2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
36 #include <linux/smp_lock.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease
= 90; /* default lease time */
49 time_t nfsd4_grace
= 90;
50 static time_t boot_time
;
51 static u32 current_ownerid
= 1;
52 static u32 current_fileid
= 1;
53 static u32 current_delegid
= 1;
54 static stateid_t zerostateid
; /* bits all 0 */
55 static stateid_t onestateid
; /* bits all 1 */
56 static u64 current_sessionid
= 1;
58 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
61 /* forward declarations */
62 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
63 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
64 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
65 static void nfs4_set_recdir(char *recdir
);
69 /* Currently used for almost all code touching nfsv4 state: */
70 static DEFINE_MUTEX(client_mutex
);
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
77 static DEFINE_SPINLOCK(recall_lock
);
79 static struct kmem_cache
*stateowner_slab
= NULL
;
80 static struct kmem_cache
*file_slab
= NULL
;
81 static struct kmem_cache
*stateid_slab
= NULL
;
82 static struct kmem_cache
*deleg_slab
= NULL
;
87 mutex_lock(&client_mutex
);
91 nfs4_unlock_state(void)
93 mutex_unlock(&client_mutex
);
97 opaque_hashval(const void *ptr
, int nbytes
)
99 unsigned char *cptr
= (unsigned char *) ptr
;
109 static struct list_head del_recall_lru
;
112 put_nfs4_file(struct nfs4_file
*fi
)
114 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
115 list_del(&fi
->fi_hash
);
116 spin_unlock(&recall_lock
);
118 kmem_cache_free(file_slab
, fi
);
123 get_nfs4_file(struct nfs4_file
*fi
)
125 atomic_inc(&fi
->fi_ref
);
128 static int num_delegations
;
129 unsigned int max_delegations
;
132 * Open owner state (share locks)
135 /* hash tables for nfs4_stateowner */
136 #define OWNER_HASH_BITS 8
137 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140 #define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142 #define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
146 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
148 /* hash table for nfs4_file */
149 #define FILE_HASH_BITS 8
150 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
152 /* hash table for (open)nfs4_stateid */
153 #define STATEID_HASH_BITS 10
154 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157 #define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159 #define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
163 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
165 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
167 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
168 atomic_inc(&fp
->fi_access
[oflag
]);
171 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
173 if (oflag
== O_RDWR
) {
174 __nfs4_file_get_access(fp
, O_RDONLY
);
175 __nfs4_file_get_access(fp
, O_WRONLY
);
177 __nfs4_file_get_access(fp
, oflag
);
180 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
182 if (fp
->fi_fds
[oflag
]) {
183 fput(fp
->fi_fds
[oflag
]);
184 fp
->fi_fds
[oflag
] = NULL
;
188 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
190 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
191 nfs4_file_put_fd(fp
, O_RDWR
);
192 nfs4_file_put_fd(fp
, oflag
);
196 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
198 if (oflag
== O_RDWR
) {
199 __nfs4_file_put_access(fp
, O_RDONLY
);
200 __nfs4_file_put_access(fp
, O_WRONLY
);
202 __nfs4_file_put_access(fp
, oflag
);
205 static struct nfs4_delegation
*
206 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
208 struct nfs4_delegation
*dp
;
209 struct nfs4_file
*fp
= stp
->st_file
;
210 struct nfs4_cb_conn
*cb
= &stp
->st_stateowner
->so_client
->cl_cb_conn
;
212 dprintk("NFSD alloc_init_deleg\n");
214 * Major work on the lease subsystem (for example, to support
215 * calbacks on stat) will be required before we can support
216 * write delegations properly.
218 if (type
!= NFS4_OPEN_DELEGATE_READ
)
220 if (fp
->fi_had_conflict
)
222 if (num_delegations
> max_delegations
)
224 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
228 INIT_LIST_HEAD(&dp
->dl_perfile
);
229 INIT_LIST_HEAD(&dp
->dl_perclnt
);
230 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
234 nfs4_file_get_access(fp
, O_RDONLY
);
237 dp
->dl_ident
= cb
->cb_ident
;
238 dp
->dl_stateid
.si_boot
= boot_time
;
239 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
240 dp
->dl_stateid
.si_fileid
= 0;
241 dp
->dl_stateid
.si_generation
= 0;
242 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
244 atomic_set(&dp
->dl_count
, 1);
245 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
246 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
247 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
252 nfs4_put_delegation(struct nfs4_delegation
*dp
)
254 if (atomic_dec_and_test(&dp
->dl_count
)) {
255 dprintk("NFSD: freeing dp %p\n",dp
);
256 put_nfs4_file(dp
->dl_file
);
257 kmem_cache_free(deleg_slab
, dp
);
262 /* Remove the associated file_lock first, then remove the delegation.
263 * lease_modify() is called to remove the FS_LEASE file_lock from
264 * the i_flock list, eventually calling nfsd's lock_manager
265 * fl_release_callback.
268 nfs4_close_delegation(struct nfs4_delegation
*dp
)
270 struct file
*filp
= find_readable_file(dp
->dl_file
);
272 dprintk("NFSD: close_delegation dp %p\n",dp
);
274 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
275 nfs4_file_put_access(dp
->dl_file
, O_RDONLY
);
278 /* Called under the state lock. */
280 unhash_delegation(struct nfs4_delegation
*dp
)
282 list_del_init(&dp
->dl_perfile
);
283 list_del_init(&dp
->dl_perclnt
);
284 spin_lock(&recall_lock
);
285 list_del_init(&dp
->dl_recall_lru
);
286 spin_unlock(&recall_lock
);
287 nfs4_close_delegation(dp
);
288 nfs4_put_delegation(dp
);
295 /* client_lock protects the client lru list and session hash table */
296 static DEFINE_SPINLOCK(client_lock
);
298 /* Hash tables for nfs4_clientid state */
299 #define CLIENT_HASH_BITS 4
300 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
301 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
303 #define clientid_hashval(id) \
304 ((id) & CLIENT_HASH_MASK)
305 #define clientstr_hashval(name) \
306 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
308 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
309 * used in reboot/reset lease grace period processing
311 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
312 * setclientid_confirmed info.
314 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
317 * client_lru holds client queue ordered by nfs4_client.cl_time
320 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
321 * for last close replay.
323 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
324 static int reclaim_str_hashtbl_size
= 0;
325 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
326 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
327 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
328 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
329 static struct list_head client_lru
;
330 static struct list_head close_lru
;
332 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
334 list_del(&stp
->st_hash
);
335 list_del(&stp
->st_perfile
);
336 list_del(&stp
->st_perstateowner
);
339 static void free_generic_stateid(struct nfs4_stateid
*stp
)
341 put_nfs4_file(stp
->st_file
);
342 kmem_cache_free(stateid_slab
, stp
);
345 static void release_lock_stateid(struct nfs4_stateid
*stp
)
349 unhash_generic_stateid(stp
);
350 file
= find_any_file(stp
->st_file
);
352 locks_remove_posix(file
, (fl_owner_t
)stp
->st_stateowner
);
353 free_generic_stateid(stp
);
356 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
358 struct nfs4_stateid
*stp
;
360 list_del(&sop
->so_idhash
);
361 list_del(&sop
->so_strhash
);
362 list_del(&sop
->so_perstateid
);
363 while (!list_empty(&sop
->so_stateids
)) {
364 stp
= list_first_entry(&sop
->so_stateids
,
365 struct nfs4_stateid
, st_perstateowner
);
366 release_lock_stateid(stp
);
370 static void release_lockowner(struct nfs4_stateowner
*sop
)
372 unhash_lockowner(sop
);
373 nfs4_put_stateowner(sop
);
377 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
379 struct nfs4_stateowner
*lock_sop
;
381 while (!list_empty(&open_stp
->st_lockowners
)) {
382 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
383 struct nfs4_stateowner
, so_perstateid
);
384 /* list_del(&open_stp->st_lockowners); */
385 BUG_ON(lock_sop
->so_is_open_owner
);
386 release_lockowner(lock_sop
);
391 set_access(unsigned int *access
, unsigned long bmap
) {
395 for (i
= 1; i
< 4; i
++) {
396 if (test_bit(i
, &bmap
))
402 set_deny(unsigned int *deny
, unsigned long bmap
) {
406 for (i
= 0; i
< 4; i
++) {
407 if (test_bit(i
, &bmap
))
413 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
414 unsigned int access
, deny
;
416 set_access(&access
, stp
->st_access_bmap
);
417 set_deny(&deny
, stp
->st_deny_bmap
);
418 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
423 static int nfs4_access_to_omode(u32 access
)
425 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
426 case NFS4_SHARE_ACCESS_READ
:
428 case NFS4_SHARE_ACCESS_WRITE
:
430 case NFS4_SHARE_ACCESS_BOTH
:
436 static int nfs4_access_bmap_to_omode(struct nfs4_stateid
*stp
)
440 set_access(&access
, stp
->st_access_bmap
);
441 return nfs4_access_to_omode(access
);
444 static void release_open_stateid(struct nfs4_stateid
*stp
)
446 int oflag
= nfs4_access_bmap_to_omode(stp
);
448 unhash_generic_stateid(stp
);
449 release_stateid_lockowners(stp
);
450 nfs4_file_put_access(stp
->st_file
, oflag
);
451 free_generic_stateid(stp
);
454 static void unhash_openowner(struct nfs4_stateowner
*sop
)
456 struct nfs4_stateid
*stp
;
458 list_del(&sop
->so_idhash
);
459 list_del(&sop
->so_strhash
);
460 list_del(&sop
->so_perclient
);
461 list_del(&sop
->so_perstateid
);
462 while (!list_empty(&sop
->so_stateids
)) {
463 stp
= list_first_entry(&sop
->so_stateids
,
464 struct nfs4_stateid
, st_perstateowner
);
465 release_open_stateid(stp
);
469 static void release_openowner(struct nfs4_stateowner
*sop
)
471 unhash_openowner(sop
);
472 list_del(&sop
->so_close_lru
);
473 nfs4_put_stateowner(sop
);
476 #define SESSION_HASH_SIZE 512
477 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
480 hash_sessionid(struct nfs4_sessionid
*sessionid
)
482 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
484 return sid
->sequence
% SESSION_HASH_SIZE
;
488 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
490 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
491 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
495 gen_sessionid(struct nfsd4_session
*ses
)
497 struct nfs4_client
*clp
= ses
->se_client
;
498 struct nfsd4_sessionid
*sid
;
500 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
501 sid
->clientid
= clp
->cl_clientid
;
502 sid
->sequence
= current_sessionid
++;
507 * The protocol defines ca_maxresponssize_cached to include the size of
508 * the rpc header, but all we need to cache is the data starting after
509 * the end of the initial SEQUENCE operation--the rest we regenerate
510 * each time. Therefore we can advertise a ca_maxresponssize_cached
511 * value that is the number of bytes in our cache plus a few additional
512 * bytes. In order to stay on the safe side, and not promise more than
513 * we can cache, those additional bytes must be the minimum possible: 24
514 * bytes of rpc header (xid through accept state, with AUTH_NULL
515 * verifier), 12 for the compound header (with zero-length tag), and 44
516 * for the SEQUENCE op response:
518 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
521 * Give the client the number of ca_maxresponsesize_cached slots it
522 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
523 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
524 * than NFSD_MAX_SLOTS_PER_SESSION.
526 * If we run out of reserved DRC memory we should (up to a point)
527 * re-negotiate active sessions and reduce their slot usage to make
528 * rooom for new connections. For now we just fail the create session.
530 static int set_forechannel_drc_size(struct nfsd4_channel_attrs
*fchan
)
532 int mem
, size
= fchan
->maxresp_cached
;
534 if (fchan
->maxreqs
< 1)
537 if (size
< NFSD_MIN_HDR_SEQ_SZ
)
538 size
= NFSD_MIN_HDR_SEQ_SZ
;
539 size
-= NFSD_MIN_HDR_SEQ_SZ
;
540 if (size
> NFSD_SLOT_CACHE_SIZE
)
541 size
= NFSD_SLOT_CACHE_SIZE
;
543 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
544 mem
= fchan
->maxreqs
* size
;
545 if (mem
> NFSD_MAX_MEM_PER_SESSION
) {
546 fchan
->maxreqs
= NFSD_MAX_MEM_PER_SESSION
/ size
;
547 if (fchan
->maxreqs
> NFSD_MAX_SLOTS_PER_SESSION
)
548 fchan
->maxreqs
= NFSD_MAX_SLOTS_PER_SESSION
;
549 mem
= fchan
->maxreqs
* size
;
552 spin_lock(&nfsd_drc_lock
);
553 /* bound the total session drc memory ussage */
554 if (mem
+ nfsd_drc_mem_used
> nfsd_drc_max_mem
) {
555 fchan
->maxreqs
= (nfsd_drc_max_mem
- nfsd_drc_mem_used
) / size
;
556 mem
= fchan
->maxreqs
* size
;
558 nfsd_drc_mem_used
+= mem
;
559 spin_unlock(&nfsd_drc_lock
);
561 if (fchan
->maxreqs
== 0)
562 return nfserr_jukebox
;
564 fchan
->maxresp_cached
= size
+ NFSD_MIN_HDR_SEQ_SZ
;
569 * fchan holds the client values on input, and the server values on output
570 * sv_max_mesg is the maximum payload plus one page for overhead.
572 static int init_forechannel_attrs(struct svc_rqst
*rqstp
,
573 struct nfsd4_channel_attrs
*session_fchan
,
574 struct nfsd4_channel_attrs
*fchan
)
577 __u32 maxcount
= nfsd_serv
->sv_max_mesg
;
579 /* headerpadsz set to zero in encode routine */
581 /* Use the client's max request and max response size if possible */
582 if (fchan
->maxreq_sz
> maxcount
)
583 fchan
->maxreq_sz
= maxcount
;
584 session_fchan
->maxreq_sz
= fchan
->maxreq_sz
;
586 if (fchan
->maxresp_sz
> maxcount
)
587 fchan
->maxresp_sz
= maxcount
;
588 session_fchan
->maxresp_sz
= fchan
->maxresp_sz
;
590 /* Use the client's maxops if possible */
591 if (fchan
->maxops
> NFSD_MAX_OPS_PER_COMPOUND
)
592 fchan
->maxops
= NFSD_MAX_OPS_PER_COMPOUND
;
593 session_fchan
->maxops
= fchan
->maxops
;
595 status
= set_forechannel_drc_size(fchan
);
597 session_fchan
->maxresp_cached
= fchan
->maxresp_cached
;
598 session_fchan
->maxreqs
= fchan
->maxreqs
;
600 dprintk("%s status %d\n", __func__
, status
);
605 free_session_slots(struct nfsd4_session
*ses
)
609 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
610 kfree(ses
->se_slots
[i
]);
614 * We don't actually need to cache the rpc and session headers, so we
615 * can allocate a little less for each slot:
617 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
619 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
623 alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
,
624 struct nfsd4_create_session
*cses
)
626 struct nfsd4_session
*new, tmp
;
627 struct nfsd4_slot
*sp
;
628 int idx
, slotsize
, cachesize
, i
;
631 memset(&tmp
, 0, sizeof(tmp
));
633 tmp
.se_bchannel
= cses
->back_channel
;
634 status
= init_forechannel_attrs(rqstp
, &tmp
.se_fchannel
,
635 &cses
->fore_channel
);
639 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
)
640 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
642 status
= nfserr_jukebox
;
643 /* allocate struct nfsd4_session and slot table pointers in one piece */
644 slotsize
= tmp
.se_fchannel
.maxreqs
* sizeof(struct nfsd4_slot
*);
645 new = kzalloc(sizeof(*new) + slotsize
, GFP_KERNEL
);
649 memcpy(new, &tmp
, sizeof(*new));
651 /* allocate each struct nfsd4_slot and data cache in one piece */
652 cachesize
= slot_bytes(&new->se_fchannel
);
653 for (i
= 0; i
< new->se_fchannel
.maxreqs
; i
++) {
654 sp
= kzalloc(sizeof(*sp
) + cachesize
, GFP_KERNEL
);
657 new->se_slots
[i
] = sp
;
660 new->se_client
= clp
;
662 idx
= hash_sessionid(&new->se_sessionid
);
663 memcpy(clp
->cl_sessionid
.data
, new->se_sessionid
.data
,
664 NFS4_MAX_SESSIONID_LEN
);
666 new->se_flags
= cses
->flags
;
667 kref_init(&new->se_ref
);
668 spin_lock(&client_lock
);
669 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
670 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
671 spin_unlock(&client_lock
);
677 free_session_slots(new);
682 /* caller must hold client_lock */
683 static struct nfsd4_session
*
684 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
686 struct nfsd4_session
*elem
;
689 dump_sessionid(__func__
, sessionid
);
690 idx
= hash_sessionid(sessionid
);
691 /* Search in the appropriate list */
692 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
693 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
694 NFS4_MAX_SESSIONID_LEN
)) {
699 dprintk("%s: session not found\n", __func__
);
703 /* caller must hold client_lock */
705 unhash_session(struct nfsd4_session
*ses
)
707 list_del(&ses
->se_hash
);
708 list_del(&ses
->se_perclnt
);
712 free_session(struct kref
*kref
)
714 struct nfsd4_session
*ses
;
717 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
718 spin_lock(&nfsd_drc_lock
);
719 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
720 nfsd_drc_mem_used
-= mem
;
721 spin_unlock(&nfsd_drc_lock
);
722 free_session_slots(ses
);
726 /* must be called under the client_lock */
728 renew_client_locked(struct nfs4_client
*clp
)
730 if (is_client_expired(clp
)) {
731 dprintk("%s: client (clientid %08x/%08x) already expired\n",
733 clp
->cl_clientid
.cl_boot
,
734 clp
->cl_clientid
.cl_id
);
739 * Move client to the end to the LRU list.
741 dprintk("renewing client (clientid %08x/%08x)\n",
742 clp
->cl_clientid
.cl_boot
,
743 clp
->cl_clientid
.cl_id
);
744 list_move_tail(&clp
->cl_lru
, &client_lru
);
745 clp
->cl_time
= get_seconds();
749 renew_client(struct nfs4_client
*clp
)
751 spin_lock(&client_lock
);
752 renew_client_locked(clp
);
753 spin_unlock(&client_lock
);
756 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
758 STALE_CLIENTID(clientid_t
*clid
)
760 if (clid
->cl_boot
== boot_time
)
762 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
763 clid
->cl_boot
, clid
->cl_id
, boot_time
);
767 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
769 struct nfs4_client
*clp
;
771 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
774 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
775 if (clp
->cl_name
.data
== NULL
) {
779 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
780 clp
->cl_name
.len
= name
.len
;
785 free_client(struct nfs4_client
*clp
)
787 if (clp
->cl_cred
.cr_group_info
)
788 put_group_info(clp
->cl_cred
.cr_group_info
);
789 kfree(clp
->cl_principal
);
790 kfree(clp
->cl_name
.data
);
795 release_session_client(struct nfsd4_session
*session
)
797 struct nfs4_client
*clp
= session
->se_client
;
799 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
801 if (is_client_expired(clp
)) {
803 session
->se_client
= NULL
;
805 renew_client_locked(clp
);
806 spin_unlock(&client_lock
);
809 /* must be called under the client_lock */
811 unhash_client_locked(struct nfs4_client
*clp
)
813 mark_client_expired(clp
);
814 list_del(&clp
->cl_lru
);
815 while (!list_empty(&clp
->cl_sessions
)) {
816 struct nfsd4_session
*ses
;
817 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
820 nfsd4_put_session(ses
);
825 expire_client(struct nfs4_client
*clp
)
827 struct nfs4_stateowner
*sop
;
828 struct nfs4_delegation
*dp
;
829 struct list_head reaplist
;
831 INIT_LIST_HEAD(&reaplist
);
832 spin_lock(&recall_lock
);
833 while (!list_empty(&clp
->cl_delegations
)) {
834 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
835 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
837 list_del_init(&dp
->dl_perclnt
);
838 list_move(&dp
->dl_recall_lru
, &reaplist
);
840 spin_unlock(&recall_lock
);
841 while (!list_empty(&reaplist
)) {
842 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
843 list_del_init(&dp
->dl_recall_lru
);
844 unhash_delegation(dp
);
846 while (!list_empty(&clp
->cl_openowners
)) {
847 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
848 release_openowner(sop
);
850 nfsd4_set_callback_client(clp
, NULL
);
851 if (clp
->cl_cb_conn
.cb_xprt
)
852 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
853 list_del(&clp
->cl_idhash
);
854 list_del(&clp
->cl_strhash
);
855 spin_lock(&client_lock
);
856 unhash_client_locked(clp
);
857 if (atomic_read(&clp
->cl_refcount
) == 0)
859 spin_unlock(&client_lock
);
862 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
864 memcpy(target
->cl_verifier
.data
, source
->data
,
865 sizeof(target
->cl_verifier
.data
));
868 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
870 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
871 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
874 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
876 target
->cr_uid
= source
->cr_uid
;
877 target
->cr_gid
= source
->cr_gid
;
878 target
->cr_group_info
= source
->cr_group_info
;
879 get_group_info(target
->cr_group_info
);
882 static int same_name(const char *n1
, const char *n2
)
884 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
888 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
890 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
894 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
896 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
900 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
902 return cr1
->cr_uid
== cr2
->cr_uid
;
905 static void gen_clid(struct nfs4_client
*clp
)
907 static u32 current_clientid
= 1;
909 clp
->cl_clientid
.cl_boot
= boot_time
;
910 clp
->cl_clientid
.cl_id
= current_clientid
++;
913 static void gen_confirm(struct nfs4_client
*clp
)
918 p
= (u32
*)clp
->cl_confirm
.data
;
919 *p
++ = get_seconds();
923 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
924 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
926 struct nfs4_client
*clp
;
927 struct sockaddr
*sa
= svc_addr(rqstp
);
930 clp
= alloc_client(name
);
934 princ
= svc_gss_principal(rqstp
);
936 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
937 if (clp
->cl_principal
== NULL
) {
943 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
944 atomic_set(&clp
->cl_refcount
, 0);
945 atomic_set(&clp
->cl_cb_set
, 0);
946 INIT_LIST_HEAD(&clp
->cl_idhash
);
947 INIT_LIST_HEAD(&clp
->cl_strhash
);
948 INIT_LIST_HEAD(&clp
->cl_openowners
);
949 INIT_LIST_HEAD(&clp
->cl_delegations
);
950 INIT_LIST_HEAD(&clp
->cl_sessions
);
951 INIT_LIST_HEAD(&clp
->cl_lru
);
952 clp
->cl_time
= get_seconds();
953 clear_bit(0, &clp
->cl_cb_slot_busy
);
954 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
955 copy_verf(clp
, verf
);
956 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
957 clp
->cl_flavor
= rqstp
->rq_flavor
;
958 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
964 static int check_name(struct xdr_netobj name
)
968 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
969 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
976 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
978 unsigned int idhashval
;
980 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
981 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
982 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
987 move_to_confirmed(struct nfs4_client
*clp
)
989 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
990 unsigned int strhashval
;
992 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
993 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
994 strhashval
= clientstr_hashval(clp
->cl_recdir
);
995 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
999 static struct nfs4_client
*
1000 find_confirmed_client(clientid_t
*clid
)
1002 struct nfs4_client
*clp
;
1003 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1005 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1006 if (same_clid(&clp
->cl_clientid
, clid
))
1012 static struct nfs4_client
*
1013 find_unconfirmed_client(clientid_t
*clid
)
1015 struct nfs4_client
*clp
;
1016 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1018 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1019 if (same_clid(&clp
->cl_clientid
, clid
))
1026 match_clientid_establishment(struct nfs4_client
*clp
, bool use_exchange_id
)
1028 bool has_exchange_flags
= (clp
->cl_exchange_flags
!= 0);
1029 return use_exchange_id
== has_exchange_flags
;
1032 static struct nfs4_client
*
1033 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
,
1034 bool use_exchange_id
)
1036 struct nfs4_client
*clp
;
1038 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1039 if (same_name(clp
->cl_recdir
, dname
) &&
1040 match_clientid_establishment(clp
, use_exchange_id
))
1046 static struct nfs4_client
*
1047 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
,
1048 bool use_exchange_id
)
1050 struct nfs4_client
*clp
;
1052 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1053 if (same_name(clp
->cl_recdir
, dname
) &&
1054 match_clientid_establishment(clp
, use_exchange_id
))
1061 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, u32 scopeid
)
1063 struct nfs4_cb_conn
*cb
= &clp
->cl_cb_conn
;
1064 unsigned short expected_family
;
1066 /* Currently, we only support tcp and tcp6 for the callback channel */
1067 if (se
->se_callback_netid_len
== 3 &&
1068 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1069 expected_family
= AF_INET
;
1070 else if (se
->se_callback_netid_len
== 4 &&
1071 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1072 expected_family
= AF_INET6
;
1076 cb
->cb_addrlen
= rpc_uaddr2sockaddr(se
->se_callback_addr_val
,
1077 se
->se_callback_addr_len
,
1078 (struct sockaddr
*) &cb
->cb_addr
,
1079 sizeof(cb
->cb_addr
));
1081 if (!cb
->cb_addrlen
|| cb
->cb_addr
.ss_family
!= expected_family
)
1084 if (cb
->cb_addr
.ss_family
== AF_INET6
)
1085 ((struct sockaddr_in6
*) &cb
->cb_addr
)->sin6_scope_id
= scopeid
;
1087 cb
->cb_minorversion
= 0;
1088 cb
->cb_prog
= se
->se_callback_prog
;
1089 cb
->cb_ident
= se
->se_callback_ident
;
1092 cb
->cb_addr
.ss_family
= AF_UNSPEC
;
1094 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1095 "will not receive delegations\n",
1096 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1102 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1105 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1107 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1110 dprintk("--> %s slot %p\n", __func__
, slot
);
1112 slot
->sl_opcnt
= resp
->opcnt
;
1113 slot
->sl_status
= resp
->cstate
.status
;
1115 if (nfsd4_not_cached(resp
)) {
1116 slot
->sl_datalen
= 0;
1119 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1120 base
= (char *)resp
->cstate
.datap
-
1121 (char *)resp
->xbuf
->head
[0].iov_base
;
1122 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1124 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1129 * Encode the replay sequence operation from the slot values.
1130 * If cachethis is FALSE encode the uncached rep error on the next
1131 * operation which sets resp->p and increments resp->opcnt for
1132 * nfs4svc_encode_compoundres.
1136 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1137 struct nfsd4_compoundres
*resp
)
1139 struct nfsd4_op
*op
;
1140 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1142 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__
,
1143 resp
->opcnt
, resp
->cstate
.slot
->sl_cachethis
);
1145 /* Encode the replayed sequence operation */
1146 op
= &args
->ops
[resp
->opcnt
- 1];
1147 nfsd4_encode_operation(resp
, op
);
1149 /* Return nfserr_retry_uncached_rep in next operation. */
1150 if (args
->opcnt
> 1 && slot
->sl_cachethis
== 0) {
1151 op
= &args
->ops
[resp
->opcnt
++];
1152 op
->status
= nfserr_retry_uncached_rep
;
1153 nfsd4_encode_operation(resp
, op
);
1159 * The sequence operation is not cached because we can use the slot and
1163 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1164 struct nfsd4_sequence
*seq
)
1166 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1169 dprintk("--> %s slot %p\n", __func__
, slot
);
1171 /* Either returns 0 or nfserr_retry_uncached */
1172 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1173 if (status
== nfserr_retry_uncached_rep
)
1176 /* The sequence operation has been encoded, cstate->datap set. */
1177 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1179 resp
->opcnt
= slot
->sl_opcnt
;
1180 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1181 status
= slot
->sl_status
;
1187 * Set the exchange_id flags returned by the server.
1190 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1192 /* pNFS is not supported */
1193 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1195 /* Referrals are supported, Migration is not. */
1196 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1198 /* set the wire flags to return to client. */
1199 clid
->flags
= new->cl_exchange_flags
;
1203 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1204 struct nfsd4_compound_state
*cstate
,
1205 struct nfsd4_exchange_id
*exid
)
1207 struct nfs4_client
*unconf
, *conf
, *new;
1209 unsigned int strhashval
;
1210 char dname
[HEXDIR_LEN
];
1211 char addr_str
[INET6_ADDRSTRLEN
];
1212 nfs4_verifier verf
= exid
->verifier
;
1213 struct sockaddr
*sa
= svc_addr(rqstp
);
1215 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1216 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1217 "ip_addr=%s flags %x, spa_how %d\n",
1218 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1219 addr_str
, exid
->flags
, exid
->spa_how
);
1221 if (!check_name(exid
->clname
) || (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
))
1222 return nfserr_inval
;
1224 /* Currently only support SP4_NONE */
1225 switch (exid
->spa_how
) {
1229 return nfserr_encr_alg_unsupp
;
1231 BUG(); /* checked by xdr code */
1233 return nfserr_serverfault
; /* no excuse :-/ */
1236 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1241 strhashval
= clientstr_hashval(dname
);
1246 conf
= find_confirmed_client_by_str(dname
, strhashval
, true);
1248 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1249 /* 18.35.4 case 8 */
1250 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1251 status
= nfserr_not_same
;
1254 /* Client reboot: destroy old state */
1255 expire_client(conf
);
1258 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1259 /* 18.35.4 case 9 */
1260 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1261 status
= nfserr_perm
;
1264 expire_client(conf
);
1268 * Set bit when the owner id and verifier map to an already
1269 * confirmed client id (18.35.3).
1271 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1274 * Falling into 18.35.4 case 2, possible router replay.
1275 * Leave confirmed record intact and return same result.
1277 copy_verf(conf
, &verf
);
1282 /* 18.35.4 case 7 */
1283 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1284 status
= nfserr_noent
;
1288 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, true);
1291 * Possible retry or client restart. Per 18.35.4 case 4,
1292 * a new unconfirmed record should be generated regardless
1293 * of whether any properties have changed.
1295 expire_client(unconf
);
1300 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1302 status
= nfserr_jukebox
;
1307 add_to_unconfirmed(new, strhashval
);
1309 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1310 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1313 nfsd4_set_ex_flags(new, exid
);
1315 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1316 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1320 nfs4_unlock_state();
1322 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status
));
1327 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1329 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1332 /* The slot is in use, and no response has been sent. */
1334 if (seqid
== slot_seqid
)
1335 return nfserr_jukebox
;
1337 return nfserr_seq_misordered
;
1340 if (likely(seqid
== slot_seqid
+ 1))
1343 if (seqid
== slot_seqid
)
1344 return nfserr_replay_cache
;
1346 if (seqid
== 1 && (slot_seqid
+ 1) == 0)
1348 /* Misordered replay or misordered new request */
1349 return nfserr_seq_misordered
;
1353 * Cache the create session result into the create session single DRC
1354 * slot cache by saving the xdr structure. sl_seqid has been set.
1355 * Do this for solo or embedded create session operations.
1358 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1359 struct nfsd4_clid_slot
*slot
, int nfserr
)
1361 slot
->sl_status
= nfserr
;
1362 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1366 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1367 struct nfsd4_clid_slot
*slot
)
1369 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1370 return slot
->sl_status
;
1374 nfsd4_create_session(struct svc_rqst
*rqstp
,
1375 struct nfsd4_compound_state
*cstate
,
1376 struct nfsd4_create_session
*cr_ses
)
1378 struct sockaddr
*sa
= svc_addr(rqstp
);
1379 struct nfs4_client
*conf
, *unconf
;
1380 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1384 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1385 conf
= find_confirmed_client(&cr_ses
->clientid
);
1388 cs_slot
= &conf
->cl_cs_slot
;
1389 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1390 if (status
== nfserr_replay_cache
) {
1391 dprintk("Got a create_session replay! seqid= %d\n",
1393 /* Return the cached reply status */
1394 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1396 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1397 status
= nfserr_seq_misordered
;
1398 dprintk("Sequence misordered!\n");
1399 dprintk("Expected seqid= %d but got seqid= %d\n",
1400 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1403 cs_slot
->sl_seqid
++;
1404 } else if (unconf
) {
1405 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1406 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1407 status
= nfserr_clid_inuse
;
1411 cs_slot
= &unconf
->cl_cs_slot
;
1412 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1414 /* an unconfirmed replay returns misordered */
1415 status
= nfserr_seq_misordered
;
1419 cs_slot
->sl_seqid
++; /* from 0 to 1 */
1420 move_to_confirmed(unconf
);
1422 if (cr_ses
->flags
& SESSION4_BACK_CHAN
) {
1423 unconf
->cl_cb_conn
.cb_xprt
= rqstp
->rq_xprt
;
1424 svc_xprt_get(rqstp
->rq_xprt
);
1426 (struct sockaddr
*)&unconf
->cl_cb_conn
.cb_addr
,
1428 unconf
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
1429 unconf
->cl_cb_conn
.cb_minorversion
=
1430 cstate
->minorversion
;
1431 unconf
->cl_cb_conn
.cb_prog
= cr_ses
->callback_prog
;
1432 unconf
->cl_cb_seq_nr
= 1;
1433 nfsd4_probe_callback(unconf
, &unconf
->cl_cb_conn
);
1437 status
= nfserr_stale_clientid
;
1442 * We do not support RDMA or persistent sessions
1444 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1445 cr_ses
->flags
&= ~SESSION4_RDMA
;
1447 status
= alloc_init_session(rqstp
, conf
, cr_ses
);
1451 memcpy(cr_ses
->sessionid
.data
, conf
->cl_sessionid
.data
,
1452 NFS4_MAX_SESSIONID_LEN
);
1453 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1456 /* cache solo and embedded create sessions under the state lock */
1457 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1459 nfs4_unlock_state();
1460 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1464 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1466 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1467 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1469 return argp
->opcnt
== resp
->opcnt
;
1472 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1476 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1480 nfsd4_destroy_session(struct svc_rqst
*r
,
1481 struct nfsd4_compound_state
*cstate
,
1482 struct nfsd4_destroy_session
*sessionid
)
1484 struct nfsd4_session
*ses
;
1485 u32 status
= nfserr_badsession
;
1488 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1489 * - Should we return nfserr_back_chan_busy if waiting for
1490 * callbacks on to-be-destroyed session?
1491 * - Do we need to clear any callback info from previous session?
1494 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1495 if (!nfsd4_last_compound_op(r
))
1496 return nfserr_not_only_op
;
1498 dump_sessionid(__func__
, &sessionid
->sessionid
);
1499 spin_lock(&client_lock
);
1500 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1502 spin_unlock(&client_lock
);
1506 unhash_session(ses
);
1507 spin_unlock(&client_lock
);
1510 /* wait for callbacks */
1511 nfsd4_set_callback_client(ses
->se_client
, NULL
);
1512 nfs4_unlock_state();
1513 nfsd4_put_session(ses
);
1516 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1521 nfsd4_sequence(struct svc_rqst
*rqstp
,
1522 struct nfsd4_compound_state
*cstate
,
1523 struct nfsd4_sequence
*seq
)
1525 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1526 struct nfsd4_session
*session
;
1527 struct nfsd4_slot
*slot
;
1530 if (resp
->opcnt
!= 1)
1531 return nfserr_sequence_pos
;
1533 spin_lock(&client_lock
);
1534 status
= nfserr_badsession
;
1535 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1539 status
= nfserr_badslot
;
1540 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1543 slot
= session
->se_slots
[seq
->slotid
];
1544 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1546 /* We do not negotiate the number of slots yet, so set the
1547 * maxslots to the session maxreqs which is used to encode
1548 * sr_highest_slotid and the sr_target_slot id to maxslots */
1549 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1551 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
, slot
->sl_inuse
);
1552 if (status
== nfserr_replay_cache
) {
1553 cstate
->slot
= slot
;
1554 cstate
->session
= session
;
1555 /* Return the cached reply status and set cstate->status
1556 * for nfsd4_proc_compound processing */
1557 status
= nfsd4_replay_cache_entry(resp
, seq
);
1558 cstate
->status
= nfserr_replay_cache
;
1564 /* Success! bump slot seqid */
1565 slot
->sl_inuse
= true;
1566 slot
->sl_seqid
= seq
->seqid
;
1567 slot
->sl_cachethis
= seq
->cachethis
;
1569 cstate
->slot
= slot
;
1570 cstate
->session
= session
;
1573 /* Hold a session reference until done processing the compound. */
1574 if (cstate
->session
) {
1575 nfsd4_get_session(cstate
->session
);
1576 atomic_inc(&session
->se_client
->cl_refcount
);
1578 spin_unlock(&client_lock
);
1579 dprintk("%s: return %d\n", __func__
, ntohl(status
));
1584 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
1586 if (rc
->rca_one_fs
) {
1587 if (!cstate
->current_fh
.fh_dentry
)
1588 return nfserr_nofilehandle
;
1590 * We don't take advantage of the rca_one_fs case.
1591 * That's OK, it's optional, we can safely ignore it.
1596 if (is_client_expired(cstate
->session
->se_client
)) {
1597 nfs4_unlock_state();
1599 * The following error isn't really legal.
1600 * But we only get here if the client just explicitly
1601 * destroyed the client. Surely it no longer cares what
1602 * error it gets back on an operation for the dead
1605 return nfserr_stale_clientid
;
1607 nfsd4_create_clid_dir(cstate
->session
->se_client
);
1608 nfs4_unlock_state();
1613 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1614 struct nfsd4_setclientid
*setclid
)
1616 struct sockaddr
*sa
= svc_addr(rqstp
);
1617 struct xdr_netobj clname
= {
1618 .len
= setclid
->se_namelen
,
1619 .data
= setclid
->se_name
,
1621 nfs4_verifier clverifier
= setclid
->se_verf
;
1622 unsigned int strhashval
;
1623 struct nfs4_client
*conf
, *unconf
, *new;
1625 char dname
[HEXDIR_LEN
];
1627 if (!check_name(clname
))
1628 return nfserr_inval
;
1630 status
= nfs4_make_rec_clidname(dname
, &clname
);
1635 strhashval
= clientstr_hashval(dname
);
1638 conf
= find_confirmed_client_by_str(dname
, strhashval
, false);
1640 /* RFC 3530 14.2.33 CASE 0: */
1641 status
= nfserr_clid_inuse
;
1642 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1643 char addr_str
[INET6_ADDRSTRLEN
];
1644 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
1646 dprintk("NFSD: setclientid: string in use by client "
1647 "at %s\n", addr_str
);
1652 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1653 * has a description of SETCLIENTID request processing consisting
1654 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1656 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
, false);
1657 status
= nfserr_resource
;
1660 * RFC 3530 14.2.33 CASE 4:
1661 * placed first, because it is the normal case
1664 expire_client(unconf
);
1665 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1669 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
1671 * RFC 3530 14.2.33 CASE 1:
1672 * probable callback update
1675 /* Note this is removing unconfirmed {*x***},
1676 * which is stronger than RFC recommended {vxc**}.
1677 * This has the advantage that there is at most
1678 * one {*x***} in either list at any time.
1680 expire_client(unconf
);
1682 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1685 copy_clid(new, conf
);
1686 } else if (!unconf
) {
1688 * RFC 3530 14.2.33 CASE 2:
1689 * probable client reboot; state will be removed if
1692 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1698 * RFC 3530 14.2.33 CASE 3:
1699 * probable client reboot; state will be removed if
1702 expire_client(unconf
);
1703 new = create_client(clname
, dname
, rqstp
, &clverifier
);
1708 gen_callback(new, setclid
, rpc_get_scope_id(sa
));
1709 add_to_unconfirmed(new, strhashval
);
1710 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1711 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
1712 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
1715 nfs4_unlock_state();
1721 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1722 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1723 * bullets, labeled as CASE1 - CASE4 below.
1726 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
1727 struct nfsd4_compound_state
*cstate
,
1728 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
1730 struct sockaddr
*sa
= svc_addr(rqstp
);
1731 struct nfs4_client
*conf
, *unconf
;
1732 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
1733 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
1736 if (STALE_CLIENTID(clid
))
1737 return nfserr_stale_clientid
;
1741 conf
= find_confirmed_client(clid
);
1742 unconf
= find_unconfirmed_client(clid
);
1744 status
= nfserr_clid_inuse
;
1745 if (conf
&& !rpc_cmp_addr((struct sockaddr
*) &conf
->cl_addr
, sa
))
1747 if (unconf
&& !rpc_cmp_addr((struct sockaddr
*) &unconf
->cl_addr
, sa
))
1751 * section 14.2.34 of RFC 3530 has a description of
1752 * SETCLIENTID_CONFIRM request processing consisting
1753 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1755 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
1757 * RFC 3530 14.2.34 CASE 1:
1760 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
1761 status
= nfserr_clid_inuse
;
1763 atomic_set(&conf
->cl_cb_set
, 0);
1764 nfsd4_probe_callback(conf
, &unconf
->cl_cb_conn
);
1765 expire_client(unconf
);
1769 } else if (conf
&& !unconf
) {
1771 * RFC 3530 14.2.34 CASE 2:
1772 * probable retransmitted request; play it safe and
1775 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
1776 status
= nfserr_clid_inuse
;
1779 } else if (!conf
&& unconf
1780 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
1782 * RFC 3530 14.2.34 CASE 3:
1783 * Normal case; new or rebooted client:
1785 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
1786 status
= nfserr_clid_inuse
;
1789 clientstr_hashval(unconf
->cl_recdir
);
1790 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
1793 nfsd4_remove_clid_dir(conf
);
1794 expire_client(conf
);
1796 move_to_confirmed(unconf
);
1798 nfsd4_probe_callback(conf
, &conf
->cl_cb_conn
);
1801 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
1802 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
1805 * RFC 3530 14.2.34 CASE 4:
1806 * Client probably hasn't noticed that we rebooted yet.
1808 status
= nfserr_stale_clientid
;
1810 /* check that we have hit one of the cases...*/
1811 status
= nfserr_clid_inuse
;
1814 nfs4_unlock_state();
1818 /* OPEN Share state helper functions */
1819 static inline struct nfs4_file
*
1820 alloc_init_file(struct inode
*ino
)
1822 struct nfs4_file
*fp
;
1823 unsigned int hashval
= file_hashval(ino
);
1825 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1827 atomic_set(&fp
->fi_ref
, 1);
1828 INIT_LIST_HEAD(&fp
->fi_hash
);
1829 INIT_LIST_HEAD(&fp
->fi_stateids
);
1830 INIT_LIST_HEAD(&fp
->fi_delegations
);
1831 fp
->fi_inode
= igrab(ino
);
1832 fp
->fi_id
= current_fileid
++;
1833 fp
->fi_had_conflict
= false;
1834 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
1835 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
1836 spin_lock(&recall_lock
);
1837 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1838 spin_unlock(&recall_lock
);
1845 nfsd4_free_slab(struct kmem_cache
**slab
)
1849 kmem_cache_destroy(*slab
);
1854 nfsd4_free_slabs(void)
1856 nfsd4_free_slab(&stateowner_slab
);
1857 nfsd4_free_slab(&file_slab
);
1858 nfsd4_free_slab(&stateid_slab
);
1859 nfsd4_free_slab(&deleg_slab
);
1863 nfsd4_init_slabs(void)
1865 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1866 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1867 if (stateowner_slab
== NULL
)
1869 file_slab
= kmem_cache_create("nfsd4_files",
1870 sizeof(struct nfs4_file
), 0, 0, NULL
);
1871 if (file_slab
== NULL
)
1873 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1874 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1875 if (stateid_slab
== NULL
)
1877 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1878 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1879 if (deleg_slab
== NULL
)
1884 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1889 nfs4_free_stateowner(struct kref
*kref
)
1891 struct nfs4_stateowner
*sop
=
1892 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1893 kfree(sop
->so_owner
.data
);
1894 kmem_cache_free(stateowner_slab
, sop
);
1897 static inline struct nfs4_stateowner
*
1898 alloc_stateowner(struct xdr_netobj
*owner
)
1900 struct nfs4_stateowner
*sop
;
1902 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1903 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1904 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1905 sop
->so_owner
.len
= owner
->len
;
1906 kref_init(&sop
->so_ref
);
1909 kmem_cache_free(stateowner_slab
, sop
);
1914 static struct nfs4_stateowner
*
1915 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1916 struct nfs4_stateowner
*sop
;
1917 struct nfs4_replay
*rp
;
1918 unsigned int idhashval
;
1920 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1922 idhashval
= ownerid_hashval(current_ownerid
);
1923 INIT_LIST_HEAD(&sop
->so_idhash
);
1924 INIT_LIST_HEAD(&sop
->so_strhash
);
1925 INIT_LIST_HEAD(&sop
->so_perclient
);
1926 INIT_LIST_HEAD(&sop
->so_stateids
);
1927 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1928 INIT_LIST_HEAD(&sop
->so_close_lru
);
1930 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1931 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1932 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1933 sop
->so_is_open_owner
= 1;
1934 sop
->so_id
= current_ownerid
++;
1935 sop
->so_client
= clp
;
1936 sop
->so_seqid
= open
->op_seqid
;
1937 sop
->so_confirmed
= 0;
1938 rp
= &sop
->so_replay
;
1939 rp
->rp_status
= nfserr_serverfault
;
1941 rp
->rp_buf
= rp
->rp_ibuf
;
1946 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1947 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1948 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1950 INIT_LIST_HEAD(&stp
->st_hash
);
1951 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1952 INIT_LIST_HEAD(&stp
->st_lockowners
);
1953 INIT_LIST_HEAD(&stp
->st_perfile
);
1954 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1955 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1956 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1957 stp
->st_stateowner
= sop
;
1960 stp
->st_stateid
.si_boot
= boot_time
;
1961 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1962 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1963 stp
->st_stateid
.si_generation
= 0;
1964 stp
->st_access_bmap
= 0;
1965 stp
->st_deny_bmap
= 0;
1966 __set_bit(open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
,
1967 &stp
->st_access_bmap
);
1968 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1969 stp
->st_openstp
= NULL
;
1973 move_to_close_lru(struct nfs4_stateowner
*sop
)
1975 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1977 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1978 sop
->so_time
= get_seconds();
1982 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1985 return (sop
->so_owner
.len
== owner
->len
) &&
1986 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1987 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1990 static struct nfs4_stateowner
*
1991 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1993 struct nfs4_stateowner
*so
= NULL
;
1995 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1996 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
2002 /* search file_hashtbl[] for file */
2003 static struct nfs4_file
*
2004 find_file(struct inode
*ino
)
2006 unsigned int hashval
= file_hashval(ino
);
2007 struct nfs4_file
*fp
;
2009 spin_lock(&recall_lock
);
2010 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2011 if (fp
->fi_inode
== ino
) {
2013 spin_unlock(&recall_lock
);
2017 spin_unlock(&recall_lock
);
2021 static inline int access_valid(u32 x
, u32 minorversion
)
2023 if ((x
& NFS4_SHARE_ACCESS_MASK
) < NFS4_SHARE_ACCESS_READ
)
2025 if ((x
& NFS4_SHARE_ACCESS_MASK
) > NFS4_SHARE_ACCESS_BOTH
)
2027 x
&= ~NFS4_SHARE_ACCESS_MASK
;
2028 if (minorversion
&& x
) {
2029 if ((x
& NFS4_SHARE_WANT_MASK
) > NFS4_SHARE_WANT_CANCEL
)
2031 if ((x
& NFS4_SHARE_WHEN_MASK
) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED
)
2033 x
&= ~(NFS4_SHARE_WANT_MASK
| NFS4_SHARE_WHEN_MASK
);
2040 static inline int deny_valid(u32 x
)
2042 /* Note: unlike access bits, deny bits may be zero. */
2043 return x
<= NFS4_SHARE_DENY_BOTH
;
2047 * Called to check deny when READ with all zero stateid or
2048 * WRITE with all zero or all one stateid
2051 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2053 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2054 struct nfs4_file
*fp
;
2055 struct nfs4_stateid
*stp
;
2058 dprintk("NFSD: nfs4_share_conflict\n");
2060 fp
= find_file(ino
);
2063 ret
= nfserr_locked
;
2064 /* Search for conflicting share reservations */
2065 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2066 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
2067 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
2077 nfs4_file_downgrade(struct nfs4_file
*fp
, unsigned int share_access
)
2079 if (share_access
& NFS4_SHARE_ACCESS_WRITE
)
2080 nfs4_file_put_access(fp
, O_WRONLY
);
2081 if (share_access
& NFS4_SHARE_ACCESS_READ
)
2082 nfs4_file_put_access(fp
, O_RDONLY
);
2086 * Spawn a thread to perform a recall on the delegation represented
2087 * by the lease (file_lock)
2089 * Called from break_lease() with lock_kernel() held.
2090 * Note: we assume break_lease will only call this *once* for any given
2094 void nfsd_break_deleg_cb(struct file_lock
*fl
)
2096 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2098 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
2102 /* We're assuming the state code never drops its reference
2103 * without first removing the lease. Since we're in this lease
2104 * callback (and since the lease code is serialized by the kernel
2105 * lock) we know the server hasn't removed the lease yet, we know
2106 * it's safe to take a reference: */
2107 atomic_inc(&dp
->dl_count
);
2109 spin_lock(&recall_lock
);
2110 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2111 spin_unlock(&recall_lock
);
2113 /* only place dl_time is set. protected by lock_kernel*/
2114 dp
->dl_time
= get_seconds();
2117 * We don't want the locks code to timeout the lease for us;
2118 * we'll remove it ourself if the delegation isn't returned
2121 fl
->fl_break_time
= 0;
2123 dp
->dl_file
->fi_had_conflict
= true;
2124 nfsd4_cb_recall(dp
);
2128 * The file_lock is being reapd.
2130 * Called by locks_free_lock() with lock_kernel() held.
2133 void nfsd_release_deleg_cb(struct file_lock
*fl
)
2135 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
2137 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
2139 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
2141 dp
->dl_flock
= NULL
;
2145 * Set the delegation file_lock back pointer.
2147 * Called from setlease() with lock_kernel() held.
2150 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
2152 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
2154 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
2161 * Called from setlease() with lock_kernel() held
2164 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
2166 struct nfs4_delegation
*onlistd
=
2167 (struct nfs4_delegation
*)onlist
->fl_owner
;
2168 struct nfs4_delegation
*tryd
=
2169 (struct nfs4_delegation
*)try->fl_owner
;
2171 if (onlist
->fl_lmops
!= try->fl_lmops
)
2174 return onlistd
->dl_client
== tryd
->dl_client
;
2179 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2182 return lease_modify(onlist
, arg
);
2187 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2188 .fl_break
= nfsd_break_deleg_cb
,
2189 .fl_release_private
= nfsd_release_deleg_cb
,
2190 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
2191 .fl_mylease
= nfsd_same_client_deleg_cb
,
2192 .fl_change
= nfsd_change_deleg_cb
,
2197 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2198 struct nfsd4_open
*open
)
2200 clientid_t
*clientid
= &open
->op_clientid
;
2201 struct nfs4_client
*clp
= NULL
;
2202 unsigned int strhashval
;
2203 struct nfs4_stateowner
*sop
= NULL
;
2205 if (!check_name(open
->op_owner
))
2206 return nfserr_inval
;
2208 if (STALE_CLIENTID(&open
->op_clientid
))
2209 return nfserr_stale_clientid
;
2211 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
2212 sop
= find_openstateowner_str(strhashval
, open
);
2213 open
->op_stateowner
= sop
;
2215 /* Make sure the client's lease hasn't expired. */
2216 clp
= find_confirmed_client(clientid
);
2218 return nfserr_expired
;
2221 /* When sessions are used, skip open sequenceid processing */
2222 if (nfsd4_has_session(cstate
))
2224 if (!sop
->so_confirmed
) {
2225 /* Replace unconfirmed owners without checking for replay. */
2226 clp
= sop
->so_client
;
2227 release_openowner(sop
);
2228 open
->op_stateowner
= NULL
;
2231 if (open
->op_seqid
== sop
->so_seqid
- 1) {
2232 if (sop
->so_replay
.rp_buflen
)
2233 return nfserr_replay_me
;
2234 /* The original OPEN failed so spectacularly
2235 * that we don't even have replay data saved!
2236 * Therefore, we have no choice but to continue
2237 * processing this OPEN; presumably, we'll
2238 * fail again for the same reason.
2240 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2243 if (open
->op_seqid
!= sop
->so_seqid
)
2244 return nfserr_bad_seqid
;
2246 if (open
->op_stateowner
== NULL
) {
2247 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2249 return nfserr_resource
;
2250 open
->op_stateowner
= sop
;
2252 list_del_init(&sop
->so_close_lru
);
2253 renew_client(sop
->so_client
);
2257 static inline __be32
2258 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2260 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2261 return nfserr_openmode
;
2266 static struct nfs4_delegation
*
2267 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
2269 struct nfs4_delegation
*dp
;
2271 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
2272 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
2278 int share_access_to_flags(u32 share_access
)
2280 share_access
&= ~NFS4_SHARE_WANT_MASK
;
2282 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2286 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2287 struct nfs4_delegation
**dp
)
2290 __be32 status
= nfserr_bad_stateid
;
2292 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
2295 flags
= share_access_to_flags(open
->op_share_access
);
2296 status
= nfs4_check_delegmode(*dp
, flags
);
2300 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2304 open
->op_stateowner
->so_confirmed
= 1;
2309 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
2311 struct nfs4_stateid
*local
;
2312 __be32 status
= nfserr_share_denied
;
2313 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
2315 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2316 /* ignore lock owners */
2317 if (local
->st_stateowner
->so_is_open_owner
== 0)
2319 /* remember if we have seen this open owner */
2320 if (local
->st_stateowner
== sop
)
2322 /* check for conflicting share reservations */
2323 if (!test_share(local
, open
))
2331 static inline struct nfs4_stateid
*
2332 nfs4_alloc_stateid(void)
2334 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
2337 static inline int nfs4_access_to_access(u32 nfs4_access
)
2341 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2342 flags
|= NFSD_MAY_READ
;
2343 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2344 flags
|= NFSD_MAY_WRITE
;
2348 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
2349 *fp
, struct svc_fh
*cur_fh
, u32 nfs4_access
)
2352 int oflag
= nfs4_access_to_omode(nfs4_access
);
2353 int access
= nfs4_access_to_access(nfs4_access
);
2355 if (!fp
->fi_fds
[oflag
]) {
2356 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2357 &fp
->fi_fds
[oflag
]);
2358 if (status
== nfserr_dropit
)
2359 status
= nfserr_jukebox
;
2363 nfs4_file_get_access(fp
, oflag
);
2369 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
2370 struct nfs4_file
*fp
, struct svc_fh
*cur_fh
,
2371 struct nfsd4_open
*open
)
2373 struct nfs4_stateid
*stp
;
2376 stp
= nfs4_alloc_stateid();
2378 return nfserr_resource
;
2380 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
->op_share_access
);
2382 kmem_cache_free(stateid_slab
, stp
);
2389 static inline __be32
2390 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2391 struct nfsd4_open
*open
)
2393 struct iattr iattr
= {
2394 .ia_valid
= ATTR_SIZE
,
2397 if (!open
->op_truncate
)
2399 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2400 return nfserr_inval
;
2401 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2405 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
2407 u32 op_share_access
= open
->op_share_access
& ~NFS4_SHARE_WANT_MASK
;
2411 new_access
= !test_bit(op_share_access
, &stp
->st_access_bmap
);
2413 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, op_share_access
);
2417 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2420 int oflag
= nfs4_access_to_omode(new_access
);
2421 nfs4_file_put_access(fp
, oflag
);
2425 /* remember the open */
2426 __set_bit(op_share_access
, &stp
->st_access_bmap
);
2427 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
2434 nfs4_set_claim_prev(struct nfsd4_open
*open
)
2436 open
->op_stateowner
->so_confirmed
= 1;
2437 open
->op_stateowner
->so_client
->cl_firststate
= 1;
2441 * Attempt to hand out a delegation.
2444 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
2446 struct nfs4_delegation
*dp
;
2447 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
2448 int cb_up
= atomic_read(&sop
->so_client
->cl_cb_set
);
2449 struct file_lock fl
, *flp
= &fl
;
2450 int status
, flag
= 0;
2452 flag
= NFS4_OPEN_DELEGATE_NONE
;
2453 open
->op_recall
= 0;
2454 switch (open
->op_claim_type
) {
2455 case NFS4_OPEN_CLAIM_PREVIOUS
:
2457 open
->op_recall
= 1;
2458 flag
= open
->op_delegate_type
;
2459 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2462 case NFS4_OPEN_CLAIM_NULL
:
2463 /* Let's not give out any delegations till everyone's
2464 * had the chance to reclaim theirs.... */
2465 if (locks_in_grace())
2467 if (!cb_up
|| !sop
->so_confirmed
)
2469 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2470 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2472 flag
= NFS4_OPEN_DELEGATE_READ
;
2478 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
2480 flag
= NFS4_OPEN_DELEGATE_NONE
;
2483 locks_init_lock(&fl
);
2484 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
2485 fl
.fl_flags
= FL_LEASE
;
2486 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2487 fl
.fl_end
= OFFSET_MAX
;
2488 fl
.fl_owner
= (fl_owner_t
)dp
;
2489 fl
.fl_file
= find_readable_file(stp
->st_file
);
2490 BUG_ON(!fl
.fl_file
);
2491 fl
.fl_pid
= current
->tgid
;
2493 /* vfs_setlease checks to see if delegation should be handed out.
2494 * the lock_manager callbacks fl_mylease and fl_change are used
2496 if ((status
= vfs_setlease(fl
.fl_file
, fl
.fl_type
, &flp
))) {
2497 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
2498 unhash_delegation(dp
);
2499 flag
= NFS4_OPEN_DELEGATE_NONE
;
2503 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
2505 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2506 STATEID_VAL(&dp
->dl_stateid
));
2508 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
2509 && flag
== NFS4_OPEN_DELEGATE_NONE
2510 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2511 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2512 open
->op_delegate_type
= flag
;
2516 * called with nfs4_lock_state() held.
2519 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2521 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2522 struct nfs4_file
*fp
= NULL
;
2523 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2524 struct nfs4_stateid
*stp
= NULL
;
2525 struct nfs4_delegation
*dp
= NULL
;
2528 status
= nfserr_inval
;
2529 if (!access_valid(open
->op_share_access
, resp
->cstate
.minorversion
)
2530 || !deny_valid(open
->op_share_deny
))
2533 * Lookup file; if found, lookup stateid and check open request,
2534 * and check for delegations in the process of being recalled.
2535 * If not found, create the nfs4_file struct
2537 fp
= find_file(ino
);
2539 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2541 status
= nfs4_check_deleg(fp
, open
, &dp
);
2545 status
= nfserr_bad_stateid
;
2546 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
2548 status
= nfserr_resource
;
2549 fp
= alloc_init_file(ino
);
2555 * OPEN the file, or upgrade an existing OPEN.
2556 * If truncate fails, the OPEN fails.
2559 /* Stateid was found, this is an OPEN upgrade */
2560 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
2563 update_stateid(&stp
->st_stateid
);
2565 status
= nfs4_new_open(rqstp
, &stp
, fp
, current_fh
, open
);
2568 init_stateid(stp
, fp
, open
);
2569 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
2571 release_open_stateid(stp
);
2574 if (nfsd4_has_session(&resp
->cstate
))
2575 update_stateid(&stp
->st_stateid
);
2577 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2579 if (nfsd4_has_session(&resp
->cstate
))
2580 open
->op_stateowner
->so_confirmed
= 1;
2583 * Attempt to hand out a delegation. No error return, because the
2584 * OPEN succeeds even if we fail.
2586 nfs4_open_delegation(current_fh
, open
, stp
);
2590 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
2591 STATEID_VAL(&stp
->st_stateid
));
2595 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
2596 nfs4_set_claim_prev(open
);
2598 * To finish the open response, we just need to set the rflags.
2600 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
2601 if (!open
->op_stateowner
->so_confirmed
&&
2602 !nfsd4_has_session(&resp
->cstate
))
2603 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
2609 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2612 struct nfs4_client
*clp
;
2616 dprintk("process_renew(%08x/%08x): starting\n",
2617 clid
->cl_boot
, clid
->cl_id
);
2618 status
= nfserr_stale_clientid
;
2619 if (STALE_CLIENTID(clid
))
2621 clp
= find_confirmed_client(clid
);
2622 status
= nfserr_expired
;
2624 /* We assume the client took too long to RENEW. */
2625 dprintk("nfsd4_renew: clientid not found!\n");
2629 status
= nfserr_cb_path_down
;
2630 if (!list_empty(&clp
->cl_delegations
)
2631 && !atomic_read(&clp
->cl_cb_set
))
2635 nfs4_unlock_state();
2639 struct lock_manager nfsd4_manager
= {
2643 nfsd4_end_grace(void)
2645 dprintk("NFSD: end of grace period\n");
2646 nfsd4_recdir_purge_old();
2647 locks_end_grace(&nfsd4_manager
);
2649 * Now that every NFSv4 client has had the chance to recover and
2650 * to see the (possibly new, possibly shorter) lease time, we
2651 * can safely set the next grace time to the current lease time:
2653 nfsd4_grace
= nfsd4_lease
;
2657 nfs4_laundromat(void)
2659 struct nfs4_client
*clp
;
2660 struct nfs4_stateowner
*sop
;
2661 struct nfs4_delegation
*dp
;
2662 struct list_head
*pos
, *next
, reaplist
;
2663 time_t cutoff
= get_seconds() - nfsd4_lease
;
2664 time_t t
, clientid_val
= nfsd4_lease
;
2665 time_t u
, test_val
= nfsd4_lease
;
2669 dprintk("NFSD: laundromat service - starting\n");
2670 if (locks_in_grace())
2672 INIT_LIST_HEAD(&reaplist
);
2673 spin_lock(&client_lock
);
2674 list_for_each_safe(pos
, next
, &client_lru
) {
2675 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2676 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
2677 t
= clp
->cl_time
- cutoff
;
2678 if (clientid_val
> t
)
2682 if (atomic_read(&clp
->cl_refcount
)) {
2683 dprintk("NFSD: client in use (clientid %08x)\n",
2684 clp
->cl_clientid
.cl_id
);
2687 unhash_client_locked(clp
);
2688 list_add(&clp
->cl_lru
, &reaplist
);
2690 spin_unlock(&client_lock
);
2691 list_for_each_safe(pos
, next
, &reaplist
) {
2692 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
2693 dprintk("NFSD: purging unused client (clientid %08x)\n",
2694 clp
->cl_clientid
.cl_id
);
2695 nfsd4_remove_clid_dir(clp
);
2698 spin_lock(&recall_lock
);
2699 list_for_each_safe(pos
, next
, &del_recall_lru
) {
2700 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2701 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
2702 u
= dp
->dl_time
- cutoff
;
2707 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2709 list_move(&dp
->dl_recall_lru
, &reaplist
);
2711 spin_unlock(&recall_lock
);
2712 list_for_each_safe(pos
, next
, &reaplist
) {
2713 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
2714 list_del_init(&dp
->dl_recall_lru
);
2715 unhash_delegation(dp
);
2717 test_val
= nfsd4_lease
;
2718 list_for_each_safe(pos
, next
, &close_lru
) {
2719 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
2720 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
2721 u
= sop
->so_time
- cutoff
;
2726 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2728 release_openowner(sop
);
2730 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
2731 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
2732 nfs4_unlock_state();
2733 return clientid_val
;
2736 static struct workqueue_struct
*laundry_wq
;
2737 static void laundromat_main(struct work_struct
*);
2738 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
2741 laundromat_main(struct work_struct
*not_used
)
2745 t
= nfs4_laundromat();
2746 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
2747 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
2750 static struct nfs4_stateowner
*
2751 search_close_lru(u32 st_id
, int flags
)
2753 struct nfs4_stateowner
*local
= NULL
;
2755 if (flags
& CLOSE_STATE
) {
2756 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
2757 if (local
->so_id
== st_id
)
2765 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
2767 return fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
;
2771 STALE_STATEID(stateid_t
*stateid
)
2773 if (stateid
->si_boot
== boot_time
)
2775 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
2776 STATEID_VAL(stateid
));
2781 access_permit_read(unsigned long access_bmap
)
2783 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
2784 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
2785 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
2789 access_permit_write(unsigned long access_bmap
)
2791 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
2792 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
2796 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
2798 __be32 status
= nfserr_openmode
;
2800 /* For lock stateid's, we test the parent open, not the lock: */
2801 if (stp
->st_openstp
)
2802 stp
= stp
->st_openstp
;
2803 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
2805 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2812 static inline __be32
2813 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2815 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2817 else if (locks_in_grace()) {
2818 /* Answer in remaining cases depends on existance of
2819 * conflicting state; so we must wait out the grace period. */
2820 return nfserr_grace
;
2821 } else if (flags
& WR_STATE
)
2822 return nfs4_share_conflict(current_fh
,
2823 NFS4_SHARE_DENY_WRITE
);
2824 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2825 return nfs4_share_conflict(current_fh
,
2826 NFS4_SHARE_DENY_READ
);
2830 * Allow READ/WRITE during grace period on recovered state only for files
2831 * that are not able to provide mandatory locking.
2834 grace_disallows_io(struct inode
*inode
)
2836 return locks_in_grace() && mandatory_lock(inode
);
2839 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, int flags
)
2842 * When sessions are used the stateid generation number is ignored
2845 if ((flags
& HAS_SESSION
) && in
->si_generation
== 0)
2848 /* If the client sends us a stateid from the future, it's buggy: */
2849 if (in
->si_generation
> ref
->si_generation
)
2850 return nfserr_bad_stateid
;
2852 * The following, however, can happen. For example, if the
2853 * client sends an open and some IO at the same time, the open
2854 * may bump si_generation while the IO is still in flight.
2855 * Thanks to hard links and renames, the client never knows what
2856 * file an open will affect. So it could avoid that situation
2857 * only by serializing all opens and IO from the same open
2858 * owner. To recover from the old_stateid error, the client
2859 * will just have to retry the IO:
2861 if (in
->si_generation
< ref
->si_generation
)
2862 return nfserr_old_stateid
;
2867 static int is_delegation_stateid(stateid_t
*stateid
)
2869 return stateid
->si_fileid
== 0;
2873 * Checks for stateid operations
2876 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
2877 stateid_t
*stateid
, int flags
, struct file
**filpp
)
2879 struct nfs4_stateid
*stp
= NULL
;
2880 struct nfs4_delegation
*dp
= NULL
;
2881 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2882 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2888 if (grace_disallows_io(ino
))
2889 return nfserr_grace
;
2891 if (nfsd4_has_session(cstate
))
2892 flags
|= HAS_SESSION
;
2894 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2895 return check_special_stateids(current_fh
, stateid
, flags
);
2897 status
= nfserr_stale_stateid
;
2898 if (STALE_STATEID(stateid
))
2901 status
= nfserr_bad_stateid
;
2902 if (is_delegation_stateid(stateid
)) {
2903 dp
= find_delegation_stateid(ino
, stateid
);
2906 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
,
2910 status
= nfs4_check_delegmode(dp
, flags
);
2913 renew_client(dp
->dl_client
);
2915 *filpp
= find_readable_file(dp
->dl_file
);
2917 } else { /* open or lock stateid */
2918 stp
= find_stateid(stateid
, flags
);
2921 if (nfs4_check_fh(current_fh
, stp
))
2923 if (!stp
->st_stateowner
->so_confirmed
)
2925 status
= check_stateid_generation(stateid
, &stp
->st_stateid
,
2929 status
= nfs4_check_openmode(stp
, flags
);
2932 renew_client(stp
->st_stateowner
->so_client
);
2934 if (flags
& RD_STATE
)
2935 *filpp
= find_readable_file(stp
->st_file
);
2937 *filpp
= find_writeable_file(stp
->st_file
);
2948 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2949 RD_STATE
: WR_STATE
;
2953 * Checks for sequence id mutating operations.
2956 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
2957 stateid_t
*stateid
, int flags
,
2958 struct nfs4_stateowner
**sopp
,
2959 struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2961 struct nfs4_stateid
*stp
;
2962 struct nfs4_stateowner
*sop
;
2963 struct svc_fh
*current_fh
= &cstate
->current_fh
;
2966 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
2967 seqid
, STATEID_VAL(stateid
));
2972 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2973 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2974 return nfserr_bad_stateid
;
2977 if (STALE_STATEID(stateid
))
2978 return nfserr_stale_stateid
;
2980 if (nfsd4_has_session(cstate
))
2981 flags
|= HAS_SESSION
;
2984 * We return BAD_STATEID if filehandle doesn't match stateid,
2985 * the confirmed flag is incorrecly set, or the generation
2986 * number is incorrect.
2988 stp
= find_stateid(stateid
, flags
);
2991 * Also, we should make sure this isn't just the result of
2994 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2996 return nfserr_bad_stateid
;
3002 *sopp
= sop
= stp
->st_stateowner
;
3005 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
3006 struct nfs4_client
*clp
= sop
->so_client
;
3010 lkflg
= setlkflg(lock
->lk_type
);
3012 if (lock
->lk_is_new
) {
3013 if (!sop
->so_is_open_owner
)
3014 return nfserr_bad_stateid
;
3015 if (!(flags
& HAS_SESSION
) &&
3016 !same_clid(&clp
->cl_clientid
, lockclid
))
3017 return nfserr_bad_stateid
;
3018 /* stp is the open stateid */
3019 status
= nfs4_check_openmode(stp
, lkflg
);
3023 /* stp is the lock stateid */
3024 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
3030 if (nfs4_check_fh(current_fh
, stp
)) {
3031 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3032 return nfserr_bad_stateid
;
3036 * We now validate the seqid and stateid generation numbers.
3037 * For the moment, we ignore the possibility of
3038 * generation number wraparound.
3040 if (!(flags
& HAS_SESSION
) && seqid
!= sop
->so_seqid
)
3043 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
3044 dprintk("NFSD: preprocess_seqid_op: expected"
3045 " unconfirmed stateowner!\n");
3046 return nfserr_bad_stateid
;
3048 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
3049 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3050 " confirmed yet!\n");
3051 return nfserr_bad_stateid
;
3053 status
= check_stateid_generation(stateid
, &stp
->st_stateid
, flags
);
3056 renew_client(sop
->so_client
);
3060 if (seqid
== sop
->so_seqid
- 1) {
3061 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3062 /* indicate replay to calling function */
3063 return nfserr_replay_me
;
3065 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3066 sop
->so_seqid
, seqid
);
3068 return nfserr_bad_seqid
;
3072 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3073 struct nfsd4_open_confirm
*oc
)
3076 struct nfs4_stateowner
*sop
;
3077 struct nfs4_stateid
*stp
;
3079 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3080 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3081 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3083 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3089 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3090 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3091 CONFIRM
| OPEN_STATE
,
3092 &oc
->oc_stateowner
, &stp
, NULL
)))
3095 sop
= oc
->oc_stateowner
;
3096 sop
->so_confirmed
= 1;
3097 update_stateid(&stp
->st_stateid
);
3098 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3099 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3100 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stateid
));
3102 nfsd4_create_clid_dir(sop
->so_client
);
3104 if (oc
->oc_stateowner
) {
3105 nfs4_get_stateowner(oc
->oc_stateowner
);
3106 cstate
->replay_owner
= oc
->oc_stateowner
;
3108 nfs4_unlock_state();
3114 * unset all bits in union bitmap (bmap) that
3115 * do not exist in share (from successful OPEN_DOWNGRADE)
3118 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
3121 for (i
= 1; i
< 4; i
++) {
3122 if ((i
& access
) != i
)
3123 __clear_bit(i
, bmap
);
3128 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
3131 for (i
= 0; i
< 4; i
++) {
3132 if ((i
& deny
) != i
)
3133 __clear_bit(i
, bmap
);
3138 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3139 struct nfsd4_compound_state
*cstate
,
3140 struct nfsd4_open_downgrade
*od
)
3143 struct nfs4_stateid
*stp
;
3144 unsigned int share_access
;
3146 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3147 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3148 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3150 if (!access_valid(od
->od_share_access
, cstate
->minorversion
)
3151 || !deny_valid(od
->od_share_deny
))
3152 return nfserr_inval
;
3155 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3159 &od
->od_stateowner
, &stp
, NULL
)))
3162 status
= nfserr_inval
;
3163 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
3164 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3165 stp
->st_access_bmap
, od
->od_share_access
);
3168 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
3169 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3170 stp
->st_deny_bmap
, od
->od_share_deny
);
3173 set_access(&share_access
, stp
->st_access_bmap
);
3174 nfs4_file_downgrade(stp
->st_file
, share_access
& ~od
->od_share_access
);
3176 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
3177 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
3179 update_stateid(&stp
->st_stateid
);
3180 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3183 if (od
->od_stateowner
) {
3184 nfs4_get_stateowner(od
->od_stateowner
);
3185 cstate
->replay_owner
= od
->od_stateowner
;
3187 nfs4_unlock_state();
3192 * nfs4_unlock_state() called after encode
3195 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3196 struct nfsd4_close
*close
)
3199 struct nfs4_stateid
*stp
;
3201 dprintk("NFSD: nfsd4_close on file %.*s\n",
3202 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3203 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3206 /* check close_lru for replay */
3207 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3210 OPEN_STATE
| CLOSE_STATE
,
3211 &close
->cl_stateowner
, &stp
, NULL
)))
3214 update_stateid(&stp
->st_stateid
);
3215 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3217 /* release_stateid() calls nfsd_close() if needed */
3218 release_open_stateid(stp
);
3220 /* place unused nfs4_stateowners on so_close_lru list to be
3221 * released by the laundromat service after the lease period
3222 * to enable us to handle CLOSE replay
3224 if (list_empty(&close
->cl_stateowner
->so_stateids
))
3225 move_to_close_lru(close
->cl_stateowner
);
3227 if (close
->cl_stateowner
) {
3228 nfs4_get_stateowner(close
->cl_stateowner
);
3229 cstate
->replay_owner
= close
->cl_stateowner
;
3231 nfs4_unlock_state();
3236 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3237 struct nfsd4_delegreturn
*dr
)
3239 struct nfs4_delegation
*dp
;
3240 stateid_t
*stateid
= &dr
->dr_stateid
;
3241 struct inode
*inode
;
3245 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3247 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3249 if (nfsd4_has_session(cstate
))
3250 flags
|= HAS_SESSION
;
3252 status
= nfserr_bad_stateid
;
3253 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3255 status
= nfserr_stale_stateid
;
3256 if (STALE_STATEID(stateid
))
3258 status
= nfserr_bad_stateid
;
3259 if (!is_delegation_stateid(stateid
))
3261 dp
= find_delegation_stateid(inode
, stateid
);
3264 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
, flags
);
3267 renew_client(dp
->dl_client
);
3269 unhash_delegation(dp
);
3271 nfs4_unlock_state();
3278 * Lock owner state (byte-range locks)
3280 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3281 #define LOCK_HASH_BITS 8
3282 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3283 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3286 end_offset(u64 start
, u64 len
)
3291 return end
>= start
? end
: NFS4_MAX_UINT64
;
3294 /* last octet in a range */
3296 last_byte_offset(u64 start
, u64 len
)
3302 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3305 #define lockownerid_hashval(id) \
3306 ((id) & LOCK_HASH_MASK)
3308 static inline unsigned int
3309 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
3310 struct xdr_netobj
*ownername
)
3312 return (file_hashval(inode
) + cl_id
3313 + opaque_hashval(ownername
->data
, ownername
->len
))
3317 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
3318 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
3319 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
3321 static struct nfs4_stateid
*
3322 find_stateid(stateid_t
*stid
, int flags
)
3324 struct nfs4_stateid
*local
;
3325 u32 st_id
= stid
->si_stateownerid
;
3326 u32 f_id
= stid
->si_fileid
;
3327 unsigned int hashval
;
3329 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
3330 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
3331 hashval
= stateid_hashval(st_id
, f_id
);
3332 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
3333 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3334 (local
->st_stateid
.si_fileid
== f_id
))
3339 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
3340 hashval
= stateid_hashval(st_id
, f_id
);
3341 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
3342 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
3343 (local
->st_stateid
.si_fileid
== f_id
))
3350 static struct nfs4_delegation
*
3351 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
3353 struct nfs4_file
*fp
;
3354 struct nfs4_delegation
*dl
;
3356 dprintk("NFSD: %s: stateid=" STATEID_FMT
"\n", __func__
,
3359 fp
= find_file(ino
);
3362 dl
= find_delegation_file(fp
, stid
);
3368 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3369 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3370 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3371 * locking, this prevents us from being completely protocol-compliant. The
3372 * real solution to this problem is to start using unsigned file offsets in
3373 * the VFS, but this is a very deep change!
3376 nfs4_transform_lock_offset(struct file_lock
*lock
)
3378 if (lock
->fl_start
< 0)
3379 lock
->fl_start
= OFFSET_MAX
;
3380 if (lock
->fl_end
< 0)
3381 lock
->fl_end
= OFFSET_MAX
;
3384 /* Hack!: For now, we're defining this just so we can use a pointer to it
3385 * as a unique cookie to identify our (NFSv4's) posix locks. */
3386 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3390 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3392 struct nfs4_stateowner
*sop
;
3394 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3395 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
3396 kref_get(&sop
->so_ref
);
3398 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
3400 deny
->ld_sop
= NULL
;
3401 deny
->ld_clientid
.cl_boot
= 0;
3402 deny
->ld_clientid
.cl_id
= 0;
3404 deny
->ld_start
= fl
->fl_start
;
3405 deny
->ld_length
= NFS4_MAX_UINT64
;
3406 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3407 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3408 deny
->ld_type
= NFS4_READ_LT
;
3409 if (fl
->fl_type
!= F_RDLCK
)
3410 deny
->ld_type
= NFS4_WRITE_LT
;
3413 static struct nfs4_stateowner
*
3414 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
3415 struct xdr_netobj
*owner
)
3417 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
3418 struct nfs4_stateowner
*op
;
3420 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
3421 if (same_owner_str(op
, owner
, clid
))
3428 * Alloc a lock owner structure.
3429 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3432 * strhashval = lock_ownerstr_hashval
3435 static struct nfs4_stateowner
*
3436 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3437 struct nfs4_stateowner
*sop
;
3438 struct nfs4_replay
*rp
;
3439 unsigned int idhashval
;
3441 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
3443 idhashval
= lockownerid_hashval(current_ownerid
);
3444 INIT_LIST_HEAD(&sop
->so_idhash
);
3445 INIT_LIST_HEAD(&sop
->so_strhash
);
3446 INIT_LIST_HEAD(&sop
->so_perclient
);
3447 INIT_LIST_HEAD(&sop
->so_stateids
);
3448 INIT_LIST_HEAD(&sop
->so_perstateid
);
3449 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
3451 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
3452 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
3453 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
3454 sop
->so_is_open_owner
= 0;
3455 sop
->so_id
= current_ownerid
++;
3456 sop
->so_client
= clp
;
3457 /* It is the openowner seqid that will be incremented in encode in the
3458 * case of new lockowners; so increment the lock seqid manually: */
3459 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3460 sop
->so_confirmed
= 1;
3461 rp
= &sop
->so_replay
;
3462 rp
->rp_status
= nfserr_serverfault
;
3464 rp
->rp_buf
= rp
->rp_ibuf
;
3468 static struct nfs4_stateid
*
3469 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
3471 struct nfs4_stateid
*stp
;
3472 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
3474 stp
= nfs4_alloc_stateid();
3477 INIT_LIST_HEAD(&stp
->st_hash
);
3478 INIT_LIST_HEAD(&stp
->st_perfile
);
3479 INIT_LIST_HEAD(&stp
->st_perstateowner
);
3480 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
3481 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
3482 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3483 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
3484 stp
->st_stateowner
= sop
;
3487 stp
->st_stateid
.si_boot
= boot_time
;
3488 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
3489 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
3490 stp
->st_stateid
.si_generation
= 0;
3491 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3492 stp
->st_openstp
= open_stp
;
3499 check_lock_length(u64 offset
, u64 length
)
3501 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3502 LOFF_OVERFLOW(offset
, length
)));
3509 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3510 struct nfsd4_lock
*lock
)
3512 struct nfs4_stateowner
*open_sop
= NULL
;
3513 struct nfs4_stateowner
*lock_sop
= NULL
;
3514 struct nfs4_stateid
*lock_stp
;
3515 struct nfs4_file
*fp
;
3516 struct file
*filp
= NULL
;
3517 struct file_lock file_lock
;
3518 struct file_lock conflock
;
3520 unsigned int strhashval
;
3524 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3525 (long long) lock
->lk_offset
,
3526 (long long) lock
->lk_length
);
3528 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
3529 return nfserr_inval
;
3531 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
3532 S_IFREG
, NFSD_MAY_LOCK
))) {
3533 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3539 if (lock
->lk_is_new
) {
3541 * Client indicates that this is a new lockowner.
3542 * Use open owner and open stateid to create lock owner and
3545 struct nfs4_stateid
*open_stp
= NULL
;
3547 status
= nfserr_stale_clientid
;
3548 if (!nfsd4_has_session(cstate
) &&
3549 STALE_CLIENTID(&lock
->lk_new_clientid
))
3552 /* validate and update open stateid and open seqid */
3553 status
= nfs4_preprocess_seqid_op(cstate
,
3554 lock
->lk_new_open_seqid
,
3555 &lock
->lk_new_open_stateid
,
3557 &lock
->lk_replay_owner
, &open_stp
,
3561 open_sop
= lock
->lk_replay_owner
;
3562 /* create lockowner and lock stateid */
3563 fp
= open_stp
->st_file
;
3564 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
3565 open_sop
->so_client
->cl_clientid
.cl_id
,
3566 &lock
->v
.new.owner
);
3567 status
= nfserr_resource
;
3568 lock_sop
= alloc_init_lock_stateowner(strhashval
,
3569 open_sop
->so_client
, open_stp
, lock
);
3570 if (lock_sop
== NULL
)
3572 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
3573 if (lock_stp
== NULL
)
3576 /* lock (lock owner + lock stateid) already exists */
3577 status
= nfs4_preprocess_seqid_op(cstate
,
3578 lock
->lk_old_lock_seqid
,
3579 &lock
->lk_old_lock_stateid
,
3581 &lock
->lk_replay_owner
, &lock_stp
, lock
);
3584 lock_sop
= lock
->lk_replay_owner
;
3585 fp
= lock_stp
->st_file
;
3587 /* lock->lk_replay_owner and lock_stp have been created or found */
3589 status
= nfserr_grace
;
3590 if (locks_in_grace() && !lock
->lk_reclaim
)
3592 status
= nfserr_no_grace
;
3593 if (!locks_in_grace() && lock
->lk_reclaim
)
3596 locks_init_lock(&file_lock
);
3597 switch (lock
->lk_type
) {
3600 if (find_readable_file(lock_stp
->st_file
)) {
3601 nfs4_get_vfs_file(rqstp
, fp
, &cstate
->current_fh
, NFS4_SHARE_ACCESS_READ
);
3602 filp
= find_readable_file(lock_stp
->st_file
);
3604 file_lock
.fl_type
= F_RDLCK
;
3608 case NFS4_WRITEW_LT
:
3609 if (find_writeable_file(lock_stp
->st_file
)) {
3610 nfs4_get_vfs_file(rqstp
, fp
, &cstate
->current_fh
, NFS4_SHARE_ACCESS_WRITE
);
3611 filp
= find_writeable_file(lock_stp
->st_file
);
3613 file_lock
.fl_type
= F_WRLCK
;
3617 status
= nfserr_inval
;
3621 status
= nfserr_openmode
;
3624 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
3625 file_lock
.fl_pid
= current
->tgid
;
3626 file_lock
.fl_file
= filp
;
3627 file_lock
.fl_flags
= FL_POSIX
;
3628 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3630 file_lock
.fl_start
= lock
->lk_offset
;
3631 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
3632 nfs4_transform_lock_offset(&file_lock
);
3635 * Try to lock the file in the VFS.
3636 * Note: locks.c uses the BKL to protect the inode's lock list.
3639 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
3641 case 0: /* success! */
3642 update_stateid(&lock_stp
->st_stateid
);
3643 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
3647 case (EAGAIN
): /* conflock holds conflicting lock */
3648 status
= nfserr_denied
;
3649 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3650 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
3653 status
= nfserr_deadlock
;
3656 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
3657 status
= nfserr_resource
;
3661 if (status
&& lock
->lk_is_new
&& lock_sop
)
3662 release_lockowner(lock_sop
);
3663 if (lock
->lk_replay_owner
) {
3664 nfs4_get_stateowner(lock
->lk_replay_owner
);
3665 cstate
->replay_owner
= lock
->lk_replay_owner
;
3667 nfs4_unlock_state();
3672 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3673 * so we do a temporary open here just to get an open file to pass to
3674 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3677 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
3682 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
3685 err
= vfs_test_lock(file
, lock
);
3694 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3695 struct nfsd4_lockt
*lockt
)
3697 struct inode
*inode
;
3698 struct file_lock file_lock
;
3702 if (locks_in_grace())
3703 return nfserr_grace
;
3705 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
3706 return nfserr_inval
;
3708 lockt
->lt_stateowner
= NULL
;
3711 status
= nfserr_stale_clientid
;
3712 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
3715 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
3716 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3717 if (status
== nfserr_symlink
)
3718 status
= nfserr_inval
;
3722 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3723 locks_init_lock(&file_lock
);
3724 switch (lockt
->lt_type
) {
3727 file_lock
.fl_type
= F_RDLCK
;
3730 case NFS4_WRITEW_LT
:
3731 file_lock
.fl_type
= F_WRLCK
;
3734 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3735 status
= nfserr_inval
;
3739 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
3740 &lockt
->lt_clientid
, &lockt
->lt_owner
);
3741 if (lockt
->lt_stateowner
)
3742 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
3743 file_lock
.fl_pid
= current
->tgid
;
3744 file_lock
.fl_flags
= FL_POSIX
;
3746 file_lock
.fl_start
= lockt
->lt_offset
;
3747 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
3749 nfs4_transform_lock_offset(&file_lock
);
3752 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
3754 status
= nfserrno(error
);
3757 if (file_lock
.fl_type
!= F_UNLCK
) {
3758 status
= nfserr_denied
;
3759 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
3762 nfs4_unlock_state();
3767 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3768 struct nfsd4_locku
*locku
)
3770 struct nfs4_stateid
*stp
;
3771 struct file
*filp
= NULL
;
3772 struct file_lock file_lock
;
3776 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3777 (long long) locku
->lu_offset
,
3778 (long long) locku
->lu_length
);
3780 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
3781 return nfserr_inval
;
3785 if ((status
= nfs4_preprocess_seqid_op(cstate
,
3789 &locku
->lu_stateowner
, &stp
, NULL
)))
3792 filp
= find_any_file(stp
->st_file
);
3794 status
= nfserr_lock_range
;
3798 locks_init_lock(&file_lock
);
3799 file_lock
.fl_type
= F_UNLCK
;
3800 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
3801 file_lock
.fl_pid
= current
->tgid
;
3802 file_lock
.fl_file
= filp
;
3803 file_lock
.fl_flags
= FL_POSIX
;
3804 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
3805 file_lock
.fl_start
= locku
->lu_offset
;
3807 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
3808 nfs4_transform_lock_offset(&file_lock
);
3811 * Try to unlock the file in the VFS.
3813 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
3815 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3819 * OK, unlock succeeded; the only thing left to do is update the stateid.
3821 update_stateid(&stp
->st_stateid
);
3822 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
3825 if (locku
->lu_stateowner
) {
3826 nfs4_get_stateowner(locku
->lu_stateowner
);
3827 cstate
->replay_owner
= locku
->lu_stateowner
;
3829 nfs4_unlock_state();
3833 status
= nfserrno(err
);
3839 * 1: locks held by lockowner
3840 * 0: no locks held by lockowner
3843 check_for_locks(struct nfs4_file
*filp
, struct nfs4_stateowner
*lowner
)
3845 struct file_lock
**flpp
;
3846 struct inode
*inode
= filp
->fi_inode
;
3850 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
3851 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3862 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3863 struct nfsd4_compound_state
*cstate
,
3864 struct nfsd4_release_lockowner
*rlockowner
)
3866 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3867 struct nfs4_stateowner
*sop
;
3868 struct nfs4_stateid
*stp
;
3869 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3870 struct list_head matches
;
3874 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3875 clid
->cl_boot
, clid
->cl_id
);
3878 status
= nfserr_stale_clientid
;
3879 if (STALE_CLIENTID(clid
))
3884 status
= nfserr_locks_held
;
3885 INIT_LIST_HEAD(&matches
);
3886 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3887 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3888 if (!same_owner_str(sop
, owner
, clid
))
3890 list_for_each_entry(stp
, &sop
->so_stateids
,
3892 if (check_for_locks(stp
->st_file
, sop
))
3894 /* Note: so_perclient unused for lockowners,
3895 * so it's OK to fool with here. */
3896 list_add(&sop
->so_perclient
, &matches
);
3900 /* Clients probably won't expect us to return with some (but not all)
3901 * of the lockowner state released; so don't release any until all
3902 * have been checked. */
3904 while (!list_empty(&matches
)) {
3905 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3907 /* unhash_stateowner deletes so_perclient only
3908 * for openowners. */
3909 list_del(&sop
->so_perclient
);
3910 release_lockowner(sop
);
3913 nfs4_unlock_state();
3917 static inline struct nfs4_client_reclaim
*
3920 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3924 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
3926 unsigned int strhashval
= clientstr_hashval(name
);
3927 struct nfs4_client
*clp
;
3929 clp
= find_confirmed_client_by_str(name
, strhashval
, use_exchange_id
);
3934 * failure => all reset bets are off, nfserr_no_grace...
3937 nfs4_client_to_reclaim(const char *name
)
3939 unsigned int strhashval
;
3940 struct nfs4_client_reclaim
*crp
= NULL
;
3942 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3943 crp
= alloc_reclaim();
3946 strhashval
= clientstr_hashval(name
);
3947 INIT_LIST_HEAD(&crp
->cr_strhash
);
3948 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3949 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3950 reclaim_str_hashtbl_size
++;
3955 nfs4_release_reclaim(void)
3957 struct nfs4_client_reclaim
*crp
= NULL
;
3960 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3961 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3962 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3963 struct nfs4_client_reclaim
, cr_strhash
);
3964 list_del(&crp
->cr_strhash
);
3966 reclaim_str_hashtbl_size
--;
3969 BUG_ON(reclaim_str_hashtbl_size
);
3973 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3974 static struct nfs4_client_reclaim
*
3975 nfs4_find_reclaim_client(clientid_t
*clid
)
3977 unsigned int strhashval
;
3978 struct nfs4_client
*clp
;
3979 struct nfs4_client_reclaim
*crp
= NULL
;
3982 /* find clientid in conf_id_hashtbl */
3983 clp
= find_confirmed_client(clid
);
3987 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3988 clp
->cl_name
.len
, clp
->cl_name
.data
,
3991 /* find clp->cl_name in reclaim_str_hashtbl */
3992 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3993 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3994 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4002 * Called from OPEN. Look for clientid in reclaim list.
4005 nfs4_check_open_reclaim(clientid_t
*clid
)
4007 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
4010 /* initialization to perform at module load time: */
4013 nfs4_state_init(void)
4017 status
= nfsd4_init_slabs();
4020 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4021 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4022 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4023 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4024 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4025 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4027 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4028 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4029 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4030 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4032 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4033 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4034 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
4036 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
4037 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
4038 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
4040 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
4041 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
4042 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
4044 memset(&onestateid
, ~0, sizeof(stateid_t
));
4045 INIT_LIST_HEAD(&close_lru
);
4046 INIT_LIST_HEAD(&client_lru
);
4047 INIT_LIST_HEAD(&del_recall_lru
);
4048 reclaim_str_hashtbl_size
= 0;
4053 nfsd4_load_reboot_recovery_data(void)
4058 nfsd4_init_recdir(user_recovery_dirname
);
4059 status
= nfsd4_recdir_load();
4060 nfs4_unlock_state();
4062 printk("NFSD: Failure reading reboot recovery data\n");
4066 * Since the lifetime of a delegation isn't limited to that of an open, a
4067 * client may quite reasonably hang on to a delegation as long as it has
4068 * the inode cached. This becomes an obvious problem the first time a
4069 * client's inode cache approaches the size of the server's total memory.
4071 * For now we avoid this problem by imposing a hard limit on the number
4072 * of delegations, which varies according to the server's memory size.
4075 set_max_delegations(void)
4078 * Allow at most 4 delegations per megabyte of RAM. Quick
4079 * estimates suggest that in the worst case (where every delegation
4080 * is for a different inode), a delegation could take about 1.5K,
4081 * giving a worst case usage of about 6% of memory.
4083 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4086 /* initialization to perform when the nfsd service is started: */
4089 __nfs4_state_start(void)
4093 boot_time
= get_seconds();
4094 locks_start_grace(&nfsd4_manager
);
4095 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4097 ret
= set_callback_cred();
4100 laundry_wq
= create_singlethread_workqueue("nfsd4");
4101 if (laundry_wq
== NULL
)
4103 ret
= nfsd4_create_callback_queue();
4105 goto out_free_laundry
;
4106 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4107 set_max_delegations();
4110 destroy_workqueue(laundry_wq
);
4115 nfs4_state_start(void)
4117 nfsd4_load_reboot_recovery_data();
4118 return __nfs4_state_start();
4122 __nfs4_state_shutdown(void)
4125 struct nfs4_client
*clp
= NULL
;
4126 struct nfs4_delegation
*dp
= NULL
;
4127 struct list_head
*pos
, *next
, reaplist
;
4129 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4130 while (!list_empty(&conf_id_hashtbl
[i
])) {
4131 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4134 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4135 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4139 INIT_LIST_HEAD(&reaplist
);
4140 spin_lock(&recall_lock
);
4141 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4142 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4143 list_move(&dp
->dl_recall_lru
, &reaplist
);
4145 spin_unlock(&recall_lock
);
4146 list_for_each_safe(pos
, next
, &reaplist
) {
4147 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4148 list_del_init(&dp
->dl_recall_lru
);
4149 unhash_delegation(dp
);
4152 nfsd4_shutdown_recdir();
4156 nfs4_state_shutdown(void)
4158 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
4159 destroy_workqueue(laundry_wq
);
4160 locks_end_grace(&nfsd4_manager
);
4162 nfs4_release_reclaim();
4163 __nfs4_state_shutdown();
4164 nfs4_unlock_state();
4165 nfsd4_destroy_callback_queue();
4169 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4170 * accessed when nfsd is starting.
4173 nfs4_set_recdir(char *recdir
)
4175 strcpy(user_recovery_dirname
, recdir
);
4179 * Change the NFSv4 recovery directory to recdir.
4182 nfs4_reset_recoverydir(char *recdir
)
4187 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
4191 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
4192 nfs4_set_recdir(recdir
);
4200 nfs4_recoverydir(void)
4202 return user_recovery_dirname
;