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
;
53 #define all_ones {{~0,~0},~0}
54 static const stateid_t one_stateid
= {
56 .si_opaque
= all_ones
,
58 static const stateid_t zero_stateid
= {
61 static const stateid_t currentstateid
= {
65 static u64 current_sessionid
= 1;
67 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
68 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
69 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
71 /* forward declarations */
72 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
);
76 /* Currently used for almost all code touching nfsv4 state: */
77 static DEFINE_MUTEX(client_mutex
);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(recall_lock
);
86 static struct kmem_cache
*openowner_slab
= NULL
;
87 static struct kmem_cache
*lockowner_slab
= NULL
;
88 static struct kmem_cache
*file_slab
= NULL
;
89 static struct kmem_cache
*stateid_slab
= NULL
;
90 static struct kmem_cache
*deleg_slab
= NULL
;
95 mutex_lock(&client_mutex
);
99 nfs4_unlock_state(void)
101 mutex_unlock(&client_mutex
);
105 opaque_hashval(const void *ptr
, int nbytes
)
107 unsigned char *cptr
= (unsigned char *) ptr
;
117 static struct list_head del_recall_lru
;
119 static void nfsd4_free_file(struct nfs4_file
*f
)
121 kmem_cache_free(file_slab
, f
);
125 put_nfs4_file(struct nfs4_file
*fi
)
127 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
128 list_del(&fi
->fi_hash
);
129 spin_unlock(&recall_lock
);
136 get_nfs4_file(struct nfs4_file
*fi
)
138 atomic_inc(&fi
->fi_ref
);
141 static int num_delegations
;
142 unsigned int max_delegations
;
145 * Open owner state (share locks)
148 /* hash tables for lock and open owners */
149 #define OWNER_HASH_BITS 8
150 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
151 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
153 static unsigned int ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
157 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
159 return ret
& OWNER_HASH_MASK
;
162 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
164 /* hash table for nfs4_file */
165 #define FILE_HASH_BITS 8
166 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
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 struct list_head file_hashtbl
[FILE_HASH_SIZE
];
176 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
178 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
179 atomic_inc(&fp
->fi_access
[oflag
]);
182 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
184 if (oflag
== O_RDWR
) {
185 __nfs4_file_get_access(fp
, O_RDONLY
);
186 __nfs4_file_get_access(fp
, O_WRONLY
);
188 __nfs4_file_get_access(fp
, oflag
);
191 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
193 if (fp
->fi_fds
[oflag
]) {
194 fput(fp
->fi_fds
[oflag
]);
195 fp
->fi_fds
[oflag
] = NULL
;
199 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
201 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
202 nfs4_file_put_fd(fp
, oflag
);
204 * It's also safe to get rid of the RDWR open *if*
205 * we no longer have need of the other kind of access
206 * or if we already have the other kind of open:
208 if (fp
->fi_fds
[1-oflag
]
209 || atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
210 nfs4_file_put_fd(fp
, O_RDWR
);
214 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
216 if (oflag
== O_RDWR
) {
217 __nfs4_file_put_access(fp
, O_RDONLY
);
218 __nfs4_file_put_access(fp
, O_WRONLY
);
220 __nfs4_file_put_access(fp
, oflag
);
223 static inline int get_new_stid(struct nfs4_stid
*stid
)
225 static int min_stateid
= 0;
226 struct idr
*stateids
= &stid
->sc_client
->cl_stateids
;
230 error
= idr_get_new_above(stateids
, stid
, min_stateid
, &new_stid
);
232 * Note: the necessary preallocation was done in
233 * nfs4_alloc_stateid(). The idr code caps the number of
234 * preallocations that can exist at a time, but the state lock
235 * prevents anyone from using ours before we get here:
239 * It shouldn't be a problem to reuse an opaque stateid value.
240 * I don't think it is for 4.1. But with 4.0 I worry that, for
241 * example, a stray write retransmission could be accepted by
242 * the server when it should have been rejected. Therefore,
243 * adopt a trick from the sctp code to attempt to maximize the
244 * amount of time until an id is reused, by ensuring they always
245 * "increase" (mod INT_MAX):
248 min_stateid
= new_stid
+1;
249 if (min_stateid
== INT_MAX
)
254 static void init_stid(struct nfs4_stid
*stid
, struct nfs4_client
*cl
, unsigned char type
)
256 stateid_t
*s
= &stid
->sc_stateid
;
259 stid
->sc_type
= type
;
260 stid
->sc_client
= cl
;
261 s
->si_opaque
.so_clid
= cl
->cl_clientid
;
262 new_id
= get_new_stid(stid
);
263 s
->si_opaque
.so_id
= (u32
)new_id
;
264 /* Will be incremented before return to client: */
265 s
->si_generation
= 0;
268 static struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
)
270 struct idr
*stateids
= &cl
->cl_stateids
;
272 if (!idr_pre_get(stateids
, GFP_KERNEL
))
275 * Note: if we fail here (or any time between now and the time
276 * we actually get the new idr), we won't need to undo the idr
277 * preallocation, since the idr code caps the number of
278 * preallocated entries.
280 return kmem_cache_alloc(slab
, GFP_KERNEL
);
283 static struct nfs4_ol_stateid
* nfs4_alloc_stateid(struct nfs4_client
*clp
)
285 return openlockstateid(nfs4_alloc_stid(clp
, stateid_slab
));
288 static struct nfs4_delegation
*
289 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_ol_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
291 struct nfs4_delegation
*dp
;
292 struct nfs4_file
*fp
= stp
->st_file
;
294 dprintk("NFSD alloc_init_deleg\n");
296 * Major work on the lease subsystem (for example, to support
297 * calbacks on stat) will be required before we can support
298 * write delegations properly.
300 if (type
!= NFS4_OPEN_DELEGATE_READ
)
302 if (fp
->fi_had_conflict
)
304 if (num_delegations
> max_delegations
)
306 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
309 init_stid(&dp
->dl_stid
, clp
, NFS4_DELEG_STID
);
311 * delegation seqid's are never incremented. The 4.1 special
312 * meaning of seqid 0 isn't meaningful, really, but let's avoid
313 * 0 anyway just for consistency and use 1:
315 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
317 INIT_LIST_HEAD(&dp
->dl_perfile
);
318 INIT_LIST_HEAD(&dp
->dl_perclnt
);
319 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
323 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
325 atomic_set(&dp
->dl_count
, 1);
326 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
331 nfs4_put_delegation(struct nfs4_delegation
*dp
)
333 if (atomic_dec_and_test(&dp
->dl_count
)) {
334 dprintk("NFSD: freeing dp %p\n",dp
);
335 put_nfs4_file(dp
->dl_file
);
336 kmem_cache_free(deleg_slab
, dp
);
341 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
343 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
344 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
346 fput(fp
->fi_deleg_file
);
347 fp
->fi_deleg_file
= NULL
;
351 static void unhash_stid(struct nfs4_stid
*s
)
353 struct idr
*stateids
= &s
->sc_client
->cl_stateids
;
355 idr_remove(stateids
, s
->sc_stateid
.si_opaque
.so_id
);
358 /* Called under the state lock. */
360 unhash_delegation(struct nfs4_delegation
*dp
)
362 unhash_stid(&dp
->dl_stid
);
363 list_del_init(&dp
->dl_perclnt
);
364 spin_lock(&recall_lock
);
365 list_del_init(&dp
->dl_perfile
);
366 list_del_init(&dp
->dl_recall_lru
);
367 spin_unlock(&recall_lock
);
368 nfs4_put_deleg_lease(dp
->dl_file
);
369 nfs4_put_delegation(dp
);
376 /* client_lock protects the client lru list and session hash table */
377 static DEFINE_SPINLOCK(client_lock
);
379 /* Hash tables for nfs4_clientid state */
380 #define CLIENT_HASH_BITS 4
381 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
382 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
384 static unsigned int clientid_hashval(u32 id
)
386 return id
& CLIENT_HASH_MASK
;
389 static unsigned int clientstr_hashval(const char *name
)
391 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
395 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
396 * used in reboot/reset lease grace period processing
398 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
399 * setclientid_confirmed info.
401 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
404 * client_lru holds client queue ordered by nfs4_client.cl_time
407 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
408 * for last close replay.
410 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
411 static int reclaim_str_hashtbl_size
= 0;
412 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
413 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
414 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
415 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
416 static struct list_head client_lru
;
417 static struct list_head close_lru
;
420 * We store the NONE, READ, WRITE, and BOTH bits separately in the
421 * st_{access,deny}_bmap field of the stateid, in order to track not
422 * only what share bits are currently in force, but also what
423 * combinations of share bits previous opens have used. This allows us
424 * to enforce the recommendation of rfc 3530 14.2.19 that the server
425 * return an error if the client attempt to downgrade to a combination
426 * of share bits not explicable by closing some of its previous opens.
428 * XXX: This enforcement is actually incomplete, since we don't keep
429 * track of access/deny bit combinations; so, e.g., we allow:
431 * OPEN allow read, deny write
432 * OPEN allow both, deny none
433 * DOWNGRADE allow read, deny none
435 * which we should reject.
438 set_access(unsigned int *access
, unsigned long bmap
) {
442 for (i
= 1; i
< 4; i
++) {
443 if (test_bit(i
, &bmap
))
449 set_deny(unsigned int *deny
, unsigned long bmap
) {
453 for (i
= 0; i
< 4; i
++) {
454 if (test_bit(i
, &bmap
))
460 test_share(struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
) {
461 unsigned int access
, deny
;
463 set_access(&access
, stp
->st_access_bmap
);
464 set_deny(&deny
, stp
->st_deny_bmap
);
465 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
470 static int nfs4_access_to_omode(u32 access
)
472 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
473 case NFS4_SHARE_ACCESS_READ
:
475 case NFS4_SHARE_ACCESS_WRITE
:
477 case NFS4_SHARE_ACCESS_BOTH
:
483 static void unhash_generic_stateid(struct nfs4_ol_stateid
*stp
)
485 list_del(&stp
->st_perfile
);
486 list_del(&stp
->st_perstateowner
);
489 static void close_generic_stateid(struct nfs4_ol_stateid
*stp
)
493 if (stp
->st_access_bmap
) {
494 for (i
= 1; i
< 4; i
++) {
495 if (test_bit(i
, &stp
->st_access_bmap
))
496 nfs4_file_put_access(stp
->st_file
,
497 nfs4_access_to_omode(i
));
498 __clear_bit(i
, &stp
->st_access_bmap
);
501 put_nfs4_file(stp
->st_file
);
505 static void free_generic_stateid(struct nfs4_ol_stateid
*stp
)
507 kmem_cache_free(stateid_slab
, stp
);
510 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
514 unhash_generic_stateid(stp
);
515 unhash_stid(&stp
->st_stid
);
516 file
= find_any_file(stp
->st_file
);
518 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
519 close_generic_stateid(stp
);
520 free_generic_stateid(stp
);
523 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
525 struct nfs4_ol_stateid
*stp
;
527 list_del(&lo
->lo_owner
.so_strhash
);
528 list_del(&lo
->lo_perstateid
);
529 list_del(&lo
->lo_owner_ino_hash
);
530 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
531 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
532 struct nfs4_ol_stateid
, st_perstateowner
);
533 release_lock_stateid(stp
);
537 static void release_lockowner(struct nfs4_lockowner
*lo
)
539 unhash_lockowner(lo
);
540 nfs4_free_lockowner(lo
);
544 release_stateid_lockowners(struct nfs4_ol_stateid
*open_stp
)
546 struct nfs4_lockowner
*lo
;
548 while (!list_empty(&open_stp
->st_lockowners
)) {
549 lo
= list_entry(open_stp
->st_lockowners
.next
,
550 struct nfs4_lockowner
, lo_perstateid
);
551 release_lockowner(lo
);
555 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
)
557 unhash_generic_stateid(stp
);
558 release_stateid_lockowners(stp
);
559 close_generic_stateid(stp
);
562 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
564 unhash_open_stateid(stp
);
565 unhash_stid(&stp
->st_stid
);
566 free_generic_stateid(stp
);
569 static void unhash_openowner(struct nfs4_openowner
*oo
)
571 struct nfs4_ol_stateid
*stp
;
573 list_del(&oo
->oo_owner
.so_strhash
);
574 list_del(&oo
->oo_perclient
);
575 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
576 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
577 struct nfs4_ol_stateid
, st_perstateowner
);
578 release_open_stateid(stp
);
582 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
584 struct nfs4_ol_stateid
*s
= oo
->oo_last_closed_stid
;
587 unhash_stid(&s
->st_stid
);
588 free_generic_stateid(s
);
589 oo
->oo_last_closed_stid
= NULL
;
593 static void release_openowner(struct nfs4_openowner
*oo
)
595 unhash_openowner(oo
);
596 list_del(&oo
->oo_close_lru
);
597 release_last_closed_stateid(oo
);
598 nfs4_free_openowner(oo
);
601 #define SESSION_HASH_SIZE 512
602 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
605 hash_sessionid(struct nfs4_sessionid
*sessionid
)
607 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
609 return sid
->sequence
% SESSION_HASH_SIZE
;
613 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
615 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
616 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
620 gen_sessionid(struct nfsd4_session
*ses
)
622 struct nfs4_client
*clp
= ses
->se_client
;
623 struct nfsd4_sessionid
*sid
;
625 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
626 sid
->clientid
= clp
->cl_clientid
;
627 sid
->sequence
= current_sessionid
++;
632 * The protocol defines ca_maxresponssize_cached to include the size of
633 * the rpc header, but all we need to cache is the data starting after
634 * the end of the initial SEQUENCE operation--the rest we regenerate
635 * each time. Therefore we can advertise a ca_maxresponssize_cached
636 * value that is the number of bytes in our cache plus a few additional
637 * bytes. In order to stay on the safe side, and not promise more than
638 * we can cache, those additional bytes must be the minimum possible: 24
639 * bytes of rpc header (xid through accept state, with AUTH_NULL
640 * verifier), 12 for the compound header (with zero-length tag), and 44
641 * for the SEQUENCE op response:
643 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
646 free_session_slots(struct nfsd4_session
*ses
)
650 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
651 kfree(ses
->se_slots
[i
]);
655 * We don't actually need to cache the rpc and session headers, so we
656 * can allocate a little less for each slot:
658 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
660 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
663 static int nfsd4_sanitize_slot_size(u32 size
)
665 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
666 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
672 * XXX: If we run out of reserved DRC memory we could (up to a point)
673 * re-negotiate active sessions and reduce their slot usage to make
674 * room for new connections. For now we just fail the create session.
676 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
680 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
682 spin_lock(&nfsd_drc_lock
);
683 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
684 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
685 num
= min_t(int, num
, avail
/ slotsize
);
686 nfsd_drc_mem_used
+= num
* slotsize
;
687 spin_unlock(&nfsd_drc_lock
);
692 static void nfsd4_put_drc_mem(int slotsize
, int num
)
694 spin_lock(&nfsd_drc_lock
);
695 nfsd_drc_mem_used
-= slotsize
* num
;
696 spin_unlock(&nfsd_drc_lock
);
699 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
701 struct nfsd4_session
*new;
704 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
705 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
706 mem
= numslots
* sizeof(struct nfsd4_slot
*);
708 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
711 /* allocate each struct nfsd4_slot and data cache in one piece */
712 for (i
= 0; i
< numslots
; i
++) {
713 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
714 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
715 if (!new->se_slots
[i
])
721 kfree(new->se_slots
[i
]);
726 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
728 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
730 new->maxreqs
= numslots
;
731 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
732 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
733 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
734 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
735 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
738 static void free_conn(struct nfsd4_conn
*c
)
740 svc_xprt_put(c
->cn_xprt
);
744 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
746 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
747 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
749 spin_lock(&clp
->cl_lock
);
750 if (!list_empty(&c
->cn_persession
)) {
751 list_del(&c
->cn_persession
);
754 spin_unlock(&clp
->cl_lock
);
755 nfsd4_probe_callback(clp
);
758 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
760 struct nfsd4_conn
*conn
;
762 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
765 svc_xprt_get(rqstp
->rq_xprt
);
766 conn
->cn_xprt
= rqstp
->rq_xprt
;
767 conn
->cn_flags
= flags
;
768 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
772 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
774 conn
->cn_session
= ses
;
775 list_add(&conn
->cn_persession
, &ses
->se_conns
);
778 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
780 struct nfs4_client
*clp
= ses
->se_client
;
782 spin_lock(&clp
->cl_lock
);
783 __nfsd4_hash_conn(conn
, ses
);
784 spin_unlock(&clp
->cl_lock
);
787 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
789 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
790 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
793 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
795 struct nfsd4_conn
*conn
;
798 conn
= alloc_conn(rqstp
, dir
);
800 return nfserr_jukebox
;
801 nfsd4_hash_conn(conn
, ses
);
802 ret
= nfsd4_register_conn(conn
);
804 /* oops; xprt is already down: */
805 nfsd4_conn_lost(&conn
->cn_xpt_user
);
809 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
811 u32 dir
= NFS4_CDFC4_FORE
;
813 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
814 dir
|= NFS4_CDFC4_BACK
;
816 return nfsd4_new_conn(rqstp
, ses
, dir
);
819 /* must be called under client_lock */
820 static void nfsd4_del_conns(struct nfsd4_session
*s
)
822 struct nfs4_client
*clp
= s
->se_client
;
823 struct nfsd4_conn
*c
;
825 spin_lock(&clp
->cl_lock
);
826 while (!list_empty(&s
->se_conns
)) {
827 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
828 list_del_init(&c
->cn_persession
);
829 spin_unlock(&clp
->cl_lock
);
831 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
834 spin_lock(&clp
->cl_lock
);
836 spin_unlock(&clp
->cl_lock
);
839 void free_session(struct kref
*kref
)
841 struct nfsd4_session
*ses
;
844 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
845 nfsd4_del_conns(ses
);
846 spin_lock(&nfsd_drc_lock
);
847 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
848 nfsd_drc_mem_used
-= mem
;
849 spin_unlock(&nfsd_drc_lock
);
850 free_session_slots(ses
);
854 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
856 struct nfsd4_session
*new;
857 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
858 int numslots
, slotsize
;
863 * Note decreasing slot size below client's request may
864 * make it difficult for client to function correctly, whereas
865 * decreasing the number of slots will (just?) affect
866 * performance. When short on memory we therefore prefer to
867 * decrease number of slots instead of their size.
869 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
870 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
874 new = alloc_session(slotsize
, numslots
);
876 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
879 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
881 new->se_client
= clp
;
884 INIT_LIST_HEAD(&new->se_conns
);
886 new->se_cb_seq_nr
= 1;
887 new->se_flags
= cses
->flags
;
888 new->se_cb_prog
= cses
->callback_prog
;
889 kref_init(&new->se_ref
);
890 idx
= hash_sessionid(&new->se_sessionid
);
891 spin_lock(&client_lock
);
892 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
893 spin_lock(&clp
->cl_lock
);
894 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
895 spin_unlock(&clp
->cl_lock
);
896 spin_unlock(&client_lock
);
898 status
= nfsd4_new_conn_from_crses(rqstp
, new);
899 /* whoops: benny points out, status is ignored! (err, or bogus) */
901 free_session(&new->se_ref
);
904 if (cses
->flags
& SESSION4_BACK_CHAN
) {
905 struct sockaddr
*sa
= svc_addr(rqstp
);
907 * This is a little silly; with sessions there's no real
908 * use for the callback address. Use the peer address
909 * as a reasonable default for now, but consider fixing
910 * the rpc client not to require an address in the
913 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
914 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
916 nfsd4_probe_callback(clp
);
920 /* caller must hold client_lock */
921 static struct nfsd4_session
*
922 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
924 struct nfsd4_session
*elem
;
927 dump_sessionid(__func__
, sessionid
);
928 idx
= hash_sessionid(sessionid
);
929 /* Search in the appropriate list */
930 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
931 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
932 NFS4_MAX_SESSIONID_LEN
)) {
937 dprintk("%s: session not found\n", __func__
);
941 /* caller must hold client_lock */
943 unhash_session(struct nfsd4_session
*ses
)
945 list_del(&ses
->se_hash
);
946 spin_lock(&ses
->se_client
->cl_lock
);
947 list_del(&ses
->se_perclnt
);
948 spin_unlock(&ses
->se_client
->cl_lock
);
951 /* must be called under the client_lock */
953 renew_client_locked(struct nfs4_client
*clp
)
955 if (is_client_expired(clp
)) {
956 dprintk("%s: client (clientid %08x/%08x) already expired\n",
958 clp
->cl_clientid
.cl_boot
,
959 clp
->cl_clientid
.cl_id
);
963 dprintk("renewing client (clientid %08x/%08x)\n",
964 clp
->cl_clientid
.cl_boot
,
965 clp
->cl_clientid
.cl_id
);
966 list_move_tail(&clp
->cl_lru
, &client_lru
);
967 clp
->cl_time
= get_seconds();
971 renew_client(struct nfs4_client
*clp
)
973 spin_lock(&client_lock
);
974 renew_client_locked(clp
);
975 spin_unlock(&client_lock
);
978 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
980 STALE_CLIENTID(clientid_t
*clid
)
982 if (clid
->cl_boot
== boot_time
)
984 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
985 clid
->cl_boot
, clid
->cl_id
, boot_time
);
990 * XXX Should we use a slab cache ?
991 * This type of memory management is somewhat inefficient, but we use it
992 * anyway since SETCLIENTID is not a common operation.
994 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
996 struct nfs4_client
*clp
;
998 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1001 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1002 if (clp
->cl_name
.data
== NULL
) {
1006 clp
->cl_name
.len
= name
.len
;
1011 free_client(struct nfs4_client
*clp
)
1013 while (!list_empty(&clp
->cl_sessions
)) {
1014 struct nfsd4_session
*ses
;
1015 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1017 list_del(&ses
->se_perclnt
);
1018 nfsd4_put_session(ses
);
1020 if (clp
->cl_cred
.cr_group_info
)
1021 put_group_info(clp
->cl_cred
.cr_group_info
);
1022 kfree(clp
->cl_principal
);
1023 kfree(clp
->cl_name
.data
);
1028 release_session_client(struct nfsd4_session
*session
)
1030 struct nfs4_client
*clp
= session
->se_client
;
1032 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
1034 if (is_client_expired(clp
)) {
1036 session
->se_client
= NULL
;
1038 renew_client_locked(clp
);
1039 spin_unlock(&client_lock
);
1042 /* must be called under the client_lock */
1044 unhash_client_locked(struct nfs4_client
*clp
)
1046 struct nfsd4_session
*ses
;
1048 mark_client_expired(clp
);
1049 list_del(&clp
->cl_lru
);
1050 spin_lock(&clp
->cl_lock
);
1051 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1052 list_del_init(&ses
->se_hash
);
1053 spin_unlock(&clp
->cl_lock
);
1057 expire_client(struct nfs4_client
*clp
)
1059 struct nfs4_openowner
*oo
;
1060 struct nfs4_delegation
*dp
;
1061 struct list_head reaplist
;
1063 INIT_LIST_HEAD(&reaplist
);
1064 spin_lock(&recall_lock
);
1065 while (!list_empty(&clp
->cl_delegations
)) {
1066 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1067 list_del_init(&dp
->dl_perclnt
);
1068 list_move(&dp
->dl_recall_lru
, &reaplist
);
1070 spin_unlock(&recall_lock
);
1071 while (!list_empty(&reaplist
)) {
1072 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1073 unhash_delegation(dp
);
1075 while (!list_empty(&clp
->cl_openowners
)) {
1076 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1077 release_openowner(oo
);
1079 nfsd4_shutdown_callback(clp
);
1080 if (clp
->cl_cb_conn
.cb_xprt
)
1081 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1082 list_del(&clp
->cl_idhash
);
1083 list_del(&clp
->cl_strhash
);
1084 spin_lock(&client_lock
);
1085 unhash_client_locked(clp
);
1086 if (atomic_read(&clp
->cl_refcount
) == 0)
1088 spin_unlock(&client_lock
);
1091 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1093 memcpy(target
->cl_verifier
.data
, source
->data
,
1094 sizeof(target
->cl_verifier
.data
));
1097 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1099 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1100 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1103 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1105 target
->cr_uid
= source
->cr_uid
;
1106 target
->cr_gid
= source
->cr_gid
;
1107 target
->cr_group_info
= source
->cr_group_info
;
1108 get_group_info(target
->cr_group_info
);
1111 static int same_name(const char *n1
, const char *n2
)
1113 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1117 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1119 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1123 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1125 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1128 /* XXX what about NGROUP */
1130 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1132 return cr1
->cr_uid
== cr2
->cr_uid
;
1135 static void gen_clid(struct nfs4_client
*clp
)
1137 static u32 current_clientid
= 1;
1139 clp
->cl_clientid
.cl_boot
= boot_time
;
1140 clp
->cl_clientid
.cl_id
= current_clientid
++;
1143 static void gen_confirm(struct nfs4_client
*clp
)
1148 p
= (u32
*)clp
->cl_confirm
.data
;
1149 *p
++ = get_seconds();
1153 static struct nfs4_stid
*find_stateid(struct nfs4_client
*cl
, stateid_t
*t
)
1155 return idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1158 static struct nfs4_stid
*find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1160 struct nfs4_stid
*s
;
1162 s
= find_stateid(cl
, t
);
1165 if (typemask
& s
->sc_type
)
1170 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1171 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1173 struct nfs4_client
*clp
;
1174 struct sockaddr
*sa
= svc_addr(rqstp
);
1177 clp
= alloc_client(name
);
1181 INIT_LIST_HEAD(&clp
->cl_sessions
);
1183 princ
= svc_gss_principal(rqstp
);
1185 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1186 if (clp
->cl_principal
== NULL
) {
1192 idr_init(&clp
->cl_stateids
);
1193 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1194 atomic_set(&clp
->cl_refcount
, 0);
1195 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1196 INIT_LIST_HEAD(&clp
->cl_idhash
);
1197 INIT_LIST_HEAD(&clp
->cl_strhash
);
1198 INIT_LIST_HEAD(&clp
->cl_openowners
);
1199 INIT_LIST_HEAD(&clp
->cl_delegations
);
1200 INIT_LIST_HEAD(&clp
->cl_lru
);
1201 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1202 spin_lock_init(&clp
->cl_lock
);
1203 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1204 clp
->cl_time
= get_seconds();
1205 clear_bit(0, &clp
->cl_cb_slot_busy
);
1206 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1207 copy_verf(clp
, verf
);
1208 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1209 clp
->cl_flavor
= rqstp
->rq_flavor
;
1210 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1212 clp
->cl_cb_session
= NULL
;
1217 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1219 unsigned int idhashval
;
1221 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1222 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1223 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1228 move_to_confirmed(struct nfs4_client
*clp
)
1230 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1231 unsigned int strhashval
;
1233 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1234 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1235 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1236 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1240 static struct nfs4_client
*
1241 find_confirmed_client(clientid_t
*clid
)
1243 struct nfs4_client
*clp
;
1244 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1246 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1247 if (same_clid(&clp
->cl_clientid
, clid
)) {
1255 static struct nfs4_client
*
1256 find_unconfirmed_client(clientid_t
*clid
)
1258 struct nfs4_client
*clp
;
1259 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1261 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1262 if (same_clid(&clp
->cl_clientid
, clid
))
1268 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1270 return clp
->cl_exchange_flags
!= 0;
1273 static struct nfs4_client
*
1274 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1276 struct nfs4_client
*clp
;
1278 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1279 if (same_name(clp
->cl_recdir
, dname
))
1285 static struct nfs4_client
*
1286 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1288 struct nfs4_client
*clp
;
1290 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1291 if (same_name(clp
->cl_recdir
, dname
))
1298 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1300 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1301 struct sockaddr
*sa
= svc_addr(rqstp
);
1302 u32 scopeid
= rpc_get_scope_id(sa
);
1303 unsigned short expected_family
;
1305 /* Currently, we only support tcp and tcp6 for the callback channel */
1306 if (se
->se_callback_netid_len
== 3 &&
1307 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1308 expected_family
= AF_INET
;
1309 else if (se
->se_callback_netid_len
== 4 &&
1310 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1311 expected_family
= AF_INET6
;
1315 conn
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
1316 se
->se_callback_addr_len
,
1317 (struct sockaddr
*)&conn
->cb_addr
,
1318 sizeof(conn
->cb_addr
));
1320 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1323 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1324 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1326 conn
->cb_prog
= se
->se_callback_prog
;
1327 conn
->cb_ident
= se
->se_callback_ident
;
1328 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
1331 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1332 conn
->cb_addrlen
= 0;
1333 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1334 "will not receive delegations\n",
1335 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1341 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1344 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1346 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1349 dprintk("--> %s slot %p\n", __func__
, slot
);
1351 slot
->sl_opcnt
= resp
->opcnt
;
1352 slot
->sl_status
= resp
->cstate
.status
;
1354 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
1355 if (nfsd4_not_cached(resp
)) {
1356 slot
->sl_datalen
= 0;
1359 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1360 base
= (char *)resp
->cstate
.datap
-
1361 (char *)resp
->xbuf
->head
[0].iov_base
;
1362 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1364 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1369 * Encode the replay sequence operation from the slot values.
1370 * If cachethis is FALSE encode the uncached rep error on the next
1371 * operation which sets resp->p and increments resp->opcnt for
1372 * nfs4svc_encode_compoundres.
1376 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1377 struct nfsd4_compoundres
*resp
)
1379 struct nfsd4_op
*op
;
1380 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1382 /* Encode the replayed sequence operation */
1383 op
= &args
->ops
[resp
->opcnt
- 1];
1384 nfsd4_encode_operation(resp
, op
);
1386 /* Return nfserr_retry_uncached_rep in next operation. */
1387 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
1388 op
= &args
->ops
[resp
->opcnt
++];
1389 op
->status
= nfserr_retry_uncached_rep
;
1390 nfsd4_encode_operation(resp
, op
);
1396 * The sequence operation is not cached because we can use the slot and
1400 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1401 struct nfsd4_sequence
*seq
)
1403 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1406 dprintk("--> %s slot %p\n", __func__
, slot
);
1408 /* Either returns 0 or nfserr_retry_uncached */
1409 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1410 if (status
== nfserr_retry_uncached_rep
)
1413 /* The sequence operation has been encoded, cstate->datap set. */
1414 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1416 resp
->opcnt
= slot
->sl_opcnt
;
1417 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1418 status
= slot
->sl_status
;
1424 * Set the exchange_id flags returned by the server.
1427 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1429 /* pNFS is not supported */
1430 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1432 /* Referrals are supported, Migration is not. */
1433 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1435 /* set the wire flags to return to client. */
1436 clid
->flags
= new->cl_exchange_flags
;
1440 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1441 struct nfsd4_compound_state
*cstate
,
1442 struct nfsd4_exchange_id
*exid
)
1444 struct nfs4_client
*unconf
, *conf
, *new;
1446 unsigned int strhashval
;
1447 char dname
[HEXDIR_LEN
];
1448 char addr_str
[INET6_ADDRSTRLEN
];
1449 nfs4_verifier verf
= exid
->verifier
;
1450 struct sockaddr
*sa
= svc_addr(rqstp
);
1452 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1453 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1454 "ip_addr=%s flags %x, spa_how %d\n",
1455 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1456 addr_str
, exid
->flags
, exid
->spa_how
);
1458 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
1459 return nfserr_inval
;
1461 /* Currently only support SP4_NONE */
1462 switch (exid
->spa_how
) {
1466 return nfserr_serverfault
;
1468 BUG(); /* checked by xdr code */
1470 return nfserr_serverfault
; /* no excuse :-/ */
1473 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1478 strhashval
= clientstr_hashval(dname
);
1483 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1485 if (!clp_used_exchangeid(conf
)) {
1486 status
= nfserr_clid_inuse
; /* XXX: ? */
1489 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1490 /* 18.35.4 case 8 */
1491 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1492 status
= nfserr_not_same
;
1495 /* Client reboot: destroy old state */
1496 expire_client(conf
);
1499 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1500 /* 18.35.4 case 9 */
1501 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1502 status
= nfserr_perm
;
1505 expire_client(conf
);
1509 * Set bit when the owner id and verifier map to an already
1510 * confirmed client id (18.35.3).
1512 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1515 * Falling into 18.35.4 case 2, possible router replay.
1516 * Leave confirmed record intact and return same result.
1518 copy_verf(conf
, &verf
);
1523 /* 18.35.4 case 7 */
1524 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1525 status
= nfserr_noent
;
1529 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1532 * Possible retry or client restart. Per 18.35.4 case 4,
1533 * a new unconfirmed record should be generated regardless
1534 * of whether any properties have changed.
1536 expire_client(unconf
);
1541 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1543 status
= nfserr_jukebox
;
1548 add_to_unconfirmed(new, strhashval
);
1550 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1551 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1554 nfsd4_set_ex_flags(new, exid
);
1556 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1557 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1561 nfs4_unlock_state();
1563 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1568 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1570 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1573 /* The slot is in use, and no response has been sent. */
1575 if (seqid
== slot_seqid
)
1576 return nfserr_jukebox
;
1578 return nfserr_seq_misordered
;
1580 /* Note unsigned 32-bit arithmetic handles wraparound: */
1581 if (likely(seqid
== slot_seqid
+ 1))
1583 if (seqid
== slot_seqid
)
1584 return nfserr_replay_cache
;
1585 return nfserr_seq_misordered
;
1589 * Cache the create session result into the create session single DRC
1590 * slot cache by saving the xdr structure. sl_seqid has been set.
1591 * Do this for solo or embedded create session operations.
1594 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1595 struct nfsd4_clid_slot
*slot
, int nfserr
)
1597 slot
->sl_status
= nfserr
;
1598 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1602 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1603 struct nfsd4_clid_slot
*slot
)
1605 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1606 return slot
->sl_status
;
1609 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1610 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1611 1 + /* MIN tag is length with zero, only length */ \
1612 3 + /* version, opcount, opcode */ \
1613 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1614 /* seqid, slotID, slotID, cache */ \
1615 4 ) * sizeof(__be32))
1617 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1618 2 + /* verifier: AUTH_NULL, length 0 */\
1620 1 + /* MIN tag is length with zero, only length */ \
1621 3 + /* opcount, opcode, opstatus*/ \
1622 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1623 /* seqid, slotID, slotID, slotID, status */ \
1624 5 ) * sizeof(__be32))
1626 static __be32
check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1628 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1629 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1633 nfsd4_create_session(struct svc_rqst
*rqstp
,
1634 struct nfsd4_compound_state
*cstate
,
1635 struct nfsd4_create_session
*cr_ses
)
1637 struct sockaddr
*sa
= svc_addr(rqstp
);
1638 struct nfs4_client
*conf
, *unconf
;
1639 struct nfsd4_session
*new;
1640 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1641 bool confirm_me
= false;
1644 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1645 return nfserr_inval
;
1648 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1649 conf
= find_confirmed_client(&cr_ses
->clientid
);
1652 cs_slot
= &conf
->cl_cs_slot
;
1653 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1654 if (status
== nfserr_replay_cache
) {
1655 dprintk("Got a create_session replay! seqid= %d\n",
1657 /* Return the cached reply status */
1658 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1660 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1661 status
= nfserr_seq_misordered
;
1662 dprintk("Sequence misordered!\n");
1663 dprintk("Expected seqid= %d but got seqid= %d\n",
1664 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1667 } else if (unconf
) {
1668 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1669 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1670 status
= nfserr_clid_inuse
;
1674 cs_slot
= &unconf
->cl_cs_slot
;
1675 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1677 /* an unconfirmed replay returns misordered */
1678 status
= nfserr_seq_misordered
;
1685 status
= nfserr_stale_clientid
;
1690 * XXX: we should probably set this at creation time, and check
1691 * for consistent minorversion use throughout:
1693 conf
->cl_minorversion
= 1;
1695 * We do not support RDMA or persistent sessions
1697 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1698 cr_ses
->flags
&= ~SESSION4_RDMA
;
1700 status
= nfserr_toosmall
;
1701 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1704 status
= nfserr_jukebox
;
1705 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1709 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1710 NFS4_MAX_SESSIONID_LEN
);
1711 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1712 sizeof(struct nfsd4_channel_attrs
));
1713 cs_slot
->sl_seqid
++;
1714 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1716 /* cache solo and embedded create sessions under the state lock */
1717 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1719 move_to_confirmed(conf
);
1721 nfs4_unlock_state();
1722 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1726 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1728 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1729 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1731 return argp
->opcnt
== resp
->opcnt
;
1734 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1737 case NFS4_CDFC4_FORE
:
1738 case NFS4_CDFC4_BACK
:
1740 case NFS4_CDFC4_FORE_OR_BOTH
:
1741 case NFS4_CDFC4_BACK_OR_BOTH
:
1742 *dir
= NFS4_CDFC4_BOTH
;
1745 return nfserr_inval
;
1748 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1749 struct nfsd4_compound_state
*cstate
,
1750 struct nfsd4_bind_conn_to_session
*bcts
)
1754 if (!nfsd4_last_compound_op(rqstp
))
1755 return nfserr_not_only_op
;
1756 spin_lock(&client_lock
);
1757 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1758 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1759 * client_lock iself: */
1760 if (cstate
->session
) {
1761 nfsd4_get_session(cstate
->session
);
1762 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1764 spin_unlock(&client_lock
);
1765 if (!cstate
->session
)
1766 return nfserr_badsession
;
1768 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1770 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1774 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1778 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1782 nfsd4_destroy_session(struct svc_rqst
*r
,
1783 struct nfsd4_compound_state
*cstate
,
1784 struct nfsd4_destroy_session
*sessionid
)
1786 struct nfsd4_session
*ses
;
1787 u32 status
= nfserr_badsession
;
1790 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1791 * - Should we return nfserr_back_chan_busy if waiting for
1792 * callbacks on to-be-destroyed session?
1793 * - Do we need to clear any callback info from previous session?
1796 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1797 if (!nfsd4_last_compound_op(r
))
1798 return nfserr_not_only_op
;
1800 dump_sessionid(__func__
, &sessionid
->sessionid
);
1801 spin_lock(&client_lock
);
1802 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1804 spin_unlock(&client_lock
);
1808 unhash_session(ses
);
1809 spin_unlock(&client_lock
);
1812 nfsd4_probe_callback_sync(ses
->se_client
);
1813 nfs4_unlock_state();
1815 nfsd4_del_conns(ses
);
1817 nfsd4_put_session(ses
);
1820 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1824 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1826 struct nfsd4_conn
*c
;
1828 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1829 if (c
->cn_xprt
== xpt
) {
1836 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1838 struct nfs4_client
*clp
= ses
->se_client
;
1839 struct nfsd4_conn
*c
;
1842 spin_lock(&clp
->cl_lock
);
1843 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1845 spin_unlock(&clp
->cl_lock
);
1849 __nfsd4_hash_conn(new, ses
);
1850 spin_unlock(&clp
->cl_lock
);
1851 ret
= nfsd4_register_conn(new);
1853 /* oops; xprt is already down: */
1854 nfsd4_conn_lost(&new->cn_xpt_user
);
1858 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1860 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1862 return args
->opcnt
> session
->se_fchannel
.maxops
;
1865 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1866 struct nfsd4_session
*session
)
1868 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1870 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1874 nfsd4_sequence(struct svc_rqst
*rqstp
,
1875 struct nfsd4_compound_state
*cstate
,
1876 struct nfsd4_sequence
*seq
)
1878 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1879 struct nfsd4_session
*session
;
1880 struct nfsd4_slot
*slot
;
1881 struct nfsd4_conn
*conn
;
1884 if (resp
->opcnt
!= 1)
1885 return nfserr_sequence_pos
;
1888 * Will be either used or freed by nfsd4_sequence_check_conn
1891 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1893 return nfserr_jukebox
;
1895 spin_lock(&client_lock
);
1896 status
= nfserr_badsession
;
1897 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1901 status
= nfserr_too_many_ops
;
1902 if (nfsd4_session_too_many_ops(rqstp
, session
))
1905 status
= nfserr_req_too_big
;
1906 if (nfsd4_request_too_big(rqstp
, session
))
1909 status
= nfserr_badslot
;
1910 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1913 slot
= session
->se_slots
[seq
->slotid
];
1914 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1916 /* We do not negotiate the number of slots yet, so set the
1917 * maxslots to the session maxreqs which is used to encode
1918 * sr_highest_slotid and the sr_target_slot id to maxslots */
1919 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1921 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
1922 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
1923 if (status
== nfserr_replay_cache
) {
1924 status
= nfserr_seq_misordered
;
1925 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
1927 cstate
->slot
= slot
;
1928 cstate
->session
= session
;
1929 /* Return the cached reply status and set cstate->status
1930 * for nfsd4_proc_compound processing */
1931 status
= nfsd4_replay_cache_entry(resp
, seq
);
1932 cstate
->status
= nfserr_replay_cache
;
1938 nfsd4_sequence_check_conn(conn
, session
);
1941 /* Success! bump slot seqid */
1942 slot
->sl_seqid
= seq
->seqid
;
1943 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
1945 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
1947 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
1949 cstate
->slot
= slot
;
1950 cstate
->session
= session
;
1953 /* Hold a session reference until done processing the compound. */
1954 if (cstate
->session
) {
1955 struct nfs4_client
*clp
= session
->se_client
;
1957 nfsd4_get_session(cstate
->session
);
1958 atomic_inc(&clp
->cl_refcount
);
1959 switch (clp
->cl_cb_state
) {
1961 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
1963 case NFSD4_CB_FAULT
:
1964 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
1967 seq
->status_flags
= 0;
1971 spin_unlock(&client_lock
);
1972 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1976 static inline bool has_resources(struct nfs4_client
*clp
)
1978 return !list_empty(&clp
->cl_openowners
)
1979 || !list_empty(&clp
->cl_delegations
)
1980 || !list_empty(&clp
->cl_sessions
);
1984 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
1986 struct nfs4_client
*conf
, *unconf
, *clp
;
1990 unconf
= find_unconfirmed_client(&dc
->clientid
);
1991 conf
= find_confirmed_client(&dc
->clientid
);
1996 if (!is_client_expired(conf
) && has_resources(conf
)) {
1997 status
= nfserr_clientid_busy
;
2001 /* rfc5661 18.50.3 */
2002 if (cstate
->session
&& conf
== cstate
->session
->se_client
) {
2003 status
= nfserr_clientid_busy
;
2009 status
= nfserr_stale_clientid
;
2015 nfs4_unlock_state();
2016 dprintk("%s return %d\n", __func__
, ntohl(status
));
2021 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2025 if (rc
->rca_one_fs
) {
2026 if (!cstate
->current_fh
.fh_dentry
)
2027 return nfserr_nofilehandle
;
2029 * We don't take advantage of the rca_one_fs case.
2030 * That's OK, it's optional, we can safely ignore it.
2036 status
= nfserr_complete_already
;
2037 if (cstate
->session
->se_client
->cl_firststate
)
2040 status
= nfserr_stale_clientid
;
2041 if (is_client_expired(cstate
->session
->se_client
))
2043 * The following error isn't really legal.
2044 * But we only get here if the client just explicitly
2045 * destroyed the client. Surely it no longer cares what
2046 * error it gets back on an operation for the dead
2052 nfsd4_create_clid_dir(cstate
->session
->se_client
);
2054 nfs4_unlock_state();
2059 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2060 struct nfsd4_setclientid
*setclid
)
2062 struct xdr_netobj clname
= setclid
->se_name
;
2063 nfs4_verifier clverifier
= setclid
->se_verf
;
2064 unsigned int strhashval
;
2065 struct nfs4_client
*conf
, *unconf
, *new;
2067 char dname
[HEXDIR_LEN
];
2069 status
= nfs4_make_rec_clidname(dname
, &clname
);
2074 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2075 * We get here on a DRC miss.
2078 strhashval
= clientstr_hashval(dname
);
2081 conf
= find_confirmed_client_by_str(dname
, strhashval
);
2083 /* RFC 3530 14.2.33 CASE 0: */
2084 status
= nfserr_clid_inuse
;
2085 if (clp_used_exchangeid(conf
))
2087 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
2088 char addr_str
[INET6_ADDRSTRLEN
];
2089 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
2091 dprintk("NFSD: setclientid: string in use by client "
2092 "at %s\n", addr_str
);
2097 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2098 * has a description of SETCLIENTID request processing consisting
2099 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2101 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
2102 status
= nfserr_jukebox
;
2105 * RFC 3530 14.2.33 CASE 4:
2106 * placed first, because it is the normal case
2109 expire_client(unconf
);
2110 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2114 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
2116 * RFC 3530 14.2.33 CASE 1:
2117 * probable callback update
2120 /* Note this is removing unconfirmed {*x***},
2121 * which is stronger than RFC recommended {vxc**}.
2122 * This has the advantage that there is at most
2123 * one {*x***} in either list at any time.
2125 expire_client(unconf
);
2127 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2130 copy_clid(new, conf
);
2131 } else if (!unconf
) {
2133 * RFC 3530 14.2.33 CASE 2:
2134 * probable client reboot; state will be removed if
2137 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2143 * RFC 3530 14.2.33 CASE 3:
2144 * probable client reboot; state will be removed if
2147 expire_client(unconf
);
2148 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2154 * XXX: we should probably set this at creation time, and check
2155 * for consistent minorversion use throughout:
2157 new->cl_minorversion
= 0;
2158 gen_callback(new, setclid
, rqstp
);
2159 add_to_unconfirmed(new, strhashval
);
2160 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2161 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2162 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2165 nfs4_unlock_state();
2171 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2172 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2173 * bullets, labeled as CASE1 - CASE4 below.
2176 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2177 struct nfsd4_compound_state
*cstate
,
2178 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2180 struct sockaddr
*sa
= svc_addr(rqstp
);
2181 struct nfs4_client
*conf
, *unconf
;
2182 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2183 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2186 if (STALE_CLIENTID(clid
))
2187 return nfserr_stale_clientid
;
2189 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2190 * We get here on a DRC miss.
2195 conf
= find_confirmed_client(clid
);
2196 unconf
= find_unconfirmed_client(clid
);
2198 status
= nfserr_clid_inuse
;
2199 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
2201 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
2205 * section 14.2.34 of RFC 3530 has a description of
2206 * SETCLIENTID_CONFIRM request processing consisting
2207 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2209 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
2211 * RFC 3530 14.2.34 CASE 1:
2214 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
2215 status
= nfserr_clid_inuse
;
2217 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2218 nfsd4_probe_callback(conf
);
2219 expire_client(unconf
);
2223 } else if (conf
&& !unconf
) {
2225 * RFC 3530 14.2.34 CASE 2:
2226 * probable retransmitted request; play it safe and
2229 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2230 status
= nfserr_clid_inuse
;
2233 } else if (!conf
&& unconf
2234 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
2236 * RFC 3530 14.2.34 CASE 3:
2237 * Normal case; new or rebooted client:
2239 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
2240 status
= nfserr_clid_inuse
;
2243 clientstr_hashval(unconf
->cl_recdir
);
2244 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
2247 nfsd4_remove_clid_dir(conf
);
2248 expire_client(conf
);
2250 move_to_confirmed(unconf
);
2252 nfsd4_probe_callback(conf
);
2255 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
2256 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
2259 * RFC 3530 14.2.34 CASE 4:
2260 * Client probably hasn't noticed that we rebooted yet.
2262 status
= nfserr_stale_clientid
;
2264 /* check that we have hit one of the cases...*/
2265 status
= nfserr_clid_inuse
;
2268 nfs4_unlock_state();
2272 static struct nfs4_file
*nfsd4_alloc_file(void)
2274 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2277 /* OPEN Share state helper functions */
2278 static void nfsd4_init_file(struct nfs4_file
*fp
, struct inode
*ino
)
2280 unsigned int hashval
= file_hashval(ino
);
2282 atomic_set(&fp
->fi_ref
, 1);
2283 INIT_LIST_HEAD(&fp
->fi_hash
);
2284 INIT_LIST_HEAD(&fp
->fi_stateids
);
2285 INIT_LIST_HEAD(&fp
->fi_delegations
);
2286 fp
->fi_inode
= igrab(ino
);
2287 fp
->fi_had_conflict
= false;
2288 fp
->fi_lease
= NULL
;
2289 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2290 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2291 spin_lock(&recall_lock
);
2292 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2293 spin_unlock(&recall_lock
);
2297 nfsd4_free_slab(struct kmem_cache
**slab
)
2301 kmem_cache_destroy(*slab
);
2306 nfsd4_free_slabs(void)
2308 nfsd4_free_slab(&openowner_slab
);
2309 nfsd4_free_slab(&lockowner_slab
);
2310 nfsd4_free_slab(&file_slab
);
2311 nfsd4_free_slab(&stateid_slab
);
2312 nfsd4_free_slab(&deleg_slab
);
2316 nfsd4_init_slabs(void)
2318 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2319 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2320 if (openowner_slab
== NULL
)
2322 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2323 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2324 if (lockowner_slab
== NULL
)
2326 file_slab
= kmem_cache_create("nfsd4_files",
2327 sizeof(struct nfs4_file
), 0, 0, NULL
);
2328 if (file_slab
== NULL
)
2330 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2331 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
2332 if (stateid_slab
== NULL
)
2334 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2335 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2336 if (deleg_slab
== NULL
)
2341 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2345 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2347 kfree(oo
->oo_owner
.so_owner
.data
);
2348 kmem_cache_free(openowner_slab
, oo
);
2351 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2353 kfree(lo
->lo_owner
.so_owner
.data
);
2354 kmem_cache_free(lockowner_slab
, lo
);
2357 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2359 rp
->rp_status
= nfserr_serverfault
;
2361 rp
->rp_buf
= rp
->rp_ibuf
;
2364 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2366 struct nfs4_stateowner
*sop
;
2368 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2372 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2373 if (!sop
->so_owner
.data
) {
2374 kmem_cache_free(slab
, sop
);
2377 sop
->so_owner
.len
= owner
->len
;
2379 INIT_LIST_HEAD(&sop
->so_stateids
);
2380 sop
->so_client
= clp
;
2381 init_nfs4_replay(&sop
->so_replay
);
2385 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2387 list_add(&oo
->oo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2388 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2391 static struct nfs4_openowner
*
2392 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2393 struct nfs4_openowner
*oo
;
2395 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2398 oo
->oo_owner
.so_is_open_owner
= 1;
2399 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2400 oo
->oo_flags
= NFS4_OO_NEW
;
2402 oo
->oo_last_closed_stid
= NULL
;
2403 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2404 hash_openowner(oo
, clp
, strhashval
);
2408 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2409 struct nfs4_openowner
*oo
= open
->op_openowner
;
2410 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
2412 init_stid(&stp
->st_stid
, clp
, NFS4_OPEN_STID
);
2413 INIT_LIST_HEAD(&stp
->st_lockowners
);
2414 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2415 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2416 stp
->st_stateowner
= &oo
->oo_owner
;
2419 stp
->st_access_bmap
= 0;
2420 stp
->st_deny_bmap
= 0;
2421 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
2422 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2423 stp
->st_openstp
= NULL
;
2427 move_to_close_lru(struct nfs4_openowner
*oo
)
2429 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2431 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2432 oo
->oo_time
= get_seconds();
2436 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2439 return (sop
->so_owner
.len
== owner
->len
) &&
2440 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2441 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2444 static struct nfs4_openowner
*
2445 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2447 struct nfs4_stateowner
*so
;
2448 struct nfs4_openowner
*oo
;
2450 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2451 if (!so
->so_is_open_owner
)
2453 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
)) {
2455 renew_client(oo
->oo_owner
.so_client
);
2462 /* search file_hashtbl[] for file */
2463 static struct nfs4_file
*
2464 find_file(struct inode
*ino
)
2466 unsigned int hashval
= file_hashval(ino
);
2467 struct nfs4_file
*fp
;
2469 spin_lock(&recall_lock
);
2470 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2471 if (fp
->fi_inode
== ino
) {
2473 spin_unlock(&recall_lock
);
2477 spin_unlock(&recall_lock
);
2482 * Called to check deny when READ with all zero stateid or
2483 * WRITE with all zero or all one stateid
2486 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2488 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2489 struct nfs4_file
*fp
;
2490 struct nfs4_ol_stateid
*stp
;
2493 dprintk("NFSD: nfs4_share_conflict\n");
2495 fp
= find_file(ino
);
2498 ret
= nfserr_locked
;
2499 /* Search for conflicting share reservations */
2500 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2501 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2502 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2511 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2513 /* We're assuming the state code never drops its reference
2514 * without first removing the lease. Since we're in this lease
2515 * callback (and since the lease code is serialized by the kernel
2516 * lock) we know the server hasn't removed the lease yet, we know
2517 * it's safe to take a reference: */
2518 atomic_inc(&dp
->dl_count
);
2520 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2522 /* only place dl_time is set. protected by lock_flocks*/
2523 dp
->dl_time
= get_seconds();
2525 nfsd4_cb_recall(dp
);
2528 /* Called from break_lease() with lock_flocks() held. */
2529 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2531 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2532 struct nfs4_delegation
*dp
;
2535 /* We assume break_lease is only called once per lease: */
2536 BUG_ON(fp
->fi_had_conflict
);
2538 * We don't want the locks code to timeout the lease for us;
2539 * we'll remove it ourself if a delegation isn't returned
2542 fl
->fl_break_time
= 0;
2544 spin_lock(&recall_lock
);
2545 fp
->fi_had_conflict
= true;
2546 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2547 nfsd_break_one_deleg(dp
);
2548 spin_unlock(&recall_lock
);
2552 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2555 return lease_modify(onlist
, arg
);
2560 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2561 .lm_break
= nfsd_break_deleg_cb
,
2562 .lm_change
= nfsd_change_deleg_cb
,
2565 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2567 if (nfsd4_has_session(cstate
))
2569 if (seqid
== so
->so_seqid
- 1)
2570 return nfserr_replay_me
;
2571 if (seqid
== so
->so_seqid
)
2573 return nfserr_bad_seqid
;
2577 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2578 struct nfsd4_open
*open
)
2580 clientid_t
*clientid
= &open
->op_clientid
;
2581 struct nfs4_client
*clp
= NULL
;
2582 unsigned int strhashval
;
2583 struct nfs4_openowner
*oo
= NULL
;
2586 if (STALE_CLIENTID(&open
->op_clientid
))
2587 return nfserr_stale_clientid
;
2589 * In case we need it later, after we've already created the
2590 * file and don't want to risk a further failure:
2592 open
->op_file
= nfsd4_alloc_file();
2593 if (open
->op_file
== NULL
)
2594 return nfserr_jukebox
;
2596 strhashval
= ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2597 oo
= find_openstateowner_str(strhashval
, open
);
2598 open
->op_openowner
= oo
;
2600 clp
= find_confirmed_client(clientid
);
2602 return nfserr_expired
;
2605 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
2606 /* Replace unconfirmed owners without checking for replay. */
2607 clp
= oo
->oo_owner
.so_client
;
2608 release_openowner(oo
);
2609 open
->op_openowner
= NULL
;
2612 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2615 clp
= oo
->oo_owner
.so_client
;
2618 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2620 return nfserr_jukebox
;
2621 open
->op_openowner
= oo
;
2623 open
->op_stp
= nfs4_alloc_stateid(clp
);
2625 return nfserr_jukebox
;
2629 static inline __be32
2630 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2632 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2633 return nfserr_openmode
;
2638 static int share_access_to_flags(u32 share_access
)
2640 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2643 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
2645 struct nfs4_stid
*ret
;
2647 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
2650 return delegstateid(ret
);
2653 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
2655 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
2656 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
2660 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2661 struct nfs4_delegation
**dp
)
2664 __be32 status
= nfserr_bad_stateid
;
2666 *dp
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
2669 flags
= share_access_to_flags(open
->op_share_access
);
2670 status
= nfs4_check_delegmode(*dp
, flags
);
2674 if (!nfsd4_is_deleg_cur(open
))
2678 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2683 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
**stpp
)
2685 struct nfs4_ol_stateid
*local
;
2686 struct nfs4_openowner
*oo
= open
->op_openowner
;
2688 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2689 /* ignore lock owners */
2690 if (local
->st_stateowner
->so_is_open_owner
== 0)
2692 /* remember if we have seen this open owner */
2693 if (local
->st_stateowner
== &oo
->oo_owner
)
2695 /* check for conflicting share reservations */
2696 if (!test_share(local
, open
))
2697 return nfserr_share_denied
;
2702 static void nfs4_free_stateid(struct nfs4_ol_stateid
*s
)
2704 kmem_cache_free(stateid_slab
, s
);
2707 static inline int nfs4_access_to_access(u32 nfs4_access
)
2711 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2712 flags
|= NFSD_MAY_READ
;
2713 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2714 flags
|= NFSD_MAY_WRITE
;
2718 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2719 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2722 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2723 int access
= nfs4_access_to_access(open
->op_share_access
);
2725 if (!fp
->fi_fds
[oflag
]) {
2726 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2727 &fp
->fi_fds
[oflag
]);
2731 nfs4_file_get_access(fp
, oflag
);
2736 static inline __be32
2737 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2738 struct nfsd4_open
*open
)
2740 struct iattr iattr
= {
2741 .ia_valid
= ATTR_SIZE
,
2744 if (!open
->op_truncate
)
2746 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2747 return nfserr_inval
;
2748 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2752 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
)
2754 u32 op_share_access
= open
->op_share_access
;
2758 new_access
= !test_bit(op_share_access
, &stp
->st_access_bmap
);
2760 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2764 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2767 int oflag
= nfs4_access_to_omode(op_share_access
);
2768 nfs4_file_put_access(fp
, oflag
);
2772 /* remember the open */
2773 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2774 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2781 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2783 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2784 open
->op_openowner
->oo_owner
.so_client
->cl_firststate
= 1;
2787 /* Should we give out recallable state?: */
2788 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2790 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2793 * In the sessions case, since we don't have to establish a
2794 * separate connection for callbacks, we assume it's OK
2795 * until we hear otherwise:
2797 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2800 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2802 struct file_lock
*fl
;
2804 fl
= locks_alloc_lock();
2807 locks_init_lock(fl
);
2808 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2809 fl
->fl_flags
= FL_LEASE
;
2810 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2811 fl
->fl_end
= OFFSET_MAX
;
2812 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2813 fl
->fl_pid
= current
->tgid
;
2817 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2819 struct nfs4_file
*fp
= dp
->dl_file
;
2820 struct file_lock
*fl
;
2823 fl
= nfs4_alloc_init_lease(dp
, flag
);
2826 fl
->fl_file
= find_readable_file(fp
);
2827 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2828 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2830 list_del_init(&dp
->dl_perclnt
);
2831 locks_free_lock(fl
);
2835 fp
->fi_deleg_file
= fl
->fl_file
;
2836 get_file(fp
->fi_deleg_file
);
2837 atomic_set(&fp
->fi_delegees
, 1);
2838 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2842 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2844 struct nfs4_file
*fp
= dp
->dl_file
;
2847 return nfs4_setlease(dp
, flag
);
2848 spin_lock(&recall_lock
);
2849 if (fp
->fi_had_conflict
) {
2850 spin_unlock(&recall_lock
);
2853 atomic_inc(&fp
->fi_delegees
);
2854 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2855 spin_unlock(&recall_lock
);
2856 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2861 * Attempt to hand out a delegation.
2864 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
*stp
)
2866 struct nfs4_delegation
*dp
;
2867 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2869 int status
, flag
= 0;
2871 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2872 flag
= NFS4_OPEN_DELEGATE_NONE
;
2873 open
->op_recall
= 0;
2874 switch (open
->op_claim_type
) {
2875 case NFS4_OPEN_CLAIM_PREVIOUS
:
2877 open
->op_recall
= 1;
2878 flag
= open
->op_delegate_type
;
2879 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2882 case NFS4_OPEN_CLAIM_NULL
:
2883 /* Let's not give out any delegations till everyone's
2884 * had the chance to reclaim theirs.... */
2885 if (locks_in_grace())
2887 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
2889 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2890 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2892 flag
= NFS4_OPEN_DELEGATE_READ
;
2898 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2901 status
= nfs4_set_delegation(dp
, flag
);
2905 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
2907 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2908 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
2910 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2911 && flag
== NFS4_OPEN_DELEGATE_NONE
2912 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2913 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2914 open
->op_delegate_type
= flag
;
2917 nfs4_put_delegation(dp
);
2919 flag
= NFS4_OPEN_DELEGATE_NONE
;
2924 * called with nfs4_lock_state() held.
2927 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2929 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2930 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
2931 struct nfs4_file
*fp
= NULL
;
2932 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2933 struct nfs4_ol_stateid
*stp
= NULL
;
2934 struct nfs4_delegation
*dp
= NULL
;
2938 * Lookup file; if found, lookup stateid and check open request,
2939 * and check for delegations in the process of being recalled.
2940 * If not found, create the nfs4_file struct
2942 fp
= find_file(ino
);
2944 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2946 status
= nfs4_check_deleg(cl
, fp
, open
, &dp
);
2950 status
= nfserr_bad_stateid
;
2951 if (nfsd4_is_deleg_cur(open
))
2953 status
= nfserr_jukebox
;
2955 open
->op_file
= NULL
;
2956 nfsd4_init_file(fp
, ino
);
2960 * OPEN the file, or upgrade an existing OPEN.
2961 * If truncate fails, the OPEN fails.
2964 /* Stateid was found, this is an OPEN upgrade */
2965 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
2969 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, open
);
2973 open
->op_stp
= NULL
;
2974 init_open_stateid(stp
, fp
, open
);
2975 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2977 release_open_stateid(stp
);
2981 update_stateid(&stp
->st_stid
.sc_stateid
);
2982 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
2984 if (nfsd4_has_session(&resp
->cstate
))
2985 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2988 * Attempt to hand out a delegation. No error return, because the
2989 * OPEN succeeds even if we fail.
2991 nfs4_open_delegation(current_fh
, open
, stp
);
2995 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2996 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3000 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
3001 nfs4_set_claim_prev(open
);
3003 * To finish the open response, we just need to set the rflags.
3005 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
3006 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
3007 !nfsd4_has_session(&resp
->cstate
))
3008 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
3013 void nfsd4_cleanup_open_state(struct nfsd4_open
*open
, __be32 status
)
3015 if (open
->op_openowner
) {
3016 struct nfs4_openowner
*oo
= open
->op_openowner
;
3018 if (!list_empty(&oo
->oo_owner
.so_stateids
))
3019 list_del_init(&oo
->oo_close_lru
);
3020 if (oo
->oo_flags
& NFS4_OO_NEW
) {
3022 release_openowner(oo
);
3023 open
->op_openowner
= NULL
;
3025 oo
->oo_flags
&= ~NFS4_OO_NEW
;
3029 nfsd4_free_file(open
->op_file
);
3031 nfs4_free_stateid(open
->op_stp
);
3035 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3038 struct nfs4_client
*clp
;
3042 dprintk("process_renew(%08x/%08x): starting\n",
3043 clid
->cl_boot
, clid
->cl_id
);
3044 status
= nfserr_stale_clientid
;
3045 if (STALE_CLIENTID(clid
))
3047 clp
= find_confirmed_client(clid
);
3048 status
= nfserr_expired
;
3050 /* We assume the client took too long to RENEW. */
3051 dprintk("nfsd4_renew: clientid not found!\n");
3054 status
= nfserr_cb_path_down
;
3055 if (!list_empty(&clp
->cl_delegations
)
3056 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
3060 nfs4_unlock_state();
3064 static struct lock_manager nfsd4_manager
= {
3068 nfsd4_end_grace(void)
3070 dprintk("NFSD: end of grace period\n");
3071 nfsd4_recdir_purge_old();
3072 locks_end_grace(&nfsd4_manager
);
3074 * Now that every NFSv4 client has had the chance to recover and
3075 * to see the (possibly new, possibly shorter) lease time, we
3076 * can safely set the next grace time to the current lease time:
3078 nfsd4_grace
= nfsd4_lease
;
3082 nfs4_laundromat(void)
3084 struct nfs4_client
*clp
;
3085 struct nfs4_openowner
*oo
;
3086 struct nfs4_delegation
*dp
;
3087 struct list_head
*pos
, *next
, reaplist
;
3088 time_t cutoff
= get_seconds() - nfsd4_lease
;
3089 time_t t
, clientid_val
= nfsd4_lease
;
3090 time_t u
, test_val
= nfsd4_lease
;
3094 dprintk("NFSD: laundromat service - starting\n");
3095 if (locks_in_grace())
3097 INIT_LIST_HEAD(&reaplist
);
3098 spin_lock(&client_lock
);
3099 list_for_each_safe(pos
, next
, &client_lru
) {
3100 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3101 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
3102 t
= clp
->cl_time
- cutoff
;
3103 if (clientid_val
> t
)
3107 if (atomic_read(&clp
->cl_refcount
)) {
3108 dprintk("NFSD: client in use (clientid %08x)\n",
3109 clp
->cl_clientid
.cl_id
);
3112 unhash_client_locked(clp
);
3113 list_add(&clp
->cl_lru
, &reaplist
);
3115 spin_unlock(&client_lock
);
3116 list_for_each_safe(pos
, next
, &reaplist
) {
3117 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3118 dprintk("NFSD: purging unused client (clientid %08x)\n",
3119 clp
->cl_clientid
.cl_id
);
3120 nfsd4_remove_clid_dir(clp
);
3123 spin_lock(&recall_lock
);
3124 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3125 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3126 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3127 u
= dp
->dl_time
- cutoff
;
3132 list_move(&dp
->dl_recall_lru
, &reaplist
);
3134 spin_unlock(&recall_lock
);
3135 list_for_each_safe(pos
, next
, &reaplist
) {
3136 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3137 unhash_delegation(dp
);
3139 test_val
= nfsd4_lease
;
3140 list_for_each_safe(pos
, next
, &close_lru
) {
3141 oo
= container_of(pos
, struct nfs4_openowner
, oo_close_lru
);
3142 if (time_after((unsigned long)oo
->oo_time
, (unsigned long)cutoff
)) {
3143 u
= oo
->oo_time
- cutoff
;
3148 release_openowner(oo
);
3150 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3151 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3152 nfs4_unlock_state();
3153 return clientid_val
;
3156 static struct workqueue_struct
*laundry_wq
;
3157 static void laundromat_main(struct work_struct
*);
3158 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3161 laundromat_main(struct work_struct
*not_used
)
3165 t
= nfs4_laundromat();
3166 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3167 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3170 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*stp
)
3172 if (fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
)
3173 return nfserr_bad_stateid
;
3178 STALE_STATEID(stateid_t
*stateid
)
3180 if (stateid
->si_opaque
.so_clid
.cl_boot
== boot_time
)
3182 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3183 STATEID_VAL(stateid
));
3188 access_permit_read(unsigned long access_bmap
)
3190 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
3191 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
3192 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
3196 access_permit_write(unsigned long access_bmap
)
3198 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
3199 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
3203 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
3205 __be32 status
= nfserr_openmode
;
3207 /* For lock stateid's, we test the parent open, not the lock: */
3208 if (stp
->st_openstp
)
3209 stp
= stp
->st_openstp
;
3210 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
3212 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
3219 static inline __be32
3220 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3222 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3224 else if (locks_in_grace()) {
3225 /* Answer in remaining cases depends on existence of
3226 * conflicting state; so we must wait out the grace period. */
3227 return nfserr_grace
;
3228 } else if (flags
& WR_STATE
)
3229 return nfs4_share_conflict(current_fh
,
3230 NFS4_SHARE_DENY_WRITE
);
3231 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3232 return nfs4_share_conflict(current_fh
,
3233 NFS4_SHARE_DENY_READ
);
3237 * Allow READ/WRITE during grace period on recovered state only for files
3238 * that are not able to provide mandatory locking.
3241 grace_disallows_io(struct inode
*inode
)
3243 return locks_in_grace() && mandatory_lock(inode
);
3246 /* Returns true iff a is later than b: */
3247 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3249 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3252 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3255 * When sessions are used the stateid generation number is ignored
3258 if (has_session
&& in
->si_generation
== 0)
3261 if (in
->si_generation
== ref
->si_generation
)
3264 /* If the client sends us a stateid from the future, it's buggy: */
3265 if (stateid_generation_after(in
, ref
))
3266 return nfserr_bad_stateid
;
3268 * However, we could see a stateid from the past, even from a
3269 * non-buggy client. For example, if the client sends a lock
3270 * while some IO is outstanding, the lock may bump si_generation
3271 * while the IO is still in flight. The client could avoid that
3272 * situation by waiting for responses on all the IO requests,
3273 * but better performance may result in retrying IO that
3274 * receives an old_stateid error if requests are rarely
3275 * reordered in flight:
3277 return nfserr_old_stateid
;
3280 __be32
nfs4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
3282 struct nfs4_stid
*s
;
3283 struct nfs4_ol_stateid
*ols
;
3286 if (STALE_STATEID(stateid
))
3287 return nfserr_stale_stateid
;
3289 s
= find_stateid(cl
, stateid
);
3291 return nfserr_stale_stateid
;
3292 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3295 if (!(s
->sc_type
& (NFS4_OPEN_STID
| NFS4_LOCK_STID
)))
3297 ols
= openlockstateid(s
);
3298 if (ols
->st_stateowner
->so_is_open_owner
3299 && !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3300 return nfserr_bad_stateid
;
3304 static __be32
nfsd4_lookup_stateid(stateid_t
*stateid
, unsigned char typemask
, struct nfs4_stid
**s
)
3306 struct nfs4_client
*cl
;
3308 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3309 return nfserr_bad_stateid
;
3310 if (STALE_STATEID(stateid
))
3311 return nfserr_stale_stateid
;
3312 cl
= find_confirmed_client(&stateid
->si_opaque
.so_clid
);
3314 return nfserr_expired
;
3315 *s
= find_stateid_by_type(cl
, stateid
, typemask
);
3317 return nfserr_bad_stateid
;
3323 * Checks for stateid operations
3326 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3327 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3329 struct nfs4_stid
*s
;
3330 struct nfs4_ol_stateid
*stp
= NULL
;
3331 struct nfs4_delegation
*dp
= NULL
;
3332 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3333 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3339 if (grace_disallows_io(ino
))
3340 return nfserr_grace
;
3342 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3343 return check_special_stateids(current_fh
, stateid
, flags
);
3345 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
, &s
);
3348 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, nfsd4_has_session(cstate
));
3351 switch (s
->sc_type
) {
3352 case NFS4_DELEG_STID
:
3353 dp
= delegstateid(s
);
3354 status
= nfs4_check_delegmode(dp
, flags
);
3358 *filpp
= dp
->dl_file
->fi_deleg_file
;
3362 case NFS4_OPEN_STID
:
3363 case NFS4_LOCK_STID
:
3364 stp
= openlockstateid(s
);
3365 status
= nfs4_check_fh(current_fh
, stp
);
3368 if (stp
->st_stateowner
->so_is_open_owner
3369 && !(openowner(stp
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3371 status
= nfs4_check_openmode(stp
, flags
);
3375 if (flags
& RD_STATE
)
3376 *filpp
= find_readable_file(stp
->st_file
);
3378 *filpp
= find_writeable_file(stp
->st_file
);
3382 return nfserr_bad_stateid
;
3390 nfsd4_free_lock_stateid(struct nfs4_ol_stateid
*stp
)
3392 if (check_for_locks(stp
->st_file
, lockowner(stp
->st_stateowner
)))
3393 return nfserr_locks_held
;
3394 release_lock_stateid(stp
);
3399 * Test if the stateid is valid
3402 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3403 struct nfsd4_test_stateid
*test_stateid
)
3405 /* real work is done during encoding */
3410 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3411 struct nfsd4_free_stateid
*free_stateid
)
3413 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3414 struct nfs4_stid
*s
;
3415 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3416 __be32 ret
= nfserr_bad_stateid
;
3419 s
= find_stateid(cl
, stateid
);
3422 switch (s
->sc_type
) {
3423 case NFS4_DELEG_STID
:
3424 ret
= nfserr_locks_held
;
3426 case NFS4_OPEN_STID
:
3427 case NFS4_LOCK_STID
:
3428 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3431 if (s
->sc_type
== NFS4_LOCK_STID
)
3432 ret
= nfsd4_free_lock_stateid(openlockstateid(s
));
3434 ret
= nfserr_locks_held
;
3437 ret
= nfserr_bad_stateid
;
3440 nfs4_unlock_state();
3447 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3448 RD_STATE
: WR_STATE
;
3451 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
3453 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3454 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
3457 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
3460 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
)
3462 * "Closed" stateid's exist *only* to return
3463 * nfserr_replay_me from the previous step.
3465 return nfserr_bad_stateid
;
3466 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3469 return nfs4_check_fh(current_fh
, stp
);
3473 * Checks for sequence id mutating operations.
3476 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3477 stateid_t
*stateid
, char typemask
,
3478 struct nfs4_ol_stateid
**stpp
)
3481 struct nfs4_stid
*s
;
3483 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3484 seqid
, STATEID_VAL(stateid
));
3487 status
= nfsd4_lookup_stateid(stateid
, typemask
, &s
);
3490 *stpp
= openlockstateid(s
);
3491 cstate
->replay_owner
= (*stpp
)->st_stateowner
;
3493 return nfs4_seqid_op_checks(cstate
, stateid
, seqid
, *stpp
);
3496 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
, stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
)
3499 struct nfs4_openowner
*oo
;
3501 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
3502 NFS4_OPEN_STID
, stpp
);
3505 oo
= openowner((*stpp
)->st_stateowner
);
3506 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
3507 return nfserr_bad_stateid
;
3512 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3513 struct nfsd4_open_confirm
*oc
)
3516 struct nfs4_openowner
*oo
;
3517 struct nfs4_ol_stateid
*stp
;
3519 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3520 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3521 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3523 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3529 status
= nfs4_preprocess_seqid_op(cstate
,
3530 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3531 NFS4_OPEN_STID
, &stp
);
3534 oo
= openowner(stp
->st_stateowner
);
3535 status
= nfserr_bad_stateid
;
3536 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
)
3538 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3539 update_stateid(&stp
->st_stid
.sc_stateid
);
3540 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3541 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3542 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3544 nfsd4_create_clid_dir(oo
->oo_owner
.so_client
);
3547 if (!cstate
->replay_owner
)
3548 nfs4_unlock_state();
3552 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
3554 if (!test_bit(access
, &stp
->st_access_bmap
))
3556 nfs4_file_put_access(stp
->st_file
, nfs4_access_to_omode(access
));
3557 __clear_bit(access
, &stp
->st_access_bmap
);
3560 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
3562 switch (to_access
) {
3563 case NFS4_SHARE_ACCESS_READ
:
3564 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
3565 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3567 case NFS4_SHARE_ACCESS_WRITE
:
3568 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
3569 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3571 case NFS4_SHARE_ACCESS_BOTH
:
3579 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3582 for (i
= 0; i
< 4; i
++) {
3583 if ((i
& deny
) != i
)
3584 __clear_bit(i
, bmap
);
3589 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3590 struct nfsd4_compound_state
*cstate
,
3591 struct nfsd4_open_downgrade
*od
)
3594 struct nfs4_ol_stateid
*stp
;
3596 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3597 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3598 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3600 /* We don't yet support WANT bits: */
3601 if (od
->od_deleg_want
)
3602 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
3606 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
3607 &od
->od_stateid
, &stp
);
3610 status
= nfserr_inval
;
3611 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3612 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3613 stp
->st_access_bmap
, od
->od_share_access
);
3616 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3617 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3618 stp
->st_deny_bmap
, od
->od_share_deny
);
3621 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
3623 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3625 update_stateid(&stp
->st_stid
.sc_stateid
);
3626 memcpy(&od
->od_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3629 if (!cstate
->replay_owner
)
3630 nfs4_unlock_state();
3634 void nfsd4_purge_closed_stateid(struct nfs4_stateowner
*so
)
3636 struct nfs4_openowner
*oo
;
3637 struct nfs4_ol_stateid
*s
;
3639 if (!so
->so_is_open_owner
)
3642 s
= oo
->oo_last_closed_stid
;
3645 if (!(oo
->oo_flags
& NFS4_OO_PURGE_CLOSE
)) {
3646 /* Release the last_closed_stid on the next seqid bump: */
3647 oo
->oo_flags
|= NFS4_OO_PURGE_CLOSE
;
3650 oo
->oo_flags
&= ~NFS4_OO_PURGE_CLOSE
;
3651 release_last_closed_stateid(oo
);
3654 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
3656 unhash_open_stateid(s
);
3657 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
3661 * nfs4_unlock_state() called after encode
3664 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3665 struct nfsd4_close
*close
)
3668 struct nfs4_openowner
*oo
;
3669 struct nfs4_ol_stateid
*stp
;
3671 dprintk("NFSD: nfsd4_close on file %.*s\n",
3672 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3673 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3676 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3678 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
3682 oo
= openowner(stp
->st_stateowner
);
3684 update_stateid(&stp
->st_stid
.sc_stateid
);
3685 memcpy(&close
->cl_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3687 nfsd4_close_open_stateid(stp
);
3688 oo
->oo_last_closed_stid
= stp
;
3690 /* place unused nfs4_stateowners on so_close_lru list to be
3691 * released by the laundromat service after the lease period
3692 * to enable us to handle CLOSE replay
3694 if (list_empty(&oo
->oo_owner
.so_stateids
))
3695 move_to_close_lru(oo
);
3697 if (!cstate
->replay_owner
)
3698 nfs4_unlock_state();
3703 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3704 struct nfsd4_delegreturn
*dr
)
3706 struct nfs4_delegation
*dp
;
3707 stateid_t
*stateid
= &dr
->dr_stateid
;
3708 struct nfs4_stid
*s
;
3709 struct inode
*inode
;
3712 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3714 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3717 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
, &s
);
3720 dp
= delegstateid(s
);
3721 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3725 unhash_delegation(dp
);
3727 nfs4_unlock_state();
3733 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3735 #define LOCKOWNER_INO_HASH_BITS 8
3736 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3737 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3740 end_offset(u64 start
, u64 len
)
3745 return end
>= start
? end
: NFS4_MAX_UINT64
;
3748 /* last octet in a range */
3750 last_byte_offset(u64 start
, u64 len
)
3756 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3759 static unsigned int lockowner_ino_hashval(struct inode
*inode
, u32 cl_id
, struct xdr_netobj
*ownername
)
3761 return (file_hashval(inode
) + cl_id
3762 + opaque_hashval(ownername
->data
, ownername
->len
))
3763 & LOCKOWNER_INO_HASH_MASK
;
3766 static struct list_head lockowner_ino_hashtbl
[LOCKOWNER_INO_HASH_SIZE
];
3769 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3770 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3771 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3772 * locking, this prevents us from being completely protocol-compliant. The
3773 * real solution to this problem is to start using unsigned file offsets in
3774 * the VFS, but this is a very deep change!
3777 nfs4_transform_lock_offset(struct file_lock
*lock
)
3779 if (lock
->fl_start
< 0)
3780 lock
->fl_start
= OFFSET_MAX
;
3781 if (lock
->fl_end
< 0)
3782 lock
->fl_end
= OFFSET_MAX
;
3785 /* Hack!: For now, we're defining this just so we can use a pointer to it
3786 * as a unique cookie to identify our (NFSv4's) posix locks. */
3787 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3791 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3793 struct nfs4_lockowner
*lo
;
3795 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3796 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3797 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3798 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3799 if (!deny
->ld_owner
.data
)
3800 /* We just don't care that much */
3802 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3803 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3806 deny
->ld_owner
.len
= 0;
3807 deny
->ld_owner
.data
= NULL
;
3808 deny
->ld_clientid
.cl_boot
= 0;
3809 deny
->ld_clientid
.cl_id
= 0;
3811 deny
->ld_start
= fl
->fl_start
;
3812 deny
->ld_length
= NFS4_MAX_UINT64
;
3813 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3814 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3815 deny
->ld_type
= NFS4_READ_LT
;
3816 if (fl
->fl_type
!= F_RDLCK
)
3817 deny
->ld_type
= NFS4_WRITE_LT
;
3820 static bool same_lockowner_ino(struct nfs4_lockowner
*lo
, struct inode
*inode
, clientid_t
*clid
, struct xdr_netobj
*owner
)
3822 struct nfs4_ol_stateid
*lst
;
3824 if (!same_owner_str(&lo
->lo_owner
, owner
, clid
))
3826 lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3827 struct nfs4_ol_stateid
, st_perstateowner
);
3828 return lst
->st_file
->fi_inode
== inode
;
3831 static struct nfs4_lockowner
*
3832 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3833 struct xdr_netobj
*owner
)
3835 unsigned int hashval
= lockowner_ino_hashval(inode
, clid
->cl_id
, owner
);
3836 struct nfs4_lockowner
*lo
;
3838 list_for_each_entry(lo
, &lockowner_ino_hashtbl
[hashval
], lo_owner_ino_hash
) {
3839 if (same_lockowner_ino(lo
, inode
, clid
, owner
))
3845 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
)
3847 struct inode
*inode
= open_stp
->st_file
->fi_inode
;
3848 unsigned int inohash
= lockowner_ino_hashval(inode
,
3849 clp
->cl_clientid
.cl_id
, &lo
->lo_owner
.so_owner
);
3851 list_add(&lo
->lo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
3852 list_add(&lo
->lo_owner_ino_hash
, &lockowner_ino_hashtbl
[inohash
]);
3853 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3857 * Alloc a lock owner structure.
3858 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3861 * strhashval = ownerstr_hashval
3864 static struct nfs4_lockowner
*
3865 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3866 struct nfs4_lockowner
*lo
;
3868 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
3871 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
3872 lo
->lo_owner
.so_is_open_owner
= 0;
3873 /* It is the openowner seqid that will be incremented in encode in the
3874 * case of new lockowners; so increment the lock seqid manually: */
3875 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3876 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
3880 static struct nfs4_ol_stateid
*
3881 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_ol_stateid
*open_stp
)
3883 struct nfs4_ol_stateid
*stp
;
3884 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
3886 stp
= nfs4_alloc_stateid(clp
);
3889 init_stid(&stp
->st_stid
, clp
, NFS4_LOCK_STID
);
3890 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3891 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
3892 stp
->st_stateowner
= &lo
->lo_owner
;
3895 stp
->st_access_bmap
= 0;
3896 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3897 stp
->st_openstp
= open_stp
;
3902 check_lock_length(u64 offset
, u64 length
)
3904 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3905 LOFF_OVERFLOW(offset
, length
)));
3908 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
3910 struct nfs4_file
*fp
= lock_stp
->st_file
;
3911 int oflag
= nfs4_access_to_omode(access
);
3913 if (test_bit(access
, &lock_stp
->st_access_bmap
))
3915 nfs4_file_get_access(fp
, oflag
);
3916 __set_bit(access
, &lock_stp
->st_access_bmap
);
3919 __be32
lookup_or_create_lock_state(struct nfsd4_compound_state
*cstate
, struct nfs4_ol_stateid
*ost
, struct nfsd4_lock
*lock
, struct nfs4_ol_stateid
**lst
, bool *new)
3921 struct nfs4_file
*fi
= ost
->st_file
;
3922 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
3923 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
3924 struct nfs4_lockowner
*lo
;
3925 unsigned int strhashval
;
3927 lo
= find_lockowner_str(fi
->fi_inode
, &cl
->cl_clientid
, &lock
->v
.new.owner
);
3929 if (!cstate
->minorversion
)
3930 return nfserr_bad_seqid
;
3931 /* XXX: a lockowner always has exactly one stateid: */
3932 *lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3933 struct nfs4_ol_stateid
, st_perstateowner
);
3936 strhashval
= ownerstr_hashval(cl
->cl_clientid
.cl_id
,
3937 &lock
->v
.new.owner
);
3938 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
3940 return nfserr_jukebox
;
3941 *lst
= alloc_init_lock_stateid(lo
, fi
, ost
);
3943 release_lockowner(lo
);
3944 return nfserr_jukebox
;
3954 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3955 struct nfsd4_lock
*lock
)
3957 struct nfs4_openowner
*open_sop
= NULL
;
3958 struct nfs4_lockowner
*lock_sop
= NULL
;
3959 struct nfs4_ol_stateid
*lock_stp
;
3960 struct nfs4_file
*fp
;
3961 struct file
*filp
= NULL
;
3962 struct file_lock file_lock
;
3963 struct file_lock conflock
;
3965 bool new_state
= false;
3969 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3970 (long long) lock
->lk_offset
,
3971 (long long) lock
->lk_length
);
3973 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3974 return nfserr_inval
;
3976 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3977 S_IFREG
, NFSD_MAY_LOCK
))) {
3978 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3984 if (lock
->lk_is_new
) {
3986 * Client indicates that this is a new lockowner.
3987 * Use open owner and open stateid to create lock owner and
3990 struct nfs4_ol_stateid
*open_stp
= NULL
;
3992 if (nfsd4_has_session(cstate
))
3993 /* See rfc 5661 18.10.3: given clientid is ignored: */
3994 memcpy(&lock
->v
.new.clientid
,
3995 &cstate
->session
->se_client
->cl_clientid
,
3996 sizeof(clientid_t
));
3998 status
= nfserr_stale_clientid
;
3999 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
4002 /* validate and update open stateid and open seqid */
4003 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
4004 lock
->lk_new_open_seqid
,
4005 &lock
->lk_new_open_stateid
,
4009 open_sop
= openowner(open_stp
->st_stateowner
);
4010 status
= nfserr_bad_stateid
;
4011 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
4012 &lock
->v
.new.clientid
))
4014 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
4015 &lock_stp
, &new_state
);
4019 /* lock (lock owner + lock stateid) already exists */
4020 status
= nfs4_preprocess_seqid_op(cstate
,
4021 lock
->lk_old_lock_seqid
,
4022 &lock
->lk_old_lock_stateid
,
4023 NFS4_LOCK_STID
, &lock_stp
);
4027 lock_sop
= lockowner(lock_stp
->st_stateowner
);
4028 fp
= lock_stp
->st_file
;
4030 lkflg
= setlkflg(lock
->lk_type
);
4031 status
= nfs4_check_openmode(lock_stp
, lkflg
);
4035 status
= nfserr_grace
;
4036 if (locks_in_grace() && !lock
->lk_reclaim
)
4038 status
= nfserr_no_grace
;
4039 if (!locks_in_grace() && lock
->lk_reclaim
)
4042 locks_init_lock(&file_lock
);
4043 switch (lock
->lk_type
) {
4046 filp
= find_readable_file(lock_stp
->st_file
);
4048 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4049 file_lock
.fl_type
= F_RDLCK
;
4052 case NFS4_WRITEW_LT
:
4053 filp
= find_writeable_file(lock_stp
->st_file
);
4055 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4056 file_lock
.fl_type
= F_WRLCK
;
4059 status
= nfserr_inval
;
4063 status
= nfserr_openmode
;
4066 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
4067 file_lock
.fl_pid
= current
->tgid
;
4068 file_lock
.fl_file
= filp
;
4069 file_lock
.fl_flags
= FL_POSIX
;
4070 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4072 file_lock
.fl_start
= lock
->lk_offset
;
4073 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4074 nfs4_transform_lock_offset(&file_lock
);
4077 * Try to lock the file in the VFS.
4078 * Note: locks.c uses the BKL to protect the inode's lock list.
4081 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
4083 case 0: /* success! */
4084 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
4085 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
4089 case (EAGAIN
): /* conflock holds conflicting lock */
4090 status
= nfserr_denied
;
4091 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4092 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
4095 status
= nfserr_deadlock
;
4098 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4099 status
= nfserrno(err
);
4103 if (status
&& new_state
)
4104 release_lockowner(lock_sop
);
4105 if (!cstate
->replay_owner
)
4106 nfs4_unlock_state();
4111 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4112 * so we do a temporary open here just to get an open file to pass to
4113 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4116 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4121 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4124 err
= vfs_test_lock(file
, lock
);
4133 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4134 struct nfsd4_lockt
*lockt
)
4136 struct inode
*inode
;
4137 struct file_lock file_lock
;
4138 struct nfs4_lockowner
*lo
;
4142 if (locks_in_grace())
4143 return nfserr_grace
;
4145 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4146 return nfserr_inval
;
4150 status
= nfserr_stale_clientid
;
4151 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
4154 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4157 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4158 locks_init_lock(&file_lock
);
4159 switch (lockt
->lt_type
) {
4162 file_lock
.fl_type
= F_RDLCK
;
4165 case NFS4_WRITEW_LT
:
4166 file_lock
.fl_type
= F_WRLCK
;
4169 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4170 status
= nfserr_inval
;
4174 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4176 file_lock
.fl_owner
= (fl_owner_t
)lo
;
4177 file_lock
.fl_pid
= current
->tgid
;
4178 file_lock
.fl_flags
= FL_POSIX
;
4180 file_lock
.fl_start
= lockt
->lt_offset
;
4181 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4183 nfs4_transform_lock_offset(&file_lock
);
4186 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4188 status
= nfserrno(error
);
4191 if (file_lock
.fl_type
!= F_UNLCK
) {
4192 status
= nfserr_denied
;
4193 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4196 nfs4_unlock_state();
4201 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4202 struct nfsd4_locku
*locku
)
4204 struct nfs4_ol_stateid
*stp
;
4205 struct file
*filp
= NULL
;
4206 struct file_lock file_lock
;
4210 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4211 (long long) locku
->lu_offset
,
4212 (long long) locku
->lu_length
);
4214 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4215 return nfserr_inval
;
4219 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4220 &locku
->lu_stateid
, NFS4_LOCK_STID
, &stp
);
4223 filp
= find_any_file(stp
->st_file
);
4225 status
= nfserr_lock_range
;
4229 locks_init_lock(&file_lock
);
4230 file_lock
.fl_type
= F_UNLCK
;
4231 file_lock
.fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4232 file_lock
.fl_pid
= current
->tgid
;
4233 file_lock
.fl_file
= filp
;
4234 file_lock
.fl_flags
= FL_POSIX
;
4235 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4236 file_lock
.fl_start
= locku
->lu_offset
;
4238 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4239 nfs4_transform_lock_offset(&file_lock
);
4242 * Try to unlock the file in the VFS.
4244 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4246 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4250 * OK, unlock succeeded; the only thing left to do is update the stateid.
4252 update_stateid(&stp
->st_stid
.sc_stateid
);
4253 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4256 if (!cstate
->replay_owner
)
4257 nfs4_unlock_state();
4261 status
= nfserrno(err
);
4267 * 1: locks held by lockowner
4268 * 0: no locks held by lockowner
4271 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4273 struct file_lock
**flpp
;
4274 struct inode
*inode
= filp
->fi_inode
;
4278 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4279 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4290 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4291 struct nfsd4_compound_state
*cstate
,
4292 struct nfsd4_release_lockowner
*rlockowner
)
4294 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4295 struct nfs4_stateowner
*sop
;
4296 struct nfs4_lockowner
*lo
;
4297 struct nfs4_ol_stateid
*stp
;
4298 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4299 struct list_head matches
;
4300 unsigned int hashval
= ownerstr_hashval(clid
->cl_id
, owner
);
4303 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4304 clid
->cl_boot
, clid
->cl_id
);
4306 /* XXX check for lease expiration */
4308 status
= nfserr_stale_clientid
;
4309 if (STALE_CLIENTID(clid
))
4314 status
= nfserr_locks_held
;
4315 INIT_LIST_HEAD(&matches
);
4317 list_for_each_entry(sop
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
4318 if (sop
->so_is_open_owner
)
4320 if (!same_owner_str(sop
, owner
, clid
))
4322 list_for_each_entry(stp
, &sop
->so_stateids
,
4324 lo
= lockowner(sop
);
4325 if (check_for_locks(stp
->st_file
, lo
))
4327 list_add(&lo
->lo_list
, &matches
);
4330 /* Clients probably won't expect us to return with some (but not all)
4331 * of the lockowner state released; so don't release any until all
4332 * have been checked. */
4334 while (!list_empty(&matches
)) {
4335 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4337 /* unhash_stateowner deletes so_perclient only
4338 * for openowners. */
4339 list_del(&lo
->lo_list
);
4340 release_lockowner(lo
);
4343 nfs4_unlock_state();
4347 static inline struct nfs4_client_reclaim
*
4350 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4354 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4356 unsigned int strhashval
= clientstr_hashval(name
);
4357 struct nfs4_client
*clp
;
4359 clp
= find_confirmed_client_by_str(name
, strhashval
);
4364 * failure => all reset bets are off, nfserr_no_grace...
4367 nfs4_client_to_reclaim(const char *name
)
4369 unsigned int strhashval
;
4370 struct nfs4_client_reclaim
*crp
= NULL
;
4372 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4373 crp
= alloc_reclaim();
4376 strhashval
= clientstr_hashval(name
);
4377 INIT_LIST_HEAD(&crp
->cr_strhash
);
4378 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4379 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4380 reclaim_str_hashtbl_size
++;
4385 nfs4_release_reclaim(void)
4387 struct nfs4_client_reclaim
*crp
= NULL
;
4390 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4391 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4392 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4393 struct nfs4_client_reclaim
, cr_strhash
);
4394 list_del(&crp
->cr_strhash
);
4396 reclaim_str_hashtbl_size
--;
4399 BUG_ON(reclaim_str_hashtbl_size
);
4403 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4404 static struct nfs4_client_reclaim
*
4405 nfs4_find_reclaim_client(clientid_t
*clid
)
4407 unsigned int strhashval
;
4408 struct nfs4_client
*clp
;
4409 struct nfs4_client_reclaim
*crp
= NULL
;
4412 /* find clientid in conf_id_hashtbl */
4413 clp
= find_confirmed_client(clid
);
4417 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4418 clp
->cl_name
.len
, clp
->cl_name
.data
,
4421 /* find clp->cl_name in reclaim_str_hashtbl */
4422 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4423 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4424 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4432 * Called from OPEN. Look for clientid in reclaim list.
4435 nfs4_check_open_reclaim(clientid_t
*clid
)
4437 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4440 #ifdef CONFIG_NFSD_FAULT_INJECTION
4442 void nfsd_forget_clients(u64 num
)
4444 struct nfs4_client
*clp
, *next
;
4448 list_for_each_entry_safe(clp
, next
, &client_lru
, cl_lru
) {
4449 nfsd4_remove_clid_dir(clp
);
4454 nfs4_unlock_state();
4456 printk(KERN_INFO
"NFSD: Forgot %d clients", count
);
4459 static void release_lockowner_sop(struct nfs4_stateowner
*sop
)
4461 release_lockowner(lockowner(sop
));
4464 static void release_openowner_sop(struct nfs4_stateowner
*sop
)
4466 release_openowner(openowner(sop
));
4469 static int nfsd_release_n_owners(u64 num
, bool is_open_owner
,
4470 void (*release_sop
)(struct nfs4_stateowner
*))
4473 struct nfs4_stateowner
*sop
, *next
;
4475 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4476 list_for_each_entry_safe(sop
, next
, &ownerstr_hashtbl
[i
], so_strhash
) {
4477 if (sop
->so_is_open_owner
!= is_open_owner
)
4487 void nfsd_forget_locks(u64 num
)
4492 count
= nfsd_release_n_owners(num
, false, release_lockowner_sop
);
4493 nfs4_unlock_state();
4495 printk(KERN_INFO
"NFSD: Forgot %d locks", count
);
4498 void nfsd_forget_openowners(u64 num
)
4503 count
= nfsd_release_n_owners(num
, true, release_openowner_sop
);
4504 nfs4_unlock_state();
4506 printk(KERN_INFO
"NFSD: Forgot %d open owners", count
);
4509 int nfsd_process_n_delegations(u64 num
, void (*deleg_func
)(struct nfs4_delegation
*))
4512 struct nfs4_file
*fp
, *fnext
;
4513 struct nfs4_delegation
*dp
, *dnext
;
4515 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4516 list_for_each_entry_safe(fp
, fnext
, &file_hashtbl
[i
], fi_hash
) {
4517 list_for_each_entry_safe(dp
, dnext
, &fp
->fi_delegations
, dl_perfile
) {
4528 void nfsd_forget_delegations(u64 num
)
4533 count
= nfsd_process_n_delegations(num
, unhash_delegation
);
4534 nfs4_unlock_state();
4536 printk(KERN_INFO
"NFSD: Forgot %d delegations", count
);
4539 void nfsd_recall_delegations(u64 num
)
4544 spin_lock(&recall_lock
);
4545 count
= nfsd_process_n_delegations(num
, nfsd_break_one_deleg
);
4546 spin_unlock(&recall_lock
);
4547 nfs4_unlock_state();
4549 printk(KERN_INFO
"NFSD: Recalled %d delegations", count
);
4552 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4554 /* initialization to perform at module load time: */
4557 nfs4_state_init(void)
4561 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4562 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4563 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4564 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4565 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4566 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4568 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4569 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4570 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4571 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4573 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4574 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4576 for (i
= 0; i
< LOCKOWNER_INO_HASH_SIZE
; i
++)
4577 INIT_LIST_HEAD(&lockowner_ino_hashtbl
[i
]);
4578 INIT_LIST_HEAD(&close_lru
);
4579 INIT_LIST_HEAD(&client_lru
);
4580 INIT_LIST_HEAD(&del_recall_lru
);
4581 reclaim_str_hashtbl_size
= 0;
4585 nfsd4_load_reboot_recovery_data(void)
4590 nfsd4_init_recdir();
4591 status
= nfsd4_recdir_load();
4592 nfs4_unlock_state();
4594 printk("NFSD: Failure reading reboot recovery data\n");
4598 * Since the lifetime of a delegation isn't limited to that of an open, a
4599 * client may quite reasonably hang on to a delegation as long as it has
4600 * the inode cached. This becomes an obvious problem the first time a
4601 * client's inode cache approaches the size of the server's total memory.
4603 * For now we avoid this problem by imposing a hard limit on the number
4604 * of delegations, which varies according to the server's memory size.
4607 set_max_delegations(void)
4610 * Allow at most 4 delegations per megabyte of RAM. Quick
4611 * estimates suggest that in the worst case (where every delegation
4612 * is for a different inode), a delegation could take about 1.5K,
4613 * giving a worst case usage of about 6% of memory.
4615 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4618 /* initialization to perform when the nfsd service is started: */
4621 __nfs4_state_start(void)
4625 boot_time
= get_seconds();
4626 locks_start_grace(&nfsd4_manager
);
4627 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4629 ret
= set_callback_cred();
4632 laundry_wq
= create_singlethread_workqueue("nfsd4");
4633 if (laundry_wq
== NULL
)
4635 ret
= nfsd4_create_callback_queue();
4637 goto out_free_laundry
;
4638 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4639 set_max_delegations();
4642 destroy_workqueue(laundry_wq
);
4647 nfs4_state_start(void)
4649 nfsd4_load_reboot_recovery_data();
4650 return __nfs4_state_start();
4654 __nfs4_state_shutdown(void)
4657 struct nfs4_client
*clp
= NULL
;
4658 struct nfs4_delegation
*dp
= NULL
;
4659 struct list_head
*pos
, *next
, reaplist
;
4661 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4662 while (!list_empty(&conf_id_hashtbl
[i
])) {
4663 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4666 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4667 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4671 INIT_LIST_HEAD(&reaplist
);
4672 spin_lock(&recall_lock
);
4673 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4674 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4675 list_move(&dp
->dl_recall_lru
, &reaplist
);
4677 spin_unlock(&recall_lock
);
4678 list_for_each_safe(pos
, next
, &reaplist
) {
4679 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4680 unhash_delegation(dp
);
4683 nfsd4_shutdown_recdir();
4687 nfs4_state_shutdown(void)
4689 cancel_delayed_work_sync(&laundromat_work
);
4690 destroy_workqueue(laundry_wq
);
4691 locks_end_grace(&nfsd4_manager
);
4693 nfs4_release_reclaim();
4694 __nfs4_state_shutdown();
4695 nfs4_unlock_state();
4696 nfsd4_destroy_callback_queue();
4700 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4702 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
4703 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
4707 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4709 if (cstate
->minorversion
) {
4710 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
4711 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4716 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
4718 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4722 * functions to set current state id
4725 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4727 put_stateid(cstate
, &odp
->od_stateid
);
4731 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
4733 put_stateid(cstate
, &open
->op_stateid
);
4737 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4739 put_stateid(cstate
, &close
->cl_stateid
);
4743 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
4745 put_stateid(cstate
, &lock
->lk_resp_stateid
);
4749 * functions to consume current state id
4753 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4755 get_stateid(cstate
, &odp
->od_stateid
);
4759 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
4761 get_stateid(cstate
, &drp
->dr_stateid
);
4765 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
4767 get_stateid(cstate
, &fsp
->fr_stateid
);
4771 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
4773 get_stateid(cstate
, &setattr
->sa_stateid
);
4777 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4779 get_stateid(cstate
, &close
->cl_stateid
);
4783 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
4785 get_stateid(cstate
, &locku
->lu_stateid
);
4789 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
4791 get_stateid(cstate
, &read
->rd_stateid
);
4795 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
4797 get_stateid(cstate
, &write
->wr_stateid
);