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_lockowner
*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
*openowner_slab
= NULL
;
81 static struct kmem_cache
*lockowner_slab
= NULL
;
82 static struct kmem_cache
*file_slab
= NULL
;
83 static struct kmem_cache
*stateid_slab
= NULL
;
84 static struct kmem_cache
*deleg_slab
= NULL
;
89 mutex_lock(&client_mutex
);
93 nfs4_unlock_state(void)
95 mutex_unlock(&client_mutex
);
99 opaque_hashval(const void *ptr
, int nbytes
)
101 unsigned char *cptr
= (unsigned char *) ptr
;
111 static struct list_head del_recall_lru
;
114 put_nfs4_file(struct nfs4_file
*fi
)
116 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
117 list_del(&fi
->fi_hash
);
118 spin_unlock(&recall_lock
);
120 kmem_cache_free(file_slab
, fi
);
125 get_nfs4_file(struct nfs4_file
*fi
)
127 atomic_inc(&fi
->fi_ref
);
130 static int num_delegations
;
131 unsigned int max_delegations
;
134 * Open owner state (share locks)
137 /* hash tables for open owners */
138 #define OPEN_OWNER_HASH_BITS 8
139 #define OPEN_OWNER_HASH_SIZE (1 << OPEN_OWNER_HASH_BITS)
140 #define OPEN_OWNER_HASH_MASK (OPEN_OWNER_HASH_SIZE - 1)
142 static unsigned int open_ownerid_hashval(const u32 id
)
144 return id
& OPEN_OWNER_HASH_MASK
;
147 static unsigned int open_ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
151 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
153 return ret
& OPEN_OWNER_HASH_MASK
;
156 static struct list_head open_ownerid_hashtbl
[OPEN_OWNER_HASH_SIZE
];
157 static struct list_head open_ownerstr_hashtbl
[OPEN_OWNER_HASH_SIZE
];
159 /* hash table for nfs4_file */
160 #define FILE_HASH_BITS 8
161 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
163 /* hash table for (open)nfs4_stateid */
164 #define STATEID_HASH_BITS 10
165 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
166 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
168 static unsigned int file_hashval(struct inode
*ino
)
170 /* XXX: why are we hashing on inode pointer, anyway? */
171 return hash_ptr(ino
, FILE_HASH_BITS
);
174 static unsigned int stateid_hashval(u32 owner_id
, u32 file_id
)
176 return (owner_id
+ file_id
) & STATEID_HASH_MASK
;
179 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
180 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
182 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
184 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
185 atomic_inc(&fp
->fi_access
[oflag
]);
188 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
190 if (oflag
== O_RDWR
) {
191 __nfs4_file_get_access(fp
, O_RDONLY
);
192 __nfs4_file_get_access(fp
, O_WRONLY
);
194 __nfs4_file_get_access(fp
, oflag
);
197 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
199 if (fp
->fi_fds
[oflag
]) {
200 fput(fp
->fi_fds
[oflag
]);
201 fp
->fi_fds
[oflag
] = NULL
;
205 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
207 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
208 nfs4_file_put_fd(fp
, O_RDWR
);
209 nfs4_file_put_fd(fp
, oflag
);
213 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
215 if (oflag
== O_RDWR
) {
216 __nfs4_file_put_access(fp
, O_RDONLY
);
217 __nfs4_file_put_access(fp
, O_WRONLY
);
219 __nfs4_file_put_access(fp
, oflag
);
222 static struct nfs4_delegation
*
223 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
225 struct nfs4_delegation
*dp
;
226 struct nfs4_file
*fp
= stp
->st_file
;
228 dprintk("NFSD alloc_init_deleg\n");
230 * Major work on the lease subsystem (for example, to support
231 * calbacks on stat) will be required before we can support
232 * write delegations properly.
234 if (type
!= NFS4_OPEN_DELEGATE_READ
)
236 if (fp
->fi_had_conflict
)
238 if (num_delegations
> max_delegations
)
240 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
244 INIT_LIST_HEAD(&dp
->dl_perfile
);
245 INIT_LIST_HEAD(&dp
->dl_perclnt
);
246 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
251 dp
->dl_stateid
.si_boot
= boot_time
;
252 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
253 dp
->dl_stateid
.si_fileid
= 0;
254 dp
->dl_stateid
.si_generation
= 1;
255 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
257 atomic_set(&dp
->dl_count
, 1);
258 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
263 nfs4_put_delegation(struct nfs4_delegation
*dp
)
265 if (atomic_dec_and_test(&dp
->dl_count
)) {
266 dprintk("NFSD: freeing dp %p\n",dp
);
267 put_nfs4_file(dp
->dl_file
);
268 kmem_cache_free(deleg_slab
, dp
);
273 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
275 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
276 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
278 fput(fp
->fi_deleg_file
);
279 fp
->fi_deleg_file
= NULL
;
283 /* Called under the state lock. */
285 unhash_delegation(struct nfs4_delegation
*dp
)
287 list_del_init(&dp
->dl_perclnt
);
288 spin_lock(&recall_lock
);
289 list_del_init(&dp
->dl_perfile
);
290 list_del_init(&dp
->dl_recall_lru
);
291 spin_unlock(&recall_lock
);
292 nfs4_put_deleg_lease(dp
->dl_file
);
293 nfs4_put_delegation(dp
);
300 /* client_lock protects the client lru list and session hash table */
301 static DEFINE_SPINLOCK(client_lock
);
303 /* Hash tables for nfs4_clientid state */
304 #define CLIENT_HASH_BITS 4
305 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
306 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
308 static unsigned int clientid_hashval(u32 id
)
310 return id
& CLIENT_HASH_MASK
;
313 static unsigned int clientstr_hashval(const char *name
)
315 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
319 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
320 * used in reboot/reset lease grace period processing
322 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
323 * setclientid_confirmed info.
325 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
328 * client_lru holds client queue ordered by nfs4_client.cl_time
331 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
332 * for last close replay.
334 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
335 static int reclaim_str_hashtbl_size
= 0;
336 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
337 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
338 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
339 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
340 static struct list_head client_lru
;
341 static struct list_head close_lru
;
344 * We store the NONE, READ, WRITE, and BOTH bits separately in the
345 * st_{access,deny}_bmap field of the stateid, in order to track not
346 * only what share bits are currently in force, but also what
347 * combinations of share bits previous opens have used. This allows us
348 * to enforce the recommendation of rfc 3530 14.2.19 that the server
349 * return an error if the client attempt to downgrade to a combination
350 * of share bits not explicable by closing some of its previous opens.
352 * XXX: This enforcement is actually incomplete, since we don't keep
353 * track of access/deny bit combinations; so, e.g., we allow:
355 * OPEN allow read, deny write
356 * OPEN allow both, deny none
357 * DOWNGRADE allow read, deny none
359 * which we should reject.
362 set_access(unsigned int *access
, unsigned long bmap
) {
366 for (i
= 1; i
< 4; i
++) {
367 if (test_bit(i
, &bmap
))
373 set_deny(unsigned int *deny
, unsigned long bmap
) {
377 for (i
= 0; i
< 4; i
++) {
378 if (test_bit(i
, &bmap
))
384 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
385 unsigned int access
, deny
;
387 set_access(&access
, stp
->st_access_bmap
);
388 set_deny(&deny
, stp
->st_deny_bmap
);
389 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
394 static int nfs4_access_to_omode(u32 access
)
396 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
397 case NFS4_SHARE_ACCESS_READ
:
399 case NFS4_SHARE_ACCESS_WRITE
:
401 case NFS4_SHARE_ACCESS_BOTH
:
407 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
409 list_del(&stp
->st_hash
);
410 list_del(&stp
->st_perfile
);
411 list_del(&stp
->st_perstateowner
);
414 static void free_generic_stateid(struct nfs4_stateid
*stp
)
418 if (stp
->st_access_bmap
) {
419 for (i
= 1; i
< 4; i
++) {
420 if (test_bit(i
, &stp
->st_access_bmap
))
421 nfs4_file_put_access(stp
->st_file
,
422 nfs4_access_to_omode(i
));
425 put_nfs4_file(stp
->st_file
);
426 kmem_cache_free(stateid_slab
, stp
);
429 static void release_lock_stateid(struct nfs4_stateid
*stp
)
433 unhash_generic_stateid(stp
);
434 file
= find_any_file(stp
->st_file
);
436 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
437 free_generic_stateid(stp
);
440 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
442 struct nfs4_stateid
*stp
;
444 list_del(&lo
->lo_owner
.so_idhash
);
445 list_del(&lo
->lo_owner
.so_strhash
);
446 list_del(&lo
->lo_perstateid
);
447 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
448 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
449 struct nfs4_stateid
, st_perstateowner
);
450 release_lock_stateid(stp
);
454 static void release_lockowner(struct nfs4_lockowner
*lo
)
456 unhash_lockowner(lo
);
457 nfs4_free_lockowner(lo
);
461 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
463 struct nfs4_lockowner
*lo
;
465 while (!list_empty(&open_stp
->st_lockowners
)) {
466 lo
= list_entry(open_stp
->st_lockowners
.next
,
467 struct nfs4_lockowner
, lo_perstateid
);
468 release_lockowner(lo
);
472 static void release_open_stateid(struct nfs4_stateid
*stp
)
474 unhash_generic_stateid(stp
);
475 release_stateid_lockowners(stp
);
476 free_generic_stateid(stp
);
479 static void unhash_openowner(struct nfs4_openowner
*oo
)
481 struct nfs4_stateid
*stp
;
483 list_del(&oo
->oo_owner
.so_idhash
);
484 list_del(&oo
->oo_owner
.so_strhash
);
485 list_del(&oo
->oo_perclient
);
486 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
487 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
488 struct nfs4_stateid
, st_perstateowner
);
489 release_open_stateid(stp
);
493 static void release_openowner(struct nfs4_openowner
*oo
)
495 unhash_openowner(oo
);
496 list_del(&oo
->oo_close_lru
);
497 nfs4_free_openowner(oo
);
500 #define SESSION_HASH_SIZE 512
501 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
504 hash_sessionid(struct nfs4_sessionid
*sessionid
)
506 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
508 return sid
->sequence
% SESSION_HASH_SIZE
;
512 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
514 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
515 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
519 gen_sessionid(struct nfsd4_session
*ses
)
521 struct nfs4_client
*clp
= ses
->se_client
;
522 struct nfsd4_sessionid
*sid
;
524 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
525 sid
->clientid
= clp
->cl_clientid
;
526 sid
->sequence
= current_sessionid
++;
531 * The protocol defines ca_maxresponssize_cached to include the size of
532 * the rpc header, but all we need to cache is the data starting after
533 * the end of the initial SEQUENCE operation--the rest we regenerate
534 * each time. Therefore we can advertise a ca_maxresponssize_cached
535 * value that is the number of bytes in our cache plus a few additional
536 * bytes. In order to stay on the safe side, and not promise more than
537 * we can cache, those additional bytes must be the minimum possible: 24
538 * bytes of rpc header (xid through accept state, with AUTH_NULL
539 * verifier), 12 for the compound header (with zero-length tag), and 44
540 * for the SEQUENCE op response:
542 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
545 free_session_slots(struct nfsd4_session
*ses
)
549 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
550 kfree(ses
->se_slots
[i
]);
554 * We don't actually need to cache the rpc and session headers, so we
555 * can allocate a little less for each slot:
557 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
559 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
562 static int nfsd4_sanitize_slot_size(u32 size
)
564 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
565 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
571 * XXX: If we run out of reserved DRC memory we could (up to a point)
572 * re-negotiate active sessions and reduce their slot usage to make
573 * rooom for new connections. For now we just fail the create session.
575 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
579 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
581 spin_lock(&nfsd_drc_lock
);
582 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
583 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
584 num
= min_t(int, num
, avail
/ slotsize
);
585 nfsd_drc_mem_used
+= num
* slotsize
;
586 spin_unlock(&nfsd_drc_lock
);
591 static void nfsd4_put_drc_mem(int slotsize
, int num
)
593 spin_lock(&nfsd_drc_lock
);
594 nfsd_drc_mem_used
-= slotsize
* num
;
595 spin_unlock(&nfsd_drc_lock
);
598 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
600 struct nfsd4_session
*new;
603 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
604 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
605 mem
= numslots
* sizeof(struct nfsd4_slot
*);
607 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
610 /* allocate each struct nfsd4_slot and data cache in one piece */
611 for (i
= 0; i
< numslots
; i
++) {
612 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
613 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
614 if (!new->se_slots
[i
])
620 kfree(new->se_slots
[i
]);
625 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
627 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
629 new->maxreqs
= numslots
;
630 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
631 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
632 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
633 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
634 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
637 static void free_conn(struct nfsd4_conn
*c
)
639 svc_xprt_put(c
->cn_xprt
);
643 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
645 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
646 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
648 spin_lock(&clp
->cl_lock
);
649 if (!list_empty(&c
->cn_persession
)) {
650 list_del(&c
->cn_persession
);
653 spin_unlock(&clp
->cl_lock
);
654 nfsd4_probe_callback(clp
);
657 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
659 struct nfsd4_conn
*conn
;
661 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
664 svc_xprt_get(rqstp
->rq_xprt
);
665 conn
->cn_xprt
= rqstp
->rq_xprt
;
666 conn
->cn_flags
= flags
;
667 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
671 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
673 conn
->cn_session
= ses
;
674 list_add(&conn
->cn_persession
, &ses
->se_conns
);
677 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
679 struct nfs4_client
*clp
= ses
->se_client
;
681 spin_lock(&clp
->cl_lock
);
682 __nfsd4_hash_conn(conn
, ses
);
683 spin_unlock(&clp
->cl_lock
);
686 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
688 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
689 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
692 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
694 struct nfsd4_conn
*conn
;
697 conn
= alloc_conn(rqstp
, dir
);
699 return nfserr_jukebox
;
700 nfsd4_hash_conn(conn
, ses
);
701 ret
= nfsd4_register_conn(conn
);
703 /* oops; xprt is already down: */
704 nfsd4_conn_lost(&conn
->cn_xpt_user
);
708 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
710 u32 dir
= NFS4_CDFC4_FORE
;
712 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
713 dir
|= NFS4_CDFC4_BACK
;
715 return nfsd4_new_conn(rqstp
, ses
, dir
);
718 /* must be called under client_lock */
719 static void nfsd4_del_conns(struct nfsd4_session
*s
)
721 struct nfs4_client
*clp
= s
->se_client
;
722 struct nfsd4_conn
*c
;
724 spin_lock(&clp
->cl_lock
);
725 while (!list_empty(&s
->se_conns
)) {
726 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
727 list_del_init(&c
->cn_persession
);
728 spin_unlock(&clp
->cl_lock
);
730 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
733 spin_lock(&clp
->cl_lock
);
735 spin_unlock(&clp
->cl_lock
);
738 void free_session(struct kref
*kref
)
740 struct nfsd4_session
*ses
;
743 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
744 nfsd4_del_conns(ses
);
745 spin_lock(&nfsd_drc_lock
);
746 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
747 nfsd_drc_mem_used
-= mem
;
748 spin_unlock(&nfsd_drc_lock
);
749 free_session_slots(ses
);
753 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
755 struct nfsd4_session
*new;
756 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
757 int numslots
, slotsize
;
762 * Note decreasing slot size below client's request may
763 * make it difficult for client to function correctly, whereas
764 * decreasing the number of slots will (just?) affect
765 * performance. When short on memory we therefore prefer to
766 * decrease number of slots instead of their size.
768 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
769 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
773 new = alloc_session(slotsize
, numslots
);
775 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
778 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
780 new->se_client
= clp
;
783 INIT_LIST_HEAD(&new->se_conns
);
785 new->se_cb_seq_nr
= 1;
786 new->se_flags
= cses
->flags
;
787 new->se_cb_prog
= cses
->callback_prog
;
788 kref_init(&new->se_ref
);
789 idx
= hash_sessionid(&new->se_sessionid
);
790 spin_lock(&client_lock
);
791 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
792 spin_lock(&clp
->cl_lock
);
793 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
794 spin_unlock(&clp
->cl_lock
);
795 spin_unlock(&client_lock
);
797 status
= nfsd4_new_conn_from_crses(rqstp
, new);
798 /* whoops: benny points out, status is ignored! (err, or bogus) */
800 free_session(&new->se_ref
);
803 if (cses
->flags
& SESSION4_BACK_CHAN
) {
804 struct sockaddr
*sa
= svc_addr(rqstp
);
806 * This is a little silly; with sessions there's no real
807 * use for the callback address. Use the peer address
808 * as a reasonable default for now, but consider fixing
809 * the rpc client not to require an address in the
812 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
813 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
815 nfsd4_probe_callback(clp
);
819 /* caller must hold client_lock */
820 static struct nfsd4_session
*
821 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
823 struct nfsd4_session
*elem
;
826 dump_sessionid(__func__
, sessionid
);
827 idx
= hash_sessionid(sessionid
);
828 /* Search in the appropriate list */
829 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
830 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
831 NFS4_MAX_SESSIONID_LEN
)) {
836 dprintk("%s: session not found\n", __func__
);
840 /* caller must hold client_lock */
842 unhash_session(struct nfsd4_session
*ses
)
844 list_del(&ses
->se_hash
);
845 spin_lock(&ses
->se_client
->cl_lock
);
846 list_del(&ses
->se_perclnt
);
847 spin_unlock(&ses
->se_client
->cl_lock
);
850 /* must be called under the client_lock */
852 renew_client_locked(struct nfs4_client
*clp
)
854 if (is_client_expired(clp
)) {
855 dprintk("%s: client (clientid %08x/%08x) already expired\n",
857 clp
->cl_clientid
.cl_boot
,
858 clp
->cl_clientid
.cl_id
);
863 * Move client to the end to the LRU list.
865 dprintk("renewing client (clientid %08x/%08x)\n",
866 clp
->cl_clientid
.cl_boot
,
867 clp
->cl_clientid
.cl_id
);
868 list_move_tail(&clp
->cl_lru
, &client_lru
);
869 clp
->cl_time
= get_seconds();
873 renew_client(struct nfs4_client
*clp
)
875 spin_lock(&client_lock
);
876 renew_client_locked(clp
);
877 spin_unlock(&client_lock
);
880 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
882 STALE_CLIENTID(clientid_t
*clid
)
884 if (clid
->cl_boot
== boot_time
)
886 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
887 clid
->cl_boot
, clid
->cl_id
, boot_time
);
892 * XXX Should we use a slab cache ?
893 * This type of memory management is somewhat inefficient, but we use it
894 * anyway since SETCLIENTID is not a common operation.
896 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
898 struct nfs4_client
*clp
;
900 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
903 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
904 if (clp
->cl_name
.data
== NULL
) {
908 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
909 clp
->cl_name
.len
= name
.len
;
914 free_client(struct nfs4_client
*clp
)
916 while (!list_empty(&clp
->cl_sessions
)) {
917 struct nfsd4_session
*ses
;
918 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
920 list_del(&ses
->se_perclnt
);
921 nfsd4_put_session(ses
);
923 if (clp
->cl_cred
.cr_group_info
)
924 put_group_info(clp
->cl_cred
.cr_group_info
);
925 kfree(clp
->cl_principal
);
926 kfree(clp
->cl_name
.data
);
931 release_session_client(struct nfsd4_session
*session
)
933 struct nfs4_client
*clp
= session
->se_client
;
935 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
937 if (is_client_expired(clp
)) {
939 session
->se_client
= NULL
;
941 renew_client_locked(clp
);
942 spin_unlock(&client_lock
);
945 /* must be called under the client_lock */
947 unhash_client_locked(struct nfs4_client
*clp
)
949 struct nfsd4_session
*ses
;
951 mark_client_expired(clp
);
952 list_del(&clp
->cl_lru
);
953 spin_lock(&clp
->cl_lock
);
954 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
955 list_del_init(&ses
->se_hash
);
956 spin_unlock(&clp
->cl_lock
);
960 expire_client(struct nfs4_client
*clp
)
962 struct nfs4_openowner
*oo
;
963 struct nfs4_delegation
*dp
;
964 struct list_head reaplist
;
966 INIT_LIST_HEAD(&reaplist
);
967 spin_lock(&recall_lock
);
968 while (!list_empty(&clp
->cl_delegations
)) {
969 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
970 list_del_init(&dp
->dl_perclnt
);
971 list_move(&dp
->dl_recall_lru
, &reaplist
);
973 spin_unlock(&recall_lock
);
974 while (!list_empty(&reaplist
)) {
975 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
976 list_del_init(&dp
->dl_recall_lru
);
977 unhash_delegation(dp
);
979 while (!list_empty(&clp
->cl_openowners
)) {
980 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
981 release_openowner(oo
);
983 nfsd4_shutdown_callback(clp
);
984 if (clp
->cl_cb_conn
.cb_xprt
)
985 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
986 list_del(&clp
->cl_idhash
);
987 list_del(&clp
->cl_strhash
);
988 spin_lock(&client_lock
);
989 unhash_client_locked(clp
);
990 if (atomic_read(&clp
->cl_refcount
) == 0)
992 spin_unlock(&client_lock
);
995 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
997 memcpy(target
->cl_verifier
.data
, source
->data
,
998 sizeof(target
->cl_verifier
.data
));
1001 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1003 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1004 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1007 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1009 target
->cr_uid
= source
->cr_uid
;
1010 target
->cr_gid
= source
->cr_gid
;
1011 target
->cr_group_info
= source
->cr_group_info
;
1012 get_group_info(target
->cr_group_info
);
1015 static int same_name(const char *n1
, const char *n2
)
1017 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1021 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1023 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1027 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1029 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1032 /* XXX what about NGROUP */
1034 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1036 return cr1
->cr_uid
== cr2
->cr_uid
;
1039 static void gen_clid(struct nfs4_client
*clp
)
1041 static u32 current_clientid
= 1;
1043 clp
->cl_clientid
.cl_boot
= boot_time
;
1044 clp
->cl_clientid
.cl_id
= current_clientid
++;
1047 static void gen_confirm(struct nfs4_client
*clp
)
1052 p
= (u32
*)clp
->cl_confirm
.data
;
1053 *p
++ = get_seconds();
1057 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1058 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1060 struct nfs4_client
*clp
;
1061 struct sockaddr
*sa
= svc_addr(rqstp
);
1064 clp
= alloc_client(name
);
1068 INIT_LIST_HEAD(&clp
->cl_sessions
);
1070 princ
= svc_gss_principal(rqstp
);
1072 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1073 if (clp
->cl_principal
== NULL
) {
1079 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1080 atomic_set(&clp
->cl_refcount
, 0);
1081 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1082 INIT_LIST_HEAD(&clp
->cl_idhash
);
1083 INIT_LIST_HEAD(&clp
->cl_strhash
);
1084 INIT_LIST_HEAD(&clp
->cl_openowners
);
1085 INIT_LIST_HEAD(&clp
->cl_delegations
);
1086 INIT_LIST_HEAD(&clp
->cl_lru
);
1087 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1088 spin_lock_init(&clp
->cl_lock
);
1089 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1090 clp
->cl_time
= get_seconds();
1091 clear_bit(0, &clp
->cl_cb_slot_busy
);
1092 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1093 copy_verf(clp
, verf
);
1094 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1095 clp
->cl_flavor
= rqstp
->rq_flavor
;
1096 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1098 clp
->cl_cb_session
= NULL
;
1102 static int check_name(struct xdr_netobj name
)
1106 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
1107 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
1114 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1116 unsigned int idhashval
;
1118 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1119 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1120 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1125 move_to_confirmed(struct nfs4_client
*clp
)
1127 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1128 unsigned int strhashval
;
1130 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1131 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1132 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1133 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1137 static struct nfs4_client
*
1138 find_confirmed_client(clientid_t
*clid
)
1140 struct nfs4_client
*clp
;
1141 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1143 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1144 if (same_clid(&clp
->cl_clientid
, clid
))
1150 static struct nfs4_client
*
1151 find_unconfirmed_client(clientid_t
*clid
)
1153 struct nfs4_client
*clp
;
1154 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1156 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1157 if (same_clid(&clp
->cl_clientid
, clid
))
1163 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1165 return clp
->cl_exchange_flags
!= 0;
1168 static struct nfs4_client
*
1169 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1171 struct nfs4_client
*clp
;
1173 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1174 if (same_name(clp
->cl_recdir
, dname
))
1180 static struct nfs4_client
*
1181 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1183 struct nfs4_client
*clp
;
1185 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1186 if (same_name(clp
->cl_recdir
, dname
))
1192 static void rpc_svcaddr2sockaddr(struct sockaddr
*sa
, unsigned short family
, union svc_addr_u
*svcaddr
)
1196 ((struct sockaddr_in
*)sa
)->sin_family
= AF_INET
;
1197 ((struct sockaddr_in
*)sa
)->sin_addr
= svcaddr
->addr
;
1200 ((struct sockaddr_in6
*)sa
)->sin6_family
= AF_INET6
;
1201 ((struct sockaddr_in6
*)sa
)->sin6_addr
= svcaddr
->addr6
;
1207 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1209 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1210 struct sockaddr
*sa
= svc_addr(rqstp
);
1211 u32 scopeid
= rpc_get_scope_id(sa
);
1212 unsigned short expected_family
;
1214 /* Currently, we only support tcp and tcp6 for the callback channel */
1215 if (se
->se_callback_netid_len
== 3 &&
1216 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1217 expected_family
= AF_INET
;
1218 else if (se
->se_callback_netid_len
== 4 &&
1219 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1220 expected_family
= AF_INET6
;
1224 conn
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
1225 se
->se_callback_addr_len
,
1226 (struct sockaddr
*)&conn
->cb_addr
,
1227 sizeof(conn
->cb_addr
));
1229 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1232 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1233 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1235 conn
->cb_prog
= se
->se_callback_prog
;
1236 conn
->cb_ident
= se
->se_callback_ident
;
1237 rpc_svcaddr2sockaddr((struct sockaddr
*)&conn
->cb_saddr
, expected_family
, &rqstp
->rq_daddr
);
1240 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1241 conn
->cb_addrlen
= 0;
1242 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1243 "will not receive delegations\n",
1244 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1250 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1253 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1255 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1258 dprintk("--> %s slot %p\n", __func__
, slot
);
1260 slot
->sl_opcnt
= resp
->opcnt
;
1261 slot
->sl_status
= resp
->cstate
.status
;
1263 if (nfsd4_not_cached(resp
)) {
1264 slot
->sl_datalen
= 0;
1267 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1268 base
= (char *)resp
->cstate
.datap
-
1269 (char *)resp
->xbuf
->head
[0].iov_base
;
1270 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1272 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1277 * Encode the replay sequence operation from the slot values.
1278 * If cachethis is FALSE encode the uncached rep error on the next
1279 * operation which sets resp->p and increments resp->opcnt for
1280 * nfs4svc_encode_compoundres.
1284 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1285 struct nfsd4_compoundres
*resp
)
1287 struct nfsd4_op
*op
;
1288 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1290 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1291 resp
->opcnt
, resp
->cstate
.slot
->sl_cachethis
);
1293 /* Encode the replayed sequence operation */
1294 op
= &args
->ops
[resp
->opcnt
- 1];
1295 nfsd4_encode_operation(resp
, op
);
1297 /* Return nfserr_retry_uncached_rep in next operation. */
1298 if (args
->opcnt
> 1 && slot
->sl_cachethis
== 0) {
1299 op
= &args
->ops
[resp
->opcnt
++];
1300 op
->status
= nfserr_retry_uncached_rep
;
1301 nfsd4_encode_operation(resp
, op
);
1307 * The sequence operation is not cached because we can use the slot and
1311 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1312 struct nfsd4_sequence
*seq
)
1314 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1317 dprintk("--> %s slot %p\n", __func__
, slot
);
1319 /* Either returns 0 or nfserr_retry_uncached */
1320 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1321 if (status
== nfserr_retry_uncached_rep
)
1324 /* The sequence operation has been encoded, cstate->datap set. */
1325 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1327 resp
->opcnt
= slot
->sl_opcnt
;
1328 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1329 status
= slot
->sl_status
;
1335 * Set the exchange_id flags returned by the server.
1338 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1340 /* pNFS is not supported */
1341 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1343 /* Referrals are supported, Migration is not. */
1344 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1346 /* set the wire flags to return to client. */
1347 clid
->flags
= new->cl_exchange_flags
;
1351 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1352 struct nfsd4_compound_state
*cstate
,
1353 struct nfsd4_exchange_id
*exid
)
1355 struct nfs4_client
*unconf
, *conf
, *new;
1357 unsigned int strhashval
;
1358 char dname
[HEXDIR_LEN
];
1359 char addr_str
[INET6_ADDRSTRLEN
];
1360 nfs4_verifier verf
= exid
->verifier
;
1361 struct sockaddr
*sa
= svc_addr(rqstp
);
1363 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1364 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1365 "ip_addr=%s flags %x, spa_how %d\n",
1366 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1367 addr_str
, exid
->flags
, exid
->spa_how
);
1369 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1370 return nfserr_inval
;
1372 /* Currently only support SP4_NONE */
1373 switch (exid
->spa_how
) {
1377 return nfserr_serverfault
;
1379 BUG(); /* checked by xdr code */
1381 return nfserr_serverfault
; /* no excuse :-/ */
1384 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1389 strhashval
= clientstr_hashval(dname
);
1394 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1396 if (!clp_used_exchangeid(conf
)) {
1397 status
= nfserr_clid_inuse
; /* XXX: ? */
1400 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1401 /* 18.35.4 case 8 */
1402 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1403 status
= nfserr_not_same
;
1406 /* Client reboot: destroy old state */
1407 expire_client(conf
);
1410 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1411 /* 18.35.4 case 9 */
1412 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1413 status
= nfserr_perm
;
1416 expire_client(conf
);
1420 * Set bit when the owner id and verifier map to an already
1421 * confirmed client id (18.35.3).
1423 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1426 * Falling into 18.35.4 case 2, possible router replay.
1427 * Leave confirmed record intact and return same result.
1429 copy_verf(conf
, &verf
);
1434 /* 18.35.4 case 7 */
1435 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1436 status
= nfserr_noent
;
1440 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1443 * Possible retry or client restart. Per 18.35.4 case 4,
1444 * a new unconfirmed record should be generated regardless
1445 * of whether any properties have changed.
1447 expire_client(unconf
);
1452 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1454 status
= nfserr_jukebox
;
1459 add_to_unconfirmed(new, strhashval
);
1461 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1462 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1465 nfsd4_set_ex_flags(new, exid
);
1467 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1468 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1472 nfs4_unlock_state();
1474 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1479 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1481 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1484 /* The slot is in use, and no response has been sent. */
1486 if (seqid
== slot_seqid
)
1487 return nfserr_jukebox
;
1489 return nfserr_seq_misordered
;
1492 if (likely(seqid
== slot_seqid
+ 1))
1495 if (seqid
== slot_seqid
)
1496 return nfserr_replay_cache
;
1498 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1500 /* Misordered replay or misordered new request */
1501 return nfserr_seq_misordered
;
1505 * Cache the create session result into the create session single DRC
1506 * slot cache by saving the xdr structure. sl_seqid has been set.
1507 * Do this for solo or embedded create session operations.
1510 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1511 struct nfsd4_clid_slot
*slot
, int nfserr
)
1513 slot
->sl_status
= nfserr
;
1514 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1518 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1519 struct nfsd4_clid_slot
*slot
)
1521 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1522 return slot
->sl_status
;
1525 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1526 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1527 1 + /* MIN tag is length with zero, only length */ \
1528 3 + /* version, opcount, opcode */ \
1529 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1530 /* seqid, slotID, slotID, cache */ \
1531 4 ) * sizeof(__be32))
1533 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1534 2 + /* verifier: AUTH_NULL, length 0 */\
1536 1 + /* MIN tag is length with zero, only length */ \
1537 3 + /* opcount, opcode, opstatus*/ \
1538 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1539 /* seqid, slotID, slotID, slotID, status */ \
1540 5 ) * sizeof(__be32))
1542 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1544 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1545 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1549 nfsd4_create_session(struct svc_rqst
*rqstp
,
1550 struct nfsd4_compound_state
*cstate
,
1551 struct nfsd4_create_session
*cr_ses
)
1553 struct sockaddr
*sa
= svc_addr(rqstp
);
1554 struct nfs4_client
*conf
, *unconf
;
1555 struct nfsd4_session
*new;
1556 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1557 bool confirm_me
= false;
1560 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1561 return nfserr_inval
;
1564 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1565 conf
= find_confirmed_client(&cr_ses
->clientid
);
1568 cs_slot
= &conf
->cl_cs_slot
;
1569 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1570 if (status
== nfserr_replay_cache
) {
1571 dprintk("Got a create_session replay! seqid= %d\n",
1573 /* Return the cached reply status */
1574 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1576 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1577 status
= nfserr_seq_misordered
;
1578 dprintk("Sequence misordered!\n");
1579 dprintk("Expected seqid= %d but got seqid= %d\n",
1580 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1583 } else if (unconf
) {
1584 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1585 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1586 status
= nfserr_clid_inuse
;
1590 cs_slot
= &unconf
->cl_cs_slot
;
1591 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1593 /* an unconfirmed replay returns misordered */
1594 status
= nfserr_seq_misordered
;
1601 status
= nfserr_stale_clientid
;
1606 * XXX: we should probably set this at creation time, and check
1607 * for consistent minorversion use throughout:
1609 conf
->cl_minorversion
= 1;
1611 * We do not support RDMA or persistent sessions
1613 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1614 cr_ses
->flags
&= ~SESSION4_RDMA
;
1616 status
= nfserr_toosmall
;
1617 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1620 status
= nfserr_jukebox
;
1621 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1625 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1626 NFS4_MAX_SESSIONID_LEN
);
1627 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1628 sizeof(struct nfsd4_channel_attrs
));
1629 cs_slot
->sl_seqid
++;
1630 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1632 /* cache solo and embedded create sessions under the state lock */
1633 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1635 move_to_confirmed(conf
);
1637 nfs4_unlock_state();
1638 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1642 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1644 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1645 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1647 return argp
->opcnt
== resp
->opcnt
;
1650 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1653 case NFS4_CDFC4_FORE
:
1654 case NFS4_CDFC4_BACK
:
1656 case NFS4_CDFC4_FORE_OR_BOTH
:
1657 case NFS4_CDFC4_BACK_OR_BOTH
:
1658 *dir
= NFS4_CDFC4_BOTH
;
1661 return nfserr_inval
;
1664 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1665 struct nfsd4_compound_state
*cstate
,
1666 struct nfsd4_bind_conn_to_session
*bcts
)
1670 if (!nfsd4_last_compound_op(rqstp
))
1671 return nfserr_not_only_op
;
1672 spin_lock(&client_lock
);
1673 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1674 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1675 * client_lock iself: */
1676 if (cstate
->session
) {
1677 nfsd4_get_session(cstate
->session
);
1678 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1680 spin_unlock(&client_lock
);
1681 if (!cstate
->session
)
1682 return nfserr_badsession
;
1684 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1686 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1690 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1694 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1698 nfsd4_destroy_session(struct svc_rqst
*r
,
1699 struct nfsd4_compound_state
*cstate
,
1700 struct nfsd4_destroy_session
*sessionid
)
1702 struct nfsd4_session
*ses
;
1703 u32 status
= nfserr_badsession
;
1706 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1707 * - Should we return nfserr_back_chan_busy if waiting for
1708 * callbacks on to-be-destroyed session?
1709 * - Do we need to clear any callback info from previous session?
1712 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1713 if (!nfsd4_last_compound_op(r
))
1714 return nfserr_not_only_op
;
1716 dump_sessionid(__func__
, &sessionid
->sessionid
);
1717 spin_lock(&client_lock
);
1718 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1720 spin_unlock(&client_lock
);
1724 unhash_session(ses
);
1725 spin_unlock(&client_lock
);
1728 nfsd4_probe_callback_sync(ses
->se_client
);
1729 nfs4_unlock_state();
1731 nfsd4_del_conns(ses
);
1733 nfsd4_put_session(ses
);
1736 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1740 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1742 struct nfsd4_conn
*c
;
1744 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1745 if (c
->cn_xprt
== xpt
) {
1752 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1754 struct nfs4_client
*clp
= ses
->se_client
;
1755 struct nfsd4_conn
*c
;
1758 spin_lock(&clp
->cl_lock
);
1759 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1761 spin_unlock(&clp
->cl_lock
);
1765 __nfsd4_hash_conn(new, ses
);
1766 spin_unlock(&clp
->cl_lock
);
1767 ret
= nfsd4_register_conn(new);
1769 /* oops; xprt is already down: */
1770 nfsd4_conn_lost(&new->cn_xpt_user
);
1774 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1776 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1778 return args
->opcnt
> session
->se_fchannel
.maxops
;
1781 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1782 struct nfsd4_session
*session
)
1784 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1786 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1790 nfsd4_sequence(struct svc_rqst
*rqstp
,
1791 struct nfsd4_compound_state
*cstate
,
1792 struct nfsd4_sequence
*seq
)
1794 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1795 struct nfsd4_session
*session
;
1796 struct nfsd4_slot
*slot
;
1797 struct nfsd4_conn
*conn
;
1800 if (resp
->opcnt
!= 1)
1801 return nfserr_sequence_pos
;
1804 * Will be either used or freed by nfsd4_sequence_check_conn
1807 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1809 return nfserr_jukebox
;
1811 spin_lock(&client_lock
);
1812 status
= nfserr_badsession
;
1813 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1817 status
= nfserr_too_many_ops
;
1818 if (nfsd4_session_too_many_ops(rqstp
, session
))
1821 status
= nfserr_req_too_big
;
1822 if (nfsd4_request_too_big(rqstp
, session
))
1825 status
= nfserr_badslot
;
1826 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1829 slot
= session
->se_slots
[seq
->slotid
];
1830 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1832 /* We do not negotiate the number of slots yet, so set the
1833 * maxslots to the session maxreqs which is used to encode
1834 * sr_highest_slotid and the sr_target_slot id to maxslots */
1835 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1837 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1838 if (status
== nfserr_replay_cache
) {
1839 cstate
->slot
= slot
;
1840 cstate
->session
= session
;
1841 /* Return the cached reply status and set cstate->status
1842 * for nfsd4_proc_compound processing */
1843 status
= nfsd4_replay_cache_entry(resp
, seq
);
1844 cstate
->status
= nfserr_replay_cache
;
1850 nfsd4_sequence_check_conn(conn
, session
);
1853 /* Success! bump slot seqid */
1854 slot
->sl_inuse
= true;
1855 slot
->sl_seqid
= seq
->seqid
;
1856 slot
->sl_cachethis
= seq
->cachethis
;
1858 cstate
->slot
= slot
;
1859 cstate
->session
= session
;
1862 /* Hold a session reference until done processing the compound. */
1863 if (cstate
->session
) {
1864 struct nfs4_client
*clp
= session
->se_client
;
1866 nfsd4_get_session(cstate
->session
);
1867 atomic_inc(&clp
->cl_refcount
);
1868 if (clp
->cl_cb_state
== NFSD4_CB_DOWN
)
1869 seq
->status_flags
|= SEQ4_STATUS_CB_PATH_DOWN
;
1872 spin_unlock(&client_lock
);
1873 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1878 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
1882 if (rc
->rca_one_fs
) {
1883 if (!cstate
->current_fh
.fh_dentry
)
1884 return nfserr_nofilehandle
;
1886 * We don't take advantage of the rca_one_fs case.
1887 * That's OK, it's optional, we can safely ignore it.
1893 status
= nfserr_complete_already
;
1894 if (cstate
->session
->se_client
->cl_firststate
)
1897 status
= nfserr_stale_clientid
;
1898 if (is_client_expired(cstate
->session
->se_client
))
1900 * The following error isn't really legal.
1901 * But we only get here if the client just explicitly
1902 * destroyed the client. Surely it no longer cares what
1903 * error it gets back on an operation for the dead
1909 nfsd4_create_clid_dir(cstate
->session
->se_client
);
1911 nfs4_unlock_state();
1916 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1917 struct nfsd4_setclientid
*setclid
)
1919 struct xdr_netobj clname
= {
1920 .len
= setclid
->se_namelen
,
1921 .data
= setclid
->se_name
,
1923 nfs4_verifier clverifier
= setclid
->se_verf
;
1924 unsigned int strhashval
;
1925 struct nfs4_client
*conf
, *unconf
, *new;
1927 char dname
[HEXDIR_LEN
];
1929 if (!check_name(clname
))
1930 return nfserr_inval
;
1932 status
= nfs4_make_rec_clidname(dname
, &clname
);
1937 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1938 * We get here on a DRC miss.
1941 strhashval
= clientstr_hashval(dname
);
1944 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1946 /* RFC 3530 14.2.33 CASE 0: */
1947 status
= nfserr_clid_inuse
;
1948 if (clp_used_exchangeid(conf
))
1950 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1951 char addr_str
[INET6_ADDRSTRLEN
];
1952 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1954 dprintk("NFSD: setclientid: string in use by client "
1955 "at %s\n", addr_str
);
1960 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1961 * has a description of SETCLIENTID request processing consisting
1962 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1964 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1965 status
= nfserr_jukebox
;
1968 * RFC 3530 14.2.33 CASE 4:
1969 * placed first, because it is the normal case
1972 expire_client(unconf
);
1973 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1977 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1979 * RFC 3530 14.2.33 CASE 1:
1980 * probable callback update
1983 /* Note this is removing unconfirmed {*x***},
1984 * which is stronger than RFC recommended {vxc**}.
1985 * This has the advantage that there is at most
1986 * one {*x***} in either list at any time.
1988 expire_client(unconf
);
1990 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1993 copy_clid(new, conf
);
1994 } else if (!unconf
) {
1996 * RFC 3530 14.2.33 CASE 2:
1997 * probable client reboot; state will be removed if
2000 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2006 * RFC 3530 14.2.33 CASE 3:
2007 * probable client reboot; state will be removed if
2010 expire_client(unconf
);
2011 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2017 * XXX: we should probably set this at creation time, and check
2018 * for consistent minorversion use throughout:
2020 new->cl_minorversion
= 0;
2021 gen_callback(new, setclid
, rqstp
);
2022 add_to_unconfirmed(new, strhashval
);
2023 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2024 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2025 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2028 nfs4_unlock_state();
2034 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2035 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2036 * bullets, labeled as CASE1 - CASE4 below.
2039 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2040 struct nfsd4_compound_state
*cstate
,
2041 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2043 struct sockaddr
*sa
= svc_addr(rqstp
);
2044 struct nfs4_client
*conf
, *unconf
;
2045 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2046 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2049 if (STALE_CLIENTID(clid
))
2050 return nfserr_stale_clientid
;
2052 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2053 * We get here on a DRC miss.
2058 conf
= find_confirmed_client(clid
);
2059 unconf
= find_unconfirmed_client(clid
);
2061 status
= nfserr_clid_inuse
;
2062 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
2064 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
2068 * section 14.2.34 of RFC 3530 has a description of
2069 * SETCLIENTID_CONFIRM request processing consisting
2070 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2072 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
2074 * RFC 3530 14.2.34 CASE 1:
2077 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
2078 status
= nfserr_clid_inuse
;
2080 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2081 nfsd4_probe_callback(conf
);
2082 expire_client(unconf
);
2086 } else if (conf
&& !unconf
) {
2088 * RFC 3530 14.2.34 CASE 2:
2089 * probable retransmitted request; play it safe and
2092 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2093 status
= nfserr_clid_inuse
;
2096 } else if (!conf
&& unconf
2097 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
2099 * RFC 3530 14.2.34 CASE 3:
2100 * Normal case; new or rebooted client:
2102 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
2103 status
= nfserr_clid_inuse
;
2106 clientstr_hashval(unconf
->cl_recdir
);
2107 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
2110 nfsd4_remove_clid_dir(conf
);
2111 expire_client(conf
);
2113 move_to_confirmed(unconf
);
2115 nfsd4_probe_callback(conf
);
2118 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
2119 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
2122 * RFC 3530 14.2.34 CASE 4:
2123 * Client probably hasn't noticed that we rebooted yet.
2125 status
= nfserr_stale_clientid
;
2127 /* check that we have hit one of the cases...*/
2128 status
= nfserr_clid_inuse
;
2131 nfs4_unlock_state();
2135 /* OPEN Share state helper functions */
2136 static inline struct nfs4_file
*
2137 alloc_init_file(struct inode
*ino
)
2139 struct nfs4_file
*fp
;
2140 unsigned int hashval
= file_hashval(ino
);
2142 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2144 atomic_set(&fp
->fi_ref
, 1);
2145 INIT_LIST_HEAD(&fp
->fi_hash
);
2146 INIT_LIST_HEAD(&fp
->fi_stateids
);
2147 INIT_LIST_HEAD(&fp
->fi_delegations
);
2148 fp
->fi_inode
= igrab(ino
);
2149 fp
->fi_id
= current_fileid
++;
2150 fp
->fi_had_conflict
= false;
2151 fp
->fi_lease
= NULL
;
2152 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2153 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2154 spin_lock(&recall_lock
);
2155 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2156 spin_unlock(&recall_lock
);
2163 nfsd4_free_slab(struct kmem_cache
**slab
)
2167 kmem_cache_destroy(*slab
);
2172 nfsd4_free_slabs(void)
2174 nfsd4_free_slab(&openowner_slab
);
2175 nfsd4_free_slab(&lockowner_slab
);
2176 nfsd4_free_slab(&file_slab
);
2177 nfsd4_free_slab(&stateid_slab
);
2178 nfsd4_free_slab(&deleg_slab
);
2182 nfsd4_init_slabs(void)
2184 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2185 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2186 if (openowner_slab
== NULL
)
2188 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2189 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2190 if (lockowner_slab
== NULL
)
2192 file_slab
= kmem_cache_create("nfsd4_files",
2193 sizeof(struct nfs4_file
), 0, 0, NULL
);
2194 if (file_slab
== NULL
)
2196 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2197 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
2198 if (stateid_slab
== NULL
)
2200 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2201 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2202 if (deleg_slab
== NULL
)
2207 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2211 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2213 kfree(oo
->oo_owner
.so_owner
.data
);
2214 kmem_cache_free(openowner_slab
, oo
);
2217 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2219 kfree(lo
->lo_owner
.so_owner
.data
);
2220 kmem_cache_free(lockowner_slab
, lo
);
2223 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2225 rp
->rp_status
= nfserr_serverfault
;
2227 rp
->rp_buf
= rp
->rp_ibuf
;
2230 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2232 struct nfs4_stateowner
*sop
;
2234 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2238 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2239 if (!sop
->so_owner
.data
) {
2240 kmem_cache_free(slab
, sop
);
2243 sop
->so_owner
.len
= owner
->len
;
2245 INIT_LIST_HEAD(&sop
->so_stateids
);
2246 sop
->so_id
= current_ownerid
++;
2247 sop
->so_client
= clp
;
2248 init_nfs4_replay(&sop
->so_replay
);
2252 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2254 unsigned int idhashval
;
2256 idhashval
= open_ownerid_hashval(oo
->oo_owner
.so_id
);
2257 list_add(&oo
->oo_owner
.so_idhash
, &open_ownerid_hashtbl
[idhashval
]);
2258 list_add(&oo
->oo_owner
.so_strhash
, &open_ownerstr_hashtbl
[strhashval
]);
2259 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2262 static struct nfs4_openowner
*
2263 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2264 struct nfs4_openowner
*oo
;
2266 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2269 oo
->oo_owner
.so_is_open_owner
= 1;
2270 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2271 oo
->oo_confirmed
= 0;
2273 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2274 hash_openowner(oo
, clp
, strhashval
);
2279 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2280 struct nfs4_openowner
*oo
= open
->op_openowner
;
2281 unsigned int hashval
= stateid_hashval(oo
->oo_owner
.so_id
, fp
->fi_id
);
2283 INIT_LIST_HEAD(&stp
->st_hash
);
2284 INIT_LIST_HEAD(&stp
->st_perstateowner
);
2285 INIT_LIST_HEAD(&stp
->st_lockowners
);
2286 INIT_LIST_HEAD(&stp
->st_perfile
);
2287 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
2288 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2289 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2290 stp
->st_type
= NFS4_OPEN_STID
;
2291 stp
->st_stateowner
= &oo
->oo_owner
;
2294 stp
->st_stateid
.si_boot
= boot_time
;
2295 stp
->st_stateid
.si_stateownerid
= oo
->oo_owner
.so_id
;
2296 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
2297 /* note will be incremented before first return to client: */
2298 stp
->st_stateid
.si_generation
= 0;
2299 stp
->st_access_bmap
= 0;
2300 stp
->st_deny_bmap
= 0;
2301 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
2302 &stp
->st_access_bmap
);
2303 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2304 stp
->st_openstp
= NULL
;
2308 move_to_close_lru(struct nfs4_openowner
*oo
)
2310 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2312 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2313 oo
->oo_time
= get_seconds();
2317 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2320 return (sop
->so_owner
.len
== owner
->len
) &&
2321 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2322 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2325 static struct nfs4_openowner
*
2326 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2328 struct nfs4_stateowner
*so
= NULL
;
2330 list_for_each_entry(so
, &open_ownerstr_hashtbl
[hashval
], so_strhash
) {
2331 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
2332 return container_of(so
, struct nfs4_openowner
, oo_owner
);
2337 /* search file_hashtbl[] for file */
2338 static struct nfs4_file
*
2339 find_file(struct inode
*ino
)
2341 unsigned int hashval
= file_hashval(ino
);
2342 struct nfs4_file
*fp
;
2344 spin_lock(&recall_lock
);
2345 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2346 if (fp
->fi_inode
== ino
) {
2348 spin_unlock(&recall_lock
);
2352 spin_unlock(&recall_lock
);
2356 static inline int access_valid(u32 x
, u32 minorversion
)
2358 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
2360 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
2362 x
&= ~NFS4_SHARE_ACCESS_MASK
;
2363 if (minorversion
&& x
) {
2364 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
2366 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
2368 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
2375 static inline int deny_valid(u32 x
)
2377 /* Note: unlike access bits, deny bits may be zero. */
2378 return x
<= NFS4_SHARE_DENY_BOTH
;
2382 * Called to check deny when READ with all zero stateid or
2383 * WRITE with all zero or all one stateid
2386 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2388 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2389 struct nfs4_file
*fp
;
2390 struct nfs4_stateid
*stp
;
2393 dprintk("NFSD: nfs4_share_conflict\n");
2395 fp
= find_file(ino
);
2398 ret
= nfserr_locked
;
2399 /* Search for conflicting share reservations */
2400 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2401 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2402 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2411 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2413 /* We're assuming the state code never drops its reference
2414 * without first removing the lease. Since we're in this lease
2415 * callback (and since the lease code is serialized by the kernel
2416 * lock) we know the server hasn't removed the lease yet, we know
2417 * it's safe to take a reference: */
2418 atomic_inc(&dp
->dl_count
);
2420 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2422 /* only place dl_time is set. protected by lock_flocks*/
2423 dp
->dl_time
= get_seconds();
2425 nfsd4_cb_recall(dp
);
2428 /* Called from break_lease() with lock_flocks() held. */
2429 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2431 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2432 struct nfs4_delegation
*dp
;
2435 /* We assume break_lease is only called once per lease: */
2436 BUG_ON(fp
->fi_had_conflict
);
2438 * We don't want the locks code to timeout the lease for us;
2439 * we'll remove it ourself if a delegation isn't returned
2442 fl
->fl_break_time
= 0;
2444 spin_lock(&recall_lock
);
2445 fp
->fi_had_conflict
= true;
2446 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2447 nfsd_break_one_deleg(dp
);
2448 spin_unlock(&recall_lock
);
2452 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2455 return lease_modify(onlist
, arg
);
2460 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2461 .lm_break
= nfsd_break_deleg_cb
,
2462 .lm_change
= nfsd_change_deleg_cb
,
2465 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2467 if (nfsd4_has_session(cstate
))
2469 if (seqid
== so
->so_seqid
- 1)
2470 return nfserr_replay_me
;
2471 if (seqid
== so
->so_seqid
)
2473 return nfserr_bad_seqid
;
2477 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2478 struct nfsd4_open
*open
)
2480 clientid_t
*clientid
= &open
->op_clientid
;
2481 struct nfs4_client
*clp
= NULL
;
2482 unsigned int strhashval
;
2483 struct nfs4_openowner
*oo
= NULL
;
2486 if (!check_name(open
->op_owner
))
2487 return nfserr_inval
;
2489 if (STALE_CLIENTID(&open
->op_clientid
))
2490 return nfserr_stale_clientid
;
2492 strhashval
= open_ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2493 oo
= find_openstateowner_str(strhashval
, open
);
2494 open
->op_openowner
= oo
;
2496 /* Make sure the client's lease hasn't expired. */
2497 clp
= find_confirmed_client(clientid
);
2499 return nfserr_expired
;
2502 if (!oo
->oo_confirmed
) {
2503 /* Replace unconfirmed owners without checking for replay. */
2504 clp
= oo
->oo_owner
.so_client
;
2505 release_openowner(oo
);
2506 open
->op_openowner
= NULL
;
2509 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2513 if (open
->op_openowner
== NULL
) {
2514 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2516 return nfserr_jukebox
;
2517 open
->op_openowner
= oo
;
2519 list_del_init(&oo
->oo_close_lru
);
2520 renew_client(oo
->oo_owner
.so_client
);
2524 static inline __be32
2525 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2527 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2528 return nfserr_openmode
;
2533 static struct nfs4_delegation
*
2534 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2536 struct nfs4_delegation
*dp
;
2538 spin_lock(&recall_lock
);
2539 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2540 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
) {
2541 spin_unlock(&recall_lock
);
2544 spin_unlock(&recall_lock
);
2548 static int share_access_to_flags(u32 share_access
)
2550 share_access
&= ~NFS4_SHARE_WANT_MASK
;
2552 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2556 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2557 struct nfs4_delegation
**dp
)
2560 __be32 status
= nfserr_bad_stateid
;
2562 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2565 flags
= share_access_to_flags(open
->op_share_access
);
2566 status
= nfs4_check_delegmode(*dp
, flags
);
2570 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2574 open
->op_openowner
->oo_confirmed
= 1;
2579 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2581 struct nfs4_stateid
*local
;
2582 struct nfs4_openowner
*oo
= open
->op_openowner
;
2584 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2585 /* ignore lock owners */
2586 if (local
->st_stateowner
->so_is_open_owner
== 0)
2588 /* remember if we have seen this open owner */
2589 if (local
->st_stateowner
== &oo
->oo_owner
)
2591 /* check for conflicting share reservations */
2592 if (!test_share(local
, open
))
2593 return nfserr_share_denied
;
2598 static inline struct nfs4_stateid
*
2599 nfs4_alloc_stateid(void)
2601 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2604 static inline int nfs4_access_to_access(u32 nfs4_access
)
2608 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2609 flags
|= NFSD_MAY_READ
;
2610 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2611 flags
|= NFSD_MAY_WRITE
;
2615 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2616 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2619 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2620 int access
= nfs4_access_to_access(open
->op_share_access
);
2622 /* CLAIM_DELEGATE_CUR is used in response to a broken lease;
2623 * allowing it to break the lease and return EAGAIN leaves the
2624 * client unable to make progress in returning the delegation */
2625 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2626 access
|= NFSD_MAY_NOT_BREAK_LEASE
;
2628 if (!fp
->fi_fds
[oflag
]) {
2629 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2630 &fp
->fi_fds
[oflag
]);
2634 nfs4_file_get_access(fp
, oflag
);
2640 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2641 struct nfs4_file
*fp
, struct svc_fh
*cur_fh
,
2642 struct nfsd4_open
*open
)
2644 struct nfs4_stateid
*stp
;
2647 stp
= nfs4_alloc_stateid();
2649 return nfserr_jukebox
;
2651 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2653 kmem_cache_free(stateid_slab
, stp
);
2660 static inline __be32
2661 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2662 struct nfsd4_open
*open
)
2664 struct iattr iattr
= {
2665 .ia_valid
= ATTR_SIZE
,
2668 if (!open
->op_truncate
)
2670 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2671 return nfserr_inval
;
2672 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2676 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2678 u32 op_share_access
= open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
;
2682 new_access
= !test_bit(op_share_access
, &stp
->st_access_bmap
);
2684 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2688 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2691 int oflag
= nfs4_access_to_omode(op_share_access
);
2692 nfs4_file_put_access(fp
, oflag
);
2696 /* remember the open */
2697 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2698 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2705 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2707 open
->op_openowner
->oo_confirmed
= 1;
2708 open
->op_openowner
->oo_owner
.so_client
->cl_firststate
= 1;
2711 /* Should we give out recallable state?: */
2712 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2714 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2717 * In the sessions case, since we don't have to establish a
2718 * separate connection for callbacks, we assume it's OK
2719 * until we hear otherwise:
2721 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2724 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2726 struct file_lock
*fl
;
2728 fl
= locks_alloc_lock();
2731 locks_init_lock(fl
);
2732 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2733 fl
->fl_flags
= FL_LEASE
;
2734 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2735 fl
->fl_end
= OFFSET_MAX
;
2736 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2737 fl
->fl_pid
= current
->tgid
;
2741 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2743 struct nfs4_file
*fp
= dp
->dl_file
;
2744 struct file_lock
*fl
;
2747 fl
= nfs4_alloc_init_lease(dp
, flag
);
2750 fl
->fl_file
= find_readable_file(fp
);
2751 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2752 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2754 list_del_init(&dp
->dl_perclnt
);
2755 locks_free_lock(fl
);
2759 fp
->fi_deleg_file
= fl
->fl_file
;
2760 get_file(fp
->fi_deleg_file
);
2761 atomic_set(&fp
->fi_delegees
, 1);
2762 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2766 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2768 struct nfs4_file
*fp
= dp
->dl_file
;
2771 return nfs4_setlease(dp
, flag
);
2772 spin_lock(&recall_lock
);
2773 if (fp
->fi_had_conflict
) {
2774 spin_unlock(&recall_lock
);
2777 atomic_inc(&fp
->fi_delegees
);
2778 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2779 spin_unlock(&recall_lock
);
2780 list_add(&dp
->dl_perclnt
, &dp
->dl_client
->cl_delegations
);
2785 * Attempt to hand out a delegation.
2788 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2790 struct nfs4_delegation
*dp
;
2791 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2793 int status
, flag
= 0;
2795 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2796 flag
= NFS4_OPEN_DELEGATE_NONE
;
2797 open
->op_recall
= 0;
2798 switch (open
->op_claim_type
) {
2799 case NFS4_OPEN_CLAIM_PREVIOUS
:
2801 open
->op_recall
= 1;
2802 flag
= open
->op_delegate_type
;
2803 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2806 case NFS4_OPEN_CLAIM_NULL
:
2807 /* Let's not give out any delegations till everyone's
2808 * had the chance to reclaim theirs.... */
2809 if (locks_in_grace())
2811 if (!cb_up
|| !oo
->oo_confirmed
)
2813 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2814 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2816 flag
= NFS4_OPEN_DELEGATE_READ
;
2822 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2825 status
= nfs4_set_delegation(dp
, flag
);
2829 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2831 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2832 STATEID_VAL(&dp
->dl_stateid
));
2834 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2835 && flag
== NFS4_OPEN_DELEGATE_NONE
2836 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2837 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2838 open
->op_delegate_type
= flag
;
2841 nfs4_put_delegation(dp
);
2843 flag
= NFS4_OPEN_DELEGATE_NONE
;
2848 * called with nfs4_lock_state() held.
2851 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2853 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2854 struct nfs4_file
*fp
= NULL
;
2855 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2856 struct nfs4_stateid
*stp
= NULL
;
2857 struct nfs4_delegation
*dp
= NULL
;
2860 status
= nfserr_inval
;
2861 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2862 || !deny_valid(open
->op_share_deny
))
2865 * Lookup file; if found, lookup stateid and check open request,
2866 * and check for delegations in the process of being recalled.
2867 * If not found, create the nfs4_file struct
2869 fp
= find_file(ino
);
2871 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2873 status
= nfs4_check_deleg(fp
, open
, &dp
);
2877 status
= nfserr_bad_stateid
;
2878 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2880 status
= nfserr_jukebox
;
2881 fp
= alloc_init_file(ino
);
2887 * OPEN the file, or upgrade an existing OPEN.
2888 * If truncate fails, the OPEN fails.
2891 /* Stateid was found, this is an OPEN upgrade */
2892 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
2896 status
= nfs4_new_open(rqstp
, &stp
, fp
, current_fh
, open
);
2899 init_stateid(stp
, fp
, open
);
2900 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2902 release_open_stateid(stp
);
2906 update_stateid(&stp
->st_stateid
);
2907 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2909 if (nfsd4_has_session(&resp
->cstate
))
2910 open
->op_openowner
->oo_confirmed
= 1;
2913 * Attempt to hand out a delegation. No error return, because the
2914 * OPEN succeeds even if we fail.
2916 nfs4_open_delegation(current_fh
, open
, stp
);
2920 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2921 STATEID_VAL(&stp
->st_stateid
));
2925 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2926 nfs4_set_claim_prev(open
);
2928 * To finish the open response, we just need to set the rflags.
2930 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2931 if (!open
->op_openowner
->oo_confirmed
&&
2932 !nfsd4_has_session(&resp
->cstate
))
2933 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2939 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2942 struct nfs4_client
*clp
;
2946 dprintk("process_renew(%08x/%08x): starting\n",
2947 clid
->cl_boot
, clid
->cl_id
);
2948 status
= nfserr_stale_clientid
;
2949 if (STALE_CLIENTID(clid
))
2951 clp
= find_confirmed_client(clid
);
2952 status
= nfserr_expired
;
2954 /* We assume the client took too long to RENEW. */
2955 dprintk("nfsd4_renew: clientid not found!\n");
2959 status
= nfserr_cb_path_down
;
2960 if (!list_empty(&clp
->cl_delegations
)
2961 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
2965 nfs4_unlock_state();
2969 static struct lock_manager nfsd4_manager
= {
2973 nfsd4_end_grace(void)
2975 dprintk("NFSD: end of grace period\n");
2976 nfsd4_recdir_purge_old();
2977 locks_end_grace(&nfsd4_manager
);
2979 * Now that every NFSv4 client has had the chance to recover and
2980 * to see the (possibly new, possibly shorter) lease time, we
2981 * can safely set the next grace time to the current lease time:
2983 nfsd4_grace
= nfsd4_lease
;
2987 nfs4_laundromat(void)
2989 struct nfs4_client
*clp
;
2990 struct nfs4_openowner
*oo
;
2991 struct nfs4_delegation
*dp
;
2992 struct list_head
*pos
, *next
, reaplist
;
2993 time_t cutoff
= get_seconds() - nfsd4_lease
;
2994 time_t t
, clientid_val
= nfsd4_lease
;
2995 time_t u
, test_val
= nfsd4_lease
;
2999 dprintk("NFSD: laundromat service - starting\n");
3000 if (locks_in_grace())
3002 INIT_LIST_HEAD(&reaplist
);
3003 spin_lock(&client_lock
);
3004 list_for_each_safe(pos
, next
, &client_lru
) {
3005 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3006 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
3007 t
= clp
->cl_time
- cutoff
;
3008 if (clientid_val
> t
)
3012 if (atomic_read(&clp
->cl_refcount
)) {
3013 dprintk("NFSD: client in use (clientid %08x)\n",
3014 clp
->cl_clientid
.cl_id
);
3017 unhash_client_locked(clp
);
3018 list_add(&clp
->cl_lru
, &reaplist
);
3020 spin_unlock(&client_lock
);
3021 list_for_each_safe(pos
, next
, &reaplist
) {
3022 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3023 dprintk("NFSD: purging unused client (clientid %08x)\n",
3024 clp
->cl_clientid
.cl_id
);
3025 nfsd4_remove_clid_dir(clp
);
3028 spin_lock(&recall_lock
);
3029 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3030 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3031 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3032 u
= dp
->dl_time
- cutoff
;
3037 list_move(&dp
->dl_recall_lru
, &reaplist
);
3039 spin_unlock(&recall_lock
);
3040 list_for_each_safe(pos
, next
, &reaplist
) {
3041 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3042 list_del_init(&dp
->dl_recall_lru
);
3043 unhash_delegation(dp
);
3045 test_val
= nfsd4_lease
;
3046 list_for_each_safe(pos
, next
, &close_lru
) {
3047 oo
= container_of(pos
, struct nfs4_openowner
, oo_close_lru
);
3048 if (time_after((unsigned long)oo
->oo_time
, (unsigned long)cutoff
)) {
3049 u
= oo
->oo_time
- cutoff
;
3054 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
3055 oo
->oo_owner
.so_id
);
3056 release_openowner(oo
);
3058 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3059 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3060 nfs4_unlock_state();
3061 return clientid_val
;
3064 static struct workqueue_struct
*laundry_wq
;
3065 static void laundromat_main(struct work_struct
*);
3066 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3069 laundromat_main(struct work_struct
*not_used
)
3073 t
= nfs4_laundromat();
3074 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3075 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3078 static struct nfs4_openowner
* search_close_lru(u32 st_id
)
3080 struct nfs4_openowner
*local
;
3082 list_for_each_entry(local
, &close_lru
, oo_close_lru
) {
3083 if (local
->oo_owner
.so_id
== st_id
)
3090 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
3092 return fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
;
3096 STALE_STATEID(stateid_t
*stateid
)
3098 if (stateid
->si_boot
== boot_time
)
3100 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3101 STATEID_VAL(stateid
));
3106 access_permit_read(unsigned long access_bmap
)
3108 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
3109 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
3110 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
3114 access_permit_write(unsigned long access_bmap
)
3116 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
3117 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
3121 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
3123 __be32 status
= nfserr_openmode
;
3125 /* For lock stateid's, we test the parent open, not the lock: */
3126 if (stp
->st_openstp
)
3127 stp
= stp
->st_openstp
;
3128 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
3130 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
3137 static inline __be32
3138 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3140 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3142 else if (locks_in_grace()) {
3143 /* Answer in remaining cases depends on existence of
3144 * conflicting state; so we must wait out the grace period. */
3145 return nfserr_grace
;
3146 } else if (flags
& WR_STATE
)
3147 return nfs4_share_conflict(current_fh
,
3148 NFS4_SHARE_DENY_WRITE
);
3149 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3150 return nfs4_share_conflict(current_fh
,
3151 NFS4_SHARE_DENY_READ
);
3155 * Allow READ/WRITE during grace period on recovered state only for files
3156 * that are not able to provide mandatory locking.
3159 grace_disallows_io(struct inode
*inode
)
3161 return locks_in_grace() && mandatory_lock(inode
);
3164 /* Returns true iff a is later than b: */
3165 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3167 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3170 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3173 * When sessions are used the stateid generation number is ignored
3176 if (has_session
&& in
->si_generation
== 0)
3179 if (in
->si_generation
== ref
->si_generation
)
3182 /* If the client sends us a stateid from the future, it's buggy: */
3183 if (stateid_generation_after(in
, ref
))
3184 return nfserr_bad_stateid
;
3186 * However, we could see a stateid from the past, even from a
3187 * non-buggy client. For example, if the client sends a lock
3188 * while some IO is outstanding, the lock may bump si_generation
3189 * while the IO is still in flight. The client could avoid that
3190 * situation by waiting for responses on all the IO requests,
3191 * but better performance may result in retrying IO that
3192 * receives an old_stateid error if requests are rarely
3193 * reordered in flight:
3195 return nfserr_old_stateid
;
3198 static int is_delegation_stateid(stateid_t
*stateid
)
3200 return stateid
->si_fileid
== 0;
3203 __be32
nfs4_validate_stateid(stateid_t
*stateid
, bool has_session
)
3205 struct nfs4_stateid
*stp
= NULL
;
3206 __be32 status
= nfserr_stale_stateid
;
3208 if (STALE_STATEID(stateid
))
3211 status
= nfserr_expired
;
3212 stp
= find_stateid(stateid
, 0);
3215 status
= nfserr_bad_stateid
;
3217 if (stp
->st_stateowner
->so_is_open_owner
3218 && !openowner(stp
->st_stateowner
)->oo_confirmed
)
3221 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, has_session
);
3231 * Checks for stateid operations
3234 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3235 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3237 struct nfs4_stateid
*stp
= NULL
;
3238 struct nfs4_delegation
*dp
= NULL
;
3239 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3240 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3246 if (grace_disallows_io(ino
))
3247 return nfserr_grace
;
3249 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3250 return check_special_stateids(current_fh
, stateid
, flags
);
3252 status
= nfserr_stale_stateid
;
3253 if (STALE_STATEID(stateid
))
3257 * We assume that any stateid that has the current boot time,
3258 * but that we can't find, is expired:
3260 status
= nfserr_expired
;
3261 if (is_delegation_stateid(stateid
)) {
3262 dp
= find_delegation_stateid(ino
, stateid
);
3265 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, nfsd4_has_session(cstate
));
3268 status
= nfs4_check_delegmode(dp
, flags
);
3271 renew_client(dp
->dl_client
);
3273 *filpp
= dp
->dl_file
->fi_deleg_file
;
3276 } else { /* open or lock stateid */
3277 stp
= find_stateid(stateid
, flags
);
3280 status
= nfserr_bad_stateid
;
3281 if (nfs4_check_fh(current_fh
, stp
))
3283 if (stp
->st_stateowner
->so_is_open_owner
3284 && !openowner(stp
->st_stateowner
)->oo_confirmed
)
3286 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
3287 nfsd4_has_session(cstate
));
3290 status
= nfs4_check_openmode(stp
, flags
);
3293 renew_client(stp
->st_stateowner
->so_client
);
3295 if (flags
& RD_STATE
)
3296 *filpp
= find_readable_file(stp
->st_file
);
3298 *filpp
= find_writeable_file(stp
->st_file
);
3307 nfsd4_free_delegation_stateid(stateid_t
*stateid
)
3309 struct nfs4_delegation
*dp
= search_for_delegation(stateid
);
3311 return nfserr_locks_held
;
3312 return nfserr_bad_stateid
;
3316 nfsd4_free_lock_stateid(struct nfs4_stateid
*stp
)
3318 if (check_for_locks(stp
->st_file
, lockowner(stp
->st_stateowner
)))
3319 return nfserr_locks_held
;
3320 release_lock_stateid(stp
);
3325 * Test if the stateid is valid
3328 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3329 struct nfsd4_test_stateid
*test_stateid
)
3331 test_stateid
->ts_has_session
= nfsd4_has_session(cstate
);
3339 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3340 struct nfsd4_free_stateid
*free_stateid
)
3342 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3343 struct nfs4_stateid
*stp
;
3347 if (is_delegation_stateid(stateid
)) {
3348 ret
= nfsd4_free_delegation_stateid(stateid
);
3352 stp
= find_stateid(stateid
, 0);
3354 ret
= nfserr_bad_stateid
;
3357 ret
= check_stateid_generation(stateid
, &stp
->st_stateid
, 1);
3361 if (stp
->st_type
== NFS4_OPEN_STID
) {
3362 ret
= nfserr_locks_held
;
3365 ret
= nfsd4_free_lock_stateid(stp
);
3370 nfs4_unlock_state();
3377 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3378 RD_STATE
: WR_STATE
;
3382 * Checks for sequence id mutating operations.
3385 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3386 stateid_t
*stateid
, int flags
,
3387 struct nfs4_stateid
**stpp
)
3389 struct nfs4_stateowner
*sop
;
3390 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3393 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3394 seqid
, STATEID_VAL(stateid
));
3398 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
3399 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3400 return nfserr_bad_stateid
;
3403 if (STALE_STATEID(stateid
))
3404 return nfserr_stale_stateid
;
3407 * We return BAD_STATEID if filehandle doesn't match stateid,
3408 * the confirmed flag is incorrecly set, or the generation
3409 * number is incorrect.
3411 *stpp
= find_stateid(stateid
, flags
);
3413 return nfserr_expired
;
3415 sop
= (*stpp
)->st_stateowner
;
3416 cstate
->replay_owner
= sop
;
3418 if (nfs4_check_fh(current_fh
, *stpp
)) {
3419 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3420 return nfserr_bad_stateid
;
3423 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
3427 if (sop
->so_is_open_owner
&& !openowner(sop
)->oo_confirmed
3428 && !(flags
& CONFIRM
)) {
3429 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3430 " confirmed yet!\n");
3431 return nfserr_bad_stateid
;
3433 status
= check_stateid_generation(stateid
, &(*stpp
)->st_stateid
, nfsd4_has_session(cstate
));
3436 renew_client(sop
->so_client
);
3441 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3442 struct nfsd4_open_confirm
*oc
)
3445 struct nfs4_openowner
*oo
;
3446 struct nfs4_stateid
*stp
;
3448 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3449 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3450 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3452 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3458 status
= nfs4_preprocess_seqid_op(cstate
,
3459 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3460 CONFIRM
| OPEN_STATE
, &stp
);
3463 oo
= openowner(stp
->st_stateowner
);
3464 status
= nfserr_bad_stateid
;
3465 if (oo
->oo_confirmed
)
3467 oo
->oo_confirmed
= 1;
3468 update_stateid(&stp
->st_stateid
);
3469 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3470 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3471 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stateid
));
3473 nfsd4_create_clid_dir(oo
->oo_owner
.so_client
);
3476 if (!cstate
->replay_owner
)
3477 nfs4_unlock_state();
3481 static inline void nfs4_file_downgrade(struct nfs4_stateid
*stp
, unsigned int to_access
)
3485 for (i
= 1; i
< 4; i
++) {
3486 if (test_bit(i
, &stp
->st_access_bmap
) && !(i
& to_access
)) {
3487 nfs4_file_put_access(stp
->st_file
, i
);
3488 __clear_bit(i
, &stp
->st_access_bmap
);
3494 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3497 for (i
= 0; i
< 4; i
++) {
3498 if ((i
& deny
) != i
)
3499 __clear_bit(i
, bmap
);
3504 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3505 struct nfsd4_compound_state
*cstate
,
3506 struct nfsd4_open_downgrade
*od
)
3509 struct nfs4_stateid
*stp
;
3511 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3512 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3513 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3515 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3516 || !deny_valid(od
->od_share_deny
))
3517 return nfserr_inval
;
3520 status
= nfs4_preprocess_seqid_op(cstate
, od
->od_seqid
,
3521 &od
->od_stateid
, OPEN_STATE
, &stp
);
3524 status
= nfserr_inval
;
3525 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3526 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3527 stp
->st_access_bmap
, od
->od_share_access
);
3530 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3531 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3532 stp
->st_deny_bmap
, od
->od_share_deny
);
3535 nfs4_file_downgrade(stp
, od
->od_share_access
);
3537 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3539 update_stateid(&stp
->st_stateid
);
3540 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3543 if (!cstate
->replay_owner
)
3544 nfs4_unlock_state();
3549 * nfs4_unlock_state() called after encode
3552 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3553 struct nfsd4_close
*close
)
3556 struct nfs4_openowner
*oo
;
3557 struct nfs4_stateid
*stp
;
3559 dprintk("NFSD: nfsd4_close on file %.*s\n",
3560 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3561 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3564 /* check close_lru for replay */
3565 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3568 if (stp
== NULL
&& status
== nfserr_expired
) {
3570 * Also, we should make sure this isn't just the result of
3573 oo
= search_close_lru(close
->cl_stateid
.si_stateownerid
);
3574 /* It's not stale; let's assume it's expired: */
3577 cstate
->replay_owner
= &oo
->oo_owner
;
3578 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, close
->cl_seqid
);
3581 status
= nfserr_bad_seqid
;
3585 oo
= openowner(stp
->st_stateowner
);
3587 update_stateid(&stp
->st_stateid
);
3588 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3590 /* release_stateid() calls nfsd_close() if needed */
3591 release_open_stateid(stp
);
3593 /* place unused nfs4_stateowners on so_close_lru list to be
3594 * released by the laundromat service after the lease period
3595 * to enable us to handle CLOSE replay
3597 if (list_empty(&oo
->oo_owner
.so_stateids
))
3598 move_to_close_lru(oo
);
3600 if (!cstate
->replay_owner
)
3601 nfs4_unlock_state();
3606 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3607 struct nfsd4_delegreturn
*dr
)
3609 struct nfs4_delegation
*dp
;
3610 stateid_t
*stateid
= &dr
->dr_stateid
;
3611 struct inode
*inode
;
3614 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3616 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3619 status
= nfserr_bad_stateid
;
3620 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3622 status
= nfserr_stale_stateid
;
3623 if (STALE_STATEID(stateid
))
3625 status
= nfserr_bad_stateid
;
3626 if (!is_delegation_stateid(stateid
))
3628 status
= nfserr_expired
;
3629 dp
= find_delegation_stateid(inode
, stateid
);
3632 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, nfsd4_has_session(cstate
));
3635 renew_client(dp
->dl_client
);
3637 unhash_delegation(dp
);
3639 nfs4_unlock_state();
3646 * Lock owner state (byte-range locks)
3648 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3649 #define LOCK_HASH_BITS 8
3650 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3651 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3654 end_offset(u64 start
, u64 len
)
3659 return end
>= start
? end
: NFS4_MAX_UINT64
;
3662 /* last octet in a range */
3664 last_byte_offset(u64 start
, u64 len
)
3670 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3673 static unsigned int lockownerid_hashval(u32 id
)
3675 return id
& LOCK_HASH_MASK
;
3678 static inline unsigned int
3679 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3680 struct xdr_netobj
*ownername
)
3682 return (file_hashval(inode
) + cl_id
3683 + opaque_hashval(ownername
->data
, ownername
->len
))
3687 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3688 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3691 same_stateid(stateid_t
*id_one
, stateid_t
*id_two
)
3693 if (id_one
->si_stateownerid
!= id_two
->si_stateownerid
)
3695 return id_one
->si_fileid
== id_two
->si_fileid
;
3698 static struct nfs4_stateid
*find_stateid(stateid_t
*t
, int flags
)
3700 struct nfs4_stateid
*s
;
3701 unsigned int hashval
;
3703 hashval
= stateid_hashval(t
->si_stateownerid
, t
->si_fileid
);
3704 list_for_each_entry(s
, &stateid_hashtbl
[hashval
], st_hash
) {
3705 if (!same_stateid(&s
->st_stateid
, t
))
3707 if (flags
& LOCK_STATE
&& s
->st_type
!= NFS4_LOCK_STID
)
3709 if (flags
& OPEN_STATE
&& s
->st_type
!= NFS4_OPEN_STID
)
3716 static struct nfs4_delegation
*
3717 search_for_delegation(stateid_t
*stid
)
3719 struct nfs4_file
*fp
;
3720 struct nfs4_delegation
*dp
;
3721 struct list_head
*pos
;
3724 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3725 list_for_each_entry(fp
, &file_hashtbl
[i
], fi_hash
) {
3726 list_for_each(pos
, &fp
->fi_delegations
) {
3727 dp
= list_entry(pos
, struct nfs4_delegation
, dl_perfile
);
3728 if (same_stateid(&dp
->dl_stateid
, stid
))
3736 static struct nfs4_delegation
*
3737 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3739 struct nfs4_file
*fp
;
3740 struct nfs4_delegation
*dl
;
3742 dprintk("NFSD: %s: stateid=" STATEID_FMT
"\n", __func__
,
3745 fp
= find_file(ino
);
3748 dl
= find_delegation_file(fp
, stid
);
3754 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3755 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3756 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3757 * locking, this prevents us from being completely protocol-compliant. The
3758 * real solution to this problem is to start using unsigned file offsets in
3759 * the VFS, but this is a very deep change!
3762 nfs4_transform_lock_offset(struct file_lock
*lock
)
3764 if (lock
->fl_start
< 0)
3765 lock
->fl_start
= OFFSET_MAX
;
3766 if (lock
->fl_end
< 0)
3767 lock
->fl_end
= OFFSET_MAX
;
3770 /* Hack!: For now, we're defining this just so we can use a pointer to it
3771 * as a unique cookie to identify our (NFSv4's) posix locks. */
3772 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3776 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3778 struct nfs4_lockowner
*lo
;
3780 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3781 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3782 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3783 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3784 if (!deny
->ld_owner
.data
)
3785 /* We just don't care that much */
3787 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3788 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3791 deny
->ld_owner
.len
= 0;
3792 deny
->ld_owner
.data
= NULL
;
3793 deny
->ld_clientid
.cl_boot
= 0;
3794 deny
->ld_clientid
.cl_id
= 0;
3796 deny
->ld_start
= fl
->fl_start
;
3797 deny
->ld_length
= NFS4_MAX_UINT64
;
3798 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3799 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3800 deny
->ld_type
= NFS4_READ_LT
;
3801 if (fl
->fl_type
!= F_RDLCK
)
3802 deny
->ld_type
= NFS4_WRITE_LT
;
3805 static struct nfs4_lockowner
*
3806 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3807 struct xdr_netobj
*owner
)
3809 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3810 struct nfs4_stateowner
*op
;
3812 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3813 if (same_owner_str(op
, owner
, clid
))
3814 return lockowner(op
);
3819 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
)
3821 unsigned int idhashval
;
3823 idhashval
= lockownerid_hashval(lo
->lo_owner
.so_id
);
3824 list_add(&lo
->lo_owner
.so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3825 list_add(&lo
->lo_owner
.so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3826 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3830 * Alloc a lock owner structure.
3831 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3834 * strhashval = lock_ownerstr_hashval
3837 static struct nfs4_lockowner
*
3838 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3839 struct nfs4_lockowner
*lo
;
3841 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
3844 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
3845 lo
->lo_owner
.so_is_open_owner
= 0;
3846 /* It is the openowner seqid that will be incremented in encode in the
3847 * case of new lockowners; so increment the lock seqid manually: */
3848 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3849 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
3853 static struct nfs4_stateid
*
3854 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3856 struct nfs4_stateid
*stp
;
3857 unsigned int hashval
= stateid_hashval(lo
->lo_owner
.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
, &lo
->lo_owner
.so_stateids
);
3869 stp
->st_stateowner
= &lo
->lo_owner
;
3870 stp
->st_type
= NFS4_LOCK_STID
;
3873 stp
->st_stateid
.si_boot
= boot_time
;
3874 stp
->st_stateid
.si_stateownerid
= lo
->lo_owner
.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_openowner
*open_sop
= NULL
;
3912 struct nfs4_lockowner
*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 open_sop
= openowner(open_stp
->st_stateowner
);
3959 status
= nfserr_bad_stateid
;
3960 if (!nfsd4_has_session(cstate
) &&
3961 !same_clid(&open_sop
->oo_owner
.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
->oo_owner
.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
->oo_owner
.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
= lockowner(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_lockowner
*lo
;
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 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4139 file_lock
.fl_owner
= (fl_owner_t
)lo
;
4140 file_lock
.fl_pid
= current
->tgid
;
4141 file_lock
.fl_flags
= FL_POSIX
;
4143 file_lock
.fl_start
= lockt
->lt_offset
;
4144 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4146 nfs4_transform_lock_offset(&file_lock
);
4149 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4151 status
= nfserrno(error
);
4154 if (file_lock
.fl_type
!= F_UNLCK
) {
4155 status
= nfserr_denied
;
4156 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4159 nfs4_unlock_state();
4164 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4165 struct nfsd4_locku
*locku
)
4167 struct nfs4_stateid
*stp
;
4168 struct file
*filp
= NULL
;
4169 struct file_lock file_lock
;
4173 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4174 (long long) locku
->lu_offset
,
4175 (long long) locku
->lu_length
);
4177 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4178 return nfserr_inval
;
4182 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4183 &locku
->lu_stateid
, LOCK_STATE
, &stp
);
4186 filp
= find_any_file(stp
->st_file
);
4188 status
= nfserr_lock_range
;
4192 locks_init_lock(&file_lock
);
4193 file_lock
.fl_type
= F_UNLCK
;
4194 file_lock
.fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4195 file_lock
.fl_pid
= current
->tgid
;
4196 file_lock
.fl_file
= filp
;
4197 file_lock
.fl_flags
= FL_POSIX
;
4198 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4199 file_lock
.fl_start
= locku
->lu_offset
;
4201 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4202 nfs4_transform_lock_offset(&file_lock
);
4205 * Try to unlock the file in the VFS.
4207 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4209 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4213 * OK, unlock succeeded; the only thing left to do is update the stateid.
4215 update_stateid(&stp
->st_stateid
);
4216 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
4219 nfs4_unlock_state();
4223 status
= nfserrno(err
);
4229 * 1: locks held by lockowner
4230 * 0: no locks held by lockowner
4233 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4235 struct file_lock
**flpp
;
4236 struct inode
*inode
= filp
->fi_inode
;
4240 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4241 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4252 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4253 struct nfsd4_compound_state
*cstate
,
4254 struct nfsd4_release_lockowner
*rlockowner
)
4256 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4257 struct nfs4_stateowner
*sop
;
4258 struct nfs4_lockowner
*lo
;
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 lo
= lockowner(sop
);
4289 if (check_for_locks(stp
->st_file
, lo
))
4291 list_add(&lo
->lo_list
, &matches
);
4295 /* Clients probably won't expect us to return with some (but not all)
4296 * of the lockowner state released; so don't release any until all
4297 * have been checked. */
4299 while (!list_empty(&matches
)) {
4300 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4302 /* unhash_stateowner deletes so_perclient only
4303 * for openowners. */
4304 list_del(&lo
->lo_list
);
4305 release_lockowner(lo
);
4308 nfs4_unlock_state();
4312 static inline struct nfs4_client_reclaim
*
4315 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4319 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4321 unsigned int strhashval
= clientstr_hashval(name
);
4322 struct nfs4_client
*clp
;
4324 clp
= find_confirmed_client_by_str(name
, strhashval
);
4329 * failure => all reset bets are off, nfserr_no_grace...
4332 nfs4_client_to_reclaim(const char *name
)
4334 unsigned int strhashval
;
4335 struct nfs4_client_reclaim
*crp
= NULL
;
4337 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4338 crp
= alloc_reclaim();
4341 strhashval
= clientstr_hashval(name
);
4342 INIT_LIST_HEAD(&crp
->cr_strhash
);
4343 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4344 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4345 reclaim_str_hashtbl_size
++;
4350 nfs4_release_reclaim(void)
4352 struct nfs4_client_reclaim
*crp
= NULL
;
4355 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4356 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4357 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4358 struct nfs4_client_reclaim
, cr_strhash
);
4359 list_del(&crp
->cr_strhash
);
4361 reclaim_str_hashtbl_size
--;
4364 BUG_ON(reclaim_str_hashtbl_size
);
4368 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4369 static struct nfs4_client_reclaim
*
4370 nfs4_find_reclaim_client(clientid_t
*clid
)
4372 unsigned int strhashval
;
4373 struct nfs4_client
*clp
;
4374 struct nfs4_client_reclaim
*crp
= NULL
;
4377 /* find clientid in conf_id_hashtbl */
4378 clp
= find_confirmed_client(clid
);
4382 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4383 clp
->cl_name
.len
, clp
->cl_name
.data
,
4386 /* find clp->cl_name in reclaim_str_hashtbl */
4387 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4388 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4389 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4397 * Called from OPEN. Look for clientid in reclaim list.
4400 nfs4_check_open_reclaim(clientid_t
*clid
)
4402 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4405 /* initialization to perform at module load time: */
4408 nfs4_state_init(void)
4412 status
= nfsd4_init_slabs();
4415 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4416 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4417 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4418 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4419 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4420 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4422 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4423 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4424 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4425 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4427 for (i
= 0; i
< OPEN_OWNER_HASH_SIZE
; i
++) {
4428 INIT_LIST_HEAD(&open_ownerstr_hashtbl
[i
]);
4429 INIT_LIST_HEAD(&open_ownerid_hashtbl
[i
]);
4431 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++)
4432 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4433 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4434 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4435 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4437 memset(&onestateid
, ~0, sizeof(stateid_t
));
4438 INIT_LIST_HEAD(&close_lru
);
4439 INIT_LIST_HEAD(&client_lru
);
4440 INIT_LIST_HEAD(&del_recall_lru
);
4441 reclaim_str_hashtbl_size
= 0;
4446 nfsd4_load_reboot_recovery_data(void)
4451 nfsd4_init_recdir();
4452 status
= nfsd4_recdir_load();
4453 nfs4_unlock_state();
4455 printk("NFSD: Failure reading reboot recovery data\n");
4459 * Since the lifetime of a delegation isn't limited to that of an open, a
4460 * client may quite reasonably hang on to a delegation as long as it has
4461 * the inode cached. This becomes an obvious problem the first time a
4462 * client's inode cache approaches the size of the server's total memory.
4464 * For now we avoid this problem by imposing a hard limit on the number
4465 * of delegations, which varies according to the server's memory size.
4468 set_max_delegations(void)
4471 * Allow at most 4 delegations per megabyte of RAM. Quick
4472 * estimates suggest that in the worst case (where every delegation
4473 * is for a different inode), a delegation could take about 1.5K,
4474 * giving a worst case usage of about 6% of memory.
4476 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4479 /* initialization to perform when the nfsd service is started: */
4482 __nfs4_state_start(void)
4486 boot_time
= get_seconds();
4487 locks_start_grace(&nfsd4_manager
);
4488 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4490 ret
= set_callback_cred();
4493 laundry_wq
= create_singlethread_workqueue("nfsd4");
4494 if (laundry_wq
== NULL
)
4496 ret
= nfsd4_create_callback_queue();
4498 goto out_free_laundry
;
4499 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4500 set_max_delegations();
4503 destroy_workqueue(laundry_wq
);
4508 nfs4_state_start(void)
4510 nfsd4_load_reboot_recovery_data();
4511 return __nfs4_state_start();
4515 __nfs4_state_shutdown(void)
4518 struct nfs4_client
*clp
= NULL
;
4519 struct nfs4_delegation
*dp
= NULL
;
4520 struct list_head
*pos
, *next
, reaplist
;
4522 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4523 while (!list_empty(&conf_id_hashtbl
[i
])) {
4524 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4527 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4528 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4532 INIT_LIST_HEAD(&reaplist
);
4533 spin_lock(&recall_lock
);
4534 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4535 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4536 list_move(&dp
->dl_recall_lru
, &reaplist
);
4538 spin_unlock(&recall_lock
);
4539 list_for_each_safe(pos
, next
, &reaplist
) {
4540 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4541 list_del_init(&dp
->dl_recall_lru
);
4542 unhash_delegation(dp
);
4545 nfsd4_shutdown_recdir();
4549 nfs4_state_shutdown(void)
4551 cancel_delayed_work_sync(&laundromat_work
);
4552 destroy_workqueue(laundry_wq
);
4553 locks_end_grace(&nfsd4_manager
);
4555 nfs4_release_reclaim();
4556 __nfs4_state_shutdown();
4557 nfs4_unlock_state();
4558 nfsd4_destroy_callback_queue();