(struct nfs4_stateowner
*sop
)
326 unhash_lockowner(sop
);
327 nfs4_put_stateowner(sop
);
331 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
333 struct nfs4_stateowner
*lock_sop
;
335 while (!list_empty(&open_stp
->st_lockowners
)) {
336 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
337 struct nfs4_stateowner
, so_perstateid
);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop
->so_is_open_owner
);
340 release_lockowner(lock_sop
);
344 static void release_open_stateid(struct nfs4_stateid
*stp
)
346 unhash_generic_stateid(stp
);
347 release_stateid_lockowners(stp
);
348 nfsd_close(stp
->st_vfs_file
);
349 free_generic_stateid(stp
);
352 static void unhash_openowner(struct nfs4_stateowner
*sop
)
354 struct nfs4_stateid
*stp
;
356 list_del(&sop
->so_idhash
);
357 list_del(&sop
->so_strhash
);
358 list_del(&sop
->so_perclient
);
359 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
360 while (!list_empty(&sop
->so_stateids
)) {
361 stp
= list_first_entry(&sop
->so_stateids
,
362 struct nfs4_stateid
, st_perstateowner
);
363 release_open_stateid(stp
);
367 static void release_openowner(struct nfs4_stateowner
*sop
)
369 unhash_openowner(sop
);
370 list_del(&sop
->so_close_lru
);
371 nfs4_put_stateowner(sop
);
374 #define SESSION_HASH_SIZE 512
375 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
378 hash_sessionid(struct nfs4_sessionid
*sessionid
)
380 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
382 return sid
->sequence
% SESSION_HASH_SIZE
;
386 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
388 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
393 gen_sessionid(struct nfsd4_session
*ses
)
395 struct nfs4_client
*clp
= ses
->se_client
;
396 struct nfsd4_sessionid
*sid
;
398 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
399 sid
->clientid
= clp
->cl_clientid
;
400 sid
->sequence
= current_sessionid
++;
405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
416 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
428 static int set_forechannel_drc_size(struct nfsd4_channel_attrs
*fchan
)
430 int mem
, size
= fchan
->maxresp_cached
;
432 if (fchan
->maxreqs
< 1)
435 if (size
< NFSD_MIN_HDR_SEQ_SZ
)
436 size
= NFSD_MIN_HDR_SEQ_SZ
;
437 size
-= NFSD_MIN_HDR_SEQ_SZ
;
438 if (size
> NFSD_SLOT_CACHE_SIZE
)
439 size
= NFSD_SLOT_CACHE_SIZE
;
441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem
= fchan
->maxreqs
* size
;
443 if (mem
> NFSD_MAX_MEM_PER_SESSION
) {
444 fchan
->maxreqs
= NFSD_MAX_MEM_PER_SESSION
/ size
;
445 if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
446 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
447 mem
= fchan
->maxreqs
* size
;
450 spin_lock(&nfsd_drc_lock
);
451 /* bound the total session drc memory ussage */
452 if (mem
+ nfsd_drc_mem_used
> nfsd_drc_max_mem
) {
453 fchan
->maxreqs
= (nfsd_drc_max_mem
- nfsd_drc_mem_used
) / size
;
454 mem
= fchan
->maxreqs
* size
;
456 nfsd_drc_mem_used
+= mem
;
457 spin_unlock(&nfsd_drc_lock
);
459 if (fchan
->maxreqs
== 0)
460 return nfserr_serverfault
;
462 fchan
->maxresp_cached
= size
+ NFSD_MIN_HDR_SEQ_SZ
;
467 * fchan holds the client values on input, and the server values on output
468 * sv_max_mesg is the maximum payload plus one page for overhead.
470 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
471 struct nfsd4_channel_attrs
*session_fchan
,
472 struct nfsd4_channel_attrs
*fchan
)
475 __u32 maxcount
= nfsd_serv
->sv_max_mesg
;
477 /* headerpadsz set to zero in encode routine */
479 /* Use the client's max request and max response size if possible */
480 if (fchan
->maxreq_sz
> maxcount
)
481 fchan
->maxreq_sz
= maxcount
;
482 session_fchan
->maxreq_sz
= fchan
->maxreq_sz
;
484 if (fchan
->maxresp_sz
> maxcount
)
485 fchan
->maxresp_sz
= maxcount
;
486 session_fchan
->maxresp_sz
= fchan
->maxresp_sz
;
488 /* Use the client's maxops if possible */
489 if (fchan
->maxops
> NFSD_MAX_OPS_PER_COMPOUND
)
490 fchan
->maxops
= NFSD_MAX_OPS_PER_COMPOUND
;
491 session_fchan
->maxops
= fchan
->maxops
;
493 /* FIXME: Error means no more DRC pages so the server should
494 * recover pages from existing sessions. For now fail session
497 status
= set_forechannel_drc_size(fchan
);
499 session_fchan
->maxresp_cached
= fchan
->maxresp_cached
;
500 session_fchan
->maxreqs
= fchan
->maxreqs
;
502 dprintk("%s status %d\n", __func__
, status
);
507 free_session_slots(struct nfsd4_session
*ses
)
511 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
512 kfree(ses
->se_slots
[i
]);
516 * We don't actually need to cache the rpc and session headers, so we
517 * can allocate a little less for each slot:
519 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
521 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
525 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
526 struct nfsd4_create_session
*cses
)
528 struct nfsd4_session
*new, tmp
;
529 struct nfsd4_slot
*sp
;
530 int idx
, slotsize
, cachesize
, i
;
533 memset(&tmp
, 0, sizeof(tmp
));
535 /* FIXME: For now, we just accept the client back channel attributes. */
536 tmp
.se_bchannel
= cses
->back_channel
;
537 status
= init_forechannel_attrs(rqstp
, &tmp
.se_fchannel
,
538 &cses
->fore_channel
);
542 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
)
543 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
545 status
= nfserr_serverfault
;
546 /* allocate struct nfsd4_session and slot table pointers in one piece */
547 slotsize
= tmp
.se_fchannel
.maxreqs
* sizeof(struct nfsd4_slot
*);
548 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
552 memcpy(new, &tmp
, sizeof(*new));
554 /* allocate each struct nfsd4_slot and data cache in one piece */
555 cachesize
= slot_bytes(&new->se_fchannel
);
556 for (i
= 0; i
< new->se_fchannel
.maxreqs
; i
++) {
557 sp
= kzalloc(sizeof(*sp
) + cachesize
, GFP_KERNEL
);
560 new->se_slots
[i
] = sp
;
563 new->se_client
= clp
;
565 idx
= hash_sessionid(&new->se_sessionid
);
566 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
567 NFS4_MAX_SESSIONID_LEN
);
569 new->se_flags
= cses
->flags
;
570 kref_init(&new->se_ref
);
571 spin_lock(&client_lock
);
572 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
573 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
574 spin_unlock(&client_lock
);
580 free_session_slots(new);
585 /* caller must hold client_lock */
586 static struct nfsd4_session
*
587 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
589 struct nfsd4_session
*elem
;
592 dump_sessionid(__func__
, sessionid
);
593 idx
= hash_sessionid(sessionid
);
594 dprintk("%s: idx is %d\n", __func__
, idx
);
595 /* Search in the appropriate list */
596 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
597 dump_sessionid("list traversal", &elem
->se_sessionid
);
598 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
599 NFS4_MAX_SESSIONID_LEN
)) {
604 dprintk("%s: session not found\n", __func__
);
608 /* caller must hold client_lock */
610 unhash_session(struct nfsd4_session
*ses
)
612 list_del(&ses
->se_hash
);
613 list_del(&ses
->se_perclnt
);
617 free_session(struct kref
*kref
)
619 struct nfsd4_session
*ses
;
622 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
623 spin_lock(&nfsd_drc_lock
);
624 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
625 nfsd_drc_mem_used
-= mem
;
626 spin_unlock(&nfsd_drc_lock
);
627 free_session_slots(ses
);
631 /* must be called under the client_lock */
633 renew_client_locked(struct nfs4_client
*clp
)
635 if (is_client_expired(clp
)) {
636 dprintk("%s: client (clientid %08x/%08x) already expired\n",
638 clp
->cl_clientid
.cl_boot
,
639 clp
->cl_clientid
.cl_id
);
644 * Move client to the end to the LRU list.
646 dprintk("renewing client (clientid %08x/%08x)\n",
647 clp
->cl_clientid
.cl_boot
,
648 clp
->cl_clientid
.cl_id
);
649 list_move_tail(&clp
->cl_lru
, &client_lru
);
650 clp
->cl_time
= get_seconds();
654 renew_client(struct nfs4_client
*clp
)
656 spin_lock(&client_lock
);
657 renew_client_locked(clp
);
658 spin_unlock(&client_lock
);
661 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
663 STALE_CLIENTID(clientid_t
*clid
)
665 if (clid
->cl_boot
== boot_time
)
667 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
668 clid
->cl_boot
, clid
->cl_id
, boot_time
);
673 * XXX Should we use a slab cache ?
674 * This type of memory management is somewhat inefficient, but we use it
675 * anyway since SETCLIENTID is not a common operation.
677 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
679 struct nfs4_client
*clp
;
681 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
684 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
685 if (clp
->cl_name
.data
== NULL
) {
689 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
690 clp
->cl_name
.len
= name
.len
;
695 free_client(struct nfs4_client
*clp
)
697 if (clp
->cl_cred
.cr_group_info
)
698 put_group_info(clp
->cl_cred
.cr_group_info
);
699 kfree(clp
->cl_principal
);
700 kfree(clp
->cl_name
.data
);
705 release_session_client(struct nfsd4_session
*session
)
707 struct nfs4_client
*clp
= session
->se_client
;
709 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
711 if (is_client_expired(clp
)) {
713 session
->se_client
= NULL
;
715 renew_client_locked(clp
);
716 spin_unlock(&client_lock
);
719 /* must be called under the client_lock */
721 unhash_client_locked(struct nfs4_client
*clp
)
723 mark_client_expired(clp
);
724 list_del(&clp
->cl_lru
);
725 while (!list_empty(&clp
->cl_sessions
)) {
726 struct nfsd4_session
*ses
;
727 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
730 nfsd4_put_session(ses
);
735 expire_client(struct nfs4_client
*clp
)
737 struct nfs4_stateowner
*sop
;
738 struct nfs4_delegation
*dp
;
739 struct list_head reaplist
;
741 INIT_LIST_HEAD(&reaplist
);
742 spin_lock(&recall_lock
);
743 while (!list_empty(&clp
->cl_delegations
)) {
744 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
745 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
747 list_del_init(&dp
->dl_perclnt
);
748 list_move(&dp
->dl_recall_lru
, &reaplist
);
750 spin_unlock(&recall_lock
);
751 while (!list_empty(&reaplist
)) {
752 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
753 list_del_init(&dp
->dl_recall_lru
);
754 unhash_delegation(dp
);
756 while (!list_empty(&clp
->cl_openowners
)) {
757 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
758 release_openowner(sop
);
760 nfsd4_set_callback_client(clp
, NULL
);
761 if (clp
->cl_cb_conn
.cb_xprt
)
762 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
763 list_del(&clp
->cl_idhash
);
764 list_del(&clp
->cl_strhash
);
765 spin_lock(&client_lock
);
766 unhash_client_locked(clp
);
767 if (atomic_read(&clp
->cl_refcount
) == 0)
769 spin_unlock(&client_lock
);
772 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
774 memcpy(target
->cl_verifier
.data
, source
->data
,
775 sizeof(target
->cl_verifier
.data
));
778 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
780 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
781 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
784 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
786 target
->cr_uid
= source
->cr_uid
;
787 target
->cr_gid
= source
->cr_gid
;
788 target
->cr_group_info
= source
->cr_group_info
;
789 get_group_info(target
->cr_group_info
);
792 static int same_name(const char *n1
, const char *n2
)
794 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
798 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
800 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
804 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
806 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
809 /* XXX what about NGROUP */
811 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
813 return cr1
->cr_uid
== cr2
->cr_uid
;
816 static void gen_clid(struct nfs4_client
*clp
)
818 static u32 current_clientid
= 1;
820 clp
->cl_clientid
.cl_boot
= boot_time
;
821 clp
->cl_clientid
.cl_id
= current_clientid
++;
824 static void gen_confirm(struct nfs4_client
*clp
)
829 p
= (u32
*)clp
->cl_confirm
.data
;
830 *p
++ = get_seconds();
834 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
835 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
837 struct nfs4_client
*clp
;
838 struct sockaddr
*sa
= svc_addr(rqstp
);
841 clp
= alloc_client(name
);
845 princ
= svc_gss_principal(rqstp
);
847 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
848 if (clp
->cl_principal
== NULL
) {
854 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
855 atomic_set(&clp
->cl_refcount
, 0);
856 atomic_set(&clp
->cl_cb_set
, 0);
857 INIT_LIST_HEAD(&clp
->cl_idhash
);
858 INIT_LIST_HEAD(&clp
->cl_strhash
);
859 INIT_LIST_HEAD(&clp
->cl_openowners
);
860 INIT_LIST_HEAD(&clp
->cl_delegations
);
861 INIT_LIST_HEAD(&clp
->cl_sessions
);
862 INIT_LIST_HEAD(&clp
->cl_lru
);
863 clp
->cl_time
= get_seconds();
864 clear_bit(0, &clp
->cl_cb_slot_busy
);
865 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
866 copy_verf(clp
, verf
);
867 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
868 clp
->cl_flavor
= rqstp
->rq_flavor
;
869 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
875 static int check_name(struct xdr_netobj name
)
879 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
880 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
887 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
889 unsigned int idhashval
;
891 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
892 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
893 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
898 move_to_confirmed(struct nfs4_client
*clp
)
900 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
901 unsigned int strhashval
;
903 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
904 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
905 strhashval
= clientstr_hashval(clp
->cl_recdir
);
906 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
910 static struct nfs4_client
*
911 find_confirmed_client(clientid_t
*clid
)
913 struct nfs4_client
*clp
;
914 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
916 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
917 if (same_clid(&clp
->cl_clientid
, clid
))
923 static struct nfs4_client
*
924 find_unconfirmed_client(clientid_t
*clid
)
926 struct nfs4_client
*clp
;
927 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
929 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
930 if (same_clid(&clp
->cl_clientid
, clid
))
937 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
938 * parameter. Matching is based on the fact the at least one of the
939 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
941 * FIXME: we need to unify the clientid namespaces for nfsv4.x
942 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
943 * and SET_CLIENTID{,_CONFIRM}
946 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
948 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
949 return use_exchange_id
== has_exchange_flags
;
952 static struct nfs4_client
*
953 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
954 bool use_exchange_id
)
956 struct nfs4_client
*clp
;
958 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
959 if (same_name(clp
->cl_recdir
, dname
) &&
960 match_clientid_establishment(clp
, use_exchange_id
))
966 static struct nfs4_client
*
967 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
968 bool use_exchange_id
)
970 struct nfs4_client
*clp
;
972 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
973 if (same_name(clp
->cl_recdir
, dname
) &&
974 match_clientid_establishment(clp
, use_exchange_id
))
981 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, u32 scopeid
)
983 struct nfs4_cb_conn
*cb
= &clp
->cl_cb_conn
;
984 unsigned short expected_family
;
986 /* Currently, we only support tcp and tcp6 for the callback channel */
987 if (se
->se_callback_netid_len
== 3 &&
988 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
989 expected_family
= AF_INET
;
990 else if (se
->se_callback_netid_len
== 4 &&
991 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
992 expected_family
= AF_INET6
;
996 cb
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
997 se
->se_callback_addr_len
,
998 (struct sockaddr
*) &cb
->cb_addr
,
999 sizeof(cb
->cb_addr
));
1001 if (!cb
->cb_addrlen
|| cb
->cb_addr
.ss_family
!= expected_family
)
1004 if (cb
->cb_addr
.ss_family
== AF_INET6
)
1005 ((struct sockaddr_in6
*) &cb
->cb_addr
)->sin6_scope_id
= scopeid
;
1007 cb
->cb_minorversion
= 0;
1008 cb
->cb_prog
= se
->se_callback_prog
;
1009 cb
->cb_ident
= se
->se_callback_ident
;
1012 cb
->cb_addr
.ss_family
= AF_UNSPEC
;
1014 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1015 "will not receive delegations\n",
1016 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1022 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1025 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1027 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1030 dprintk("--> %s slot %p\n", __func__
, slot
);
1032 slot
->sl_opcnt
= resp
->opcnt
;
1033 slot
->sl_status
= resp
->cstate
.status
;
1035 if (nfsd4_not_cached(resp
)) {
1036 slot
->sl_datalen
= 0;
1039 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1040 base
= (char *)resp
->cstate
.datap
-
1041 (char *)resp
->xbuf
->head
[0].iov_base
;
1042 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1044 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1049 * Encode the replay sequence operation from the slot values.
1050 * If cachethis is FALSE encode the uncached rep error on the next
1051 * operation which sets resp->p and increments resp->opcnt for
1052 * nfs4svc_encode_compoundres.
1056 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1057 struct nfsd4_compoundres
*resp
)
1059 struct nfsd4_op
*op
;
1060 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1062 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1063 resp
->opcnt
, resp
->cstate
.slot
->sl_cachethis
);
1065 /* Encode the replayed sequence operation */
1066 op
= &args
->ops
[resp
->opcnt
- 1];
1067 nfsd4_encode_operation(resp
, op
);
1069 /* Return nfserr_retry_uncached_rep in next operation. */
1070 if (args
->opcnt
> 1 && slot
->sl_cachethis
== 0) {
1071 op
= &args
->ops
[resp
->opcnt
++];
1072 op
->status
= nfserr_retry_uncached_rep
;
1073 nfsd4_encode_operation(resp
, op
);
1079 * The sequence operation is not cached because we can use the slot and
1083 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1084 struct nfsd4_sequence
*seq
)
1086 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1089 dprintk("--> %s slot %p\n", __func__
, slot
);
1091 /* Either returns 0 or nfserr_retry_uncached */
1092 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1093 if (status
== nfserr_retry_uncached_rep
)
1096 /* The sequence operation has been encoded, cstate->datap set. */
1097 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1099 resp
->opcnt
= slot
->sl_opcnt
;
1100 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1101 status
= slot
->sl_status
;
1107 * Set the exchange_id flags returned by the server.
1110 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1112 /* pNFS is not supported */
1113 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1115 /* Referrals are supported, Migration is not. */
1116 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1118 /* set the wire flags to return to client. */
1119 clid
->flags
= new->cl_exchange_flags
;
1123 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1124 struct nfsd4_compound_state
*cstate
,
1125 struct nfsd4_exchange_id
*exid
)
1127 struct nfs4_client
*unconf
, *conf
, *new;
1129 unsigned int strhashval
;
1130 char dname
[HEXDIR_LEN
];
1131 char addr_str
[INET6_ADDRSTRLEN
];
1132 nfs4_verifier verf
= exid
->verifier
;
1133 struct sockaddr
*sa
= svc_addr(rqstp
);
1135 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1136 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1137 "ip_addr=%s flags %x, spa_how %d\n",
1138 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1139 addr_str
, exid
->flags
, exid
->spa_how
);
1141 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1142 return nfserr_inval
;
1144 /* Currently only support SP4_NONE */
1145 switch (exid
->spa_how
) {
1149 return nfserr_encr_alg_unsupp
;
1151 BUG(); /* checked by xdr code */
1153 return nfserr_serverfault
; /* no excuse :-/ */
1156 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1161 strhashval
= clientstr_hashval(dname
);
1166 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1168 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1169 /* 18.35.4 case 8 */
1170 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1171 status
= nfserr_not_same
;
1174 /* Client reboot: destroy old state */
1175 expire_client(conf
);
1178 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1179 /* 18.35.4 case 9 */
1180 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1181 status
= nfserr_perm
;
1184 expire_client(conf
);
1188 * Set bit when the owner id and verifier map to an already
1189 * confirmed client id (18.35.3).
1191 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1194 * Falling into 18.35.4 case 2, possible router replay.
1195 * Leave confirmed record intact and return same result.
1197 copy_verf(conf
, &verf
);
1202 /* 18.35.4 case 7 */
1203 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1204 status
= nfserr_noent
;
1208 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1211 * Possible retry or client restart. Per 18.35.4 case 4,
1212 * a new unconfirmed record should be generated regardless
1213 * of whether any properties have changed.
1215 expire_client(unconf
);
1220 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1222 status
= nfserr_serverfault
;
1227 add_to_unconfirmed(new, strhashval
);
1229 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1230 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1233 nfsd4_set_ex_flags(new, exid
);
1235 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1236 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1240 nfs4_unlock_state();
1242 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1247 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1249 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1252 /* The slot is in use, and no response has been sent. */
1254 if (seqid
== slot_seqid
)
1255 return nfserr_jukebox
;
1257 return nfserr_seq_misordered
;
1260 if (likely(seqid
== slot_seqid
+ 1))
1263 if (seqid
== slot_seqid
)
1264 return nfserr_replay_cache
;
1266 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1268 /* Misordered replay or misordered new request */
1269 return nfserr_seq_misordered
;
1273 * Cache the create session result into the create session single DRC
1274 * slot cache by saving the xdr structure. sl_seqid has been set.
1275 * Do this for solo or embedded create session operations.
1278 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1279 struct nfsd4_clid_slot
*slot
, int nfserr
)
1281 slot
->sl_status
= nfserr
;
1282 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1286 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1287 struct nfsd4_clid_slot
*slot
)
1289 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1290 return slot
->sl_status
;
1294 nfsd4_create_session(struct svc_rqst
*rqstp
,
1295 struct nfsd4_compound_state
*cstate
,
1296 struct nfsd4_create_session
*cr_ses
)
1298 struct sockaddr
*sa
= svc_addr(rqstp
);
1299 struct nfs4_client
*conf
, *unconf
;
1300 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1304 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1305 conf
= find_confirmed_client(&cr_ses
->clientid
);
1308 cs_slot
= &conf
->cl_cs_slot
;
1309 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1310 if (status
== nfserr_replay_cache
) {
1311 dprintk("Got a create_session replay! seqid= %d\n",
1313 /* Return the cached reply status */
1314 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1316 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1317 status
= nfserr_seq_misordered
;
1318 dprintk("Sequence misordered!\n");
1319 dprintk("Expected seqid= %d but got seqid= %d\n",
1320 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1323 cs_slot
->sl_seqid
++;
1324 } else if (unconf
) {
1325 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1326 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1327 status
= nfserr_clid_inuse
;
1331 cs_slot
= &unconf
->cl_cs_slot
;
1332 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1334 /* an unconfirmed replay returns misordered */
1335 status
= nfserr_seq_misordered
;
1339 cs_slot
->sl_seqid
++; /* from 0 to 1 */
1340 move_to_confirmed(unconf
);
1342 if (cr_ses
->flags
& SESSION4_BACK_CHAN
) {
1343 unconf
->cl_cb_conn
.cb_xprt
= rqstp
->rq_xprt
;
1344 svc_xprt_get(rqstp
->rq_xprt
);
1346 (struct sockaddr
*)&unconf
->cl_cb_conn
.cb_addr
,
1348 unconf
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
1349 unconf
->cl_cb_conn
.cb_minorversion
=
1350 cstate
->minorversion
;
1351 unconf
->cl_cb_conn
.cb_prog
= cr_ses
->callback_prog
;
1352 unconf
->cl_cb_seq_nr
= 1;
1353 nfsd4_probe_callback(unconf
, &unconf
->cl_cb_conn
);
1357 status
= nfserr_stale_clientid
;
1362 * We do not support RDMA or persistent sessions
1364 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1365 cr_ses
->flags
&= ~SESSION4_RDMA
;
1367 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1371 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1372 NFS4_MAX_SESSIONID_LEN
);
1373 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1376 /* cache solo and embedded create sessions under the state lock */
1377 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1379 nfs4_unlock_state();
1380 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1384 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1386 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1387 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1389 return argp
->opcnt
== resp
->opcnt
;
1392 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1396 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1400 nfsd4_destroy_session(struct svc_rqst
*r
,
1401 struct nfsd4_compound_state
*cstate
,
1402 struct nfsd4_destroy_session
*sessionid
)
1404 struct nfsd4_session
*ses
;
1405 u32 status
= nfserr_badsession
;
1408 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1409 * - Should we return nfserr_back_chan_busy if waiting for
1410 * callbacks on to-be-destroyed session?
1411 * - Do we need to clear any callback info from previous session?
1414 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1415 if (!nfsd4_last_compound_op(r
))
1416 return nfserr_not_only_op
;
1418 dump_sessionid(__func__
, &sessionid
->sessionid
);
1419 spin_lock(&client_lock
);
1420 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1422 spin_unlock(&client_lock
);
1426 unhash_session(ses
);
1427 spin_unlock(&client_lock
);
1430 /* wait for callbacks */
1431 nfsd4_set_callback_client(ses
->se_client
, NULL
);
1432 nfs4_unlock_state();
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(&client_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 /* We do not negotiate the number of slots yet, so set the
1467 * maxslots to the session maxreqs which is used to encode
1468 * sr_highest_slotid and the sr_target_slot id to maxslots */
1469 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1471 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1472 if (status
== nfserr_replay_cache
) {
1473 cstate
->slot
= slot
;
1474 cstate
->session
= session
;
1475 /* Return the cached reply status and set cstate->status
1476 * for nfsd4_proc_compound processing */
1477 status
= nfsd4_replay_cache_entry(resp
, seq
);
1478 cstate
->status
= nfserr_replay_cache
;
1484 /* Success! bump slot seqid */
1485 slot
->sl_inuse
= true;
1486 slot
->sl_seqid
= seq
->seqid
;
1487 slot
->sl_cachethis
= seq
->cachethis
;
1489 cstate
->slot
= slot
;
1490 cstate
->session
= session
;
1493 /* Hold a session reference until done processing the compound. */
1494 if (cstate
->session
) {
1495 nfsd4_get_session(cstate
->session
);
1496 atomic_inc(&session
->se_client
->cl_refcount
);
1498 spin_unlock(&client_lock
);
1499 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1504 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
1506 if (rc
->rca_one_fs
) {
1507 if (!cstate
->current_fh
.fh_dentry
)
1508 return nfserr_nofilehandle
;
1510 * We don't take advantage of the rca_one_fs case.
1511 * That's OK, it's optional, we can safely ignore it.
1516 if (is_client_expired(cstate
->session
->se_client
)) {
1517 nfs4_unlock_state();
1519 * The following error isn't really legal.
1520 * But we only get here if the client just explicitly
1521 * destroyed the client. Surely it no longer cares what
1522 * error it gets back on an operation for the dead
1525 return nfserr_stale_clientid
;
1527 nfsd4_create_clid_dir(cstate
->session
->se_client
);
1528 nfs4_unlock_state();
1533 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1534 struct nfsd4_setclientid
*setclid
)
1536 struct sockaddr
*sa
= svc_addr(rqstp
);
1537 struct xdr_netobj clname
= {
1538 .len
= setclid
->se_namelen
,
1539 .data
= setclid
->se_name
,
1541 nfs4_verifier clverifier
= setclid
->se_verf
;
1542 unsigned int strhashval
;
1543 struct nfs4_client
*conf
, *unconf
, *new;
1545 char dname
[HEXDIR_LEN
];
1547 if (!check_name(clname
))
1548 return nfserr_inval
;
1550 status
= nfs4_make_rec_clidname(dname
, &clname
);
1555 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1556 * We get here on a DRC miss.
1559 strhashval
= clientstr_hashval(dname
);
1562 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1564 /* RFC 3530 14.2.33 CASE 0: */
1565 status
= nfserr_clid_inuse
;
1566 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1567 char addr_str
[INET6_ADDRSTRLEN
];
1568 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1570 dprintk("NFSD: setclientid: string in use by client "
1571 "at %s\n", addr_str
);
1576 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1577 * has a description of SETCLIENTID request processing consisting
1578 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1580 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1581 status
= nfserr_resource
;
1584 * RFC 3530 14.2.33 CASE 4:
1585 * placed first, because it is the normal case
1588 expire_client(unconf
);
1589 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1593 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1595 * RFC 3530 14.2.33 CASE 1:
1596 * probable callback update
1599 /* Note this is removing unconfirmed {*x***},
1600 * which is stronger than RFC recommended {vxc**}.
1601 * This has the advantage that there is at most
1602 * one {*x***} in either list at any time.
1604 expire_client(unconf
);
1606 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1609 copy_clid(new, conf
);
1610 } else if (!unconf
) {
1612 * RFC 3530 14.2.33 CASE 2:
1613 * probable client reboot; state will be removed if
1616 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1622 * RFC 3530 14.2.33 CASE 3:
1623 * probable client reboot; state will be removed if
1626 expire_client(unconf
);
1627 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1632 gen_callback(new, setclid
, rpc_get_scope_id(sa
));
1633 add_to_unconfirmed(new, strhashval
);
1634 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1635 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1636 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1639 nfs4_unlock_state();
1645 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1646 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1647 * bullets, labeled as CASE1 - CASE4 below.
1650 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1651 struct nfsd4_compound_state
*cstate
,
1652 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1654 struct sockaddr
*sa
= svc_addr(rqstp
);
1655 struct nfs4_client
*conf
, *unconf
;
1656 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1657 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1660 if (STALE_CLIENTID(clid
))
1661 return nfserr_stale_clientid
;
1663 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1664 * We get here on a DRC miss.
1669 conf
= find_confirmed_client(clid
);
1670 unconf
= find_unconfirmed_client(clid
);
1672 status
= nfserr_clid_inuse
;
1673 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
1675 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
1679 * section 14.2.34 of RFC 3530 has a description of
1680 * SETCLIENTID_CONFIRM request processing consisting
1681 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1683 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1685 * RFC 3530 14.2.34 CASE 1:
1688 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1689 status
= nfserr_clid_inuse
;
1691 atomic_set(&conf
->cl_cb_set
, 0);
1692 nfsd4_probe_callback(conf
, &unconf
->cl_cb_conn
);
1693 expire_client(unconf
);
1697 } else if (conf
&& !unconf
) {
1699 * RFC 3530 14.2.34 CASE 2:
1700 * probable retransmitted request; play it safe and
1703 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1704 status
= nfserr_clid_inuse
;
1707 } else if (!conf
&& unconf
1708 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1710 * RFC 3530 14.2.34 CASE 3:
1711 * Normal case; new or rebooted client:
1713 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1714 status
= nfserr_clid_inuse
;
1717 clientstr_hashval(unconf
->cl_recdir
);
1718 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1721 nfsd4_remove_clid_dir(conf
);
1722 expire_client(conf
);
1724 move_to_confirmed(unconf
);
1726 nfsd4_probe_callback(conf
, &conf
->cl_cb_conn
);
1729 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1730 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1733 * RFC 3530 14.2.34 CASE 4:
1734 * Client probably hasn't noticed that we rebooted yet.
1736 status
= nfserr_stale_clientid
;
1738 /* check that we have hit one of the cases...*/
1739 status
= nfserr_clid_inuse
;
1742 nfs4_unlock_state();
1746 /* OPEN Share state helper functions */
1747 static inline struct nfs4_file
*
1748 alloc_init_file(struct inode
*ino
)
1750 struct nfs4_file
*fp
;
1751 unsigned int hashval
= file_hashval(ino
);
1753 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1755 atomic_set(&fp
->fi_ref
, 1);
1756 INIT_LIST_HEAD(&fp
->fi_hash
);
1757 INIT_LIST_HEAD(&fp
->fi_stateids
);
1758 INIT_LIST_HEAD(&fp
->fi_delegations
);
1759 fp
->fi_inode
= igrab(ino
);
1760 fp
->fi_id
= current_fileid
++;
1761 fp
->fi_had_conflict
= false;
1762 spin_lock(&recall_lock
);
1763 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1764 spin_unlock(&recall_lock
);
1771 nfsd4_free_slab(struct kmem_cache
**slab
)
1775 kmem_cache_destroy(*slab
);
1780 nfsd4_free_slabs(void)
1782 nfsd4_free_slab(&stateowner_slab
);
1783 nfsd4_free_slab(&file_slab
);
1784 nfsd4_free_slab(&stateid_slab
);
1785 nfsd4_free_slab(&deleg_slab
);
1789 nfsd4_init_slabs(void)
1791 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1792 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1793 if (stateowner_slab
== NULL
)
1795 file_slab
= kmem_cache_create("nfsd4_files",
1796 sizeof(struct nfs4_file
), 0, 0, NULL
);
1797 if (file_slab
== NULL
)
1799 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1800 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1801 if (stateid_slab
== NULL
)
1803 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1804 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1805 if (deleg_slab
== NULL
)
1810 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1815 nfs4_free_stateowner(struct kref
*kref
)
1817 struct nfs4_stateowner
*sop
=
1818 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1819 kfree(sop
->so_owner
.data
);
1820 kmem_cache_free(stateowner_slab
, sop
);
1823 static inline struct nfs4_stateowner
*
1824 alloc_stateowner(struct xdr_netobj
*owner
)
1826 struct nfs4_stateowner
*sop
;
1828 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1829 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1830 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1831 sop
->so_owner
.len
= owner
->len
;
1832 kref_init(&sop
->so_ref
);
1835 kmem_cache_free(stateowner_slab
, sop
);
1840 static struct nfs4_stateowner
*
1841 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1842 struct nfs4_stateowner
*sop
;
1843 struct nfs4_replay
*rp
;
1844 unsigned int idhashval
;
1846 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1848 idhashval
= ownerid_hashval(current_ownerid
);
1849 INIT_LIST_HEAD(&sop
->so_idhash
);
1850 INIT_LIST_HEAD(&sop
->so_strhash
);
1851 INIT_LIST_HEAD(&sop
->so_perclient
);
1852 INIT_LIST_HEAD(&sop
->so_stateids
);
1853 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1854 INIT_LIST_HEAD(&sop
->so_close_lru
);
1856 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1857 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1858 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1859 sop
->so_is_open_owner
= 1;
1860 sop
->so_id
= current_ownerid
++;
1861 sop
->so_client
= clp
;
1862 sop
->so_seqid
= open
->op_seqid
;
1863 sop
->so_confirmed
= 0;
1864 rp
= &sop
->so_replay
;
1865 rp
->rp_status
= nfserr_serverfault
;
1867 rp
->rp_buf
= rp
->rp_ibuf
;
1872 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1873 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1874 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1876 INIT_LIST_HEAD(&stp
->st_hash
);
1877 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1878 INIT_LIST_HEAD(&stp
->st_lockowners
);
1879 INIT_LIST_HEAD(&stp
->st_perfile
);
1880 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1881 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1882 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1883 stp
->st_stateowner
= sop
;
1886 stp
->st_stateid
.si_boot
= boot_time
;
1887 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1888 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1889 stp
->st_stateid
.si_generation
= 0;
1890 stp
->st_access_bmap
= 0;
1891 stp
->st_deny_bmap
= 0;
1892 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
1893 &stp
->st_access_bmap
);
1894 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1895 stp
->st_openstp
= NULL
;
1899 move_to_close_lru(struct nfs4_stateowner
*sop
)
1901 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1903 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1904 sop
->so_time
= get_seconds();
1908 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1911 return (sop
->so_owner
.len
== owner
->len
) &&
1912 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1913 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1916 static struct nfs4_stateowner
*
1917 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1919 struct nfs4_stateowner
*so
= NULL
;
1921 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1922 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1928 /* search file_hashtbl[] for file */
1929 static struct nfs4_file
*
1930 find_file(struct inode
*ino
)
1932 unsigned int hashval
= file_hashval(ino
);
1933 struct nfs4_file
*fp
;
1935 spin_lock(&recall_lock
);
1936 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1937 if (fp
->fi_inode
== ino
) {
1939 spin_unlock(&recall_lock
);
1943 spin_unlock(&recall_lock
);
1947 static inline int access_valid(u32 x
, u32 minorversion
)
1949 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
1951 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
1953 x
&= ~NFS4_SHARE_ACCESS_MASK
;
1954 if (minorversion
&& x
) {
1955 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
1957 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
1959 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
1966 static inline int deny_valid(u32 x
)
1968 /* Note: unlike access bits, deny bits may be zero. */
1969 return x
<= NFS4_SHARE_DENY_BOTH
;
1973 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1974 * st_{access,deny}_bmap field of the stateid, in order to track not
1975 * only what share bits are currently in force, but also what
1976 * combinations of share bits previous opens have used. This allows us
1977 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1978 * return an error if the client attempt to downgrade to a combination
1979 * of share bits not explicable by closing some of its previous opens.
1981 * XXX: This enforcement is actually incomplete, since we don't keep
1982 * track of access/deny bit combinations; so, e.g., we allow:
1984 * OPEN allow read, deny write
1985 * OPEN allow both, deny none
1986 * DOWNGRADE allow read, deny none
1988 * which we should reject.
1991 set_access(unsigned int *access
, unsigned long bmap
) {
1995 for (i
= 1; i
< 4; i
++) {
1996 if (test_bit(i
, &bmap
))
2002 set_deny(unsigned int *deny
, unsigned long bmap
) {
2006 for (i
= 0; i
< 4; i
++) {
2007 if (test_bit(i
, &bmap
))
2013 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
2014 unsigned int access
, deny
;
2016 set_access(&access
, stp
->st_access_bmap
);
2017 set_deny(&deny
, stp
->st_deny_bmap
);
2018 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
2024 * Called to check deny when READ with all zero stateid or
2025 * WRITE with all zero or all one stateid
2028 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2030 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2031 struct nfs4_file
*fp
;
2032 struct nfs4_stateid
*stp
;
2035 dprintk("NFSD: nfs4_share_conflict\n");
2037 fp
= find_file(ino
);
2040 ret
= nfserr_locked
;
2041 /* Search for conflicting share reservations */
2042 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2043 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2044 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2054 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
2056 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
2057 drop_file_write_access(filp
);
2058 spin_lock(&filp
->f_lock
);
2059 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
2060 spin_unlock(&filp
->f_lock
);
2065 * Spawn a thread to perform a recall on the delegation represented
2066 * by the lease (file_lock)
2068 * Called from break_lease() with lock_kernel() held.
2069 * Note: we assume break_lease will only call this *once* for any given
2073 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2075 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2077 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2081 /* We're assuming the state code never drops its reference
2082 * without first removing the lease. Since we're in this lease
2083 * callback (and since the lease code is serialized by the kernel
2084 * lock) we know the server hasn't removed the lease yet, we know
2085 * it's safe to take a reference: */
2086 atomic_inc(&dp
->dl_count
);
2088 spin_lock(&recall_lock
);
2089 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2090 spin_unlock(&recall_lock
);
2092 /* only place dl_time is set. protected by lock_kernel*/
2093 dp
->dl_time
= get_seconds();
2096 * We don't want the locks code to timeout the lease for us;
2097 * we'll remove it ourself if the delegation isn't returned
2100 fl
->fl_break_time
= 0;
2102 dp
->dl_file
->fi_had_conflict
= true;
2103 nfsd4_cb_recall(dp
);
2107 * The file_lock is being reapd.
2109 * Called by locks_free_lock() with lock_kernel() held.
2112 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2114 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2116 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2118 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2120 dp
->dl_flock
= NULL
;
2124 * Set the delegation file_lock back pointer.
2126 * Called from setlease() with lock_kernel() held.
2129 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2131 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2133 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2140 * Called from setlease() with lock_kernel() held
2143 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2145 struct nfs4_delegation
*onlistd
=
2146 (struct nfs4_delegation
*)onlist
->fl_owner
;
2147 struct nfs4_delegation
*tryd
=
2148 (struct nfs4_delegation
*)try->fl_owner
;
2150 if (onlist
->fl_lmops
!= try->fl_lmops
)
2153 return onlistd
->dl_client
== tryd
->dl_client
;
2158 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2161 return lease_modify(onlist
, arg
);
2166 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2167 .fl_break
= nfsd_break_deleg_cb
,
2168 .fl_release_private
= nfsd_release_deleg_cb
,
2169 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2170 .fl_mylease
= nfsd_same_client_deleg_cb
,
2171 .fl_change
= nfsd_change_deleg_cb
,
2176 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2177 struct nfsd4_open
*open
)
2179 clientid_t
*clientid
= &open
->op_clientid
;
2180 struct nfs4_client
*clp
= NULL
;
2181 unsigned int strhashval
;
2182 struct nfs4_stateowner
*sop
= NULL
;
2184 if (!check_name(open
->op_owner
))
2185 return nfserr_inval
;
2187 if (STALE_CLIENTID(&open
->op_clientid
))
2188 return nfserr_stale_clientid
;
2190 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2191 sop
= find_openstateowner_str(strhashval
, open
);
2192 open
->op_stateowner
= sop
;
2194 /* Make sure the client's lease hasn't expired. */
2195 clp
= find_confirmed_client(clientid
);
2197 return nfserr_expired
;
2200 /* When sessions are used, skip open sequenceid processing */
2201 if (nfsd4_has_session(cstate
))
2203 if (!sop
->so_confirmed
) {
2204 /* Replace unconfirmed owners without checking for replay. */
2205 clp
= sop
->so_client
;
2206 release_openowner(sop
);
2207 open
->op_stateowner
= NULL
;
2210 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2211 if (sop
->so_replay
.rp_buflen
)
2212 return nfserr_replay_me
;
2213 /* The original OPEN failed so spectacularly
2214 * that we don't even have replay data saved!
2215 * Therefore, we have no choice but to continue
2216 * processing this OPEN; presumably, we'll
2217 * fail again for the same reason.
2219 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2222 if (open
->op_seqid
!= sop
->so_seqid
)
2223 return nfserr_bad_seqid
;
2225 if (open
->op_stateowner
== NULL
) {
2226 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2228 return nfserr_resource
;
2229 open
->op_stateowner
= sop
;
2231 list_del_init(&sop
->so_close_lru
);
2232 renew_client(sop
->so_client
);
2236 static inline __be32
2237 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2239 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2240 return nfserr_openmode
;
2245 static struct nfs4_delegation
*
2246 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2248 struct nfs4_delegation
*dp
;
2250 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2251 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2257 int share_access_to_flags(u32 share_access
)
2259 share_access
&= ~NFS4_SHARE_WANT_MASK
;
2261 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2265 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2266 struct nfs4_delegation
**dp
)
2269 __be32 status
= nfserr_bad_stateid
;
2271 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2274 flags
= share_access_to_flags(open
->op_share_access
);
2275 status
= nfs4_check_delegmode(*dp
, flags
);
2279 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2283 open
->op_stateowner
->so_confirmed
= 1;
2288 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2290 struct nfs4_stateid
*local
;
2291 __be32 status
= nfserr_share_denied
;
2292 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2294 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2295 /* ignore lock owners */
2296 if (local
->st_stateowner
->so_is_open_owner
== 0)
2298 /* remember if we have seen this open owner */
2299 if (local
->st_stateowner
== sop
)
2301 /* check for conflicting share reservations */
2302 if (!test_share(local
, open
))
2310 static inline struct nfs4_stateid
*
2311 nfs4_alloc_stateid(void)
2313 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2317 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2318 struct nfs4_delegation
*dp
,
2319 struct svc_fh
*cur_fh
, int flags
)
2321 struct nfs4_stateid
*stp
;
2323 stp
= nfs4_alloc_stateid();
2325 return nfserr_resource
;
2328 get_file(dp
->dl_vfs_file
);
2329 stp
->st_vfs_file
= dp
->dl_vfs_file
;
2332 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
2335 if (status
== nfserr_dropit
)
2336 status
= nfserr_jukebox
;
2337 kmem_cache_free(stateid_slab
, stp
);
2345 static inline __be32
2346 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2347 struct nfsd4_open
*open
)
2349 struct iattr iattr
= {
2350 .ia_valid
= ATTR_SIZE
,
2353 if (!open
->op_truncate
)
2355 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2356 return nfserr_inval
;
2357 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2361 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2363 struct file
*filp
= stp
->st_vfs_file
;
2364 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2365 unsigned int share_access
, new_writer
;
2366 u32 op_share_access
;
2369 set_access(&share_access
, stp
->st_access_bmap
);
2370 new_writer
= (~share_access
) & open
->op_share_access
2371 & NFS4_SHARE_ACCESS_WRITE
;
2374 int err
= get_write_access(inode
);
2376 return nfserrno(err
);
2377 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
2379 return nfserrno(err
);
2380 file_take_write(filp
);
2382 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2385 put_write_access(inode
);
2388 /* remember the open */
2389 op_share_access
= open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
;
2390 filp
->f_mode
|= op_share_access
;
2391 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2392 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2399 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2401 open
->op_stateowner
->so_confirmed
= 1;
2402 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2406 * Attempt to hand out a delegation.
2409 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2411 struct nfs4_delegation
*dp
;
2412 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2413 int cb_up
= atomic_read(&sop
->so_client
->cl_cb_set
);
2414 struct file_lock fl
, *flp
= &fl
;
2415 int status
, flag
= 0;
2417 flag
= NFS4_OPEN_DELEGATE_NONE
;
2418 open
->op_recall
= 0;
2419 switch (open
->op_claim_type
) {
2420 case NFS4_OPEN_CLAIM_PREVIOUS
:
2422 open
->op_recall
= 1;
2423 flag
= open
->op_delegate_type
;
2424 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2427 case NFS4_OPEN_CLAIM_NULL
:
2428 /* Let's not give out any delegations till everyone's
2429 * had the chance to reclaim theirs.... */
2430 if (locks_in_grace())
2432 if (!cb_up
|| !sop
->so_confirmed
)
2434 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2435 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2437 flag
= NFS4_OPEN_DELEGATE_READ
;
2443 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2445 flag
= NFS4_OPEN_DELEGATE_NONE
;
2448 locks_init_lock(&fl
);
2449 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2450 fl
.fl_flags
= FL_LEASE
;
2451 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2452 fl
.fl_end
= OFFSET_MAX
;
2453 fl
.fl_owner
= (fl_owner_t
)dp
;
2454 fl
.fl_file
= stp
->st_vfs_file
;
2455 fl
.fl_pid
= current
->tgid
;
2457 /* vfs_setlease checks to see if delegation should be handed out.
2458 * the lock_manager callbacks fl_mylease and fl_change are used
2460 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
2461 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2462 unhash_delegation(dp
);
2463 flag
= NFS4_OPEN_DELEGATE_NONE
;
2467 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2469 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2470 STATEID_VAL(&dp
->dl_stateid
));
2472 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2473 && flag
== NFS4_OPEN_DELEGATE_NONE
2474 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2475 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2476 open
->op_delegate_type
= flag
;
2480 * called with nfs4_lock_state() held.
2483 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2485 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2486 struct nfs4_file
*fp
= NULL
;
2487 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2488 struct nfs4_stateid
*stp
= NULL
;
2489 struct nfs4_delegation
*dp
= NULL
;
2492 status
= nfserr_inval
;
2493 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2494 || !deny_valid(open
->op_share_deny
))
2497 * Lookup file; if found, lookup stateid and check open request,
2498 * and check for delegations in the process of being recalled.
2499 * If not found, create the nfs4_file struct
2501 fp
= find_file(ino
);
2503 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2505 status
= nfs4_check_deleg(fp
, open
, &dp
);
2509 status
= nfserr_bad_stateid
;
2510 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2512 status
= nfserr_resource
;
2513 fp
= alloc_init_file(ino
);
2519 * OPEN the file, or upgrade an existing OPEN.
2520 * If truncate fails, the OPEN fails.
2523 /* Stateid was found, this is an OPEN upgrade */
2524 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
2527 update_stateid(&stp
->st_stateid
);
2529 /* Stateid was not found, this is a new OPEN */
2531 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
2532 flags
|= NFSD_MAY_READ
;
2533 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2534 flags
|= NFSD_MAY_WRITE
;
2535 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
2538 init_stateid(stp
, fp
, open
);
2539 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2541 release_open_stateid(stp
);
2544 if (nfsd4_has_session(&resp
->cstate
))
2545 update_stateid(&stp
->st_stateid
);
2547 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2549 if (nfsd4_has_session(&resp
->cstate
))
2550 open
->op_stateowner
->so_confirmed
= 1;
2553 * Attempt to hand out a delegation. No error return, because the
2554 * OPEN succeeds even if we fail.
2556 nfs4_open_delegation(current_fh
, open
, stp
);
2560 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2561 STATEID_VAL(&stp
->st_stateid
));
2565 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2566 nfs4_set_claim_prev(open
);
2568 * To finish the open response, we just need to set the rflags.
2570 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2571 if (!open
->op_stateowner
->so_confirmed
&&
2572 !nfsd4_has_session(&resp
->cstate
))
2573 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2579 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2582 struct nfs4_client
*clp
;
2586 dprintk("process_renew(%08x/%08x): starting\n",
2587 clid
->cl_boot
, clid
->cl_id
);
2588 status
= nfserr_stale_clientid
;
2589 if (STALE_CLIENTID(clid
))
2591 clp
= find_confirmed_client(clid
);
2592 status
= nfserr_expired
;
2594 /* We assume the client took too long to RENEW. */
2595 dprintk("nfsd4_renew: clientid not found!\n");
2599 status
= nfserr_cb_path_down
;
2600 if (!list_empty(&clp
->cl_delegations
)
2601 && !atomic_read(&clp
->cl_cb_set
))
2605 nfs4_unlock_state();
2609 struct lock_manager nfsd4_manager
= {
2613 nfsd4_end_grace(void)
2615 dprintk("NFSD: end of grace period\n");
2616 nfsd4_recdir_purge_old();
2617 locks_end_grace(&nfsd4_manager
);
2619 * Now that every NFSv4 client has had the chance to recover and
2620 * to see the (possibly new, possibly shorter) lease time, we
2621 * can safely set the next grace time to the current lease time:
2623 nfsd4_grace
= nfsd4_lease
;
2627 nfs4_laundromat(void)
2629 struct nfs4_client
*clp
;
2630 struct nfs4_stateowner
*sop
;
2631 struct nfs4_delegation
*dp
;
2632 struct list_head
*pos
, *next
, reaplist
;
2633 time_t cutoff
= get_seconds() - nfsd4_lease
;
2634 time_t t
, clientid_val
= nfsd4_lease
;
2635 time_t u
, test_val
= nfsd4_lease
;
2639 dprintk("NFSD: laundromat service - starting\n");
2640 if (locks_in_grace())
2642 INIT_LIST_HEAD(&reaplist
);
2643 spin_lock(&client_lock
);
2644 list_for_each_safe(pos
, next
, &client_lru
) {
2645 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2646 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2647 t
= clp
->cl_time
- cutoff
;
2648 if (clientid_val
> t
)
2652 if (atomic_read(&clp
->cl_refcount
)) {
2653 dprintk("NFSD: client in use (clientid %08x)\n",
2654 clp
->cl_clientid
.cl_id
);
2657 unhash_client_locked(clp
);
2658 list_add(&clp
->cl_lru
, &reaplist
);
2660 spin_unlock(&client_lock
);
2661 list_for_each_safe(pos
, next
, &reaplist
) {
2662 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2663 dprintk("NFSD: purging unused client (clientid %08x)\n",
2664 clp
->cl_clientid
.cl_id
);
2665 nfsd4_remove_clid_dir(clp
);
2668 spin_lock(&recall_lock
);
2669 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2670 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2671 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2672 u
= dp
->dl_time
- cutoff
;
2677 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2679 list_move(&dp
->dl_recall_lru
, &reaplist
);
2681 spin_unlock(&recall_lock
);
2682 list_for_each_safe(pos
, next
, &reaplist
) {
2683 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2684 list_del_init(&dp
->dl_recall_lru
);
2685 unhash_delegation(dp
);
2687 test_val
= nfsd4_lease
;
2688 list_for_each_safe(pos
, next
, &close_lru
) {
2689 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2690 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2691 u
= sop
->so_time
- cutoff
;
2696 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2698 release_openowner(sop
);
2700 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2701 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2702 nfs4_unlock_state();
2703 return clientid_val
;
2706 static struct workqueue_struct
*laundry_wq
;
2707 static void laundromat_main(struct work_struct
*);
2708 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2711 laundromat_main(struct work_struct
*not_used
)
2715 t
= nfs4_laundromat();
2716 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2717 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2720 static struct nfs4_stateowner
*
2721 search_close_lru(u32 st_id
, int flags
)
2723 struct nfs4_stateowner
*local
= NULL
;
2725 if (flags
& CLOSE_STATE
) {
2726 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2727 if (local
->so_id
== st_id
)
2735 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2737 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
2741 STALE_STATEID(stateid_t
*stateid
)
2743 if (stateid
->si_boot
== boot_time
)
2745 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
2746 STATEID_VAL(stateid
));
2751 access_permit_read(unsigned long access_bmap
)
2753 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2754 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2755 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2759 access_permit_write(unsigned long access_bmap
)
2761 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2762 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2766 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2768 __be32 status
= nfserr_openmode
;
2770 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2772 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2779 static inline __be32
2780 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2782 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2784 else if (locks_in_grace()) {
2785 /* Answer in remaining cases depends on existance of
2786 * conflicting state; so we must wait out the grace period. */
2787 return nfserr_grace
;
2788 } else if (flags
& WR_STATE
)
2789 return nfs4_share_conflict(current_fh
,
2790 NFS4_SHARE_DENY_WRITE
);
2791 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2792 return nfs4_share_conflict(current_fh
,
2793 NFS4_SHARE_DENY_READ
);
2797 * Allow READ/WRITE during grace period on recovered state only for files
2798 * that are not able to provide mandatory locking.
2801 grace_disallows_io(struct inode
*inode
)
2803 return locks_in_grace() && mandatory_lock(inode
);
2806 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
2809 * When sessions are used the stateid generation number is ignored
2812 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
2815 /* If the client sends us a stateid from the future, it's buggy: */
2816 if (in
->si_generation
> ref
->si_generation
)
2817 return nfserr_bad_stateid
;
2819 * The following, however, can happen. For example, if the
2820 * client sends an open and some IO at the same time, the open
2821 * may bump si_generation while the IO is still in flight.
2822 * Thanks to hard links and renames, the client never knows what
2823 * file an open will affect. So it could avoid that situation
2824 * only by serializing all opens and IO from the same open
2825 * owner. To recover from the old_stateid error, the client
2826 * will just have to retry the IO:
2828 if (in
->si_generation
< ref
->si_generation
)
2829 return nfserr_old_stateid
;
2834 static int is_delegation_stateid(stateid_t
*stateid
)
2836 return stateid
->si_fileid
== 0;
2840 * Checks for stateid operations
2843 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
2844 stateid_t
*stateid
, int flags
, struct file
**filpp
)
2846 struct nfs4_stateid
*stp
= NULL
;
2847 struct nfs4_delegation
*dp
= NULL
;
2848 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2849 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2855 if (grace_disallows_io(ino
))
2856 return nfserr_grace
;
2858 if (nfsd4_has_session(cstate
))
2859 flags
|= HAS_SESSION
;
2861 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2862 return check_special_stateids(current_fh
, stateid
, flags
);
2864 status
= nfserr_stale_stateid
;
2865 if (STALE_STATEID(stateid
))
2868 status
= nfserr_bad_stateid
;
2869 if (is_delegation_stateid(stateid
)) {
2870 dp
= find_delegation_stateid(ino
, stateid
);
2873 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
2877 status
= nfs4_check_delegmode(dp
, flags
);
2880 renew_client(dp
->dl_client
);
2882 *filpp
= dp
->dl_vfs_file
;
2883 } else { /* open or lock stateid */
2884 stp
= find_stateid(stateid
, flags
);
2887 if (nfs4_check_fh(current_fh
, stp
))
2889 if (!stp
->st_stateowner
->so_confirmed
)
2891 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
2895 status
= nfs4_check_openmode(stp
, flags
);
2898 renew_client(stp
->st_stateowner
->so_client
);
2900 *filpp
= stp
->st_vfs_file
;
2910 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2911 RD_STATE
: WR_STATE
;
2915 * Checks for sequence id mutating operations.
2918 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
2919 stateid_t
*stateid
, int flags
,
2920 struct nfs4_stateowner
**sopp
,
2921 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2923 struct nfs4_stateid
*stp
;
2924 struct nfs4_stateowner
*sop
;
2925 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2928 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
2929 seqid
, STATEID_VAL(stateid
));
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 nfserr_bad_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: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3062 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stateid
));
3064 nfsd4_create_clid_dir(sop
->so_client
);
3066 if (oc
->oc_stateowner
) {
3067 nfs4_get_stateowner(oc
->oc_stateowner
);
3068 cstate
->replay_owner
= oc
->oc_stateowner
;
3070 nfs4_unlock_state();
3076 * unset all bits in union bitmap (bmap) that
3077 * do not exist in share (from successful OPEN_DOWNGRADE)
3080 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3083 for (i
= 1; i
< 4; i
++) {
3084 if ((i
& access
) != i
)
3085 __clear_bit(i
, bmap
);
3090 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3093 for (i
= 0; i
< 4; i
++) {
3094 if ((i
& deny
) != i
)
3095 __clear_bit(i
, bmap
);
3100 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3101 struct nfsd4_compound_state
*cstate
,
3102 struct nfsd4_open_downgrade
*od
)
3105 struct nfs4_stateid
*stp
;
3106 unsigned int share_access
;
3108 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3109 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3110 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3112 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3113 || !deny_valid(od
->od_share_deny
))
3114 return nfserr_inval
;
3117 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3121 &od
->od_stateowner
, &stp
, NULL
)))
3124 status
= nfserr_inval
;
3125 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3126 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3127 stp
->st_access_bmap
, od
->od_share_access
);
3130 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3131 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3132 stp
->st_deny_bmap
, od
->od_share_deny
);
3135 set_access(&share_access
, stp
->st_access_bmap
);
3136 nfs4_file_downgrade(stp
->st_vfs_file
,
3137 share_access
& ~od
->od_share_access
);
3139 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3140 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3142 update_stateid(&stp
->st_stateid
);
3143 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3146 if (od
->od_stateowner
) {
3147 nfs4_get_stateowner(od
->od_stateowner
);
3148 cstate
->replay_owner
= od
->od_stateowner
;
3150 nfs4_unlock_state();
3155 * nfs4_unlock_state() called after encode
3158 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3159 struct nfsd4_close
*close
)
3162 struct nfs4_stateid
*stp
;
3164 dprintk("NFSD: nfsd4_close on file %.*s\n",
3165 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3166 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3169 /* check close_lru for replay */
3170 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3173 OPEN_STATE
| CLOSE_STATE
,
3174 &close
->cl_stateowner
, &stp
, NULL
)))
3177 update_stateid(&stp
->st_stateid
);
3178 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3180 /* release_stateid() calls nfsd_close() if needed */
3181 release_open_stateid(stp
);
3183 /* place unused nfs4_stateowners on so_close_lru list to be
3184 * released by the laundromat service after the lease period
3185 * to enable us to handle CLOSE replay
3187 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3188 move_to_close_lru(close
->cl_stateowner
);
3190 if (close
->cl_stateowner
) {
3191 nfs4_get_stateowner(close
->cl_stateowner
);
3192 cstate
->replay_owner
= close
->cl_stateowner
;
3194 nfs4_unlock_state();
3199 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3200 struct nfsd4_delegreturn
*dr
)
3202 struct nfs4_delegation
*dp
;
3203 stateid_t
*stateid
= &dr
->dr_stateid
;
3204 struct inode
*inode
;
3208 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3210 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3212 if (nfsd4_has_session(cstate
))
3213 flags
|= HAS_SESSION
;
3215 status
= nfserr_bad_stateid
;
3216 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3218 status
= nfserr_stale_stateid
;
3219 if (STALE_STATEID(stateid
))
3221 status
= nfserr_bad_stateid
;
3222 if (!is_delegation_stateid(stateid
))
3224 dp
= find_delegation_stateid(inode
, stateid
);
3227 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3230 renew_client(dp
->dl_client
);
3232 unhash_delegation(dp
);
3234 nfs4_unlock_state();
3241 * Lock owner state (byte-range locks)
3243 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3244 #define LOCK_HASH_BITS 8
3245 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3246 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3249 end_offset(u64 start
, u64 len
)
3254 return end
>= start
? end
: NFS4_MAX_UINT64
;
3257 /* last octet in a range */
3259 last_byte_offset(u64 start
, u64 len
)
3265 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3268 #define lockownerid_hashval(id) \
3269 ((id) & LOCK_HASH_MASK)
3271 static inline unsigned int
3272 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3273 struct xdr_netobj
*ownername
)
3275 return (file_hashval(inode
) + cl_id
3276 + opaque_hashval(ownername
->data
, ownername
->len
))
3280 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3281 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3282 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3284 static struct nfs4_stateid
*
3285 find_stateid(stateid_t
*stid
, int flags
)
3287 struct nfs4_stateid
*local
;
3288 u32 st_id
= stid
->si_stateownerid
;
3289 u32 f_id
= stid
->si_fileid
;
3290 unsigned int hashval
;
3292 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3293 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3294 hashval
= stateid_hashval(st_id
, f_id
);
3295 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3296 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3297 (local
->st_stateid
.si_fileid
== f_id
))
3302 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3303 hashval
= stateid_hashval(st_id
, f_id
);
3304 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3305 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3306 (local
->st_stateid
.si_fileid
== f_id
))
3313 static struct nfs4_delegation
*
3314 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3316 struct nfs4_file
*fp
;
3317 struct nfs4_delegation
*dl
;
3319 dprintk("NFSD: %s: stateid=" STATEID_FMT
"\n", __func__
,
3322 fp
= find_file(ino
);
3325 dl
= find_delegation_file(fp
, stid
);
3331 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3332 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3333 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3334 * locking, this prevents us from being completely protocol-compliant. The
3335 * real solution to this problem is to start using unsigned file offsets in
3336 * the VFS, but this is a very deep change!
3339 nfs4_transform_lock_offset(struct file_lock
*lock
)
3341 if (lock
->fl_start
< 0)
3342 lock
->fl_start
= OFFSET_MAX
;
3343 if (lock
->fl_end
< 0)
3344 lock
->fl_end
= OFFSET_MAX
;
3347 /* Hack!: For now, we're defining this just so we can use a pointer to it
3348 * as a unique cookie to identify our (NFSv4's) posix locks. */
3349 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3353 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3355 struct nfs4_stateowner
*sop
;
3358 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3359 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3360 hval
= lockownerid_hashval(sop
->so_id
);
3361 kref_get(&sop
->so_ref
);
3363 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3365 deny
->ld_sop
= NULL
;
3366 deny
->ld_clientid
.cl_boot
= 0;
3367 deny
->ld_clientid
.cl_id
= 0;
3369 deny
->ld_start
= fl
->fl_start
;
3370 deny
->ld_length
= NFS4_MAX_UINT64
;
3371 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3372 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3373 deny
->ld_type
= NFS4_READ_LT
;
3374 if (fl
->fl_type
!= F_RDLCK
)
3375 deny
->ld_type
= NFS4_WRITE_LT
;
3378 static struct nfs4_stateowner
*
3379 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3380 struct xdr_netobj
*owner
)
3382 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3383 struct nfs4_stateowner
*op
;
3385 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3386 if (same_owner_str(op
, owner
, clid
))
3393 * Alloc a lock owner structure.
3394 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3397 * strhashval = lock_ownerstr_hashval
3400 static struct nfs4_stateowner
*
3401 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3402 struct nfs4_stateowner
*sop
;
3403 struct nfs4_replay
*rp
;
3404 unsigned int idhashval
;
3406 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3408 idhashval
= lockownerid_hashval(current_ownerid
);
3409 INIT_LIST_HEAD(&sop
->so_idhash
);
3410 INIT_LIST_HEAD(&sop
->so_strhash
);
3411 INIT_LIST_HEAD(&sop
->so_perclient
);
3412 INIT_LIST_HEAD(&sop
->so_stateids
);
3413 INIT_LIST_HEAD(&sop
->so_perstateid
);
3414 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3416 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3417 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3418 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3419 sop
->so_is_open_owner
= 0;
3420 sop
->so_id
= current_ownerid
++;
3421 sop
->so_client
= clp
;
3422 /* It is the openowner seqid that will be incremented in encode in the
3423 * case of new lockowners; so increment the lock seqid manually: */
3424 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3425 sop
->so_confirmed
= 1;
3426 rp
= &sop
->so_replay
;
3427 rp
->rp_status
= nfserr_serverfault
;
3429 rp
->rp_buf
= rp
->rp_ibuf
;
3433 static struct nfs4_stateid
*
3434 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3436 struct nfs4_stateid
*stp
;
3437 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3439 stp
= nfs4_alloc_stateid();
3442 INIT_LIST_HEAD(&stp
->st_hash
);
3443 INIT_LIST_HEAD(&stp
->st_perfile
);
3444 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3445 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3446 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3447 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3448 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3449 stp
->st_stateowner
= sop
;
3452 stp
->st_stateid
.si_boot
= boot_time
;
3453 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3454 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3455 stp
->st_stateid
.si_generation
= 0;
3456 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
3457 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
3458 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3459 stp
->st_openstp
= open_stp
;
3466 check_lock_length(u64 offset
, u64 length
)
3468 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3469 LOFF_OVERFLOW(offset
, length
)));
3476 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3477 struct nfsd4_lock
*lock
)
3479 struct nfs4_stateowner
*open_sop
= NULL
;
3480 struct nfs4_stateowner
*lock_sop
= NULL
;
3481 struct nfs4_stateid
*lock_stp
;
3483 struct file_lock file_lock
;
3484 struct file_lock conflock
;
3486 unsigned int strhashval
;
3490 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3491 (long long) lock
->lk_offset
,
3492 (long long) lock
->lk_length
);
3494 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3495 return nfserr_inval
;
3497 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3498 S_IFREG
, NFSD_MAY_LOCK
))) {
3499 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3505 if (lock
->lk_is_new
) {
3507 * Client indicates that this is a new lockowner.
3508 * Use open owner and open stateid to create lock owner and
3511 struct nfs4_stateid
*open_stp
= NULL
;
3512 struct nfs4_file
*fp
;
3514 status
= nfserr_stale_clientid
;
3515 if (!nfsd4_has_session(cstate
) &&
3516 STALE_CLIENTID(&lock
->lk_new_clientid
))
3519 /* validate and update open stateid and open seqid */
3520 status
= nfs4_preprocess_seqid_op(cstate
,
3521 lock
->lk_new_open_seqid
,
3522 &lock
->lk_new_open_stateid
,
3524 &lock
->lk_replay_owner
, &open_stp
,
3528 open_sop
= lock
->lk_replay_owner
;
3529 /* create lockowner and lock stateid */
3530 fp
= open_stp
->st_file
;
3531 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3532 open_sop
->so_client
->cl_clientid
.cl_id
,
3533 &lock
->v
.new.owner
);
3534 /* XXX: Do we need to check for duplicate stateowners on
3535 * the same file, or should they just be allowed (and
3536 * create new stateids)? */
3537 status
= nfserr_resource
;
3538 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3539 open_sop
->so_client
, open_stp
, lock
);
3540 if (lock_sop
== NULL
)
3542 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3543 if (lock_stp
== NULL
)
3546 /* lock (lock owner + lock stateid) already exists */
3547 status
= nfs4_preprocess_seqid_op(cstate
,
3548 lock
->lk_old_lock_seqid
,
3549 &lock
->lk_old_lock_stateid
,
3551 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3554 lock_sop
= lock
->lk_replay_owner
;
3556 /* lock->lk_replay_owner and lock_stp have been created or found */
3557 filp
= lock_stp
->st_vfs_file
;
3559 status
= nfserr_grace
;
3560 if (locks_in_grace() && !lock
->lk_reclaim
)
3562 status
= nfserr_no_grace
;
3563 if (!locks_in_grace() && lock
->lk_reclaim
)
3566 locks_init_lock(&file_lock
);
3567 switch (lock
->lk_type
) {
3570 file_lock
.fl_type
= F_RDLCK
;
3574 case NFS4_WRITEW_LT
:
3575 file_lock
.fl_type
= F_WRLCK
;
3579 status
= nfserr_inval
;
3582 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3583 file_lock
.fl_pid
= current
->tgid
;
3584 file_lock
.fl_file
= filp
;
3585 file_lock
.fl_flags
= FL_POSIX
;
3586 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3588 file_lock
.fl_start
= lock
->lk_offset
;
3589 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3590 nfs4_transform_lock_offset(&file_lock
);
3593 * Try to lock the file in the VFS.
3594 * Note: locks.c uses the BKL to protect the inode's lock list.
3597 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3599 case 0: /* success! */
3600 update_stateid(&lock_stp
->st_stateid
);
3601 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3605 case (EAGAIN
): /* conflock holds conflicting lock */
3606 status
= nfserr_denied
;
3607 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3608 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3611 status
= nfserr_deadlock
;
3614 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3615 status
= nfserr_resource
;
3619 if (status
&& lock
->lk_is_new
&& lock_sop
)
3620 release_lockowner(lock_sop
);
3621 if (lock
->lk_replay_owner
) {
3622 nfs4_get_stateowner(lock
->lk_replay_owner
);
3623 cstate
->replay_owner
= lock
->lk_replay_owner
;
3625 nfs4_unlock_state();
3630 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3631 * so we do a temporary open here just to get an open file to pass to
3632 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3635 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3640 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3643 err
= vfs_test_lock(file
, lock
);
3652 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3653 struct nfsd4_lockt
*lockt
)
3655 struct inode
*inode
;
3656 struct file_lock file_lock
;
3660 if (locks_in_grace())
3661 return nfserr_grace
;
3663 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3664 return nfserr_inval
;
3666 lockt
->lt_stateowner
= NULL
;
3669 status
= nfserr_stale_clientid
;
3670 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3673 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3674 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3675 if (status
== nfserr_symlink
)
3676 status
= nfserr_inval
;
3680 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3681 locks_init_lock(&file_lock
);
3682 switch (lockt
->lt_type
) {
3685 file_lock
.fl_type
= F_RDLCK
;
3688 case NFS4_WRITEW_LT
:
3689 file_lock
.fl_type
= F_WRLCK
;
3692 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3693 status
= nfserr_inval
;
3697 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3698 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3699 if (lockt
->lt_stateowner
)
3700 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3701 file_lock
.fl_pid
= current
->tgid
;
3702 file_lock
.fl_flags
= FL_POSIX
;
3704 file_lock
.fl_start
= lockt
->lt_offset
;
3705 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3707 nfs4_transform_lock_offset(&file_lock
);
3710 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3712 status
= nfserrno(error
);
3715 if (file_lock
.fl_type
!= F_UNLCK
) {
3716 status
= nfserr_denied
;
3717 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3720 nfs4_unlock_state();
3725 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3726 struct nfsd4_locku
*locku
)
3728 struct nfs4_stateid
*stp
;
3729 struct file
*filp
= NULL
;
3730 struct file_lock file_lock
;
3734 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3735 (long long) locku
->lu_offset
,
3736 (long long) locku
->lu_length
);
3738 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3739 return nfserr_inval
;
3743 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3747 &locku
->lu_stateowner
, &stp
, NULL
)))
3750 filp
= stp
->st_vfs_file
;
3752 locks_init_lock(&file_lock
);
3753 file_lock
.fl_type
= F_UNLCK
;
3754 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3755 file_lock
.fl_pid
= current
->tgid
;
3756 file_lock
.fl_file
= filp
;
3757 file_lock
.fl_flags
= FL_POSIX
;
3758 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3759 file_lock
.fl_start
= locku
->lu_offset
;
3761 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3762 nfs4_transform_lock_offset(&file_lock
);
3765 * Try to unlock the file in the VFS.
3767 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3769 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3773 * OK, unlock succeeded; the only thing left to do is update the stateid.
3775 update_stateid(&stp
->st_stateid
);
3776 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3779 if (locku
->lu_stateowner
) {
3780 nfs4_get_stateowner(locku
->lu_stateowner
);
3781 cstate
->replay_owner
= locku
->lu_stateowner
;
3783 nfs4_unlock_state();
3787 status
= nfserrno(err
);
3793 * 1: locks held by lockowner
3794 * 0: no locks held by lockowner
3797 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
3799 struct file_lock
**flpp
;
3800 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
3804 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3805 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3816 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3817 struct nfsd4_compound_state
*cstate
,
3818 struct nfsd4_release_lockowner
*rlockowner
)
3820 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3821 struct nfs4_stateowner
*sop
;
3822 struct nfs4_stateid
*stp
;
3823 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3824 struct list_head matches
;
3828 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3829 clid
->cl_boot
, clid
->cl_id
);
3831 /* XXX check for lease expiration */
3833 status
= nfserr_stale_clientid
;
3834 if (STALE_CLIENTID(clid
))
3839 status
= nfserr_locks_held
;
3840 /* XXX: we're doing a linear search through all the lockowners.
3841 * Yipes! For now we'll just hope clients aren't really using
3842 * release_lockowner much, but eventually we have to fix these
3843 * data structures. */
3844 INIT_LIST_HEAD(&matches
);
3845 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3846 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3847 if (!same_owner_str(sop
, owner
, clid
))
3849 list_for_each_entry(stp
, &sop
->so_stateids
,
3851 if (check_for_locks(stp
->st_vfs_file
, sop
))
3853 /* Note: so_perclient unused for lockowners,
3854 * so it's OK to fool with here. */
3855 list_add(&sop
->so_perclient
, &matches
);
3859 /* Clients probably won't expect us to return with some (but not all)
3860 * of the lockowner state released; so don't release any until all
3861 * have been checked. */
3863 while (!list_empty(&matches
)) {
3864 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3866 /* unhash_stateowner deletes so_perclient only
3867 * for openowners. */
3868 list_del(&sop
->so_perclient
);
3869 release_lockowner(sop
);
3872 nfs4_unlock_state();
3876 static inline struct nfs4_client_reclaim
*
3879 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3883 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3885 unsigned int strhashval
= clientstr_hashval(name
);
3886 struct nfs4_client
*clp
;
3888 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3893 * failure => all reset bets are off, nfserr_no_grace...
3896 nfs4_client_to_reclaim(const char *name
)
3898 unsigned int strhashval
;
3899 struct nfs4_client_reclaim
*crp
= NULL
;
3901 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3902 crp
= alloc_reclaim();
3905 strhashval
= clientstr_hashval(name
);
3906 INIT_LIST_HEAD(&crp
->cr_strhash
);
3907 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3908 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3909 reclaim_str_hashtbl_size
++;
3914 nfs4_release_reclaim(void)
3916 struct nfs4_client_reclaim
*crp
= NULL
;
3919 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3920 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3921 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3922 struct nfs4_client_reclaim
, cr_strhash
);
3923 list_del(&crp
->cr_strhash
);
3925 reclaim_str_hashtbl_size
--;
3928 BUG_ON(reclaim_str_hashtbl_size
);
3932 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3933 static struct nfs4_client_reclaim
*
3934 nfs4_find_reclaim_client(clientid_t
*clid
)
3936 unsigned int strhashval
;
3937 struct nfs4_client
*clp
;
3938 struct nfs4_client_reclaim
*crp
= NULL
;
3941 /* find clientid in conf_id_hashtbl */
3942 clp
= find_confirmed_client(clid
);
3946 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3947 clp
->cl_name
.len
, clp
->cl_name
.data
,
3950 /* find clp->cl_name in reclaim_str_hashtbl */
3951 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3952 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3953 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3961 * Called from OPEN. Look for clientid in reclaim list.
3964 nfs4_check_open_reclaim(clientid_t
*clid
)
3966 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3969 /* initialization to perform at module load time: */
3972 nfs4_state_init(void)
3976 status
= nfsd4_init_slabs();
3979 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3980 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
3981 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
3982 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
3983 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
3984 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
3986 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
3987 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
3988 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3989 INIT_LIST_HEAD(&file_hashtbl
[i
]);
3991 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
3992 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
3993 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
3995 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
3996 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
3997 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
3999 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4000 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4001 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4003 memset(&onestateid
, ~0, sizeof(stateid_t
));
4004 INIT_LIST_HEAD(&close_lru
);
4005 INIT_LIST_HEAD(&client_lru
);
4006 INIT_LIST_HEAD(&del_recall_lru
);
4007 reclaim_str_hashtbl_size
= 0;
4012 nfsd4_load_reboot_recovery_data(void)
4017 nfsd4_init_recdir(user_recovery_dirname
);
4018 status
= nfsd4_recdir_load();
4019 nfs4_unlock_state();
4021 printk("NFSD: Failure reading reboot recovery data\n");
4025 * Since the lifetime of a delegation isn't limited to that of an open, a
4026 * client may quite reasonably hang on to a delegation as long as it has
4027 * the inode cached. This becomes an obvious problem the first time a
4028 * client's inode cache approaches the size of the server's total memory.
4030 * For now we avoid this problem by imposing a hard limit on the number
4031 * of delegations, which varies according to the server's memory size.
4034 set_max_delegations(void)
4037 * Allow at most 4 delegations per megabyte of RAM. Quick
4038 * estimates suggest that in the worst case (where every delegation
4039 * is for a different inode), a delegation could take about 1.5K,
4040 * giving a worst case usage of about 6% of memory.
4042 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4045 /* initialization to perform when the nfsd service is started: */
4048 __nfs4_state_start(void)
4052 boot_time
= get_seconds();
4053 locks_start_grace(&nfsd4_manager
);
4054 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4056 ret
= set_callback_cred();
4059 laundry_wq
= create_singlethread_workqueue("nfsd4");
4060 if (laundry_wq
== NULL
)
4062 ret
= nfsd4_create_callback_queue();
4064 goto out_free_laundry
;
4065 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4066 set_max_delegations();
4069 destroy_workqueue(laundry_wq
);
4074 nfs4_state_start(void)
4080 nfsd4_load_reboot_recovery_data();
4081 ret
= __nfs4_state_start();
4089 __nfs4_state_shutdown(void)
4092 struct nfs4_client
*clp
= NULL
;
4093 struct nfs4_delegation
*dp
= NULL
;
4094 struct list_head
*pos
, *next
, reaplist
;
4096 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4097 while (!list_empty(&conf_id_hashtbl
[i
])) {
4098 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4101 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4102 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4106 INIT_LIST_HEAD(&reaplist
);
4107 spin_lock(&recall_lock
);
4108 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4109 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4110 list_move(&dp
->dl_recall_lru
, &reaplist
);
4112 spin_unlock(&recall_lock
);
4113 list_for_each_safe(pos
, next
, &reaplist
) {
4114 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4115 list_del_init(&dp
->dl_recall_lru
);
4116 unhash_delegation(dp
);
4119 nfsd4_shutdown_recdir();
4124 nfs4_state_shutdown(void)
4126 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4127 destroy_workqueue(laundry_wq
);
4128 locks_end_grace(&nfsd4_manager
);
4130 nfs4_release_reclaim();
4131 __nfs4_state_shutdown();
4132 nfsd4_destroy_callback_queue();
4133 nfs4_unlock_state();
4137 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4138 * accessed when nfsd is starting.
4141 nfs4_set_recdir(char *recdir
)
4143 strcpy(user_recovery_dirname
, recdir
);
4147 * Change the NFSv4 recovery directory to recdir.
4150 nfs4_reset_recoverydir(char *recdir
)
4155 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4159 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4160 nfs4_set_recdir(recdir
);
4168 nfs4_recoverydir(void)
4170 return user_recovery_dirname
;