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>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
46 #define NFSDDBG_FACILITY NFSDDBG_PROC
49 time_t nfsd4_lease
= 90; /* default lease time */
50 time_t nfsd4_grace
= 90;
51 static time_t boot_time
;
52 static u32 current_ownerid
= 1;
53 static u32 current_fileid
= 1;
54 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
* search_for_delegation(stateid_t
*stid
);
65 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
66 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_stateowner
*lowner
);
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 open owners */
137 #define OPEN_OWNER_HASH_BITS 8
138 #define OPEN_OWNER_HASH_SIZE (1 << OPEN_OWNER_HASH_BITS)
139 #define OPEN_OWNER_HASH_MASK (OPEN_OWNER_HASH_SIZE - 1)
141 static unsigned int open_ownerid_hashval(const u32 id
)
143 return id
& OPEN_OWNER_HASH_MASK
;
146 static unsigned int open_ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
150 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
152 return ret
& OPEN_OWNER_HASH_MASK
;
155 static struct list_head open_ownerid_hashtbl
[OPEN_OWNER_HASH_SIZE
];
156 static struct list_head open_ownerstr_hashtbl
[OPEN_OWNER_HASH_SIZE
];
158 /* hash table for nfs4_file */
159 #define FILE_HASH_BITS 8
160 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
162 /* hash table for (open)nfs4_stateid */
163 #define STATEID_HASH_BITS 10
164 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
165 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
167 static unsigned int file_hashval(struct inode
*ino
)
169 /* XXX: why are we hashing on inode pointer, anyway? */
170 return hash_ptr(ino
, FILE_HASH_BITS
);
173 static unsigned int stateid_hashval(u32 owner_id
, u32 file_id
)
175 return (owner_id
+ file_id
) & STATEID_HASH_MASK
;
178 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
179 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
181 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
183 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
184 atomic_inc(&fp
->fi_access
[oflag
]);
187 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
189 if (oflag
== O_RDWR
) {
190 __nfs4_file_get_access(fp
, O_RDONLY
);
191 __nfs4_file_get_access(fp
, O_WRONLY
);
193 __nfs4_file_get_access(fp
, oflag
);
196 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
198 if (fp
->fi_fds
[oflag
]) {
199 fput(fp
->fi_fds
[oflag
]);
200 fp
->fi_fds
[oflag
] = NULL
;
204 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
206 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
207 nfs4_file_put_fd(fp
, O_RDWR
);
208 nfs4_file_put_fd(fp
, oflag
);
212 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
214 if (oflag
== O_RDWR
) {
215 __nfs4_file_put_access(fp
, O_RDONLY
);
216 __nfs4_file_put_access(fp
, O_WRONLY
);
218 __nfs4_file_put_access(fp
, oflag
);
221 static struct nfs4_delegation
*
222 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
224 struct nfs4_delegation
*dp
;
225 struct nfs4_file
*fp
= stp
->st_file
;
227 dprintk("NFSD alloc_init_deleg\n");
229 * Major work on the lease subsystem (for example, to support
230 * calbacks on stat) will be required before we can support
231 * write delegations properly.
233 if (type
!= NFS4_OPEN_DELEGATE_READ
)
235 if (fp
->fi_had_conflict
)
237 if (num_delegations
> max_delegations
)
239 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
243 INIT_LIST_HEAD(&dp
->dl_perfile
);
244 INIT_LIST_HEAD(&dp
->dl_perclnt
);
245 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
250 dp
->dl_stateid
.si_boot
= boot_time
;
251 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
252 dp
->dl_stateid
.si_fileid
= 0;
253 dp
->dl_stateid
.si_generation
= 1;
254 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
256 atomic_set(&dp
->dl_count
, 1);
257 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
262 nfs4_put_delegation(struct nfs4_delegation
*dp
)
264 if (atomic_dec_and_test(&dp
->dl_count
)) {
265 dprintk("NFSD: freeing dp %p\n",dp
);
266 put_nfs4_file(dp
->dl_file
);
267 kmem_cache_free(deleg_slab
, dp
);
272 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
274 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
275 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
277 fput(fp
->fi_deleg_file
);
278 fp
->fi_deleg_file
= NULL
;
282 /* Called under the state lock. */
284 unhash_delegation(struct nfs4_delegation
*dp
)
286 list_del_init(&dp
->dl_perclnt
);
287 spin_lock(&recall_lock
);
288 list_del_init(&dp
->dl_perfile
);
289 list_del_init(&dp
->dl_recall_lru
);
290 spin_unlock(&recall_lock
);
291 nfs4_put_deleg_lease(dp
->dl_file
);
292 nfs4_put_delegation(dp
);
299 /* client_lock protects the client lru list and session hash table */
300 static DEFINE_SPINLOCK(client_lock
);
302 /* Hash tables for nfs4_clientid state */
303 #define CLIENT_HASH_BITS 4
304 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
305 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
307 static unsigned int clientid_hashval(u32 id
)
309 return id
& CLIENT_HASH_MASK
;
312 static unsigned int clientstr_hashval(const char *name
)
314 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
318 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
319 * used in reboot/reset lease grace period processing
321 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
322 * setclientid_confirmed info.
324 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
327 * client_lru holds client queue ordered by nfs4_client.cl_time
330 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
331 * for last close replay.
333 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
334 static int reclaim_str_hashtbl_size
= 0;
335 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
336 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
337 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
338 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
339 static struct list_head client_lru
;
340 static struct list_head close_lru
;
343 * We store the NONE, READ, WRITE, and BOTH bits separately in the
344 * st_{access,deny}_bmap field of the stateid, in order to track not
345 * only what share bits are currently in force, but also what
346 * combinations of share bits previous opens have used. This allows us
347 * to enforce the recommendation of rfc 3530 14.2.19 that the server
348 * return an error if the client attempt to downgrade to a combination
349 * of share bits not explicable by closing some of its previous opens.
351 * XXX: This enforcement is actually incomplete, since we don't keep
352 * track of access/deny bit combinations; so, e.g., we allow:
354 * OPEN allow read, deny write
355 * OPEN allow both, deny none
356 * DOWNGRADE allow read, deny none
358 * which we should reject.
361 set_access(unsigned int *access
, unsigned long bmap
) {
365 for (i
= 1; i
< 4; i
++) {
366 if (test_bit(i
, &bmap
))
372 set_deny(unsigned int *deny
, unsigned long bmap
) {
376 for (i
= 0; i
< 4; i
++) {
377 if (test_bit(i
, &bmap
))
383 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
384 unsigned int access
, deny
;
386 set_access(&access
, stp
->st_access_bmap
);
387 set_deny(&deny
, stp
->st_deny_bmap
);
388 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
393 static int nfs4_access_to_omode(u32 access
)
395 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
396 case NFS4_SHARE_ACCESS_READ
:
398 case NFS4_SHARE_ACCESS_WRITE
:
400 case NFS4_SHARE_ACCESS_BOTH
:
406 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
408 list_del(&stp
->st_hash
);
409 list_del(&stp
->st_perfile
);
410 list_del(&stp
->st_perstateowner
);
413 static void free_generic_stateid(struct nfs4_stateid
*stp
)
417 if (stp
->st_access_bmap
) {
418 for (i
= 1; i
< 4; i
++) {
419 if (test_bit(i
, &stp
->st_access_bmap
))
420 nfs4_file_put_access(stp
->st_file
,
421 nfs4_access_to_omode(i
));
424 put_nfs4_file(stp
->st_file
);
425 kmem_cache_free(stateid_slab
, stp
);
428 static void release_lock_stateid(struct nfs4_stateid
*stp
)
432 unhash_generic_stateid(stp
);
433 file
= find_any_file(stp
->st_file
);
435 locks_remove_posix(file
, (fl_owner_t
)stp
->st_stateowner
);
436 free_generic_stateid(stp
);
439 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
441 struct nfs4_stateid
*stp
;
443 list_del(&sop
->so_idhash
);
444 list_del(&sop
->so_strhash
);
445 list_del(&sop
->so_perstateid
);
446 while (!list_empty(&sop
->so_stateids
)) {
447 stp
= list_first_entry(&sop
->so_stateids
,
448 struct nfs4_stateid
, st_perstateowner
);
449 release_lock_stateid(stp
);
453 static void release_lockowner(struct nfs4_stateowner
*sop
)
455 unhash_lockowner(sop
);
456 nfs4_free_stateowner(sop
);
460 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
462 struct nfs4_stateowner
*lock_sop
;
464 while (!list_empty(&open_stp
->st_lockowners
)) {
465 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
466 struct nfs4_stateowner
, so_perstateid
);
467 /* list_del(&open_stp->st_lockowners); */
468 BUG_ON(lock_sop
->so_is_open_owner
);
469 release_lockowner(lock_sop
);
473 static void release_open_stateid(struct nfs4_stateid
*stp
)
475 unhash_generic_stateid(stp
);
476 release_stateid_lockowners(stp
);
477 free_generic_stateid(stp
);
480 static void unhash_openowner(struct nfs4_stateowner
*sop
)
482 struct nfs4_stateid
*stp
;
484 list_del(&sop
->so_idhash
);
485 list_del(&sop
->so_strhash
);
486 list_del(&sop
->so_perclient
);
487 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
488 while (!list_empty(&sop
->so_stateids
)) {
489 stp
= list_first_entry(&sop
->so_stateids
,
490 struct nfs4_stateid
, st_perstateowner
);
491 release_open_stateid(stp
);
495 static void release_openowner(struct nfs4_stateowner
*sop
)
497 unhash_openowner(sop
);
498 list_del(&sop
->so_close_lru
);
499 nfs4_free_stateowner(sop
);
502 #define SESSION_HASH_SIZE 512
503 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
506 hash_sessionid(struct nfs4_sessionid
*sessionid
)
508 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
510 return sid
->sequence
% SESSION_HASH_SIZE
;
514 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
516 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
517 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
521 gen_sessionid(struct nfsd4_session
*ses
)
523 struct nfs4_client
*clp
= ses
->se_client
;
524 struct nfsd4_sessionid
*sid
;
526 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
527 sid
->clientid
= clp
->cl_clientid
;
528 sid
->sequence
= current_sessionid
++;
533 * The protocol defines ca_maxresponssize_cached to include the size of
534 * the rpc header, but all we need to cache is the data starting after
535 * the end of the initial SEQUENCE operation--the rest we regenerate
536 * each time. Therefore we can advertise a ca_maxresponssize_cached
537 * value that is the number of bytes in our cache plus a few additional
538 * bytes. In order to stay on the safe side, and not promise more than
539 * we can cache, those additional bytes must be the minimum possible: 24
540 * bytes of rpc header (xid through accept state, with AUTH_NULL
541 * verifier), 12 for the compound header (with zero-length tag), and 44
542 * for the SEQUENCE op response:
544 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
547 free_session_slots(struct nfsd4_session
*ses
)
551 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
552 kfree(ses
->se_slots
[i
]);
556 * We don't actually need to cache the rpc and session headers, so we
557 * can allocate a little less for each slot:
559 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
561 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
564 static int nfsd4_sanitize_slot_size(u32 size
)
566 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
567 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
573 * XXX: If we run out of reserved DRC memory we could (up to a point)
574 * re-negotiate active sessions and reduce their slot usage to make
575 * rooom for new connections. For now we just fail the create session.
577 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
581 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
583 spin_lock(&nfsd_drc_lock
);
584 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
585 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
586 num
= min_t(int, num
, avail
/ slotsize
);
587 nfsd_drc_mem_used
+= num
* slotsize
;
588 spin_unlock(&nfsd_drc_lock
);
593 static void nfsd4_put_drc_mem(int slotsize
, int num
)
595 spin_lock(&nfsd_drc_lock
);
596 nfsd_drc_mem_used
-= slotsize
* num
;
597 spin_unlock(&nfsd_drc_lock
);
600 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
602 struct nfsd4_session
*new;
605 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
606 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
607 mem
= numslots
* sizeof(struct nfsd4_slot
*);
609 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
612 /* allocate each struct nfsd4_slot and data cache in one piece */
613 for (i
= 0; i
< numslots
; i
++) {
614 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
615 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
616 if (!new->se_slots
[i
])
622 kfree(new->se_slots
[i
]);
627 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
629 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
631 new->maxreqs
= numslots
;
632 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
633 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
634 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
635 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
636 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
639 static void free_conn(struct nfsd4_conn
*c
)
641 svc_xprt_put(c
->cn_xprt
);
645 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
647 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
648 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
650 spin_lock(&clp
->cl_lock
);
651 if (!list_empty(&c
->cn_persession
)) {
652 list_del(&c
->cn_persession
);
655 spin_unlock(&clp
->cl_lock
);
656 nfsd4_probe_callback(clp
);
659 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
661 struct nfsd4_conn
*conn
;
663 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
666 svc_xprt_get(rqstp
->rq_xprt
);
667 conn
->cn_xprt
= rqstp
->rq_xprt
;
668 conn
->cn_flags
= flags
;
669 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
673 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
675 conn
->cn_session
= ses
;
676 list_add(&conn
->cn_persession
, &ses
->se_conns
);
679 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
681 struct nfs4_client
*clp
= ses
->se_client
;
683 spin_lock(&clp
->cl_lock
);
684 __nfsd4_hash_conn(conn
, ses
);
685 spin_unlock(&clp
->cl_lock
);
688 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
690 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
691 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
694 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
696 struct nfsd4_conn
*conn
;
699 conn
= alloc_conn(rqstp
, dir
);
701 return nfserr_jukebox
;
702 nfsd4_hash_conn(conn
, ses
);
703 ret
= nfsd4_register_conn(conn
);
705 /* oops; xprt is already down: */
706 nfsd4_conn_lost(&conn
->cn_xpt_user
);
710 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
712 u32 dir
= NFS4_CDFC4_FORE
;
714 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
715 dir
|= NFS4_CDFC4_BACK
;
717 return nfsd4_new_conn(rqstp
, ses
, dir
);
720 /* must be called under client_lock */
721 static void nfsd4_del_conns(struct nfsd4_session
*s
)
723 struct nfs4_client
*clp
= s
->se_client
;
724 struct nfsd4_conn
*c
;
726 spin_lock(&clp
->cl_lock
);
727 while (!list_empty(&s
->se_conns
)) {
728 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
729 list_del_init(&c
->cn_persession
);
730 spin_unlock(&clp
->cl_lock
);
732 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
735 spin_lock(&clp
->cl_lock
);
737 spin_unlock(&clp
->cl_lock
);
740 void free_session(struct kref
*kref
)
742 struct nfsd4_session
*ses
;
745 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
746 nfsd4_del_conns(ses
);
747 spin_lock(&nfsd_drc_lock
);
748 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
749 nfsd_drc_mem_used
-= mem
;
750 spin_unlock(&nfsd_drc_lock
);
751 free_session_slots(ses
);
755 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
757 struct nfsd4_session
*new;
758 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
759 int numslots
, slotsize
;
764 * Note decreasing slot size below client's request may
765 * make it difficult for client to function correctly, whereas
766 * decreasing the number of slots will (just?) affect
767 * performance. When short on memory we therefore prefer to
768 * decrease number of slots instead of their size.
770 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
771 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
775 new = alloc_session(slotsize
, numslots
);
777 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
780 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
782 new->se_client
= clp
;
785 INIT_LIST_HEAD(&new->se_conns
);
787 new->se_cb_seq_nr
= 1;
788 new->se_flags
= cses
->flags
;
789 new->se_cb_prog
= cses
->callback_prog
;
790 kref_init(&new->se_ref
);
791 idx
= hash_sessionid(&new->se_sessionid
);
792 spin_lock(&client_lock
);
793 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
794 spin_lock(&clp
->cl_lock
);
795 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
796 spin_unlock(&clp
->cl_lock
);
797 spin_unlock(&client_lock
);
799 status
= nfsd4_new_conn_from_crses(rqstp
, new);
800 /* whoops: benny points out, status is ignored! (err, or bogus) */
802 free_session(&new->se_ref
);
805 if (cses
->flags
& SESSION4_BACK_CHAN
) {
806 struct sockaddr
*sa
= svc_addr(rqstp
);
808 * This is a little silly; with sessions there's no real
809 * use for the callback address. Use the peer address
810 * as a reasonable default for now, but consider fixing
811 * the rpc client not to require an address in the
814 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
815 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
817 nfsd4_probe_callback(clp
);
821 /* caller must hold client_lock */
822 static struct nfsd4_session
*
823 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
825 struct nfsd4_session
*elem
;
828 dump_sessionid(__func__
, sessionid
);
829 idx
= hash_sessionid(sessionid
);
830 /* Search in the appropriate list */
831 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
832 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
833 NFS4_MAX_SESSIONID_LEN
)) {
838 dprintk("%s: session not found\n", __func__
);
842 /* caller must hold client_lock */
844 unhash_session(struct nfsd4_session
*ses
)
846 list_del(&ses
->se_hash
);
847 spin_lock(&ses
->se_client
->cl_lock
);
848 list_del(&ses
->se_perclnt
);
849 spin_unlock(&ses
->se_client
->cl_lock
);
852 /* must be called under the client_lock */
854 renew_client_locked(struct nfs4_client
*clp
)
856 if (is_client_expired(clp
)) {
857 dprintk("%s: client (clientid %08x/%08x) already expired\n",
859 clp
->cl_clientid
.cl_boot
,
860 clp
->cl_clientid
.cl_id
);
865 * Move client to the end to the LRU list.
867 dprintk("renewing client (clientid %08x/%08x)\n",
868 clp
->cl_clientid
.cl_boot
,
869 clp
->cl_clientid
.cl_id
);
870 list_move_tail(&clp
->cl_lru
, &client_lru
);
871 clp
->cl_time
= get_seconds();
875 renew_client(struct nfs4_client
*clp
)
877 spin_lock(&client_lock
);
878 renew_client_locked(clp
);
879 spin_unlock(&client_lock
);
882 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
884 STALE_CLIENTID(clientid_t
*clid
)
886 if (clid
->cl_boot
== boot_time
)
888 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
889 clid
->cl_boot
, clid
->cl_id
, boot_time
);
894 * XXX Should we use a slab cache ?
895 * This type of memory management is somewhat inefficient, but we use it
896 * anyway since SETCLIENTID is not a common operation.
898 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
900 struct nfs4_client
*clp
;
902 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
905 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
906 if (clp
->cl_name
.data
== NULL
) {
910 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
911 clp
->cl_name
.len
= name
.len
;
916 free_client(struct nfs4_client
*clp
)
918 while (!list_empty(&clp
->cl_sessions
)) {
919 struct nfsd4_session
*ses
;
920 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
922 list_del(&ses
->se_perclnt
);
923 nfsd4_put_session(ses
);
925 if (clp
->cl_cred
.cr_group_info
)
926 put_group_info(clp
->cl_cred
.cr_group_info
);
927 kfree(clp
->cl_principal
);
928 kfree(clp
->cl_name
.data
);
933 release_session_client(struct nfsd4_session
*session
)
935 struct nfs4_client
*clp
= session
->se_client
;
937 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
939 if (is_client_expired(clp
)) {
941 session
->se_client
= NULL
;
943 renew_client_locked(clp
);
944 spin_unlock(&client_lock
);
947 /* must be called under the client_lock */
949 unhash_client_locked(struct nfs4_client
*clp
)
951 struct nfsd4_session
*ses
;
953 mark_client_expired(clp
);
954 list_del(&clp
->cl_lru
);
955 spin_lock(&clp
->cl_lock
);
956 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
957 list_del_init(&ses
->se_hash
);
958 spin_unlock(&clp
->cl_lock
);
962 expire_client(struct nfs4_client
*clp
)
964 struct nfs4_stateowner
*sop
;
965 struct nfs4_delegation
*dp
;
966 struct list_head reaplist
;
968 INIT_LIST_HEAD(&reaplist
);
969 spin_lock(&recall_lock
);
970 while (!list_empty(&clp
->cl_delegations
)) {
971 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
972 list_del_init(&dp
->dl_perclnt
);
973 list_move(&dp
->dl_recall_lru
, &reaplist
);
975 spin_unlock(&recall_lock
);
976 while (!list_empty(&reaplist
)) {
977 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
978 list_del_init(&dp
->dl_recall_lru
);
979 unhash_delegation(dp
);
981 while (!list_empty(&clp
->cl_openowners
)) {
982 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
983 release_openowner(sop
);
985 nfsd4_shutdown_callback(clp
);
986 if (clp
->cl_cb_conn
.cb_xprt
)
987 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
988 list_del(&clp
->cl_idhash
);
989 list_del(&clp
->cl_strhash
);
990 spin_lock(&client_lock
);
991 unhash_client_locked(clp
);
992 if (atomic_read(&clp
->cl_refcount
) == 0)
994 spin_unlock(&client_lock
);
997 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
999 memcpy(target
->cl_verifier
.data
, source
->data
,
1000 sizeof(target
->cl_verifier
.data
));
1003 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1005 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1006 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1009 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1011 target
->cr_uid
= source
->cr_uid
;
1012 target
->cr_gid
= source
->cr_gid
;
1013 target
->cr_group_info
= source
->cr_group_info
;
1014 get_group_info(target
->cr_group_info
);
1017 static int same_name(const char *n1
, const char *n2
)
1019 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1023 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1025 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1029 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1031 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1034 /* XXX what about NGROUP */
1036 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1038 return cr1
->cr_uid
== cr2
->cr_uid
;
1041 static void gen_clid(struct nfs4_client
*clp
)
1043 static u32 current_clientid
= 1;
1045 clp
->cl_clientid
.cl_boot
= boot_time
;
1046 clp
->cl_clientid
.cl_id
= current_clientid
++;
1049 static void gen_confirm(struct nfs4_client
*clp
)
1054 p
= (u32
*)clp
->cl_confirm
.data
;
1055 *p
++ = get_seconds();
1059 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1060 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1062 struct nfs4_client
*clp
;
1063 struct sockaddr
*sa
= svc_addr(rqstp
);
1066 clp
= alloc_client(name
);
1070 INIT_LIST_HEAD(&clp
->cl_sessions
);
1072 princ
= svc_gss_principal(rqstp
);
1074 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1075 if (clp
->cl_principal
== NULL
) {
1081 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1082 atomic_set(&clp
->cl_refcount
, 0);
1083 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1084 INIT_LIST_HEAD(&clp
->cl_idhash
);
1085 INIT_LIST_HEAD(&clp
->cl_strhash
);
1086 INIT_LIST_HEAD(&clp
->cl_openowners
);
1087 INIT_LIST_HEAD(&clp
->cl_delegations
);
1088 INIT_LIST_HEAD(&clp
->cl_lru
);
1089 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1090 spin_lock_init(&clp
->cl_lock
);
1091 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1092 clp
->cl_time
= get_seconds();
1093 clear_bit(0, &clp
->cl_cb_slot_busy
);
1094 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1095 copy_verf(clp
, verf
);
1096 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1097 clp
->cl_flavor
= rqstp
->rq_flavor
;
1098 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1100 clp
->cl_cb_session
= NULL
;
1104 static int check_name(struct xdr_netobj name
)
1108 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
1109 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
1116 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1118 unsigned int idhashval
;
1120 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1121 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1122 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1127 move_to_confirmed(struct nfs4_client
*clp
)
1129 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1130 unsigned int strhashval
;
1132 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1133 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1134 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1135 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1139 static struct nfs4_client
*
1140 find_confirmed_client(clientid_t
*clid
)
1142 struct nfs4_client
*clp
;
1143 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1145 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1146 if (same_clid(&clp
->cl_clientid
, clid
))
1152 static struct nfs4_client
*
1153 find_unconfirmed_client(clientid_t
*clid
)
1155 struct nfs4_client
*clp
;
1156 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1158 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1159 if (same_clid(&clp
->cl_clientid
, clid
))
1165 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1167 return clp
->cl_exchange_flags
!= 0;
1170 static struct nfs4_client
*
1171 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1173 struct nfs4_client
*clp
;
1175 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1176 if (same_name(clp
->cl_recdir
, dname
))
1182 static struct nfs4_client
*
1183 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1185 struct nfs4_client
*clp
;
1187 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1188 if (same_name(clp
->cl_recdir
, dname
))
1194 static void rpc_svcaddr2sockaddr(struct sockaddr
*sa
, unsigned short family
, union svc_addr_u
*svcaddr
)
1198 ((struct sockaddr_in
*)sa
)->sin_family
= AF_INET
;
1199 ((struct sockaddr_in
*)sa
)->sin_addr
= svcaddr
->addr
;
1202 ((struct sockaddr_in6
*)sa
)->sin6_family
= AF_INET6
;
1203 ((struct sockaddr_in6
*)sa
)->sin6_addr
= svcaddr
->addr6
;
1209 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1211 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1212 struct sockaddr
*sa
= svc_addr(rqstp
);
1213 u32 scopeid
= rpc_get_scope_id(sa
);
1214 unsigned short expected_family
;
1216 /* Currently, we only support tcp and tcp6 for the callback channel */
1217 if (se
->se_callback_netid_len
== 3 &&
1218 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1219 expected_family
= AF_INET
;
1220 else if (se
->se_callback_netid_len
== 4 &&
1221 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1222 expected_family
= AF_INET6
;
1226 conn
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
1227 se
->se_callback_addr_len
,
1228 (struct sockaddr
*)&conn
->cb_addr
,
1229 sizeof(conn
->cb_addr
));
1231 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1234 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1235 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1237 conn
->cb_prog
= se
->se_callback_prog
;
1238 conn
->cb_ident
= se
->se_callback_ident
;
1239 rpc_svcaddr2sockaddr((struct sockaddr
*)&conn
->cb_saddr
, expected_family
, &rqstp
->rq_daddr
);
1242 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1243 conn
->cb_addrlen
= 0;
1244 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1245 "will not receive delegations\n",
1246 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1252 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1255 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1257 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1260 dprintk("--> %s slot %p\n", __func__
, slot
);
1262 slot
->sl_opcnt
= resp
->opcnt
;
1263 slot
->sl_status
= resp
->cstate
.status
;
1265 if (nfsd4_not_cached(resp
)) {
1266 slot
->sl_datalen
= 0;
1269 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1270 base
= (char *)resp
->cstate
.datap
-
1271 (char *)resp
->xbuf
->head
[0].iov_base
;
1272 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1274 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1279 * Encode the replay sequence operation from the slot values.
1280 * If cachethis is FALSE encode the uncached rep error on the next
1281 * operation which sets resp->p and increments resp->opcnt for
1282 * nfs4svc_encode_compoundres.
1286 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1287 struct nfsd4_compoundres
*resp
)
1289 struct nfsd4_op
*op
;
1290 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1292 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1293 resp
->opcnt
, resp
->cstate
.slot
->sl_cachethis
);
1295 /* Encode the replayed sequence operation */
1296 op
= &args
->ops
[resp
->opcnt
- 1];
1297 nfsd4_encode_operation(resp
, op
);
1299 /* Return nfserr_retry_uncached_rep in next operation. */
1300 if (args
->opcnt
> 1 && slot
->sl_cachethis
== 0) {
1301 op
= &args
->ops
[resp
->opcnt
++];
1302 op
->status
= nfserr_retry_uncached_rep
;
1303 nfsd4_encode_operation(resp
, op
);
1309 * The sequence operation is not cached because we can use the slot and
1313 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1314 struct nfsd4_sequence
*seq
)
1316 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1319 dprintk("--> %s slot %p\n", __func__
, slot
);
1321 /* Either returns 0 or nfserr_retry_uncached */
1322 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1323 if (status
== nfserr_retry_uncached_rep
)
1326 /* The sequence operation has been encoded, cstate->datap set. */
1327 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1329 resp
->opcnt
= slot
->sl_opcnt
;
1330 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1331 status
= slot
->sl_status
;
1337 * Set the exchange_id flags returned by the server.
1340 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1342 /* pNFS is not supported */
1343 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1345 /* Referrals are supported, Migration is not. */
1346 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1348 /* set the wire flags to return to client. */
1349 clid
->flags
= new->cl_exchange_flags
;
1353 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1354 struct nfsd4_compound_state
*cstate
,
1355 struct nfsd4_exchange_id
*exid
)
1357 struct nfs4_client
*unconf
, *conf
, *new;
1359 unsigned int strhashval
;
1360 char dname
[HEXDIR_LEN
];
1361 char addr_str
[INET6_ADDRSTRLEN
];
1362 nfs4_verifier verf
= exid
->verifier
;
1363 struct sockaddr
*sa
= svc_addr(rqstp
);
1365 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1366 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1367 "ip_addr=%s flags %x, spa_how %d\n",
1368 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1369 addr_str
, exid
->flags
, exid
->spa_how
);
1371 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1372 return nfserr_inval
;
1374 /* Currently only support SP4_NONE */
1375 switch (exid
->spa_how
) {
1379 return nfserr_serverfault
;
1381 BUG(); /* checked by xdr code */
1383 return nfserr_serverfault
; /* no excuse :-/ */
1386 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1391 strhashval
= clientstr_hashval(dname
);
1396 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1398 if (!clp_used_exchangeid(conf
)) {
1399 status
= nfserr_clid_inuse
; /* XXX: ? */
1402 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1403 /* 18.35.4 case 8 */
1404 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1405 status
= nfserr_not_same
;
1408 /* Client reboot: destroy old state */
1409 expire_client(conf
);
1412 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1413 /* 18.35.4 case 9 */
1414 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1415 status
= nfserr_perm
;
1418 expire_client(conf
);
1422 * Set bit when the owner id and verifier map to an already
1423 * confirmed client id (18.35.3).
1425 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1428 * Falling into 18.35.4 case 2, possible router replay.
1429 * Leave confirmed record intact and return same result.
1431 copy_verf(conf
, &verf
);
1436 /* 18.35.4 case 7 */
1437 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1438 status
= nfserr_noent
;
1442 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1445 * Possible retry or client restart. Per 18.35.4 case 4,
1446 * a new unconfirmed record should be generated regardless
1447 * of whether any properties have changed.
1449 expire_client(unconf
);
1454 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1456 status
= nfserr_jukebox
;
1461 add_to_unconfirmed(new, strhashval
);
1463 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1464 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1467 nfsd4_set_ex_flags(new, exid
);
1469 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1470 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1474 nfs4_unlock_state();
1476 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1481 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1483 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1486 /* The slot is in use, and no response has been sent. */
1488 if (seqid
== slot_seqid
)
1489 return nfserr_jukebox
;
1491 return nfserr_seq_misordered
;
1494 if (likely(seqid
== slot_seqid
+ 1))
1497 if (seqid
== slot_seqid
)
1498 return nfserr_replay_cache
;
1500 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1502 /* Misordered replay or misordered new request */
1503 return nfserr_seq_misordered
;
1507 * Cache the create session result into the create session single DRC
1508 * slot cache by saving the xdr structure. sl_seqid has been set.
1509 * Do this for solo or embedded create session operations.
1512 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1513 struct nfsd4_clid_slot
*slot
, int nfserr
)
1515 slot
->sl_status
= nfserr
;
1516 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1520 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1521 struct nfsd4_clid_slot
*slot
)
1523 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1524 return slot
->sl_status
;
1527 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1528 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1529 1 + /* MIN tag is length with zero, only length */ \
1530 3 + /* version, opcount, opcode */ \
1531 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1532 /* seqid, slotID, slotID, cache */ \
1533 4 ) * sizeof(__be32))
1535 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1536 2 + /* verifier: AUTH_NULL, length 0 */\
1538 1 + /* MIN tag is length with zero, only length */ \
1539 3 + /* opcount, opcode, opstatus*/ \
1540 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1541 /* seqid, slotID, slotID, slotID, status */ \
1542 5 ) * sizeof(__be32))
1544 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1546 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1547 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1551 nfsd4_create_session(struct svc_rqst
*rqstp
,
1552 struct nfsd4_compound_state
*cstate
,
1553 struct nfsd4_create_session
*cr_ses
)
1555 struct sockaddr
*sa
= svc_addr(rqstp
);
1556 struct nfs4_client
*conf
, *unconf
;
1557 struct nfsd4_session
*new;
1558 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1559 bool confirm_me
= false;
1562 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1563 return nfserr_inval
;
1566 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1567 conf
= find_confirmed_client(&cr_ses
->clientid
);
1570 cs_slot
= &conf
->cl_cs_slot
;
1571 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1572 if (status
== nfserr_replay_cache
) {
1573 dprintk("Got a create_session replay! seqid= %d\n",
1575 /* Return the cached reply status */
1576 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1578 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1579 status
= nfserr_seq_misordered
;
1580 dprintk("Sequence misordered!\n");
1581 dprintk("Expected seqid= %d but got seqid= %d\n",
1582 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1585 } else if (unconf
) {
1586 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1587 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1588 status
= nfserr_clid_inuse
;
1592 cs_slot
= &unconf
->cl_cs_slot
;
1593 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1595 /* an unconfirmed replay returns misordered */
1596 status
= nfserr_seq_misordered
;
1603 status
= nfserr_stale_clientid
;
1608 * XXX: we should probably set this at creation time, and check
1609 * for consistent minorversion use throughout:
1611 conf
->cl_minorversion
= 1;
1613 * We do not support RDMA or persistent sessions
1615 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1616 cr_ses
->flags
&= ~SESSION4_RDMA
;
1618 status
= nfserr_toosmall
;
1619 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1622 status
= nfserr_jukebox
;
1623 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1627 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1628 NFS4_MAX_SESSIONID_LEN
);
1629 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1630 sizeof(struct nfsd4_channel_attrs
));
1631 cs_slot
->sl_seqid
++;
1632 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1634 /* cache solo and embedded create sessions under the state lock */
1635 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1637 move_to_confirmed(conf
);
1639 nfs4_unlock_state();
1640 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1644 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1646 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1647 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1649 return argp
->opcnt
== resp
->opcnt
;
1652 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1655 case NFS4_CDFC4_FORE
:
1656 case NFS4_CDFC4_BACK
:
1658 case NFS4_CDFC4_FORE_OR_BOTH
:
1659 case NFS4_CDFC4_BACK_OR_BOTH
:
1660 *dir
= NFS4_CDFC4_BOTH
;
1663 return nfserr_inval
;
1666 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1667 struct nfsd4_compound_state
*cstate
,
1668 struct nfsd4_bind_conn_to_session
*bcts
)
1672 if (!nfsd4_last_compound_op(rqstp
))
1673 return nfserr_not_only_op
;
1674 spin_lock(&client_lock
);
1675 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1676 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1677 * client_lock iself: */
1678 if (cstate
->session
) {
1679 nfsd4_get_session(cstate
->session
);
1680 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1682 spin_unlock(&client_lock
);
1683 if (!cstate
->session
)
1684 return nfserr_badsession
;
1686 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1688 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1692 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1696 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1700 nfsd4_destroy_session(struct svc_rqst
*r
,
1701 struct nfsd4_compound_state
*cstate
,
1702 struct nfsd4_destroy_session
*sessionid
)
1704 struct nfsd4_session
*ses
;
1705 u32 status
= nfserr_badsession
;
1708 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1709 * - Should we return nfserr_back_chan_busy if waiting for
1710 * callbacks on to-be-destroyed session?
1711 * - Do we need to clear any callback info from previous session?
1714 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1715 if (!nfsd4_last_compound_op(r
))
1716 return nfserr_not_only_op
;
1718 dump_sessionid(__func__
, &sessionid
->sessionid
);
1719 spin_lock(&client_lock
);
1720 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1722 spin_unlock(&client_lock
);
1726 unhash_session(ses
);
1727 spin_unlock(&client_lock
);
1730 nfsd4_probe_callback_sync(ses
->se_client
);
1731 nfs4_unlock_state();
1733 nfsd4_del_conns(ses
);
1735 nfsd4_put_session(ses
);
1738 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1742 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1744 struct nfsd4_conn
*c
;
1746 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1747 if (c
->cn_xprt
== xpt
) {
1754 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1756 struct nfs4_client
*clp
= ses
->se_client
;
1757 struct nfsd4_conn
*c
;
1760 spin_lock(&clp
->cl_lock
);
1761 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1763 spin_unlock(&clp
->cl_lock
);
1767 __nfsd4_hash_conn(new, ses
);
1768 spin_unlock(&clp
->cl_lock
);
1769 ret
= nfsd4_register_conn(new);
1771 /* oops; xprt is already down: */
1772 nfsd4_conn_lost(&new->cn_xpt_user
);
1776 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1778 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1780 return args
->opcnt
> session
->se_fchannel
.maxops
;
1783 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1784 struct nfsd4_session
*session
)
1786 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1788 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1792 nfsd4_sequence(struct svc_rqst
*rqstp
,
1793 struct nfsd4_compound_state
*cstate
,
1794 struct nfsd4_sequence
*seq
)
1796 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1797 struct nfsd4_session
*session
;
1798 struct nfsd4_slot
*slot
;
1799 struct nfsd4_conn
*conn
;
1802 if (resp
->opcnt
!= 1)
1803 return nfserr_sequence_pos
;
1806 * Will be either used or freed by nfsd4_sequence_check_conn
1809 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1811 return nfserr_jukebox
;
1813 spin_lock(&client_lock
);
1814 status
= nfserr_badsession
;
1815 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1819 status
= nfserr_too_many_ops
;
1820 if (nfsd4_session_too_many_ops(rqstp
, session
))
1823 status
= nfserr_req_too_big
;
1824 if (nfsd4_request_too_big(rqstp
, session
))
1827 status
= nfserr_badslot
;
1828 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1831 slot
= session
->se_slots
[seq
->slotid
];
1832 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1834 /* We do not negotiate the number of slots yet, so set the
1835 * maxslots to the session maxreqs which is used to encode
1836 * sr_highest_slotid and the sr_target_slot id to maxslots */
1837 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1839 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1840 if (status
== nfserr_replay_cache
) {
1841 cstate
->slot
= slot
;
1842 cstate
->session
= session
;
1843 /* Return the cached reply status and set cstate->status
1844 * for nfsd4_proc_compound processing */
1845 status
= nfsd4_replay_cache_entry(resp
, seq
);
1846 cstate
->status
= nfserr_replay_cache
;
1852 nfsd4_sequence_check_conn(conn
, session
);
1855 /* Success! bump slot seqid */
1856 slot
->sl_inuse
= true;
1857 slot
->sl_seqid
= seq
->seqid
;
1858 slot
->sl_cachethis
= seq
->cachethis
;
1860 cstate
->slot
= slot
;
1861 cstate
->session
= session
;
1864 /* Hold a session reference until done processing the compound. */
1865 if (cstate
->session
) {
1866 struct nfs4_client
*clp
= session
->se_client
;
1868 nfsd4_get_session(cstate
->session
);
1869 atomic_inc(&clp
->cl_refcount
);
1870 if (clp
->cl_cb_state
== NFSD4_CB_DOWN
)
1871 seq
->status_flags
|= SEQ4_STATUS_CB_PATH_DOWN
;
1874 spin_unlock(&client_lock
);
1875 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1880 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
1884 if (rc
->rca_one_fs
) {
1885 if (!cstate
->current_fh
.fh_dentry
)
1886 return nfserr_nofilehandle
;
1888 * We don't take advantage of the rca_one_fs case.
1889 * That's OK, it's optional, we can safely ignore it.
1895 status
= nfserr_complete_already
;
1896 if (cstate
->session
->se_client
->cl_firststate
)
1899 status
= nfserr_stale_clientid
;
1900 if (is_client_expired(cstate
->session
->se_client
))
1902 * The following error isn't really legal.
1903 * But we only get here if the client just explicitly
1904 * destroyed the client. Surely it no longer cares what
1905 * error it gets back on an operation for the dead
1911 nfsd4_create_clid_dir(cstate
->session
->se_client
);
1913 nfs4_unlock_state();
1918 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1919 struct nfsd4_setclientid
*setclid
)
1921 struct xdr_netobj clname
= {
1922 .len
= setclid
->se_namelen
,
1923 .data
= setclid
->se_name
,
1925 nfs4_verifier clverifier
= setclid
->se_verf
;
1926 unsigned int strhashval
;
1927 struct nfs4_client
*conf
, *unconf
, *new;
1929 char dname
[HEXDIR_LEN
];
1931 if (!check_name(clname
))
1932 return nfserr_inval
;
1934 status
= nfs4_make_rec_clidname(dname
, &clname
);
1939 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1940 * We get here on a DRC miss.
1943 strhashval
= clientstr_hashval(dname
);
1946 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1948 /* RFC 3530 14.2.33 CASE 0: */
1949 status
= nfserr_clid_inuse
;
1950 if (clp_used_exchangeid(conf
))
1952 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1953 char addr_str
[INET6_ADDRSTRLEN
];
1954 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1956 dprintk("NFSD: setclientid: string in use by client "
1957 "at %s\n", addr_str
);
1962 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1963 * has a description of SETCLIENTID request processing consisting
1964 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1966 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1967 status
= nfserr_jukebox
;
1970 * RFC 3530 14.2.33 CASE 4:
1971 * placed first, because it is the normal case
1974 expire_client(unconf
);
1975 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1979 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1981 * RFC 3530 14.2.33 CASE 1:
1982 * probable callback update
1985 /* Note this is removing unconfirmed {*x***},
1986 * which is stronger than RFC recommended {vxc**}.
1987 * This has the advantage that there is at most
1988 * one {*x***} in either list at any time.
1990 expire_client(unconf
);
1992 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1995 copy_clid(new, conf
);
1996 } else if (!unconf
) {
1998 * RFC 3530 14.2.33 CASE 2:
1999 * probable client reboot; state will be removed if
2002 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2008 * RFC 3530 14.2.33 CASE 3:
2009 * probable client reboot; state will be removed if
2012 expire_client(unconf
);
2013 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2019 * XXX: we should probably set this at creation time, and check
2020 * for consistent minorversion use throughout:
2022 new->cl_minorversion
= 0;
2023 gen_callback(new, setclid
, rqstp
);
2024 add_to_unconfirmed(new, strhashval
);
2025 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2026 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2027 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2030 nfs4_unlock_state();
2036 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2037 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2038 * bullets, labeled as CASE1 - CASE4 below.
2041 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2042 struct nfsd4_compound_state
*cstate
,
2043 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2045 struct sockaddr
*sa
= svc_addr(rqstp
);
2046 struct nfs4_client
*conf
, *unconf
;
2047 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2048 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2051 if (STALE_CLIENTID(clid
))
2052 return nfserr_stale_clientid
;
2054 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2055 * We get here on a DRC miss.
2060 conf
= find_confirmed_client(clid
);
2061 unconf
= find_unconfirmed_client(clid
);
2063 status
= nfserr_clid_inuse
;
2064 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
2066 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
2070 * section 14.2.34 of RFC 3530 has a description of
2071 * SETCLIENTID_CONFIRM request processing consisting
2072 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2074 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
2076 * RFC 3530 14.2.34 CASE 1:
2079 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
2080 status
= nfserr_clid_inuse
;
2082 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2083 nfsd4_probe_callback(conf
);
2084 expire_client(unconf
);
2088 } else if (conf
&& !unconf
) {
2090 * RFC 3530 14.2.34 CASE 2:
2091 * probable retransmitted request; play it safe and
2094 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2095 status
= nfserr_clid_inuse
;
2098 } else if (!conf
&& unconf
2099 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
2101 * RFC 3530 14.2.34 CASE 3:
2102 * Normal case; new or rebooted client:
2104 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
2105 status
= nfserr_clid_inuse
;
2108 clientstr_hashval(unconf
->cl_recdir
);
2109 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
2112 nfsd4_remove_clid_dir(conf
);
2113 expire_client(conf
);
2115 move_to_confirmed(unconf
);
2117 nfsd4_probe_callback(conf
);
2120 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
2121 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
2124 * RFC 3530 14.2.34 CASE 4:
2125 * Client probably hasn't noticed that we rebooted yet.
2127 status
= nfserr_stale_clientid
;
2129 /* check that we have hit one of the cases...*/
2130 status
= nfserr_clid_inuse
;
2133 nfs4_unlock_state();
2137 /* OPEN Share state helper functions */
2138 static inline struct nfs4_file
*
2139 alloc_init_file(struct inode
*ino
)
2141 struct nfs4_file
*fp
;
2142 unsigned int hashval
= file_hashval(ino
);
2144 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2146 atomic_set(&fp
->fi_ref
, 1);
2147 INIT_LIST_HEAD(&fp
->fi_hash
);
2148 INIT_LIST_HEAD(&fp
->fi_stateids
);
2149 INIT_LIST_HEAD(&fp
->fi_delegations
);
2150 fp
->fi_inode
= igrab(ino
);
2151 fp
->fi_id
= current_fileid
++;
2152 fp
->fi_had_conflict
= false;
2153 fp
->fi_lease
= NULL
;
2154 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2155 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2156 spin_lock(&recall_lock
);
2157 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2158 spin_unlock(&recall_lock
);
2165 nfsd4_free_slab(struct kmem_cache
**slab
)
2169 kmem_cache_destroy(*slab
);
2174 nfsd4_free_slabs(void)
2176 nfsd4_free_slab(&stateowner_slab
);
2177 nfsd4_free_slab(&file_slab
);
2178 nfsd4_free_slab(&stateid_slab
);
2179 nfsd4_free_slab(&deleg_slab
);
2183 nfsd4_init_slabs(void)
2185 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
2186 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
2187 if (stateowner_slab
== NULL
)
2189 file_slab
= kmem_cache_create("nfsd4_files",
2190 sizeof(struct nfs4_file
), 0, 0, NULL
);
2191 if (file_slab
== NULL
)
2193 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2194 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
2195 if (stateid_slab
== NULL
)
2197 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2198 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2199 if (deleg_slab
== NULL
)
2204 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2209 nfs4_free_stateowner(struct nfs4_stateowner
*sop
)
2211 kfree(sop
->so_owner
.data
);
2212 kmem_cache_free(stateowner_slab
, sop
);
2215 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2217 rp
->rp_status
= nfserr_serverfault
;
2219 rp
->rp_buf
= rp
->rp_ibuf
;
2222 static inline struct nfs4_stateowner
*alloc_stateowner(struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2224 struct nfs4_stateowner
*sop
;
2226 sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
);
2230 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2231 if (!sop
->so_owner
.data
) {
2232 kmem_cache_free(stateowner_slab
, sop
);
2235 sop
->so_owner
.len
= owner
->len
;
2237 INIT_LIST_HEAD(&sop
->so_perclient
);
2238 INIT_LIST_HEAD(&sop
->so_stateids
);
2239 INIT_LIST_HEAD(&sop
->so_perstateid
);
2240 INIT_LIST_HEAD(&sop
->so_close_lru
);
2241 sop
->so_id
= current_ownerid
++;
2243 sop
->so_client
= clp
;
2244 init_nfs4_replay(&sop
->so_replay
);
2248 static void hash_openowner(struct nfs4_stateowner
*sop
, struct nfs4_client
*clp
, unsigned int strhashval
)
2250 unsigned int idhashval
;
2252 idhashval
= open_ownerid_hashval(sop
->so_id
);
2253 list_add(&sop
->so_idhash
, &open_ownerid_hashtbl
[idhashval
]);
2254 list_add(&sop
->so_strhash
, &open_ownerstr_hashtbl
[strhashval
]);
2255 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
2258 static struct nfs4_stateowner
*
2259 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2260 struct nfs4_stateowner
*sop
;
2262 sop
= alloc_stateowner(&open
->op_owner
, clp
);
2265 sop
->so_is_open_owner
= 1;
2266 sop
->so_seqid
= open
->op_seqid
;
2267 sop
->so_confirmed
= 0;
2268 hash_openowner(sop
, clp
, strhashval
);
2273 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2274 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2275 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
2277 INIT_LIST_HEAD(&stp
->st_hash
);
2278 INIT_LIST_HEAD(&stp
->st_perstateowner
);
2279 INIT_LIST_HEAD(&stp
->st_lockowners
);
2280 INIT_LIST_HEAD(&stp
->st_perfile
);
2281 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
2282 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
2283 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2284 stp
->st_type
= NFS4_OPEN_STID
;
2285 stp
->st_stateowner
= sop
;
2288 stp
->st_stateid
.si_boot
= boot_time
;
2289 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
2290 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
2291 /* note will be incremented before first return to client: */
2292 stp
->st_stateid
.si_generation
= 0;
2293 stp
->st_access_bmap
= 0;
2294 stp
->st_deny_bmap
= 0;
2295 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
2296 &stp
->st_access_bmap
);
2297 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2298 stp
->st_openstp
= NULL
;
2302 move_to_close_lru(struct nfs4_stateowner
*sop
)
2304 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
2306 list_move_tail(&sop
->so_close_lru
, &close_lru
);
2307 sop
->so_time
= get_seconds();
2311 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2314 return (sop
->so_owner
.len
== owner
->len
) &&
2315 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2316 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2319 static struct nfs4_stateowner
*
2320 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2322 struct nfs4_stateowner
*so
= NULL
;
2324 list_for_each_entry(so
, &open_ownerstr_hashtbl
[hashval
], so_strhash
) {
2325 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
2331 /* search file_hashtbl[] for file */
2332 static struct nfs4_file
*
2333 find_file(struct inode
*ino
)
2335 unsigned int hashval
= file_hashval(ino
);
2336 struct nfs4_file
*fp
;
2338 spin_lock(&recall_lock
);
2339 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2340 if (fp
->fi_inode
== ino
) {
2342 spin_unlock(&recall_lock
);
2346 spin_unlock(&recall_lock
);
2350 static inline int access_valid(u32 x
, u32 minorversion
)
2352 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
2354 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
2356 x
&= ~NFS4_SHARE_ACCESS_MASK
;
2357 if (minorversion
&& x
) {
2358 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
2360 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
2362 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
2369 static inline int deny_valid(u32 x
)
2371 /* Note: unlike access bits, deny bits may be zero. */
2372 return x
<= NFS4_SHARE_DENY_BOTH
;
2376 * Called to check deny when READ with all zero stateid or
2377 * WRITE with all zero or all one stateid
2380 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2382 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2383 struct nfs4_file
*fp
;
2384 struct nfs4_stateid
*stp
;
2387 dprintk("NFSD: nfs4_share_conflict\n");
2389 fp
= find_file(ino
);
2392 ret
= nfserr_locked
;
2393 /* Search for conflicting share reservations */
2394 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2395 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2396 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2405 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2407 /* We're assuming the state code never drops its reference
2408 * without first removing the lease. Since we're in this lease
2409 * callback (and since the lease code is serialized by the kernel
2410 * lock) we know the server hasn't removed the lease yet, we know
2411 * it's safe to take a reference: */
2412 atomic_inc(&dp
->dl_count
);
2414 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2416 /* only place dl_time is set. protected by lock_flocks*/
2417 dp
->dl_time
= get_seconds();
2419 nfsd4_cb_recall(dp
);
2422 /* Called from break_lease() with lock_flocks() held. */
2423 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2425 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2426 struct nfs4_delegation
*dp
;
2429 /* We assume break_lease is only called once per lease: */
2430 BUG_ON(fp
->fi_had_conflict
);
2432 * We don't want the locks code to timeout the lease for us;
2433 * we'll remove it ourself if a delegation isn't returned
2436 fl
->fl_break_time
= 0;
2438 spin_lock(&recall_lock
);
2439 fp
->fi_had_conflict
= true;
2440 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2441 nfsd_break_one_deleg(dp
);
2442 spin_unlock(&recall_lock
);
2446 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2449 return lease_modify(onlist
, arg
);
2454 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2455 .lm_break
= nfsd_break_deleg_cb
,
2456 .lm_change
= nfsd_change_deleg_cb
,
2461 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2462 struct nfsd4_open
*open
)
2464 clientid_t
*clientid
= &open
->op_clientid
;
2465 struct nfs4_client
*clp
= NULL
;
2466 unsigned int strhashval
;
2467 struct nfs4_stateowner
*sop
= NULL
;
2469 if (!check_name(open
->op_owner
))
2470 return nfserr_inval
;
2472 if (STALE_CLIENTID(&open
->op_clientid
))
2473 return nfserr_stale_clientid
;
2475 strhashval
= open_ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2476 sop
= find_openstateowner_str(strhashval
, open
);
2477 open
->op_stateowner
= sop
;
2479 /* Make sure the client's lease hasn't expired. */
2480 clp
= find_confirmed_client(clientid
);
2482 return nfserr_expired
;
2485 /* When sessions are used, skip open sequenceid processing */
2486 if (nfsd4_has_session(cstate
))
2488 if (!sop
->so_confirmed
) {
2489 /* Replace unconfirmed owners without checking for replay. */
2490 clp
= sop
->so_client
;
2491 release_openowner(sop
);
2492 open
->op_stateowner
= NULL
;
2495 if (open
->op_seqid
== sop
->so_seqid
- 1)
2496 return nfserr_replay_me
;
2497 if (open
->op_seqid
!= sop
->so_seqid
)
2498 return nfserr_bad_seqid
;
2500 if (open
->op_stateowner
== NULL
) {
2501 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2503 return nfserr_jukebox
;
2504 open
->op_stateowner
= sop
;
2506 list_del_init(&sop
->so_close_lru
);
2507 renew_client(sop
->so_client
);
2511 static inline __be32
2512 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2514 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2515 return nfserr_openmode
;
2520 static struct nfs4_delegation
*
2521 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2523 struct nfs4_delegation
*dp
;
2525 spin_lock(&recall_lock
);
2526 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2527 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
) {
2528 spin_unlock(&recall_lock
);
2531 spin_unlock(&recall_lock
);
2535 static int share_access_to_flags(u32 share_access
)
2537 share_access
&= ~NFS4_SHARE_WANT_MASK
;
2539 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2543 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2544 struct nfs4_delegation
**dp
)
2547 __be32 status
= nfserr_bad_stateid
;
2549 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2552 flags
= share_access_to_flags(open
->op_share_access
);
2553 status
= nfs4_check_delegmode(*dp
, flags
);
2557 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2561 open
->op_stateowner
->so_confirmed
= 1;
2566 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2568 struct nfs4_stateid
*local
;
2569 __be32 status
= nfserr_share_denied
;
2570 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2572 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2573 /* ignore lock owners */
2574 if (local
->st_stateowner
->so_is_open_owner
== 0)
2576 /* remember if we have seen this open owner */
2577 if (local
->st_stateowner
== sop
)
2579 /* check for conflicting share reservations */
2580 if (!test_share(local
, open
))
2588 static inline struct nfs4_stateid
*
2589 nfs4_alloc_stateid(void)
2591 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2594 static inline int nfs4_access_to_access(u32 nfs4_access
)
2598 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2599 flags
|= NFSD_MAY_READ
;
2600 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2601 flags
|= NFSD_MAY_WRITE
;
2605 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2606 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2609 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2610 int access
= nfs4_access_to_access(open
->op_share_access
);
2612 /* CLAIM_DELEGATE_CUR is used in response to a broken lease;
2613 * allowing it to break the lease and return EAGAIN leaves the
2614 * client unable to make progress in returning the delegation */
2615 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2616 access
|= NFSD_MAY_NOT_BREAK_LEASE
;
2618 if (!fp
->fi_fds
[oflag
]) {
2619 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2620 &fp
->fi_fds
[oflag
]);
2624 nfs4_file_get_access(fp
, oflag
);
2630 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2631 struct nfs4_file
*fp
, struct svc_fh
*cur_fh
,
2632 struct nfsd4_open
*open
)
2634 struct nfs4_stateid
*stp
;
2637 stp
= nfs4_alloc_stateid();
2639 return nfserr_jukebox
;
2641 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2643 kmem_cache_free(stateid_slab
, stp
);
2650 static inline __be32
2651 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2652 struct nfsd4_open
*open
)
2654 struct iattr iattr
= {
2655 .ia_valid
= ATTR_SIZE
,
2658 if (!open
->op_truncate
)
2660 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2661 return nfserr_inval
;
2662 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2666 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2668 u32 op_share_access
= open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
;
2672 new_access
= !test_bit(op_share_access
, &stp
->st_access_bmap
);
2674 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2678 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2681 int oflag
= nfs4_access_to_omode(op_share_access
);
2682 nfs4_file_put_access(fp
, oflag
);
2686 /* remember the open */
2687 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2688 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2695 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2697 open
->op_stateowner
->so_confirmed
= 1;
2698 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2701 /* Should we give out recallable state?: */
2702 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2704 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2707 * In the sessions case, since we don't have to establish a
2708 * separate connection for callbacks, we assume it's OK
2709 * until we hear otherwise:
2711 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2714 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2716 struct file_lock
*fl
;
2718 fl
= locks_alloc_lock();
2721 locks_init_lock(fl
);
2722 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2723 fl
->fl_flags
= FL_LEASE
;
2724 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2725 fl
->fl_end
= OFFSET_MAX
;
2726 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2727 fl
->fl_pid
= current
->tgid
;
2731 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2733 struct nfs4_file
*fp
= dp
->dl_file
;
2734 struct file_lock
*fl
;
2737 fl
= nfs4_alloc_init_lease(dp
, flag
);
2740 fl
->fl_file
= find_readable_file(fp
);
2741 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2742 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2744 list_del_init(&dp
->dl_perclnt
);
2745 locks_free_lock(fl
);
2749 fp
->fi_deleg_file
= fl
->fl_file
;
2750 get_file(fp
->fi_deleg_file
);
2751 atomic_set(&fp
->fi_delegees
, 1);
2752 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2756 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2758 struct nfs4_file
*fp
= dp
->dl_file
;
2761 return nfs4_setlease(dp
, flag
);
2762 spin_lock(&recall_lock
);
2763 if (fp
->fi_had_conflict
) {
2764 spin_unlock(&recall_lock
);
2767 atomic_inc(&fp
->fi_delegees
);
2768 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2769 spin_unlock(&recall_lock
);
2770 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2775 * Attempt to hand out a delegation.
2778 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2780 struct nfs4_delegation
*dp
;
2781 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2783 int status
, flag
= 0;
2785 cb_up
= nfsd4_cb_channel_good(sop
->so_client
);
2786 flag
= NFS4_OPEN_DELEGATE_NONE
;
2787 open
->op_recall
= 0;
2788 switch (open
->op_claim_type
) {
2789 case NFS4_OPEN_CLAIM_PREVIOUS
:
2791 open
->op_recall
= 1;
2792 flag
= open
->op_delegate_type
;
2793 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2796 case NFS4_OPEN_CLAIM_NULL
:
2797 /* Let's not give out any delegations till everyone's
2798 * had the chance to reclaim theirs.... */
2799 if (locks_in_grace())
2801 if (!cb_up
|| !sop
->so_confirmed
)
2803 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2804 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2806 flag
= NFS4_OPEN_DELEGATE_READ
;
2812 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2815 status
= nfs4_set_delegation(dp
, flag
);
2819 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2821 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2822 STATEID_VAL(&dp
->dl_stateid
));
2824 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2825 && flag
== NFS4_OPEN_DELEGATE_NONE
2826 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2827 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2828 open
->op_delegate_type
= flag
;
2831 nfs4_put_delegation(dp
);
2833 flag
= NFS4_OPEN_DELEGATE_NONE
;
2838 * called with nfs4_lock_state() held.
2841 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2843 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2844 struct nfs4_file
*fp
= NULL
;
2845 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2846 struct nfs4_stateid
*stp
= NULL
;
2847 struct nfs4_delegation
*dp
= NULL
;
2850 status
= nfserr_inval
;
2851 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2852 || !deny_valid(open
->op_share_deny
))
2855 * Lookup file; if found, lookup stateid and check open request,
2856 * and check for delegations in the process of being recalled.
2857 * If not found, create the nfs4_file struct
2859 fp
= find_file(ino
);
2861 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2863 status
= nfs4_check_deleg(fp
, open
, &dp
);
2867 status
= nfserr_bad_stateid
;
2868 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2870 status
= nfserr_jukebox
;
2871 fp
= alloc_init_file(ino
);
2877 * OPEN the file, or upgrade an existing OPEN.
2878 * If truncate fails, the OPEN fails.
2881 /* Stateid was found, this is an OPEN upgrade */
2882 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
2886 status
= nfs4_new_open(rqstp
, &stp
, fp
, current_fh
, open
);
2889 init_stateid(stp
, fp
, open
);
2890 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2892 release_open_stateid(stp
);
2896 update_stateid(&stp
->st_stateid
);
2897 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2899 if (nfsd4_has_session(&resp
->cstate
))
2900 open
->op_stateowner
->so_confirmed
= 1;
2903 * Attempt to hand out a delegation. No error return, because the
2904 * OPEN succeeds even if we fail.
2906 nfs4_open_delegation(current_fh
, open
, stp
);
2910 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2911 STATEID_VAL(&stp
->st_stateid
));
2915 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2916 nfs4_set_claim_prev(open
);
2918 * To finish the open response, we just need to set the rflags.
2920 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2921 if (!open
->op_stateowner
->so_confirmed
&&
2922 !nfsd4_has_session(&resp
->cstate
))
2923 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2929 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2932 struct nfs4_client
*clp
;
2936 dprintk("process_renew(%08x/%08x): starting\n",
2937 clid
->cl_boot
, clid
->cl_id
);
2938 status
= nfserr_stale_clientid
;
2939 if (STALE_CLIENTID(clid
))
2941 clp
= find_confirmed_client(clid
);
2942 status
= nfserr_expired
;
2944 /* We assume the client took too long to RENEW. */
2945 dprintk("nfsd4_renew: clientid not found!\n");
2949 status
= nfserr_cb_path_down
;
2950 if (!list_empty(&clp
->cl_delegations
)
2951 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
2955 nfs4_unlock_state();
2959 static struct lock_manager nfsd4_manager
= {
2963 nfsd4_end_grace(void)
2965 dprintk("NFSD: end of grace period\n");
2966 nfsd4_recdir_purge_old();
2967 locks_end_grace(&nfsd4_manager
);
2969 * Now that every NFSv4 client has had the chance to recover and
2970 * to see the (possibly new, possibly shorter) lease time, we
2971 * can safely set the next grace time to the current lease time:
2973 nfsd4_grace
= nfsd4_lease
;
2977 nfs4_laundromat(void)
2979 struct nfs4_client
*clp
;
2980 struct nfs4_stateowner
*sop
;
2981 struct nfs4_delegation
*dp
;
2982 struct list_head
*pos
, *next
, reaplist
;
2983 time_t cutoff
= get_seconds() - nfsd4_lease
;
2984 time_t t
, clientid_val
= nfsd4_lease
;
2985 time_t u
, test_val
= nfsd4_lease
;
2989 dprintk("NFSD: laundromat service - starting\n");
2990 if (locks_in_grace())
2992 INIT_LIST_HEAD(&reaplist
);
2993 spin_lock(&client_lock
);
2994 list_for_each_safe(pos
, next
, &client_lru
) {
2995 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2996 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2997 t
= clp
->cl_time
- cutoff
;
2998 if (clientid_val
> t
)
3002 if (atomic_read(&clp
->cl_refcount
)) {
3003 dprintk("NFSD: client in use (clientid %08x)\n",
3004 clp
->cl_clientid
.cl_id
);
3007 unhash_client_locked(clp
);
3008 list_add(&clp
->cl_lru
, &reaplist
);
3010 spin_unlock(&client_lock
);
3011 list_for_each_safe(pos
, next
, &reaplist
) {
3012 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3013 dprintk("NFSD: purging unused client (clientid %08x)\n",
3014 clp
->cl_clientid
.cl_id
);
3015 nfsd4_remove_clid_dir(clp
);
3018 spin_lock(&recall_lock
);
3019 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3020 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3021 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3022 u
= dp
->dl_time
- cutoff
;
3027 list_move(&dp
->dl_recall_lru
, &reaplist
);
3029 spin_unlock(&recall_lock
);
3030 list_for_each_safe(pos
, next
, &reaplist
) {
3031 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3032 list_del_init(&dp
->dl_recall_lru
);
3033 unhash_delegation(dp
);
3035 test_val
= nfsd4_lease
;
3036 list_for_each_safe(pos
, next
, &close_lru
) {
3037 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
3038 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
3039 u
= sop
->so_time
- cutoff
;
3044 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
3046 release_openowner(sop
);
3048 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3049 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3050 nfs4_unlock_state();
3051 return clientid_val
;
3054 static struct workqueue_struct
*laundry_wq
;
3055 static void laundromat_main(struct work_struct
*);
3056 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3059 laundromat_main(struct work_struct
*not_used
)
3063 t
= nfs4_laundromat();
3064 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3065 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3068 static struct nfs4_stateowner
*
3069 search_close_lru(u32 st_id
, int flags
)
3071 struct nfs4_stateowner
*local
= NULL
;
3073 if (flags
& CLOSE_STATE
) {
3074 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
3075 if (local
->so_id
== st_id
)
3083 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
3085 return fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
;
3089 STALE_STATEID(stateid_t
*stateid
)
3091 if (stateid
->si_boot
== boot_time
)
3093 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3094 STATEID_VAL(stateid
));
3099 access_permit_read(unsigned long access_bmap
)
3101 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
3102 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
3103 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
3107 access_permit_write(unsigned long access_bmap
)
3109 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
3110 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
3114 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
3116 __be32 status
= nfserr_openmode
;
3118 /* For lock stateid's, we test the parent open, not the lock: */
3119 if (stp
->st_openstp
)
3120 stp
= stp
->st_openstp
;
3121 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
3123 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
3130 static inline __be32
3131 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3133 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3135 else if (locks_in_grace()) {
3136 /* Answer in remaining cases depends on existence of
3137 * conflicting state; so we must wait out the grace period. */
3138 return nfserr_grace
;
3139 } else if (flags
& WR_STATE
)
3140 return nfs4_share_conflict(current_fh
,
3141 NFS4_SHARE_DENY_WRITE
);
3142 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3143 return nfs4_share_conflict(current_fh
,
3144 NFS4_SHARE_DENY_READ
);
3148 * Allow READ/WRITE during grace period on recovered state only for files
3149 * that are not able to provide mandatory locking.
3152 grace_disallows_io(struct inode
*inode
)
3154 return locks_in_grace() && mandatory_lock(inode
);
3157 /* Returns true iff a is later than b: */
3158 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3160 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3163 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3166 * When sessions are used the stateid generation number is ignored
3169 if (has_session
&& in
->si_generation
== 0)
3172 if (in
->si_generation
== ref
->si_generation
)
3175 /* If the client sends us a stateid from the future, it's buggy: */
3176 if (stateid_generation_after(in
, ref
))
3177 return nfserr_bad_stateid
;
3179 * However, we could see a stateid from the past, even from a
3180 * non-buggy client. For example, if the client sends a lock
3181 * while some IO is outstanding, the lock may bump si_generation
3182 * while the IO is still in flight. The client could avoid that
3183 * situation by waiting for responses on all the IO requests,
3184 * but better performance may result in retrying IO that
3185 * receives an old_stateid error if requests are rarely
3186 * reordered in flight:
3188 return nfserr_old_stateid
;
3191 static int is_delegation_stateid(stateid_t
*stateid
)
3193 return stateid
->si_fileid
== 0;
3196 __be32
nfs4_validate_stateid(stateid_t
*stateid
, bool has_session
)
3198 struct nfs4_stateid
*stp
= NULL
;
3199 __be32 status
= nfserr_stale_stateid
;
3201 if (STALE_STATEID(stateid
))
3204 status
= nfserr_expired
;
3205 stp
= find_stateid(stateid
, 0);
3208 status
= nfserr_bad_stateid
;
3210 if (!stp
->st_stateowner
->so_confirmed
)
3213 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, has_session
);
3223 * Checks for stateid operations
3226 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3227 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3229 struct nfs4_stateid
*stp
= NULL
;
3230 struct nfs4_delegation
*dp
= NULL
;
3231 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3232 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3238 if (grace_disallows_io(ino
))
3239 return nfserr_grace
;
3241 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3242 return check_special_stateids(current_fh
, stateid
, flags
);
3244 status
= nfserr_stale_stateid
;
3245 if (STALE_STATEID(stateid
))
3249 * We assume that any stateid that has the current boot time,
3250 * but that we can't find, is expired:
3252 status
= nfserr_expired
;
3253 if (is_delegation_stateid(stateid
)) {
3254 dp
= find_delegation_stateid(ino
, stateid
);
3257 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, nfsd4_has_session(cstate
));
3260 status
= nfs4_check_delegmode(dp
, flags
);
3263 renew_client(dp
->dl_client
);
3265 *filpp
= dp
->dl_file
->fi_deleg_file
;
3268 } else { /* open or lock stateid */
3269 stp
= find_stateid(stateid
, flags
);
3272 status
= nfserr_bad_stateid
;
3273 if (nfs4_check_fh(current_fh
, stp
))
3275 if (!stp
->st_stateowner
->so_confirmed
)
3277 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
3278 nfsd4_has_session(cstate
));
3281 status
= nfs4_check_openmode(stp
, flags
);
3284 renew_client(stp
->st_stateowner
->so_client
);
3286 if (flags
& RD_STATE
)
3287 *filpp
= find_readable_file(stp
->st_file
);
3289 *filpp
= find_writeable_file(stp
->st_file
);
3298 nfsd4_free_delegation_stateid(stateid_t
*stateid
)
3300 struct nfs4_delegation
*dp
= search_for_delegation(stateid
);
3302 return nfserr_locks_held
;
3303 return nfserr_bad_stateid
;
3307 nfsd4_free_lock_stateid(struct nfs4_stateid
*stp
)
3309 if (check_for_locks(stp
->st_file
, stp
->st_stateowner
))
3310 return nfserr_locks_held
;
3311 release_lock_stateid(stp
);
3316 * Test if the stateid is valid
3319 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3320 struct nfsd4_test_stateid
*test_stateid
)
3322 test_stateid
->ts_has_session
= nfsd4_has_session(cstate
);
3330 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3331 struct nfsd4_free_stateid
*free_stateid
)
3333 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3334 struct nfs4_stateid
*stp
;
3338 if (is_delegation_stateid(stateid
)) {
3339 ret
= nfsd4_free_delegation_stateid(stateid
);
3343 stp
= find_stateid(stateid
, 0);
3345 ret
= nfserr_bad_stateid
;
3348 ret
= check_stateid_generation(stateid
, &stp
->st_stateid
, 1);
3352 if (stp
->st_type
== NFS4_OPEN_STID
) {
3353 ret
= nfserr_locks_held
;
3356 ret
= nfsd4_free_lock_stateid(stp
);
3361 nfs4_unlock_state();
3368 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3369 RD_STATE
: WR_STATE
;
3373 * Checks for sequence id mutating operations.
3376 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3377 stateid_t
*stateid
, int flags
,
3378 struct nfs4_stateid
**stpp
)
3380 struct nfs4_stateid
*stp
;
3381 struct nfs4_stateowner
*sop
;
3382 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3385 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3386 seqid
, STATEID_VAL(stateid
));
3390 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
3391 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3392 return nfserr_bad_stateid
;
3395 if (STALE_STATEID(stateid
))
3396 return nfserr_stale_stateid
;
3399 * We return BAD_STATEID if filehandle doesn't match stateid,
3400 * the confirmed flag is incorrecly set, or the generation
3401 * number is incorrect.
3403 stp
= find_stateid(stateid
, flags
);
3406 * Also, we should make sure this isn't just the result of
3409 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
3410 /* It's not stale; let's assume it's expired: */
3412 return nfserr_expired
;
3413 cstate
->replay_owner
= sop
;
3418 sop
= stp
->st_stateowner
;
3419 cstate
->replay_owner
= sop
;
3421 if (nfs4_check_fh(current_fh
, stp
)) {
3422 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3423 return nfserr_bad_stateid
;
3426 if (!nfsd4_has_session(cstate
) && seqid
!= sop
->so_seqid
)
3429 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3430 dprintk("NFSD: preprocess_seqid_op: expected"
3431 " unconfirmed stateowner!\n");
3432 return nfserr_bad_stateid
;
3434 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3435 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3436 " confirmed yet!\n");
3437 return nfserr_bad_stateid
;
3439 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, nfsd4_has_session(cstate
));
3442 renew_client(sop
->so_client
);
3446 if (seqid
== sop
->so_seqid
- 1) {
3447 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3448 /* indicate replay to calling function */
3449 return nfserr_replay_me
;
3451 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3452 sop
->so_seqid
, seqid
);
3453 return nfserr_bad_seqid
;
3457 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3458 struct nfsd4_open_confirm
*oc
)
3461 struct nfs4_stateowner
*sop
;
3462 struct nfs4_stateid
*stp
;
3464 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3465 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3466 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3468 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3474 status
= nfs4_preprocess_seqid_op(cstate
,
3475 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3476 CONFIRM
| OPEN_STATE
, &stp
);
3480 sop
= stp
->st_stateowner
;
3481 sop
->so_confirmed
= 1;
3482 update_stateid(&stp
->st_stateid
);
3483 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3484 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3485 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stateid
));
3487 nfsd4_create_clid_dir(sop
->so_client
);
3489 if (!cstate
->replay_owner
)
3490 nfs4_unlock_state();
3494 static inline void nfs4_file_downgrade(struct nfs4_stateid
*stp
, unsigned int to_access
)
3498 for (i
= 1; i
< 4; i
++) {
3499 if (test_bit(i
, &stp
->st_access_bmap
) && !(i
& to_access
)) {
3500 nfs4_file_put_access(stp
->st_file
, i
);
3501 __clear_bit(i
, &stp
->st_access_bmap
);
3507 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3510 for (i
= 0; i
< 4; i
++) {
3511 if ((i
& deny
) != i
)
3512 __clear_bit(i
, bmap
);
3517 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3518 struct nfsd4_compound_state
*cstate
,
3519 struct nfsd4_open_downgrade
*od
)
3522 struct nfs4_stateid
*stp
;
3524 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3525 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3526 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3528 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3529 || !deny_valid(od
->od_share_deny
))
3530 return nfserr_inval
;
3533 status
= nfs4_preprocess_seqid_op(cstate
, od
->od_seqid
,
3534 &od
->od_stateid
, OPEN_STATE
, &stp
);
3538 status
= nfserr_inval
;
3539 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3540 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3541 stp
->st_access_bmap
, od
->od_share_access
);
3544 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3545 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3546 stp
->st_deny_bmap
, od
->od_share_deny
);
3549 nfs4_file_downgrade(stp
, od
->od_share_access
);
3551 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3553 update_stateid(&stp
->st_stateid
);
3554 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3557 if (!cstate
->replay_owner
)
3558 nfs4_unlock_state();
3563 * nfs4_unlock_state() called after encode
3566 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3567 struct nfsd4_close
*close
)
3570 struct nfs4_stateid
*stp
;
3571 struct nfs4_stateowner
*so
;
3573 dprintk("NFSD: nfsd4_close on file %.*s\n",
3574 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3575 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3578 /* check close_lru for replay */
3579 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3581 OPEN_STATE
| CLOSE_STATE
, &stp
);
3584 so
= stp
->st_stateowner
;
3586 update_stateid(&stp
->st_stateid
);
3587 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3589 /* release_stateid() calls nfsd_close() if needed */
3590 release_open_stateid(stp
);
3592 /* place unused nfs4_stateowners on so_close_lru list to be
3593 * released by the laundromat service after the lease period
3594 * to enable us to handle CLOSE replay
3596 if (list_empty(&so
->so_stateids
))
3597 move_to_close_lru(so
);
3599 if (!cstate
->replay_owner
)
3600 nfs4_unlock_state();
3605 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3606 struct nfsd4_delegreturn
*dr
)
3608 struct nfs4_delegation
*dp
;
3609 stateid_t
*stateid
= &dr
->dr_stateid
;
3610 struct inode
*inode
;
3613 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3615 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3618 status
= nfserr_bad_stateid
;
3619 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3621 status
= nfserr_stale_stateid
;
3622 if (STALE_STATEID(stateid
))
3624 status
= nfserr_bad_stateid
;
3625 if (!is_delegation_stateid(stateid
))
3627 status
= nfserr_expired
;
3628 dp
= find_delegation_stateid(inode
, stateid
);
3631 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, nfsd4_has_session(cstate
));
3634 renew_client(dp
->dl_client
);
3636 unhash_delegation(dp
);
3638 nfs4_unlock_state();
3645 * Lock owner state (byte-range locks)
3647 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3648 #define LOCK_HASH_BITS 8
3649 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3650 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3653 end_offset(u64 start
, u64 len
)
3658 return end
>= start
? end
: NFS4_MAX_UINT64
;
3661 /* last octet in a range */
3663 last_byte_offset(u64 start
, u64 len
)
3669 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3672 static unsigned int lockownerid_hashval(u32 id
)
3674 return id
& LOCK_HASH_MASK
;
3677 static inline unsigned int
3678 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3679 struct xdr_netobj
*ownername
)
3681 return (file_hashval(inode
) + cl_id
3682 + opaque_hashval(ownername
->data
, ownername
->len
))
3686 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3687 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3690 same_stateid(stateid_t
*id_one
, stateid_t
*id_two
)
3692 if (id_one
->si_stateownerid
!= id_two
->si_stateownerid
)
3694 return id_one
->si_fileid
== id_two
->si_fileid
;
3697 static struct nfs4_stateid
*find_stateid(stateid_t
*t
, int flags
)
3699 struct nfs4_stateid
*s
;
3700 unsigned int hashval
;
3702 hashval
= stateid_hashval(t
->si_stateownerid
, t
->si_fileid
);
3703 list_for_each_entry(s
, &stateid_hashtbl
[hashval
], st_hash
) {
3704 if (!same_stateid(&s
->st_stateid
, t
))
3706 if (flags
& LOCK_STATE
&& s
->st_type
!= NFS4_LOCK_STID
)
3708 if (flags
& OPEN_STATE
&& s
->st_type
!= NFS4_OPEN_STID
)
3715 static struct nfs4_delegation
*
3716 search_for_delegation(stateid_t
*stid
)
3718 struct nfs4_file
*fp
;
3719 struct nfs4_delegation
*dp
;
3720 struct list_head
*pos
;
3723 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3724 list_for_each_entry(fp
, &file_hashtbl
[i
], fi_hash
) {
3725 list_for_each(pos
, &fp
->fi_delegations
) {
3726 dp
= list_entry(pos
, struct nfs4_delegation
, dl_perfile
);
3727 if (same_stateid(&dp
->dl_stateid
, stid
))
3735 static struct nfs4_delegation
*
3736 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3738 struct nfs4_file
*fp
;
3739 struct nfs4_delegation
*dl
;
3741 dprintk("NFSD: %s: stateid=" STATEID_FMT
"\n", __func__
,
3744 fp
= find_file(ino
);
3747 dl
= find_delegation_file(fp
, stid
);
3753 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3754 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3755 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3756 * locking, this prevents us from being completely protocol-compliant. The
3757 * real solution to this problem is to start using unsigned file offsets in
3758 * the VFS, but this is a very deep change!
3761 nfs4_transform_lock_offset(struct file_lock
*lock
)
3763 if (lock
->fl_start
< 0)
3764 lock
->fl_start
= OFFSET_MAX
;
3765 if (lock
->fl_end
< 0)
3766 lock
->fl_end
= OFFSET_MAX
;
3769 /* Hack!: For now, we're defining this just so we can use a pointer to it
3770 * as a unique cookie to identify our (NFSv4's) posix locks. */
3771 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3775 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3777 struct nfs4_stateowner
*sop
;
3779 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3780 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3781 deny
->ld_owner
.data
= kmemdup(sop
->so_owner
.data
,
3782 sop
->so_owner
.len
, GFP_KERNEL
);
3783 if (!deny
->ld_owner
.data
)
3784 /* We just don't care that much */
3786 deny
->ld_owner
.len
= sop
->so_owner
.len
;
3787 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3790 deny
->ld_owner
.len
= 0;
3791 deny
->ld_owner
.data
= NULL
;
3792 deny
->ld_clientid
.cl_boot
= 0;
3793 deny
->ld_clientid
.cl_id
= 0;
3795 deny
->ld_start
= fl
->fl_start
;
3796 deny
->ld_length
= NFS4_MAX_UINT64
;
3797 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3798 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3799 deny
->ld_type
= NFS4_READ_LT
;
3800 if (fl
->fl_type
!= F_RDLCK
)
3801 deny
->ld_type
= NFS4_WRITE_LT
;
3804 static struct nfs4_stateowner
*
3805 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3806 struct xdr_netobj
*owner
)
3808 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3809 struct nfs4_stateowner
*op
;
3811 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3812 if (same_owner_str(op
, owner
, clid
))
3818 static void hash_lockowner(struct nfs4_stateowner
*sop
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
)
3820 unsigned int idhashval
;
3822 idhashval
= lockownerid_hashval(sop
->so_id
);
3823 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3824 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3825 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3829 * Alloc a lock owner structure.
3830 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3833 * strhashval = lock_ownerstr_hashval
3836 static struct nfs4_stateowner
*
3837 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3838 struct nfs4_stateowner
*sop
;
3840 sop
= alloc_stateowner(&lock
->lk_new_owner
, clp
);
3843 INIT_LIST_HEAD(&sop
->so_stateids
);
3844 sop
->so_is_open_owner
= 0;
3845 /* It is the openowner seqid that will be incremented in encode in the
3846 * case of new lockowners; so increment the lock seqid manually: */
3847 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3848 sop
->so_confirmed
= 1;
3849 hash_lockowner(sop
, strhashval
, clp
, open_stp
);
3853 static struct nfs4_stateid
*
3854 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3856 struct nfs4_stateid
*stp
;
3857 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3859 stp
= nfs4_alloc_stateid();
3862 INIT_LIST_HEAD(&stp
->st_hash
);
3863 INIT_LIST_HEAD(&stp
->st_perfile
);
3864 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3865 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3866 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
3867 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3868 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3869 stp
->st_stateowner
= sop
;
3870 stp
->st_type
= NFS4_LOCK_STID
;
3873 stp
->st_stateid
.si_boot
= boot_time
;
3874 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3875 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3876 /* note will be incremented before first return to client: */
3877 stp
->st_stateid
.si_generation
= 0;
3878 stp
->st_access_bmap
= 0;
3879 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3880 stp
->st_openstp
= open_stp
;
3887 check_lock_length(u64 offset
, u64 length
)
3889 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3890 LOFF_OVERFLOW(offset
, length
)));
3893 static void get_lock_access(struct nfs4_stateid
*lock_stp
, u32 access
)
3895 struct nfs4_file
*fp
= lock_stp
->st_file
;
3896 int oflag
= nfs4_access_to_omode(access
);
3898 if (test_bit(access
, &lock_stp
->st_access_bmap
))
3900 nfs4_file_get_access(fp
, oflag
);
3901 __set_bit(access
, &lock_stp
->st_access_bmap
);
3908 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3909 struct nfsd4_lock
*lock
)
3911 struct nfs4_stateowner
*open_sop
= NULL
;
3912 struct nfs4_stateowner
*lock_sop
= NULL
;
3913 struct nfs4_stateid
*lock_stp
;
3914 struct nfs4_file
*fp
;
3915 struct file
*filp
= NULL
;
3916 struct file_lock file_lock
;
3917 struct file_lock conflock
;
3919 unsigned int strhashval
;
3923 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3924 (long long) lock
->lk_offset
,
3925 (long long) lock
->lk_length
);
3927 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3928 return nfserr_inval
;
3930 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3931 S_IFREG
, NFSD_MAY_LOCK
))) {
3932 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3938 if (lock
->lk_is_new
) {
3940 * Client indicates that this is a new lockowner.
3941 * Use open owner and open stateid to create lock owner and
3944 struct nfs4_stateid
*open_stp
= NULL
;
3946 status
= nfserr_stale_clientid
;
3947 if (!nfsd4_has_session(cstate
) &&
3948 STALE_CLIENTID(&lock
->lk_new_clientid
))
3951 /* validate and update open stateid and open seqid */
3952 status
= nfs4_preprocess_seqid_op(cstate
,
3953 lock
->lk_new_open_seqid
,
3954 &lock
->lk_new_open_stateid
,
3955 OPEN_STATE
, &open_stp
);
3958 status
= nfserr_bad_stateid
;
3959 open_sop
= open_stp
->st_stateowner
;
3960 if (!nfsd4_has_session(cstate
) &&
3961 !same_clid(&open_sop
->so_client
->cl_clientid
,
3962 &lock
->v
.new.clientid
))
3964 /* create lockowner and lock stateid */
3965 fp
= open_stp
->st_file
;
3966 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3967 open_sop
->so_client
->cl_clientid
.cl_id
,
3968 &lock
->v
.new.owner
);
3969 /* XXX: Do we need to check for duplicate stateowners on
3970 * the same file, or should they just be allowed (and
3971 * create new stateids)? */
3972 status
= nfserr_jukebox
;
3973 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3974 open_sop
->so_client
, open_stp
, lock
);
3975 if (lock_sop
== NULL
)
3977 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3978 if (lock_stp
== NULL
)
3981 /* lock (lock owner + lock stateid) already exists */
3982 status
= nfs4_preprocess_seqid_op(cstate
,
3983 lock
->lk_old_lock_seqid
,
3984 &lock
->lk_old_lock_stateid
,
3985 LOCK_STATE
, &lock_stp
);
3988 lock_sop
= lock_stp
->st_stateowner
;
3989 fp
= lock_stp
->st_file
;
3991 /* lock_sop and lock_stp have been created or found */
3993 lkflg
= setlkflg(lock
->lk_type
);
3994 status
= nfs4_check_openmode(lock_stp
, lkflg
);
3998 status
= nfserr_grace
;
3999 if (locks_in_grace() && !lock
->lk_reclaim
)
4001 status
= nfserr_no_grace
;
4002 if (!locks_in_grace() && lock
->lk_reclaim
)
4005 locks_init_lock(&file_lock
);
4006 switch (lock
->lk_type
) {
4009 filp
= find_readable_file(lock_stp
->st_file
);
4011 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4012 file_lock
.fl_type
= F_RDLCK
;
4015 case NFS4_WRITEW_LT
:
4016 filp
= find_writeable_file(lock_stp
->st_file
);
4018 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4019 file_lock
.fl_type
= F_WRLCK
;
4022 status
= nfserr_inval
;
4026 status
= nfserr_openmode
;
4029 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
4030 file_lock
.fl_pid
= current
->tgid
;
4031 file_lock
.fl_file
= filp
;
4032 file_lock
.fl_flags
= FL_POSIX
;
4033 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4035 file_lock
.fl_start
= lock
->lk_offset
;
4036 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4037 nfs4_transform_lock_offset(&file_lock
);
4040 * Try to lock the file in the VFS.
4041 * Note: locks.c uses the BKL to protect the inode's lock list.
4044 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
4046 case 0: /* success! */
4047 update_stateid(&lock_stp
->st_stateid
);
4048 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
4052 case (EAGAIN
): /* conflock holds conflicting lock */
4053 status
= nfserr_denied
;
4054 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4055 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
4058 status
= nfserr_deadlock
;
4061 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4062 status
= nfserrno(err
);
4066 if (status
&& lock
->lk_is_new
&& lock_sop
)
4067 release_lockowner(lock_sop
);
4068 if (!cstate
->replay_owner
)
4069 nfs4_unlock_state();
4074 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4075 * so we do a temporary open here just to get an open file to pass to
4076 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4079 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4084 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4087 err
= vfs_test_lock(file
, lock
);
4096 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4097 struct nfsd4_lockt
*lockt
)
4099 struct inode
*inode
;
4100 struct file_lock file_lock
;
4101 struct nfs4_stateowner
*so
;
4105 if (locks_in_grace())
4106 return nfserr_grace
;
4108 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4109 return nfserr_inval
;
4113 status
= nfserr_stale_clientid
;
4114 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
4117 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4120 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4121 locks_init_lock(&file_lock
);
4122 switch (lockt
->lt_type
) {
4125 file_lock
.fl_type
= F_RDLCK
;
4128 case NFS4_WRITEW_LT
:
4129 file_lock
.fl_type
= F_WRLCK
;
4132 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4133 status
= nfserr_inval
;
4137 so
= find_lockstateowner_str(inode
,
4138 &lockt
->lt_clientid
, &lockt
->lt_owner
);
4140 file_lock
.fl_owner
= (fl_owner_t
)so
;
4141 file_lock
.fl_pid
= current
->tgid
;
4142 file_lock
.fl_flags
= FL_POSIX
;
4144 file_lock
.fl_start
= lockt
->lt_offset
;
4145 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4147 nfs4_transform_lock_offset(&file_lock
);
4150 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4152 status
= nfserrno(error
);
4155 if (file_lock
.fl_type
!= F_UNLCK
) {
4156 status
= nfserr_denied
;
4157 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4160 nfs4_unlock_state();
4165 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4166 struct nfsd4_locku
*locku
)
4168 struct nfs4_stateid
*stp
;
4169 struct file
*filp
= NULL
;
4170 struct file_lock file_lock
;
4174 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4175 (long long) locku
->lu_offset
,
4176 (long long) locku
->lu_length
);
4178 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4179 return nfserr_inval
;
4183 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4184 &locku
->lu_stateid
, LOCK_STATE
, &stp
);
4187 filp
= find_any_file(stp
->st_file
);
4189 status
= nfserr_lock_range
;
4193 locks_init_lock(&file_lock
);
4194 file_lock
.fl_type
= F_UNLCK
;
4195 file_lock
.fl_owner
= (fl_owner_t
) stp
->st_stateowner
;
4196 file_lock
.fl_pid
= current
->tgid
;
4197 file_lock
.fl_file
= filp
;
4198 file_lock
.fl_flags
= FL_POSIX
;
4199 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4200 file_lock
.fl_start
= locku
->lu_offset
;
4202 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4203 nfs4_transform_lock_offset(&file_lock
);
4206 * Try to unlock the file in the VFS.
4208 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4210 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4214 * OK, unlock succeeded; the only thing left to do is update the stateid.
4216 update_stateid(&stp
->st_stateid
);
4217 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
4220 nfs4_unlock_state();
4224 status
= nfserrno(err
);
4230 * 1: locks held by lockowner
4231 * 0: no locks held by lockowner
4234 check_for_locks(struct nfs4_file
*filp
, struct nfs4_stateowner
*lowner
)
4236 struct file_lock
**flpp
;
4237 struct inode
*inode
= filp
->fi_inode
;
4241 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4242 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4253 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4254 struct nfsd4_compound_state
*cstate
,
4255 struct nfsd4_release_lockowner
*rlockowner
)
4257 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4258 struct nfs4_stateowner
*sop
;
4259 struct nfs4_stateid
*stp
;
4260 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4261 struct list_head matches
;
4265 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4266 clid
->cl_boot
, clid
->cl_id
);
4268 /* XXX check for lease expiration */
4270 status
= nfserr_stale_clientid
;
4271 if (STALE_CLIENTID(clid
))
4276 status
= nfserr_locks_held
;
4277 /* XXX: we're doing a linear search through all the lockowners.
4278 * Yipes! For now we'll just hope clients aren't really using
4279 * release_lockowner much, but eventually we have to fix these
4280 * data structures. */
4281 INIT_LIST_HEAD(&matches
);
4282 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4283 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
4284 if (!same_owner_str(sop
, owner
, clid
))
4286 list_for_each_entry(stp
, &sop
->so_stateids
,
4288 if (check_for_locks(stp
->st_file
, sop
))
4290 /* Note: so_perclient unused for lockowners,
4291 * so it's OK to fool with here. */
4292 list_add(&sop
->so_perclient
, &matches
);
4296 /* Clients probably won't expect us to return with some (but not all)
4297 * of the lockowner state released; so don't release any until all
4298 * have been checked. */
4300 while (!list_empty(&matches
)) {
4301 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
4303 /* unhash_stateowner deletes so_perclient only
4304 * for openowners. */
4305 list_del(&sop
->so_perclient
);
4306 release_lockowner(sop
);
4309 nfs4_unlock_state();
4313 static inline struct nfs4_client_reclaim
*
4316 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4320 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4322 unsigned int strhashval
= clientstr_hashval(name
);
4323 struct nfs4_client
*clp
;
4325 clp
= find_confirmed_client_by_str(name
, strhashval
);
4330 * failure => all reset bets are off, nfserr_no_grace...
4333 nfs4_client_to_reclaim(const char *name
)
4335 unsigned int strhashval
;
4336 struct nfs4_client_reclaim
*crp
= NULL
;
4338 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4339 crp
= alloc_reclaim();
4342 strhashval
= clientstr_hashval(name
);
4343 INIT_LIST_HEAD(&crp
->cr_strhash
);
4344 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4345 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4346 reclaim_str_hashtbl_size
++;
4351 nfs4_release_reclaim(void)
4353 struct nfs4_client_reclaim
*crp
= NULL
;
4356 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4357 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4358 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4359 struct nfs4_client_reclaim
, cr_strhash
);
4360 list_del(&crp
->cr_strhash
);
4362 reclaim_str_hashtbl_size
--;
4365 BUG_ON(reclaim_str_hashtbl_size
);
4369 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4370 static struct nfs4_client_reclaim
*
4371 nfs4_find_reclaim_client(clientid_t
*clid
)
4373 unsigned int strhashval
;
4374 struct nfs4_client
*clp
;
4375 struct nfs4_client_reclaim
*crp
= NULL
;
4378 /* find clientid in conf_id_hashtbl */
4379 clp
= find_confirmed_client(clid
);
4383 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4384 clp
->cl_name
.len
, clp
->cl_name
.data
,
4387 /* find clp->cl_name in reclaim_str_hashtbl */
4388 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4389 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4390 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4398 * Called from OPEN. Look for clientid in reclaim list.
4401 nfs4_check_open_reclaim(clientid_t
*clid
)
4403 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4406 /* initialization to perform at module load time: */
4409 nfs4_state_init(void)
4413 status
= nfsd4_init_slabs();
4416 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4417 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4418 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4419 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4420 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4421 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4423 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4424 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4425 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4426 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4428 for (i
= 0; i
< OPEN_OWNER_HASH_SIZE
; i
++) {
4429 INIT_LIST_HEAD(&open_ownerstr_hashtbl
[i
]);
4430 INIT_LIST_HEAD(&open_ownerid_hashtbl
[i
]);
4432 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++)
4433 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4434 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4435 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4436 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4438 memset(&onestateid
, ~0, sizeof(stateid_t
));
4439 INIT_LIST_HEAD(&close_lru
);
4440 INIT_LIST_HEAD(&client_lru
);
4441 INIT_LIST_HEAD(&del_recall_lru
);
4442 reclaim_str_hashtbl_size
= 0;
4447 nfsd4_load_reboot_recovery_data(void)
4452 nfsd4_init_recdir();
4453 status
= nfsd4_recdir_load();
4454 nfs4_unlock_state();
4456 printk("NFSD: Failure reading reboot recovery data\n");
4460 * Since the lifetime of a delegation isn't limited to that of an open, a
4461 * client may quite reasonably hang on to a delegation as long as it has
4462 * the inode cached. This becomes an obvious problem the first time a
4463 * client's inode cache approaches the size of the server's total memory.
4465 * For now we avoid this problem by imposing a hard limit on the number
4466 * of delegations, which varies according to the server's memory size.
4469 set_max_delegations(void)
4472 * Allow at most 4 delegations per megabyte of RAM. Quick
4473 * estimates suggest that in the worst case (where every delegation
4474 * is for a different inode), a delegation could take about 1.5K,
4475 * giving a worst case usage of about 6% of memory.
4477 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4480 /* initialization to perform when the nfsd service is started: */
4483 __nfs4_state_start(void)
4487 boot_time
= get_seconds();
4488 locks_start_grace(&nfsd4_manager
);
4489 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4491 ret
= set_callback_cred();
4494 laundry_wq
= create_singlethread_workqueue("nfsd4");
4495 if (laundry_wq
== NULL
)
4497 ret
= nfsd4_create_callback_queue();
4499 goto out_free_laundry
;
4500 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4501 set_max_delegations();
4504 destroy_workqueue(laundry_wq
);
4509 nfs4_state_start(void)
4511 nfsd4_load_reboot_recovery_data();
4512 return __nfs4_state_start();
4516 __nfs4_state_shutdown(void)
4519 struct nfs4_client
*clp
= NULL
;
4520 struct nfs4_delegation
*dp
= NULL
;
4521 struct list_head
*pos
, *next
, reaplist
;
4523 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4524 while (!list_empty(&conf_id_hashtbl
[i
])) {
4525 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4528 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4529 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4533 INIT_LIST_HEAD(&reaplist
);
4534 spin_lock(&recall_lock
);
4535 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4536 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4537 list_move(&dp
->dl_recall_lru
, &reaplist
);
4539 spin_unlock(&recall_lock
);
4540 list_for_each_safe(pos
, next
, &reaplist
) {
4541 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4542 list_del_init(&dp
->dl_recall_lru
);
4543 unhash_delegation(dp
);
4546 nfsd4_shutdown_recdir();
4550 nfs4_state_shutdown(void)
4552 cancel_delayed_work_sync(&laundromat_work
);
4553 destroy_workqueue(laundry_wq
);
4554 locks_end_grace(&nfsd4_manager
);
4556 nfs4_release_reclaim();
4557 __nfs4_state_shutdown();
4558 nfs4_unlock_state();
4559 nfsd4_destroy_callback_queue();