2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/clnt.h>
46 #include "current_stateid.h"
47 #include "fault_inject.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 time_t nfsd4_lease
= 90; /* default lease time */
55 time_t nfsd4_grace
= 90;
57 #define all_ones {{~0,~0},~0}
58 static const stateid_t one_stateid
= {
60 .si_opaque
= all_ones
,
62 static const stateid_t zero_stateid
= {
65 static const stateid_t currentstateid
= {
69 static u64 current_sessionid
= 1;
71 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
72 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
73 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
);
80 /* Currently used for almost all code touching nfsv4 state: */
81 static DEFINE_MUTEX(client_mutex
);
84 * Currently used for the del_recall_lru and file hash table. In an
85 * effort to decrease the scope of the client_mutex, this spinlock may
86 * eventually cover more:
88 static DEFINE_SPINLOCK(recall_lock
);
90 static struct kmem_cache
*openowner_slab
= NULL
;
91 static struct kmem_cache
*lockowner_slab
= NULL
;
92 static struct kmem_cache
*file_slab
= NULL
;
93 static struct kmem_cache
*stateid_slab
= NULL
;
94 static struct kmem_cache
*deleg_slab
= NULL
;
99 mutex_lock(&client_mutex
);
102 static void free_session(struct kref
*);
104 /* Must be called under the client_lock */
105 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
107 kref_put(&ses
->se_ref
, free_session
);
110 static void nfsd4_get_session(struct nfsd4_session
*ses
)
112 kref_get(&ses
->se_ref
);
116 nfs4_unlock_state(void)
118 mutex_unlock(&client_mutex
);
122 opaque_hashval(const void *ptr
, int nbytes
)
124 unsigned char *cptr
= (unsigned char *) ptr
;
134 static struct list_head del_recall_lru
;
136 static void nfsd4_free_file(struct nfs4_file
*f
)
138 kmem_cache_free(file_slab
, f
);
142 put_nfs4_file(struct nfs4_file
*fi
)
144 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
145 list_del(&fi
->fi_hash
);
146 spin_unlock(&recall_lock
);
153 get_nfs4_file(struct nfs4_file
*fi
)
155 atomic_inc(&fi
->fi_ref
);
158 static int num_delegations
;
159 unsigned int max_delegations
;
162 * Open owner state (share locks)
165 /* hash tables for lock and open owners */
166 #define OWNER_HASH_BITS 8
167 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
168 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
170 static unsigned int ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
174 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
176 return ret
& OWNER_HASH_MASK
;
179 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
181 /* hash table for nfs4_file */
182 #define FILE_HASH_BITS 8
183 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
185 static unsigned int file_hashval(struct inode
*ino
)
187 /* XXX: why are we hashing on inode pointer, anyway? */
188 return hash_ptr(ino
, FILE_HASH_BITS
);
191 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
193 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
195 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
196 atomic_inc(&fp
->fi_access
[oflag
]);
199 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
201 if (oflag
== O_RDWR
) {
202 __nfs4_file_get_access(fp
, O_RDONLY
);
203 __nfs4_file_get_access(fp
, O_WRONLY
);
205 __nfs4_file_get_access(fp
, oflag
);
208 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
210 if (fp
->fi_fds
[oflag
]) {
211 fput(fp
->fi_fds
[oflag
]);
212 fp
->fi_fds
[oflag
] = NULL
;
216 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
218 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
219 nfs4_file_put_fd(fp
, oflag
);
221 * It's also safe to get rid of the RDWR open *if*
222 * we no longer have need of the other kind of access
223 * or if we already have the other kind of open:
225 if (fp
->fi_fds
[1-oflag
]
226 || atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
227 nfs4_file_put_fd(fp
, O_RDWR
);
231 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
233 if (oflag
== O_RDWR
) {
234 __nfs4_file_put_access(fp
, O_RDONLY
);
235 __nfs4_file_put_access(fp
, O_WRONLY
);
237 __nfs4_file_put_access(fp
, oflag
);
240 static inline int get_new_stid(struct nfs4_stid
*stid
)
242 static int min_stateid
= 0;
243 struct idr
*stateids
= &stid
->sc_client
->cl_stateids
;
247 error
= idr_get_new_above(stateids
, stid
, min_stateid
, &new_stid
);
249 * Note: the necessary preallocation was done in
250 * nfs4_alloc_stateid(). The idr code caps the number of
251 * preallocations that can exist at a time, but the state lock
252 * prevents anyone from using ours before we get here:
256 * It shouldn't be a problem to reuse an opaque stateid value.
257 * I don't think it is for 4.1. But with 4.0 I worry that, for
258 * example, a stray write retransmission could be accepted by
259 * the server when it should have been rejected. Therefore,
260 * adopt a trick from the sctp code to attempt to maximize the
261 * amount of time until an id is reused, by ensuring they always
262 * "increase" (mod INT_MAX):
265 min_stateid
= new_stid
+1;
266 if (min_stateid
== INT_MAX
)
271 static void init_stid(struct nfs4_stid
*stid
, struct nfs4_client
*cl
, unsigned char type
)
273 stateid_t
*s
= &stid
->sc_stateid
;
276 stid
->sc_type
= type
;
277 stid
->sc_client
= cl
;
278 s
->si_opaque
.so_clid
= cl
->cl_clientid
;
279 new_id
= get_new_stid(stid
);
280 s
->si_opaque
.so_id
= (u32
)new_id
;
281 /* Will be incremented before return to client: */
282 s
->si_generation
= 0;
285 static struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
)
287 struct idr
*stateids
= &cl
->cl_stateids
;
289 if (!idr_pre_get(stateids
, GFP_KERNEL
))
292 * Note: if we fail here (or any time between now and the time
293 * we actually get the new idr), we won't need to undo the idr
294 * preallocation, since the idr code caps the number of
295 * preallocated entries.
297 return kmem_cache_alloc(slab
, GFP_KERNEL
);
300 static struct nfs4_ol_stateid
* nfs4_alloc_stateid(struct nfs4_client
*clp
)
302 return openlockstateid(nfs4_alloc_stid(clp
, stateid_slab
));
305 static struct nfs4_delegation
*
306 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_ol_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
308 struct nfs4_delegation
*dp
;
309 struct nfs4_file
*fp
= stp
->st_file
;
311 dprintk("NFSD alloc_init_deleg\n");
313 * Major work on the lease subsystem (for example, to support
314 * calbacks on stat) will be required before we can support
315 * write delegations properly.
317 if (type
!= NFS4_OPEN_DELEGATE_READ
)
319 if (fp
->fi_had_conflict
)
321 if (num_delegations
> max_delegations
)
323 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
326 init_stid(&dp
->dl_stid
, clp
, NFS4_DELEG_STID
);
328 * delegation seqid's are never incremented. The 4.1 special
329 * meaning of seqid 0 isn't meaningful, really, but let's avoid
330 * 0 anyway just for consistency and use 1:
332 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
334 INIT_LIST_HEAD(&dp
->dl_perfile
);
335 INIT_LIST_HEAD(&dp
->dl_perclnt
);
336 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
340 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
342 atomic_set(&dp
->dl_count
, 1);
343 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
348 nfs4_put_delegation(struct nfs4_delegation
*dp
)
350 if (atomic_dec_and_test(&dp
->dl_count
)) {
351 dprintk("NFSD: freeing dp %p\n",dp
);
352 put_nfs4_file(dp
->dl_file
);
353 kmem_cache_free(deleg_slab
, dp
);
358 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
360 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
361 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
363 fput(fp
->fi_deleg_file
);
364 fp
->fi_deleg_file
= NULL
;
368 static void unhash_stid(struct nfs4_stid
*s
)
370 struct idr
*stateids
= &s
->sc_client
->cl_stateids
;
372 idr_remove(stateids
, s
->sc_stateid
.si_opaque
.so_id
);
375 /* Called under the state lock. */
377 unhash_delegation(struct nfs4_delegation
*dp
)
379 unhash_stid(&dp
->dl_stid
);
380 list_del_init(&dp
->dl_perclnt
);
381 spin_lock(&recall_lock
);
382 list_del_init(&dp
->dl_perfile
);
383 list_del_init(&dp
->dl_recall_lru
);
384 spin_unlock(&recall_lock
);
385 nfs4_put_deleg_lease(dp
->dl_file
);
386 nfs4_put_delegation(dp
);
393 /* client_lock protects the client lru list and session hash table */
394 static DEFINE_SPINLOCK(client_lock
);
396 /* Hash tables for nfs4_clientid state */
397 #define CLIENT_HASH_BITS 4
398 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
399 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
401 static unsigned int clientid_hashval(u32 id
)
403 return id
& CLIENT_HASH_MASK
;
406 static unsigned int clientstr_hashval(const char *name
)
408 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
412 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
413 * used in reboot/reset lease grace period processing
415 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
416 * setclientid_confirmed info.
418 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
421 * client_lru holds client queue ordered by nfs4_client.cl_time
424 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
425 * for last close replay.
427 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
428 static int reclaim_str_hashtbl_size
= 0;
429 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
430 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
431 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
432 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
433 static struct list_head client_lru
;
434 static struct list_head close_lru
;
437 * We store the NONE, READ, WRITE, and BOTH bits separately in the
438 * st_{access,deny}_bmap field of the stateid, in order to track not
439 * only what share bits are currently in force, but also what
440 * combinations of share bits previous opens have used. This allows us
441 * to enforce the recommendation of rfc 3530 14.2.19 that the server
442 * return an error if the client attempt to downgrade to a combination
443 * of share bits not explicable by closing some of its previous opens.
445 * XXX: This enforcement is actually incomplete, since we don't keep
446 * track of access/deny bit combinations; so, e.g., we allow:
448 * OPEN allow read, deny write
449 * OPEN allow both, deny none
450 * DOWNGRADE allow read, deny none
452 * which we should reject.
455 bmap_to_share_mode(unsigned long bmap
) {
457 unsigned int access
= 0;
459 for (i
= 1; i
< 4; i
++) {
460 if (test_bit(i
, &bmap
))
467 test_share(struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
) {
468 unsigned int access
, deny
;
470 access
= bmap_to_share_mode(stp
->st_access_bmap
);
471 deny
= bmap_to_share_mode(stp
->st_deny_bmap
);
472 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
477 /* set share access for a given stateid */
479 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
481 __set_bit(access
, &stp
->st_access_bmap
);
484 /* clear share access for a given stateid */
486 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
488 __clear_bit(access
, &stp
->st_access_bmap
);
491 /* test whether a given stateid has access */
493 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
495 return test_bit(access
, &stp
->st_access_bmap
);
498 /* set share deny for a given stateid */
500 set_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
502 __set_bit(access
, &stp
->st_deny_bmap
);
505 /* clear share deny for a given stateid */
507 clear_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
509 __clear_bit(access
, &stp
->st_deny_bmap
);
512 /* test whether a given stateid is denying specific access */
514 test_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
516 return test_bit(access
, &stp
->st_deny_bmap
);
519 static int nfs4_access_to_omode(u32 access
)
521 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
522 case NFS4_SHARE_ACCESS_READ
:
524 case NFS4_SHARE_ACCESS_WRITE
:
526 case NFS4_SHARE_ACCESS_BOTH
:
532 /* release all access and file references for a given stateid */
534 release_all_access(struct nfs4_ol_stateid
*stp
)
538 for (i
= 1; i
< 4; i
++) {
539 if (test_access(i
, stp
))
540 nfs4_file_put_access(stp
->st_file
,
541 nfs4_access_to_omode(i
));
542 clear_access(i
, stp
);
546 static void unhash_generic_stateid(struct nfs4_ol_stateid
*stp
)
548 list_del(&stp
->st_perfile
);
549 list_del(&stp
->st_perstateowner
);
552 static void close_generic_stateid(struct nfs4_ol_stateid
*stp
)
554 release_all_access(stp
);
555 put_nfs4_file(stp
->st_file
);
559 static void free_generic_stateid(struct nfs4_ol_stateid
*stp
)
561 kmem_cache_free(stateid_slab
, stp
);
564 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
568 unhash_generic_stateid(stp
);
569 unhash_stid(&stp
->st_stid
);
570 file
= find_any_file(stp
->st_file
);
572 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
573 close_generic_stateid(stp
);
574 free_generic_stateid(stp
);
577 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
579 struct nfs4_ol_stateid
*stp
;
581 list_del(&lo
->lo_owner
.so_strhash
);
582 list_del(&lo
->lo_perstateid
);
583 list_del(&lo
->lo_owner_ino_hash
);
584 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
585 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
586 struct nfs4_ol_stateid
, st_perstateowner
);
587 release_lock_stateid(stp
);
591 static void release_lockowner(struct nfs4_lockowner
*lo
)
593 unhash_lockowner(lo
);
594 nfs4_free_lockowner(lo
);
598 release_stateid_lockowners(struct nfs4_ol_stateid
*open_stp
)
600 struct nfs4_lockowner
*lo
;
602 while (!list_empty(&open_stp
->st_lockowners
)) {
603 lo
= list_entry(open_stp
->st_lockowners
.next
,
604 struct nfs4_lockowner
, lo_perstateid
);
605 release_lockowner(lo
);
609 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
)
611 unhash_generic_stateid(stp
);
612 release_stateid_lockowners(stp
);
613 close_generic_stateid(stp
);
616 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
618 unhash_open_stateid(stp
);
619 unhash_stid(&stp
->st_stid
);
620 free_generic_stateid(stp
);
623 static void unhash_openowner(struct nfs4_openowner
*oo
)
625 struct nfs4_ol_stateid
*stp
;
627 list_del(&oo
->oo_owner
.so_strhash
);
628 list_del(&oo
->oo_perclient
);
629 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
630 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
631 struct nfs4_ol_stateid
, st_perstateowner
);
632 release_open_stateid(stp
);
636 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
638 struct nfs4_ol_stateid
*s
= oo
->oo_last_closed_stid
;
641 unhash_stid(&s
->st_stid
);
642 free_generic_stateid(s
);
643 oo
->oo_last_closed_stid
= NULL
;
647 static void release_openowner(struct nfs4_openowner
*oo
)
649 unhash_openowner(oo
);
650 list_del(&oo
->oo_close_lru
);
651 release_last_closed_stateid(oo
);
652 nfs4_free_openowner(oo
);
655 #define SESSION_HASH_SIZE 512
656 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
659 hash_sessionid(struct nfs4_sessionid
*sessionid
)
661 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
663 return sid
->sequence
% SESSION_HASH_SIZE
;
668 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
670 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
671 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
675 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
682 gen_sessionid(struct nfsd4_session
*ses
)
684 struct nfs4_client
*clp
= ses
->se_client
;
685 struct nfsd4_sessionid
*sid
;
687 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
688 sid
->clientid
= clp
->cl_clientid
;
689 sid
->sequence
= current_sessionid
++;
694 * The protocol defines ca_maxresponssize_cached to include the size of
695 * the rpc header, but all we need to cache is the data starting after
696 * the end of the initial SEQUENCE operation--the rest we regenerate
697 * each time. Therefore we can advertise a ca_maxresponssize_cached
698 * value that is the number of bytes in our cache plus a few additional
699 * bytes. In order to stay on the safe side, and not promise more than
700 * we can cache, those additional bytes must be the minimum possible: 24
701 * bytes of rpc header (xid through accept state, with AUTH_NULL
702 * verifier), 12 for the compound header (with zero-length tag), and 44
703 * for the SEQUENCE op response:
705 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
708 free_session_slots(struct nfsd4_session
*ses
)
712 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
713 kfree(ses
->se_slots
[i
]);
717 * We don't actually need to cache the rpc and session headers, so we
718 * can allocate a little less for each slot:
720 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
722 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
725 static int nfsd4_sanitize_slot_size(u32 size
)
727 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
728 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
734 * XXX: If we run out of reserved DRC memory we could (up to a point)
735 * re-negotiate active sessions and reduce their slot usage to make
736 * room for new connections. For now we just fail the create session.
738 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
742 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
744 spin_lock(&nfsd_drc_lock
);
745 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
746 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
747 num
= min_t(int, num
, avail
/ slotsize
);
748 nfsd_drc_mem_used
+= num
* slotsize
;
749 spin_unlock(&nfsd_drc_lock
);
754 static void nfsd4_put_drc_mem(int slotsize
, int num
)
756 spin_lock(&nfsd_drc_lock
);
757 nfsd_drc_mem_used
-= slotsize
* num
;
758 spin_unlock(&nfsd_drc_lock
);
761 static struct nfsd4_session
*__alloc_session(int slotsize
, int numslots
)
763 struct nfsd4_session
*new;
766 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
767 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
768 mem
= numslots
* sizeof(struct nfsd4_slot
*);
770 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
773 /* allocate each struct nfsd4_slot and data cache in one piece */
774 for (i
= 0; i
< numslots
; i
++) {
775 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
776 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
777 if (!new->se_slots
[i
])
783 kfree(new->se_slots
[i
]);
788 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
790 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
792 new->maxreqs
= numslots
;
793 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
794 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
795 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
796 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
797 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
800 static void free_conn(struct nfsd4_conn
*c
)
802 svc_xprt_put(c
->cn_xprt
);
806 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
808 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
809 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
811 spin_lock(&clp
->cl_lock
);
812 if (!list_empty(&c
->cn_persession
)) {
813 list_del(&c
->cn_persession
);
816 spin_unlock(&clp
->cl_lock
);
817 nfsd4_probe_callback(clp
);
820 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
822 struct nfsd4_conn
*conn
;
824 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
827 svc_xprt_get(rqstp
->rq_xprt
);
828 conn
->cn_xprt
= rqstp
->rq_xprt
;
829 conn
->cn_flags
= flags
;
830 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
834 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
836 conn
->cn_session
= ses
;
837 list_add(&conn
->cn_persession
, &ses
->se_conns
);
840 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
842 struct nfs4_client
*clp
= ses
->se_client
;
844 spin_lock(&clp
->cl_lock
);
845 __nfsd4_hash_conn(conn
, ses
);
846 spin_unlock(&clp
->cl_lock
);
849 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
851 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
852 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
855 static void nfsd4_init_conn(struct svc_rqst
*rqstp
, struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
859 nfsd4_hash_conn(conn
, ses
);
860 ret
= nfsd4_register_conn(conn
);
862 /* oops; xprt is already down: */
863 nfsd4_conn_lost(&conn
->cn_xpt_user
);
864 if (ses
->se_client
->cl_cb_state
== NFSD4_CB_DOWN
&&
865 conn
->cn_flags
& NFS4_CDFC4_BACK
) {
866 /* callback channel may be back up */
867 nfsd4_probe_callback(ses
->se_client
);
871 static struct nfsd4_conn
*alloc_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_create_session
*cses
)
873 u32 dir
= NFS4_CDFC4_FORE
;
875 if (cses
->flags
& SESSION4_BACK_CHAN
)
876 dir
|= NFS4_CDFC4_BACK
;
877 return alloc_conn(rqstp
, dir
);
880 /* must be called under client_lock */
881 static void nfsd4_del_conns(struct nfsd4_session
*s
)
883 struct nfs4_client
*clp
= s
->se_client
;
884 struct nfsd4_conn
*c
;
886 spin_lock(&clp
->cl_lock
);
887 while (!list_empty(&s
->se_conns
)) {
888 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
889 list_del_init(&c
->cn_persession
);
890 spin_unlock(&clp
->cl_lock
);
892 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
895 spin_lock(&clp
->cl_lock
);
897 spin_unlock(&clp
->cl_lock
);
900 static void __free_session(struct nfsd4_session
*ses
)
902 nfsd4_put_drc_mem(slot_bytes(&ses
->se_fchannel
), ses
->se_fchannel
.maxreqs
);
903 free_session_slots(ses
);
907 static void free_session(struct kref
*kref
)
909 struct nfsd4_session
*ses
;
911 lockdep_assert_held(&client_lock
);
912 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
913 nfsd4_del_conns(ses
);
917 void nfsd4_put_session(struct nfsd4_session
*ses
)
919 spin_lock(&client_lock
);
920 nfsd4_put_session_locked(ses
);
921 spin_unlock(&client_lock
);
924 static struct nfsd4_session
*alloc_session(struct nfsd4_channel_attrs
*fchan
)
926 struct nfsd4_session
*new;
927 int numslots
, slotsize
;
929 * Note decreasing slot size below client's request may
930 * make it difficult for client to function correctly, whereas
931 * decreasing the number of slots will (just?) affect
932 * performance. When short on memory we therefore prefer to
933 * decrease number of slots instead of their size.
935 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
936 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
940 new = __alloc_session(slotsize
, numslots
);
942 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
945 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
949 static struct nfsd4_session
*init_session(struct svc_rqst
*rqstp
, struct nfsd4_session
*new, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
953 new->se_client
= clp
;
956 INIT_LIST_HEAD(&new->se_conns
);
958 new->se_cb_seq_nr
= 1;
959 new->se_flags
= cses
->flags
;
960 new->se_cb_prog
= cses
->callback_prog
;
961 kref_init(&new->se_ref
);
962 idx
= hash_sessionid(&new->se_sessionid
);
963 spin_lock(&client_lock
);
964 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
965 spin_lock(&clp
->cl_lock
);
966 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
967 spin_unlock(&clp
->cl_lock
);
968 spin_unlock(&client_lock
);
970 if (cses
->flags
& SESSION4_BACK_CHAN
) {
971 struct sockaddr
*sa
= svc_addr(rqstp
);
973 * This is a little silly; with sessions there's no real
974 * use for the callback address. Use the peer address
975 * as a reasonable default for now, but consider fixing
976 * the rpc client not to require an address in the
979 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
980 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
982 nfsd4_probe_callback(clp
);
986 /* caller must hold client_lock */
987 static struct nfsd4_session
*
988 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
990 struct nfsd4_session
*elem
;
993 dump_sessionid(__func__
, sessionid
);
994 idx
= hash_sessionid(sessionid
);
995 /* Search in the appropriate list */
996 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
997 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
998 NFS4_MAX_SESSIONID_LEN
)) {
1003 dprintk("%s: session not found\n", __func__
);
1007 /* caller must hold client_lock */
1009 unhash_session(struct nfsd4_session
*ses
)
1011 list_del(&ses
->se_hash
);
1012 spin_lock(&ses
->se_client
->cl_lock
);
1013 list_del(&ses
->se_perclnt
);
1014 spin_unlock(&ses
->se_client
->cl_lock
);
1017 /* must be called under the client_lock */
1019 renew_client_locked(struct nfs4_client
*clp
)
1021 if (is_client_expired(clp
)) {
1023 printk("%s: client (clientid %08x/%08x) already expired\n",
1025 clp
->cl_clientid
.cl_boot
,
1026 clp
->cl_clientid
.cl_id
);
1030 dprintk("renewing client (clientid %08x/%08x)\n",
1031 clp
->cl_clientid
.cl_boot
,
1032 clp
->cl_clientid
.cl_id
);
1033 list_move_tail(&clp
->cl_lru
, &client_lru
);
1034 clp
->cl_time
= get_seconds();
1038 renew_client(struct nfs4_client
*clp
)
1040 spin_lock(&client_lock
);
1041 renew_client_locked(clp
);
1042 spin_unlock(&client_lock
);
1045 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1047 STALE_CLIENTID(clientid_t
*clid
, struct nfsd_net
*nn
)
1049 if (clid
->cl_boot
== nn
->boot_time
)
1051 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1052 clid
->cl_boot
, clid
->cl_id
, nn
->boot_time
);
1057 * XXX Should we use a slab cache ?
1058 * This type of memory management is somewhat inefficient, but we use it
1059 * anyway since SETCLIENTID is not a common operation.
1061 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1063 struct nfs4_client
*clp
;
1065 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1068 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1069 if (clp
->cl_name
.data
== NULL
) {
1073 clp
->cl_name
.len
= name
.len
;
1078 free_client(struct nfs4_client
*clp
)
1080 lockdep_assert_held(&client_lock
);
1081 while (!list_empty(&clp
->cl_sessions
)) {
1082 struct nfsd4_session
*ses
;
1083 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1085 list_del(&ses
->se_perclnt
);
1086 nfsd4_put_session_locked(ses
);
1088 free_svc_cred(&clp
->cl_cred
);
1089 kfree(clp
->cl_name
.data
);
1094 release_session_client(struct nfsd4_session
*session
)
1096 struct nfs4_client
*clp
= session
->se_client
;
1098 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
1100 if (is_client_expired(clp
)) {
1102 session
->se_client
= NULL
;
1104 renew_client_locked(clp
);
1105 spin_unlock(&client_lock
);
1108 /* must be called under the client_lock */
1110 unhash_client_locked(struct nfs4_client
*clp
)
1112 struct nfsd4_session
*ses
;
1114 mark_client_expired(clp
);
1115 list_del(&clp
->cl_lru
);
1116 spin_lock(&clp
->cl_lock
);
1117 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1118 list_del_init(&ses
->se_hash
);
1119 spin_unlock(&clp
->cl_lock
);
1123 expire_client(struct nfs4_client
*clp
)
1125 struct nfs4_openowner
*oo
;
1126 struct nfs4_delegation
*dp
;
1127 struct list_head reaplist
;
1129 INIT_LIST_HEAD(&reaplist
);
1130 spin_lock(&recall_lock
);
1131 while (!list_empty(&clp
->cl_delegations
)) {
1132 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1133 list_del_init(&dp
->dl_perclnt
);
1134 list_move(&dp
->dl_recall_lru
, &reaplist
);
1136 spin_unlock(&recall_lock
);
1137 while (!list_empty(&reaplist
)) {
1138 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1139 unhash_delegation(dp
);
1141 while (!list_empty(&clp
->cl_openowners
)) {
1142 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1143 release_openowner(oo
);
1145 nfsd4_shutdown_callback(clp
);
1146 if (clp
->cl_cb_conn
.cb_xprt
)
1147 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1148 list_del(&clp
->cl_idhash
);
1149 list_del(&clp
->cl_strhash
);
1150 spin_lock(&client_lock
);
1151 unhash_client_locked(clp
);
1152 if (atomic_read(&clp
->cl_refcount
) == 0)
1154 spin_unlock(&client_lock
);
1157 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1159 memcpy(target
->cl_verifier
.data
, source
->data
,
1160 sizeof(target
->cl_verifier
.data
));
1163 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1165 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1166 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1169 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1171 if (source
->cr_principal
) {
1172 target
->cr_principal
=
1173 kstrdup(source
->cr_principal
, GFP_KERNEL
);
1174 if (target
->cr_principal
== NULL
)
1177 target
->cr_principal
= NULL
;
1178 target
->cr_flavor
= source
->cr_flavor
;
1179 target
->cr_uid
= source
->cr_uid
;
1180 target
->cr_gid
= source
->cr_gid
;
1181 target
->cr_group_info
= source
->cr_group_info
;
1182 get_group_info(target
->cr_group_info
);
1186 static int same_name(const char *n1
, const char *n2
)
1188 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1192 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1194 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1198 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1200 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1203 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
1207 if (g1
->ngroups
!= g2
->ngroups
)
1209 for (i
=0; i
<g1
->ngroups
; i
++)
1210 if (GROUP_AT(g1
, i
) != GROUP_AT(g2
, i
))
1216 * RFC 3530 language requires clid_inuse be returned when the
1217 * "principal" associated with a requests differs from that previously
1218 * used. We use uid, gid's, and gss principal string as our best
1219 * approximation. We also don't want to allow non-gss use of a client
1220 * established using gss: in theory cr_principal should catch that
1221 * change, but in practice cr_principal can be null even in the gss case
1222 * since gssd doesn't always pass down a principal string.
1224 static bool is_gss_cred(struct svc_cred
*cr
)
1226 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1227 return (cr
->cr_flavor
> RPC_AUTH_MAXFLAVOR
);
1232 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1234 if ((is_gss_cred(cr1
) != is_gss_cred(cr2
))
1235 || (cr1
->cr_uid
!= cr2
->cr_uid
)
1236 || (cr1
->cr_gid
!= cr2
->cr_gid
)
1237 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
1239 if (cr1
->cr_principal
== cr2
->cr_principal
)
1241 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
1243 return 0 == strcmp(cr1
->cr_principal
, cr2
->cr_principal
);
1246 static void gen_clid(struct nfs4_client
*clp
)
1248 static u32 current_clientid
= 1;
1249 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
1251 clp
->cl_clientid
.cl_boot
= nn
->boot_time
;
1252 clp
->cl_clientid
.cl_id
= current_clientid
++;
1255 static void gen_confirm(struct nfs4_client
*clp
)
1260 verf
[0] = (__be32
)get_seconds();
1261 verf
[1] = (__be32
)i
++;
1262 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
1265 static struct nfs4_stid
*find_stateid(struct nfs4_client
*cl
, stateid_t
*t
)
1267 return idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1270 static struct nfs4_stid
*find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1272 struct nfs4_stid
*s
;
1274 s
= find_stateid(cl
, t
);
1277 if (typemask
& s
->sc_type
)
1282 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1283 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1285 struct nfs4_client
*clp
;
1286 struct sockaddr
*sa
= svc_addr(rqstp
);
1289 clp
= alloc_client(name
);
1293 INIT_LIST_HEAD(&clp
->cl_sessions
);
1294 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1296 spin_lock(&client_lock
);
1298 spin_unlock(&client_lock
);
1301 idr_init(&clp
->cl_stateids
);
1302 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1303 atomic_set(&clp
->cl_refcount
, 0);
1304 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1305 INIT_LIST_HEAD(&clp
->cl_idhash
);
1306 INIT_LIST_HEAD(&clp
->cl_strhash
);
1307 INIT_LIST_HEAD(&clp
->cl_openowners
);
1308 INIT_LIST_HEAD(&clp
->cl_delegations
);
1309 INIT_LIST_HEAD(&clp
->cl_lru
);
1310 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1311 spin_lock_init(&clp
->cl_lock
);
1312 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1313 clp
->cl_time
= get_seconds();
1314 clear_bit(0, &clp
->cl_cb_slot_busy
);
1315 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1316 copy_verf(clp
, verf
);
1317 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1319 clp
->cl_cb_session
= NULL
;
1324 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1326 unsigned int idhashval
;
1328 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1329 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1330 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1335 move_to_confirmed(struct nfs4_client
*clp
)
1337 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1338 unsigned int strhashval
;
1340 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1341 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1342 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1343 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1347 static struct nfs4_client
*
1348 find_confirmed_client(clientid_t
*clid
, bool sessions
)
1350 struct nfs4_client
*clp
;
1351 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1353 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1354 if (same_clid(&clp
->cl_clientid
, clid
)) {
1355 if ((bool)clp
->cl_minorversion
!= sessions
)
1364 static struct nfs4_client
*
1365 find_unconfirmed_client(clientid_t
*clid
, bool sessions
)
1367 struct nfs4_client
*clp
;
1368 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1370 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1371 if (same_clid(&clp
->cl_clientid
, clid
)) {
1372 if ((bool)clp
->cl_minorversion
!= sessions
)
1380 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1382 return clp
->cl_exchange_flags
!= 0;
1385 static struct nfs4_client
*
1386 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1388 struct nfs4_client
*clp
;
1390 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1391 if (same_name(clp
->cl_recdir
, dname
))
1397 static struct nfs4_client
*
1398 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1400 struct nfs4_client
*clp
;
1402 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1403 if (same_name(clp
->cl_recdir
, dname
))
1410 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1412 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1413 struct sockaddr
*sa
= svc_addr(rqstp
);
1414 u32 scopeid
= rpc_get_scope_id(sa
);
1415 unsigned short expected_family
;
1417 /* Currently, we only support tcp and tcp6 for the callback channel */
1418 if (se
->se_callback_netid_len
== 3 &&
1419 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1420 expected_family
= AF_INET
;
1421 else if (se
->se_callback_netid_len
== 4 &&
1422 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1423 expected_family
= AF_INET6
;
1427 conn
->cb_addrlen
= rpc_uaddr2sockaddr(&init_net
, se
->se_callback_addr_val
,
1428 se
->se_callback_addr_len
,
1429 (struct sockaddr
*)&conn
->cb_addr
,
1430 sizeof(conn
->cb_addr
));
1432 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1435 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1436 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1438 conn
->cb_prog
= se
->se_callback_prog
;
1439 conn
->cb_ident
= se
->se_callback_ident
;
1440 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
1443 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1444 conn
->cb_addrlen
= 0;
1445 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1446 "will not receive delegations\n",
1447 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1453 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1456 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1458 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1461 dprintk("--> %s slot %p\n", __func__
, slot
);
1463 slot
->sl_opcnt
= resp
->opcnt
;
1464 slot
->sl_status
= resp
->cstate
.status
;
1466 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
1467 if (nfsd4_not_cached(resp
)) {
1468 slot
->sl_datalen
= 0;
1471 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1472 base
= (char *)resp
->cstate
.datap
-
1473 (char *)resp
->xbuf
->head
[0].iov_base
;
1474 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1476 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1481 * Encode the replay sequence operation from the slot values.
1482 * If cachethis is FALSE encode the uncached rep error on the next
1483 * operation which sets resp->p and increments resp->opcnt for
1484 * nfs4svc_encode_compoundres.
1488 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1489 struct nfsd4_compoundres
*resp
)
1491 struct nfsd4_op
*op
;
1492 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1494 /* Encode the replayed sequence operation */
1495 op
= &args
->ops
[resp
->opcnt
- 1];
1496 nfsd4_encode_operation(resp
, op
);
1498 /* Return nfserr_retry_uncached_rep in next operation. */
1499 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
1500 op
= &args
->ops
[resp
->opcnt
++];
1501 op
->status
= nfserr_retry_uncached_rep
;
1502 nfsd4_encode_operation(resp
, op
);
1508 * The sequence operation is not cached because we can use the slot and
1512 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1513 struct nfsd4_sequence
*seq
)
1515 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1518 dprintk("--> %s slot %p\n", __func__
, slot
);
1520 /* Either returns 0 or nfserr_retry_uncached */
1521 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1522 if (status
== nfserr_retry_uncached_rep
)
1525 /* The sequence operation has been encoded, cstate->datap set. */
1526 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1528 resp
->opcnt
= slot
->sl_opcnt
;
1529 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1530 status
= slot
->sl_status
;
1536 * Set the exchange_id flags returned by the server.
1539 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1541 /* pNFS is not supported */
1542 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1544 /* Referrals are supported, Migration is not. */
1545 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1547 /* set the wire flags to return to client. */
1548 clid
->flags
= new->cl_exchange_flags
;
1551 static bool client_has_state(struct nfs4_client
*clp
)
1554 * Note clp->cl_openowners check isn't quite right: there's no
1555 * need to count owners without stateid's.
1557 * Also note we should probably be using this in 4.0 case too.
1559 return !list_empty(&clp
->cl_openowners
)
1560 || !list_empty(&clp
->cl_delegations
)
1561 || !list_empty(&clp
->cl_sessions
);
1565 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1566 struct nfsd4_compound_state
*cstate
,
1567 struct nfsd4_exchange_id
*exid
)
1569 struct nfs4_client
*unconf
, *conf
, *new;
1571 unsigned int strhashval
;
1572 char dname
[HEXDIR_LEN
];
1573 char addr_str
[INET6_ADDRSTRLEN
];
1574 nfs4_verifier verf
= exid
->verifier
;
1575 struct sockaddr
*sa
= svc_addr(rqstp
);
1576 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
1578 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1579 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1580 "ip_addr=%s flags %x, spa_how %d\n",
1581 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1582 addr_str
, exid
->flags
, exid
->spa_how
);
1584 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
1585 return nfserr_inval
;
1587 /* Currently only support SP4_NONE */
1588 switch (exid
->spa_how
) {
1592 return nfserr_serverfault
;
1594 BUG(); /* checked by xdr code */
1596 return nfserr_serverfault
; /* no excuse :-/ */
1599 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1604 strhashval
= clientstr_hashval(dname
);
1606 /* Cases below refer to rfc 5661 section 18.35.4: */
1608 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1610 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
1611 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
1614 if (!clp_used_exchangeid(conf
)) { /* buggy client */
1615 status
= nfserr_inval
;
1618 if (!creds_match
) { /* case 9 */
1619 status
= nfserr_perm
;
1622 if (!verfs_match
) { /* case 8 */
1623 status
= nfserr_not_same
;
1627 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1631 if (!creds_match
) { /* case 3 */
1632 if (client_has_state(conf
)) {
1633 status
= nfserr_clid_inuse
;
1636 expire_client(conf
);
1639 if (verfs_match
) { /* case 2 */
1640 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1644 /* case 5, client reboot */
1648 if (update
) { /* case 7 */
1649 status
= nfserr_noent
;
1653 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1654 if (unconf
) /* case 4, possible retry or client restart */
1655 expire_client(unconf
);
1657 /* case 1 (normal case) */
1659 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1661 status
= nfserr_jukebox
;
1664 new->cl_minorversion
= 1;
1667 add_to_unconfirmed(new, strhashval
);
1669 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1670 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1672 exid
->seqid
= new->cl_cs_slot
.sl_seqid
+ 1;
1673 nfsd4_set_ex_flags(new, exid
);
1675 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1676 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1680 nfs4_unlock_state();
1685 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1687 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1690 /* The slot is in use, and no response has been sent. */
1692 if (seqid
== slot_seqid
)
1693 return nfserr_jukebox
;
1695 return nfserr_seq_misordered
;
1697 /* Note unsigned 32-bit arithmetic handles wraparound: */
1698 if (likely(seqid
== slot_seqid
+ 1))
1700 if (seqid
== slot_seqid
)
1701 return nfserr_replay_cache
;
1702 return nfserr_seq_misordered
;
1706 * Cache the create session result into the create session single DRC
1707 * slot cache by saving the xdr structure. sl_seqid has been set.
1708 * Do this for solo or embedded create session operations.
1711 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1712 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
1714 slot
->sl_status
= nfserr
;
1715 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1719 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1720 struct nfsd4_clid_slot
*slot
)
1722 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1723 return slot
->sl_status
;
1726 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1727 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1728 1 + /* MIN tag is length with zero, only length */ \
1729 3 + /* version, opcount, opcode */ \
1730 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1731 /* seqid, slotID, slotID, cache */ \
1732 4 ) * sizeof(__be32))
1734 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1735 2 + /* verifier: AUTH_NULL, length 0 */\
1737 1 + /* MIN tag is length with zero, only length */ \
1738 3 + /* opcount, opcode, opstatus*/ \
1739 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1740 /* seqid, slotID, slotID, slotID, status */ \
1741 5 ) * sizeof(__be32))
1743 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1745 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1746 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1750 nfsd4_create_session(struct svc_rqst
*rqstp
,
1751 struct nfsd4_compound_state
*cstate
,
1752 struct nfsd4_create_session
*cr_ses
)
1754 struct sockaddr
*sa
= svc_addr(rqstp
);
1755 struct nfs4_client
*conf
, *unconf
;
1756 struct nfsd4_session
*new;
1757 struct nfsd4_conn
*conn
;
1758 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1761 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1762 return nfserr_inval
;
1763 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1764 return nfserr_toosmall
;
1765 new = alloc_session(&cr_ses
->fore_channel
);
1767 return nfserr_jukebox
;
1768 status
= nfserr_jukebox
;
1769 conn
= alloc_conn_from_crses(rqstp
, cr_ses
);
1771 goto out_free_session
;
1774 unconf
= find_unconfirmed_client(&cr_ses
->clientid
, true);
1775 conf
= find_confirmed_client(&cr_ses
->clientid
, true);
1778 cs_slot
= &conf
->cl_cs_slot
;
1779 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1780 if (status
== nfserr_replay_cache
) {
1781 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1783 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1784 status
= nfserr_seq_misordered
;
1787 } else if (unconf
) {
1789 struct nfs4_client
*old
;
1790 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1791 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1792 status
= nfserr_clid_inuse
;
1795 cs_slot
= &unconf
->cl_cs_slot
;
1796 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1798 /* an unconfirmed replay returns misordered */
1799 status
= nfserr_seq_misordered
;
1802 hash
= clientstr_hashval(unconf
->cl_recdir
);
1803 old
= find_confirmed_client_by_str(unconf
->cl_recdir
, hash
);
1806 move_to_confirmed(unconf
);
1809 status
= nfserr_stale_clientid
;
1814 * We do not support RDMA or persistent sessions
1816 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1817 cr_ses
->flags
&= ~SESSION4_RDMA
;
1819 init_session(rqstp
, new, conf
, cr_ses
);
1820 nfsd4_init_conn(rqstp
, conn
, new);
1822 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1823 NFS4_MAX_SESSIONID_LEN
);
1824 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1825 sizeof(struct nfsd4_channel_attrs
));
1826 cs_slot
->sl_seqid
++;
1827 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1829 /* cache solo and embedded create sessions under the state lock */
1830 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1832 nfs4_unlock_state();
1833 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1838 __free_session(new);
1842 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1844 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1845 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1847 return argp
->opcnt
== resp
->opcnt
;
1850 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1853 case NFS4_CDFC4_FORE
:
1854 case NFS4_CDFC4_BACK
:
1856 case NFS4_CDFC4_FORE_OR_BOTH
:
1857 case NFS4_CDFC4_BACK_OR_BOTH
:
1858 *dir
= NFS4_CDFC4_BOTH
;
1861 return nfserr_inval
;
1864 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1865 struct nfsd4_compound_state
*cstate
,
1866 struct nfsd4_bind_conn_to_session
*bcts
)
1869 struct nfsd4_conn
*conn
;
1871 if (!nfsd4_last_compound_op(rqstp
))
1872 return nfserr_not_only_op
;
1873 spin_lock(&client_lock
);
1874 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1875 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1876 * client_lock iself: */
1877 if (cstate
->session
) {
1878 nfsd4_get_session(cstate
->session
);
1879 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1881 spin_unlock(&client_lock
);
1882 if (!cstate
->session
)
1883 return nfserr_badsession
;
1885 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1888 conn
= alloc_conn(rqstp
, bcts
->dir
);
1890 return nfserr_jukebox
;
1891 nfsd4_init_conn(rqstp
, conn
, cstate
->session
);
1895 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1899 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1903 nfsd4_destroy_session(struct svc_rqst
*r
,
1904 struct nfsd4_compound_state
*cstate
,
1905 struct nfsd4_destroy_session
*sessionid
)
1907 struct nfsd4_session
*ses
;
1908 __be32 status
= nfserr_badsession
;
1911 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1912 * - Should we return nfserr_back_chan_busy if waiting for
1913 * callbacks on to-be-destroyed session?
1914 * - Do we need to clear any callback info from previous session?
1917 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1918 if (!nfsd4_last_compound_op(r
))
1919 return nfserr_not_only_op
;
1921 dump_sessionid(__func__
, &sessionid
->sessionid
);
1922 spin_lock(&client_lock
);
1923 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1925 spin_unlock(&client_lock
);
1929 unhash_session(ses
);
1930 spin_unlock(&client_lock
);
1933 nfsd4_probe_callback_sync(ses
->se_client
);
1934 nfs4_unlock_state();
1936 spin_lock(&client_lock
);
1937 nfsd4_del_conns(ses
);
1938 nfsd4_put_session_locked(ses
);
1939 spin_unlock(&client_lock
);
1942 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1946 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1948 struct nfsd4_conn
*c
;
1950 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1951 if (c
->cn_xprt
== xpt
) {
1958 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1960 struct nfs4_client
*clp
= ses
->se_client
;
1961 struct nfsd4_conn
*c
;
1964 spin_lock(&clp
->cl_lock
);
1965 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1967 spin_unlock(&clp
->cl_lock
);
1971 __nfsd4_hash_conn(new, ses
);
1972 spin_unlock(&clp
->cl_lock
);
1973 ret
= nfsd4_register_conn(new);
1975 /* oops; xprt is already down: */
1976 nfsd4_conn_lost(&new->cn_xpt_user
);
1980 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1982 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1984 return args
->opcnt
> session
->se_fchannel
.maxops
;
1987 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1988 struct nfsd4_session
*session
)
1990 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1992 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1996 nfsd4_sequence(struct svc_rqst
*rqstp
,
1997 struct nfsd4_compound_state
*cstate
,
1998 struct nfsd4_sequence
*seq
)
2000 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2001 struct nfsd4_session
*session
;
2002 struct nfsd4_slot
*slot
;
2003 struct nfsd4_conn
*conn
;
2006 if (resp
->opcnt
!= 1)
2007 return nfserr_sequence_pos
;
2010 * Will be either used or freed by nfsd4_sequence_check_conn
2013 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
2015 return nfserr_jukebox
;
2017 spin_lock(&client_lock
);
2018 status
= nfserr_badsession
;
2019 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
2023 status
= nfserr_too_many_ops
;
2024 if (nfsd4_session_too_many_ops(rqstp
, session
))
2027 status
= nfserr_req_too_big
;
2028 if (nfsd4_request_too_big(rqstp
, session
))
2031 status
= nfserr_badslot
;
2032 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
2035 slot
= session
->se_slots
[seq
->slotid
];
2036 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
2038 /* We do not negotiate the number of slots yet, so set the
2039 * maxslots to the session maxreqs which is used to encode
2040 * sr_highest_slotid and the sr_target_slot id to maxslots */
2041 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
2043 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
2044 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
2045 if (status
== nfserr_replay_cache
) {
2046 status
= nfserr_seq_misordered
;
2047 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
2049 cstate
->slot
= slot
;
2050 cstate
->session
= session
;
2051 /* Return the cached reply status and set cstate->status
2052 * for nfsd4_proc_compound processing */
2053 status
= nfsd4_replay_cache_entry(resp
, seq
);
2054 cstate
->status
= nfserr_replay_cache
;
2060 nfsd4_sequence_check_conn(conn
, session
);
2063 /* Success! bump slot seqid */
2064 slot
->sl_seqid
= seq
->seqid
;
2065 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
2067 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
2069 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
2071 cstate
->slot
= slot
;
2072 cstate
->session
= session
;
2075 /* Hold a session reference until done processing the compound. */
2076 if (cstate
->session
) {
2077 struct nfs4_client
*clp
= session
->se_client
;
2079 nfsd4_get_session(cstate
->session
);
2080 atomic_inc(&clp
->cl_refcount
);
2081 switch (clp
->cl_cb_state
) {
2083 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
2085 case NFSD4_CB_FAULT
:
2086 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
2089 seq
->status_flags
= 0;
2093 spin_unlock(&client_lock
);
2094 dprintk("%s: return %d\n", __func__
, ntohl(status
));
2099 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
2101 struct nfs4_client
*conf
, *unconf
, *clp
;
2105 unconf
= find_unconfirmed_client(&dc
->clientid
, true);
2106 conf
= find_confirmed_client(&dc
->clientid
, true);
2111 if (!is_client_expired(conf
) && client_has_state(conf
)) {
2112 status
= nfserr_clientid_busy
;
2116 /* rfc5661 18.50.3 */
2117 if (cstate
->session
&& conf
== cstate
->session
->se_client
) {
2118 status
= nfserr_clientid_busy
;
2124 status
= nfserr_stale_clientid
;
2130 nfs4_unlock_state();
2131 dprintk("%s return %d\n", __func__
, ntohl(status
));
2136 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2140 if (rc
->rca_one_fs
) {
2141 if (!cstate
->current_fh
.fh_dentry
)
2142 return nfserr_nofilehandle
;
2144 * We don't take advantage of the rca_one_fs case.
2145 * That's OK, it's optional, we can safely ignore it.
2151 status
= nfserr_complete_already
;
2152 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
2153 &cstate
->session
->se_client
->cl_flags
))
2156 status
= nfserr_stale_clientid
;
2157 if (is_client_expired(cstate
->session
->se_client
))
2159 * The following error isn't really legal.
2160 * But we only get here if the client just explicitly
2161 * destroyed the client. Surely it no longer cares what
2162 * error it gets back on an operation for the dead
2168 nfsd4_client_record_create(cstate
->session
->se_client
);
2170 nfs4_unlock_state();
2175 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2176 struct nfsd4_setclientid
*setclid
)
2178 struct xdr_netobj clname
= setclid
->se_name
;
2179 nfs4_verifier clverifier
= setclid
->se_verf
;
2180 unsigned int strhashval
;
2181 struct nfs4_client
*conf
, *unconf
, *new;
2183 char dname
[HEXDIR_LEN
];
2185 status
= nfs4_make_rec_clidname(dname
, &clname
);
2189 strhashval
= clientstr_hashval(dname
);
2191 /* Cases below refer to rfc 3530 section 14.2.33: */
2193 conf
= find_confirmed_client_by_str(dname
, strhashval
);
2196 status
= nfserr_clid_inuse
;
2197 if (clp_used_exchangeid(conf
))
2199 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
2200 char addr_str
[INET6_ADDRSTRLEN
];
2201 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
2203 dprintk("NFSD: setclientid: string in use by client "
2204 "at %s\n", addr_str
);
2208 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
2210 expire_client(unconf
);
2211 status
= nfserr_jukebox
;
2212 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2215 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
))
2216 /* case 1: probable callback update */
2217 copy_clid(new, conf
);
2218 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2220 new->cl_minorversion
= 0;
2221 gen_callback(new, setclid
, rqstp
);
2222 add_to_unconfirmed(new, strhashval
);
2223 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2224 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2225 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2228 nfs4_unlock_state();
2234 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2235 struct nfsd4_compound_state
*cstate
,
2236 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2238 struct nfs4_client
*conf
, *unconf
;
2239 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2240 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2242 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
2244 if (STALE_CLIENTID(clid
, nn
))
2245 return nfserr_stale_clientid
;
2248 conf
= find_confirmed_client(clid
, false);
2249 unconf
= find_unconfirmed_client(clid
, false);
2251 * We try hard to give out unique clientid's, so if we get an
2252 * attempt to confirm the same clientid with a different cred,
2253 * there's a bug somewhere. Let's charitably assume it's our
2256 status
= nfserr_serverfault
;
2257 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
2259 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2261 /* cases below refer to rfc 3530 section 14.2.34: */
2262 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
2263 if (conf
&& !unconf
) /* case 2: probable retransmit */
2265 else /* case 4: client hasn't noticed we rebooted yet? */
2266 status
= nfserr_stale_clientid
;
2270 if (conf
) { /* case 1: callback update */
2271 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2272 nfsd4_probe_callback(conf
);
2273 expire_client(unconf
);
2274 } else { /* case 3: normal case; new or rebooted client */
2275 unsigned int hash
= clientstr_hashval(unconf
->cl_recdir
);
2277 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
, hash
);
2279 nfsd4_client_record_remove(conf
);
2280 expire_client(conf
);
2282 move_to_confirmed(unconf
);
2283 nfsd4_probe_callback(unconf
);
2286 nfs4_unlock_state();
2290 static struct nfs4_file
*nfsd4_alloc_file(void)
2292 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2295 /* OPEN Share state helper functions */
2296 static void nfsd4_init_file(struct nfs4_file
*fp
, struct inode
*ino
)
2298 unsigned int hashval
= file_hashval(ino
);
2300 atomic_set(&fp
->fi_ref
, 1);
2301 INIT_LIST_HEAD(&fp
->fi_hash
);
2302 INIT_LIST_HEAD(&fp
->fi_stateids
);
2303 INIT_LIST_HEAD(&fp
->fi_delegations
);
2304 fp
->fi_inode
= igrab(ino
);
2305 fp
->fi_had_conflict
= false;
2306 fp
->fi_lease
= NULL
;
2307 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2308 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2309 spin_lock(&recall_lock
);
2310 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2311 spin_unlock(&recall_lock
);
2315 nfsd4_free_slab(struct kmem_cache
**slab
)
2319 kmem_cache_destroy(*slab
);
2324 nfsd4_free_slabs(void)
2326 nfsd4_free_slab(&openowner_slab
);
2327 nfsd4_free_slab(&lockowner_slab
);
2328 nfsd4_free_slab(&file_slab
);
2329 nfsd4_free_slab(&stateid_slab
);
2330 nfsd4_free_slab(&deleg_slab
);
2334 nfsd4_init_slabs(void)
2336 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2337 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2338 if (openowner_slab
== NULL
)
2340 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2341 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2342 if (lockowner_slab
== NULL
)
2344 file_slab
= kmem_cache_create("nfsd4_files",
2345 sizeof(struct nfs4_file
), 0, 0, NULL
);
2346 if (file_slab
== NULL
)
2348 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2349 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
2350 if (stateid_slab
== NULL
)
2352 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2353 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2354 if (deleg_slab
== NULL
)
2359 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2363 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2365 kfree(oo
->oo_owner
.so_owner
.data
);
2366 kmem_cache_free(openowner_slab
, oo
);
2369 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2371 kfree(lo
->lo_owner
.so_owner
.data
);
2372 kmem_cache_free(lockowner_slab
, lo
);
2375 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2377 rp
->rp_status
= nfserr_serverfault
;
2379 rp
->rp_buf
= rp
->rp_ibuf
;
2382 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2384 struct nfs4_stateowner
*sop
;
2386 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2390 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2391 if (!sop
->so_owner
.data
) {
2392 kmem_cache_free(slab
, sop
);
2395 sop
->so_owner
.len
= owner
->len
;
2397 INIT_LIST_HEAD(&sop
->so_stateids
);
2398 sop
->so_client
= clp
;
2399 init_nfs4_replay(&sop
->so_replay
);
2403 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2405 list_add(&oo
->oo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2406 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2409 static struct nfs4_openowner
*
2410 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2411 struct nfs4_openowner
*oo
;
2413 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2416 oo
->oo_owner
.so_is_open_owner
= 1;
2417 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2418 oo
->oo_flags
= NFS4_OO_NEW
;
2420 oo
->oo_last_closed_stid
= NULL
;
2421 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2422 hash_openowner(oo
, clp
, strhashval
);
2426 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2427 struct nfs4_openowner
*oo
= open
->op_openowner
;
2428 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
2430 init_stid(&stp
->st_stid
, clp
, NFS4_OPEN_STID
);
2431 INIT_LIST_HEAD(&stp
->st_lockowners
);
2432 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2433 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2434 stp
->st_stateowner
= &oo
->oo_owner
;
2437 stp
->st_access_bmap
= 0;
2438 stp
->st_deny_bmap
= 0;
2439 set_access(open
->op_share_access
, stp
);
2440 set_deny(open
->op_share_deny
, stp
);
2441 stp
->st_openstp
= NULL
;
2445 move_to_close_lru(struct nfs4_openowner
*oo
)
2447 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2449 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2450 oo
->oo_time
= get_seconds();
2454 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2457 return (sop
->so_owner
.len
== owner
->len
) &&
2458 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2459 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2462 static struct nfs4_openowner
*
2463 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
, bool sessions
)
2465 struct nfs4_stateowner
*so
;
2466 struct nfs4_openowner
*oo
;
2467 struct nfs4_client
*clp
;
2469 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2470 if (!so
->so_is_open_owner
)
2472 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
)) {
2474 clp
= oo
->oo_owner
.so_client
;
2475 if ((bool)clp
->cl_minorversion
!= sessions
)
2477 renew_client(oo
->oo_owner
.so_client
);
2484 /* search file_hashtbl[] for file */
2485 static struct nfs4_file
*
2486 find_file(struct inode
*ino
)
2488 unsigned int hashval
= file_hashval(ino
);
2489 struct nfs4_file
*fp
;
2491 spin_lock(&recall_lock
);
2492 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2493 if (fp
->fi_inode
== ino
) {
2495 spin_unlock(&recall_lock
);
2499 spin_unlock(&recall_lock
);
2504 * Called to check deny when READ with all zero stateid or
2505 * WRITE with all zero or all one stateid
2508 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2510 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2511 struct nfs4_file
*fp
;
2512 struct nfs4_ol_stateid
*stp
;
2515 dprintk("NFSD: nfs4_share_conflict\n");
2517 fp
= find_file(ino
);
2520 ret
= nfserr_locked
;
2521 /* Search for conflicting share reservations */
2522 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2523 if (test_deny(deny_type
, stp
) ||
2524 test_deny(NFS4_SHARE_DENY_BOTH
, stp
))
2533 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2535 /* We're assuming the state code never drops its reference
2536 * without first removing the lease. Since we're in this lease
2537 * callback (and since the lease code is serialized by the kernel
2538 * lock) we know the server hasn't removed the lease yet, we know
2539 * it's safe to take a reference: */
2540 atomic_inc(&dp
->dl_count
);
2542 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2544 /* only place dl_time is set. protected by lock_flocks*/
2545 dp
->dl_time
= get_seconds();
2547 nfsd4_cb_recall(dp
);
2550 /* Called from break_lease() with lock_flocks() held. */
2551 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2553 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2554 struct nfs4_delegation
*dp
;
2557 /* We assume break_lease is only called once per lease: */
2558 BUG_ON(fp
->fi_had_conflict
);
2560 * We don't want the locks code to timeout the lease for us;
2561 * we'll remove it ourself if a delegation isn't returned
2564 fl
->fl_break_time
= 0;
2566 spin_lock(&recall_lock
);
2567 fp
->fi_had_conflict
= true;
2568 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2569 nfsd_break_one_deleg(dp
);
2570 spin_unlock(&recall_lock
);
2574 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2577 return lease_modify(onlist
, arg
);
2582 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2583 .lm_break
= nfsd_break_deleg_cb
,
2584 .lm_change
= nfsd_change_deleg_cb
,
2587 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2589 if (nfsd4_has_session(cstate
))
2591 if (seqid
== so
->so_seqid
- 1)
2592 return nfserr_replay_me
;
2593 if (seqid
== so
->so_seqid
)
2595 return nfserr_bad_seqid
;
2599 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2600 struct nfsd4_open
*open
)
2602 clientid_t
*clientid
= &open
->op_clientid
;
2603 struct nfs4_client
*clp
= NULL
;
2604 unsigned int strhashval
;
2605 struct nfs4_openowner
*oo
= NULL
;
2607 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
2609 if (STALE_CLIENTID(&open
->op_clientid
, nn
))
2610 return nfserr_stale_clientid
;
2612 * In case we need it later, after we've already created the
2613 * file and don't want to risk a further failure:
2615 open
->op_file
= nfsd4_alloc_file();
2616 if (open
->op_file
== NULL
)
2617 return nfserr_jukebox
;
2619 strhashval
= ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2620 oo
= find_openstateowner_str(strhashval
, open
, cstate
->minorversion
);
2621 open
->op_openowner
= oo
;
2623 clp
= find_confirmed_client(clientid
, cstate
->minorversion
);
2625 return nfserr_expired
;
2628 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
2629 /* Replace unconfirmed owners without checking for replay. */
2630 clp
= oo
->oo_owner
.so_client
;
2631 release_openowner(oo
);
2632 open
->op_openowner
= NULL
;
2635 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2638 clp
= oo
->oo_owner
.so_client
;
2641 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2643 return nfserr_jukebox
;
2644 open
->op_openowner
= oo
;
2646 open
->op_stp
= nfs4_alloc_stateid(clp
);
2648 return nfserr_jukebox
;
2652 static inline __be32
2653 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2655 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2656 return nfserr_openmode
;
2661 static int share_access_to_flags(u32 share_access
)
2663 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2666 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
2668 struct nfs4_stid
*ret
;
2670 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
2673 return delegstateid(ret
);
2676 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
2678 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
2679 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
2683 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2684 struct nfs4_delegation
**dp
)
2687 __be32 status
= nfserr_bad_stateid
;
2689 *dp
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
2692 flags
= share_access_to_flags(open
->op_share_access
);
2693 status
= nfs4_check_delegmode(*dp
, flags
);
2697 if (!nfsd4_is_deleg_cur(open
))
2701 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2706 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
**stpp
)
2708 struct nfs4_ol_stateid
*local
;
2709 struct nfs4_openowner
*oo
= open
->op_openowner
;
2711 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2712 /* ignore lock owners */
2713 if (local
->st_stateowner
->so_is_open_owner
== 0)
2715 /* remember if we have seen this open owner */
2716 if (local
->st_stateowner
== &oo
->oo_owner
)
2718 /* check for conflicting share reservations */
2719 if (!test_share(local
, open
))
2720 return nfserr_share_denied
;
2725 static inline int nfs4_access_to_access(u32 nfs4_access
)
2729 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2730 flags
|= NFSD_MAY_READ
;
2731 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2732 flags
|= NFSD_MAY_WRITE
;
2736 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2737 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2740 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2741 int access
= nfs4_access_to_access(open
->op_share_access
);
2743 if (!fp
->fi_fds
[oflag
]) {
2744 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2745 &fp
->fi_fds
[oflag
]);
2749 nfs4_file_get_access(fp
, oflag
);
2754 static inline __be32
2755 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2756 struct nfsd4_open
*open
)
2758 struct iattr iattr
= {
2759 .ia_valid
= ATTR_SIZE
,
2762 if (!open
->op_truncate
)
2764 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2765 return nfserr_inval
;
2766 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2770 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
, struct svc_fh
*cur_fh
, struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
)
2772 u32 op_share_access
= open
->op_share_access
;
2776 new_access
= !test_access(op_share_access
, stp
);
2778 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2782 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2785 int oflag
= nfs4_access_to_omode(op_share_access
);
2786 nfs4_file_put_access(fp
, oflag
);
2790 /* remember the open */
2791 set_access(op_share_access
, stp
);
2792 set_deny(open
->op_share_deny
, stp
);
2799 nfs4_set_claim_prev(struct nfsd4_open
*open
, bool has_session
)
2801 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2804 /* Should we give out recallable state?: */
2805 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2807 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2810 * In the sessions case, since we don't have to establish a
2811 * separate connection for callbacks, we assume it's OK
2812 * until we hear otherwise:
2814 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2817 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2819 struct file_lock
*fl
;
2821 fl
= locks_alloc_lock();
2824 locks_init_lock(fl
);
2825 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2826 fl
->fl_flags
= FL_LEASE
;
2827 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2828 fl
->fl_end
= OFFSET_MAX
;
2829 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2830 fl
->fl_pid
= current
->tgid
;
2834 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2836 struct nfs4_file
*fp
= dp
->dl_file
;
2837 struct file_lock
*fl
;
2840 fl
= nfs4_alloc_init_lease(dp
, flag
);
2843 fl
->fl_file
= find_readable_file(fp
);
2844 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2845 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2847 list_del_init(&dp
->dl_perclnt
);
2848 locks_free_lock(fl
);
2852 fp
->fi_deleg_file
= fl
->fl_file
;
2853 get_file(fp
->fi_deleg_file
);
2854 atomic_set(&fp
->fi_delegees
, 1);
2855 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2859 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2861 struct nfs4_file
*fp
= dp
->dl_file
;
2864 return nfs4_setlease(dp
, flag
);
2865 spin_lock(&recall_lock
);
2866 if (fp
->fi_had_conflict
) {
2867 spin_unlock(&recall_lock
);
2870 atomic_inc(&fp
->fi_delegees
);
2871 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2872 spin_unlock(&recall_lock
);
2873 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2877 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
2879 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2880 if (status
== -EAGAIN
)
2881 open
->op_why_no_deleg
= WND4_CONTENTION
;
2883 open
->op_why_no_deleg
= WND4_RESOURCE
;
2884 switch (open
->op_deleg_want
) {
2885 case NFS4_SHARE_WANT_READ_DELEG
:
2886 case NFS4_SHARE_WANT_WRITE_DELEG
:
2887 case NFS4_SHARE_WANT_ANY_DELEG
:
2889 case NFS4_SHARE_WANT_CANCEL
:
2890 open
->op_why_no_deleg
= WND4_CANCELLED
;
2892 case NFS4_SHARE_WANT_NO_DELEG
:
2893 BUG(); /* not supposed to get here */
2899 * Attempt to hand out a delegation.
2902 nfs4_open_delegation(struct net
*net
, struct svc_fh
*fh
,
2903 struct nfsd4_open
*open
, struct nfs4_ol_stateid
*stp
)
2905 struct nfs4_delegation
*dp
;
2906 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2908 int status
= 0, flag
= 0;
2910 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2911 flag
= NFS4_OPEN_DELEGATE_NONE
;
2912 open
->op_recall
= 0;
2913 switch (open
->op_claim_type
) {
2914 case NFS4_OPEN_CLAIM_PREVIOUS
:
2916 open
->op_recall
= 1;
2917 flag
= open
->op_delegate_type
;
2918 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2921 case NFS4_OPEN_CLAIM_NULL
:
2922 /* Let's not give out any delegations till everyone's
2923 * had the chance to reclaim theirs.... */
2924 if (locks_in_grace(net
))
2926 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
2928 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2929 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2931 flag
= NFS4_OPEN_DELEGATE_READ
;
2937 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2940 status
= nfs4_set_delegation(dp
, flag
);
2944 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
2946 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2947 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
2949 open
->op_delegate_type
= flag
;
2950 if (flag
== NFS4_OPEN_DELEGATE_NONE
) {
2951 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
2952 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2953 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2955 /* 4.1 client asking for a delegation? */
2956 if (open
->op_deleg_want
)
2957 nfsd4_open_deleg_none_ext(open
, status
);
2961 nfs4_put_delegation(dp
);
2963 flag
= NFS4_OPEN_DELEGATE_NONE
;
2967 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
2968 struct nfs4_delegation
*dp
)
2970 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
2971 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2972 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2973 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
2974 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
2975 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2976 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2977 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
2979 /* Otherwise the client must be confused wanting a delegation
2980 * it already has, therefore we don't return
2981 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2986 * called with nfs4_lock_state() held.
2989 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2991 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2992 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
2993 struct nfs4_file
*fp
= NULL
;
2994 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2995 struct nfs4_ol_stateid
*stp
= NULL
;
2996 struct nfs4_delegation
*dp
= NULL
;
3000 * Lookup file; if found, lookup stateid and check open request,
3001 * and check for delegations in the process of being recalled.
3002 * If not found, create the nfs4_file struct
3004 fp
= find_file(ino
);
3006 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
3008 status
= nfs4_check_deleg(cl
, fp
, open
, &dp
);
3012 status
= nfserr_bad_stateid
;
3013 if (nfsd4_is_deleg_cur(open
))
3015 status
= nfserr_jukebox
;
3017 open
->op_file
= NULL
;
3018 nfsd4_init_file(fp
, ino
);
3022 * OPEN the file, or upgrade an existing OPEN.
3023 * If truncate fails, the OPEN fails.
3026 /* Stateid was found, this is an OPEN upgrade */
3027 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
3031 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, open
);
3034 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
3038 open
->op_stp
= NULL
;
3039 init_open_stateid(stp
, fp
, open
);
3041 update_stateid(&stp
->st_stid
.sc_stateid
);
3042 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3044 if (nfsd4_has_session(&resp
->cstate
)) {
3045 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3047 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
3048 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3049 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
3055 * Attempt to hand out a delegation. No error return, because the
3056 * OPEN succeeds even if we fail.
3058 nfs4_open_delegation(SVC_NET(rqstp
), current_fh
, open
, stp
);
3062 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
3063 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3065 /* 4.1 client trying to upgrade/downgrade delegation? */
3066 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
3067 open
->op_deleg_want
)
3068 nfsd4_deleg_xgrade_none_ext(open
, dp
);
3072 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
3073 nfs4_set_claim_prev(open
, nfsd4_has_session(&resp
->cstate
));
3075 * To finish the open response, we just need to set the rflags.
3077 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
3078 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
3079 !nfsd4_has_session(&resp
->cstate
))
3080 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
3085 void nfsd4_cleanup_open_state(struct nfsd4_open
*open
, __be32 status
)
3087 if (open
->op_openowner
) {
3088 struct nfs4_openowner
*oo
= open
->op_openowner
;
3090 if (!list_empty(&oo
->oo_owner
.so_stateids
))
3091 list_del_init(&oo
->oo_close_lru
);
3092 if (oo
->oo_flags
& NFS4_OO_NEW
) {
3094 release_openowner(oo
);
3095 open
->op_openowner
= NULL
;
3097 oo
->oo_flags
&= ~NFS4_OO_NEW
;
3101 nfsd4_free_file(open
->op_file
);
3103 free_generic_stateid(open
->op_stp
);
3107 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3110 struct nfs4_client
*clp
;
3112 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
3115 dprintk("process_renew(%08x/%08x): starting\n",
3116 clid
->cl_boot
, clid
->cl_id
);
3117 status
= nfserr_stale_clientid
;
3118 if (STALE_CLIENTID(clid
, nn
))
3120 clp
= find_confirmed_client(clid
, cstate
->minorversion
);
3121 status
= nfserr_expired
;
3123 /* We assume the client took too long to RENEW. */
3124 dprintk("nfsd4_renew: clientid not found!\n");
3127 status
= nfserr_cb_path_down
;
3128 if (!list_empty(&clp
->cl_delegations
)
3129 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
3133 nfs4_unlock_state();
3138 nfsd4_end_grace(struct net
*net
)
3140 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
3142 /* do nothing if grace period already ended */
3143 if (nn
->grace_ended
)
3146 dprintk("NFSD: end of grace period\n");
3147 nn
->grace_ended
= true;
3148 nfsd4_record_grace_done(net
, nn
->boot_time
);
3149 locks_end_grace(&nn
->nfsd4_manager
);
3151 * Now that every NFSv4 client has had the chance to recover and
3152 * to see the (possibly new, possibly shorter) lease time, we
3153 * can safely set the next grace time to the current lease time:
3155 nfsd4_grace
= nfsd4_lease
;
3159 nfs4_laundromat(void)
3161 struct nfs4_client
*clp
;
3162 struct nfs4_openowner
*oo
;
3163 struct nfs4_delegation
*dp
;
3164 struct list_head
*pos
, *next
, reaplist
;
3165 time_t cutoff
= get_seconds() - nfsd4_lease
;
3166 time_t t
, clientid_val
= nfsd4_lease
;
3167 time_t u
, test_val
= nfsd4_lease
;
3171 dprintk("NFSD: laundromat service - starting\n");
3172 nfsd4_end_grace(&init_net
);
3173 INIT_LIST_HEAD(&reaplist
);
3174 spin_lock(&client_lock
);
3175 list_for_each_safe(pos
, next
, &client_lru
) {
3176 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3177 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
3178 t
= clp
->cl_time
- cutoff
;
3179 if (clientid_val
> t
)
3183 if (atomic_read(&clp
->cl_refcount
)) {
3184 dprintk("NFSD: client in use (clientid %08x)\n",
3185 clp
->cl_clientid
.cl_id
);
3188 unhash_client_locked(clp
);
3189 list_add(&clp
->cl_lru
, &reaplist
);
3191 spin_unlock(&client_lock
);
3192 list_for_each_safe(pos
, next
, &reaplist
) {
3193 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3194 dprintk("NFSD: purging unused client (clientid %08x)\n",
3195 clp
->cl_clientid
.cl_id
);
3196 nfsd4_client_record_remove(clp
);
3199 spin_lock(&recall_lock
);
3200 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3201 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3202 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3203 u
= dp
->dl_time
- cutoff
;
3208 list_move(&dp
->dl_recall_lru
, &reaplist
);
3210 spin_unlock(&recall_lock
);
3211 list_for_each_safe(pos
, next
, &reaplist
) {
3212 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3213 unhash_delegation(dp
);
3215 test_val
= nfsd4_lease
;
3216 list_for_each_safe(pos
, next
, &close_lru
) {
3217 oo
= container_of(pos
, struct nfs4_openowner
, oo_close_lru
);
3218 if (time_after((unsigned long)oo
->oo_time
, (unsigned long)cutoff
)) {
3219 u
= oo
->oo_time
- cutoff
;
3224 release_openowner(oo
);
3226 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3227 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3228 nfs4_unlock_state();
3229 return clientid_val
;
3232 static struct workqueue_struct
*laundry_wq
;
3233 static void laundromat_main(struct work_struct
*);
3234 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3237 laundromat_main(struct work_struct
*not_used
)
3241 t
= nfs4_laundromat();
3242 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3243 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3246 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*stp
)
3248 if (fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
)
3249 return nfserr_bad_stateid
;
3254 STALE_STATEID(stateid_t
*stateid
, struct nfsd_net
*nn
)
3256 if (stateid
->si_opaque
.so_clid
.cl_boot
== nn
->boot_time
)
3258 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3259 STATEID_VAL(stateid
));
3264 access_permit_read(struct nfs4_ol_stateid
*stp
)
3266 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
3267 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
3268 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
3272 access_permit_write(struct nfs4_ol_stateid
*stp
)
3274 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
3275 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
3279 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
3281 __be32 status
= nfserr_openmode
;
3283 /* For lock stateid's, we test the parent open, not the lock: */
3284 if (stp
->st_openstp
)
3285 stp
= stp
->st_openstp
;
3286 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
3288 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
3295 static inline __be32
3296 check_special_stateids(struct net
*net
, svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3298 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3300 else if (locks_in_grace(net
)) {
3301 /* Answer in remaining cases depends on existence of
3302 * conflicting state; so we must wait out the grace period. */
3303 return nfserr_grace
;
3304 } else if (flags
& WR_STATE
)
3305 return nfs4_share_conflict(current_fh
,
3306 NFS4_SHARE_DENY_WRITE
);
3307 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3308 return nfs4_share_conflict(current_fh
,
3309 NFS4_SHARE_DENY_READ
);
3313 * Allow READ/WRITE during grace period on recovered state only for files
3314 * that are not able to provide mandatory locking.
3317 grace_disallows_io(struct net
*net
, struct inode
*inode
)
3319 return locks_in_grace(net
) && mandatory_lock(inode
);
3322 /* Returns true iff a is later than b: */
3323 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3325 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3328 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3331 * When sessions are used the stateid generation number is ignored
3334 if (has_session
&& in
->si_generation
== 0)
3337 if (in
->si_generation
== ref
->si_generation
)
3340 /* If the client sends us a stateid from the future, it's buggy: */
3341 if (stateid_generation_after(in
, ref
))
3342 return nfserr_bad_stateid
;
3344 * However, we could see a stateid from the past, even from a
3345 * non-buggy client. For example, if the client sends a lock
3346 * while some IO is outstanding, the lock may bump si_generation
3347 * while the IO is still in flight. The client could avoid that
3348 * situation by waiting for responses on all the IO requests,
3349 * but better performance may result in retrying IO that
3350 * receives an old_stateid error if requests are rarely
3351 * reordered in flight:
3353 return nfserr_old_stateid
;
3356 static __be32
nfsd4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
3358 struct nfs4_stid
*s
;
3359 struct nfs4_ol_stateid
*ols
;
3362 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3363 return nfserr_bad_stateid
;
3364 /* Client debugging aid. */
3365 if (!same_clid(&stateid
->si_opaque
.so_clid
, &cl
->cl_clientid
)) {
3366 char addr_str
[INET6_ADDRSTRLEN
];
3367 rpc_ntop((struct sockaddr
*)&cl
->cl_addr
, addr_str
,
3369 pr_warn_ratelimited("NFSD: client %s testing state ID "
3370 "with incorrect client ID\n", addr_str
);
3371 return nfserr_bad_stateid
;
3373 s
= find_stateid(cl
, stateid
);
3375 return nfserr_bad_stateid
;
3376 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3379 if (!(s
->sc_type
& (NFS4_OPEN_STID
| NFS4_LOCK_STID
)))
3381 ols
= openlockstateid(s
);
3382 if (ols
->st_stateowner
->so_is_open_owner
3383 && !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3384 return nfserr_bad_stateid
;
3388 static __be32
nfsd4_lookup_stateid(stateid_t
*stateid
, unsigned char typemask
, struct nfs4_stid
**s
, bool sessions
)
3390 struct nfs4_client
*cl
;
3391 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
3393 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3394 return nfserr_bad_stateid
;
3395 if (STALE_STATEID(stateid
, nn
))
3396 return nfserr_stale_stateid
;
3397 cl
= find_confirmed_client(&stateid
->si_opaque
.so_clid
, sessions
);
3399 return nfserr_expired
;
3400 *s
= find_stateid_by_type(cl
, stateid
, typemask
);
3402 return nfserr_bad_stateid
;
3408 * Checks for stateid operations
3411 nfs4_preprocess_stateid_op(struct net
*net
, struct nfsd4_compound_state
*cstate
,
3412 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3414 struct nfs4_stid
*s
;
3415 struct nfs4_ol_stateid
*stp
= NULL
;
3416 struct nfs4_delegation
*dp
= NULL
;
3417 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3418 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3424 if (grace_disallows_io(net
, ino
))
3425 return nfserr_grace
;
3427 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3428 return check_special_stateids(net
, current_fh
, stateid
, flags
);
3430 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
, &s
, cstate
->minorversion
);
3433 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, nfsd4_has_session(cstate
));
3436 switch (s
->sc_type
) {
3437 case NFS4_DELEG_STID
:
3438 dp
= delegstateid(s
);
3439 status
= nfs4_check_delegmode(dp
, flags
);
3443 *filpp
= dp
->dl_file
->fi_deleg_file
;
3447 case NFS4_OPEN_STID
:
3448 case NFS4_LOCK_STID
:
3449 stp
= openlockstateid(s
);
3450 status
= nfs4_check_fh(current_fh
, stp
);
3453 if (stp
->st_stateowner
->so_is_open_owner
3454 && !(openowner(stp
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3456 status
= nfs4_check_openmode(stp
, flags
);
3460 if (flags
& RD_STATE
)
3461 *filpp
= find_readable_file(stp
->st_file
);
3463 *filpp
= find_writeable_file(stp
->st_file
);
3467 return nfserr_bad_stateid
;
3475 nfsd4_free_lock_stateid(struct nfs4_ol_stateid
*stp
)
3477 if (check_for_locks(stp
->st_file
, lockowner(stp
->st_stateowner
)))
3478 return nfserr_locks_held
;
3479 release_lock_stateid(stp
);
3484 * Test if the stateid is valid
3487 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3488 struct nfsd4_test_stateid
*test_stateid
)
3490 struct nfsd4_test_stateid_id
*stateid
;
3491 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3494 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
3495 stateid
->ts_id_status
=
3496 nfsd4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
3497 nfs4_unlock_state();
3503 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3504 struct nfsd4_free_stateid
*free_stateid
)
3506 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3507 struct nfs4_stid
*s
;
3508 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3509 __be32 ret
= nfserr_bad_stateid
;
3512 s
= find_stateid(cl
, stateid
);
3515 switch (s
->sc_type
) {
3516 case NFS4_DELEG_STID
:
3517 ret
= nfserr_locks_held
;
3519 case NFS4_OPEN_STID
:
3520 case NFS4_LOCK_STID
:
3521 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3524 if (s
->sc_type
== NFS4_LOCK_STID
)
3525 ret
= nfsd4_free_lock_stateid(openlockstateid(s
));
3527 ret
= nfserr_locks_held
;
3530 ret
= nfserr_bad_stateid
;
3533 nfs4_unlock_state();
3540 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3541 RD_STATE
: WR_STATE
;
3544 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
3546 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3547 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
3550 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
3553 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
)
3555 * "Closed" stateid's exist *only* to return
3556 * nfserr_replay_me from the previous step.
3558 return nfserr_bad_stateid
;
3559 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3562 return nfs4_check_fh(current_fh
, stp
);
3566 * Checks for sequence id mutating operations.
3569 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3570 stateid_t
*stateid
, char typemask
,
3571 struct nfs4_ol_stateid
**stpp
)
3574 struct nfs4_stid
*s
;
3576 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3577 seqid
, STATEID_VAL(stateid
));
3580 status
= nfsd4_lookup_stateid(stateid
, typemask
, &s
, cstate
->minorversion
);
3583 *stpp
= openlockstateid(s
);
3584 cstate
->replay_owner
= (*stpp
)->st_stateowner
;
3586 return nfs4_seqid_op_checks(cstate
, stateid
, seqid
, *stpp
);
3589 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
, stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
)
3592 struct nfs4_openowner
*oo
;
3594 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
3595 NFS4_OPEN_STID
, stpp
);
3598 oo
= openowner((*stpp
)->st_stateowner
);
3599 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
3600 return nfserr_bad_stateid
;
3605 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3606 struct nfsd4_open_confirm
*oc
)
3609 struct nfs4_openowner
*oo
;
3610 struct nfs4_ol_stateid
*stp
;
3612 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3613 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3614 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3616 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3622 status
= nfs4_preprocess_seqid_op(cstate
,
3623 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3624 NFS4_OPEN_STID
, &stp
);
3627 oo
= openowner(stp
->st_stateowner
);
3628 status
= nfserr_bad_stateid
;
3629 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
)
3631 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3632 update_stateid(&stp
->st_stid
.sc_stateid
);
3633 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3634 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3635 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3637 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
3640 if (!cstate
->replay_owner
)
3641 nfs4_unlock_state();
3645 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
3647 if (!test_access(access
, stp
))
3649 nfs4_file_put_access(stp
->st_file
, nfs4_access_to_omode(access
));
3650 clear_access(access
, stp
);
3653 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
3655 switch (to_access
) {
3656 case NFS4_SHARE_ACCESS_READ
:
3657 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
3658 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3660 case NFS4_SHARE_ACCESS_WRITE
:
3661 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
3662 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3664 case NFS4_SHARE_ACCESS_BOTH
:
3672 reset_union_bmap_deny(unsigned long deny
, struct nfs4_ol_stateid
*stp
)
3675 for (i
= 0; i
< 4; i
++) {
3676 if ((i
& deny
) != i
)
3682 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3683 struct nfsd4_compound_state
*cstate
,
3684 struct nfsd4_open_downgrade
*od
)
3687 struct nfs4_ol_stateid
*stp
;
3689 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3690 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3691 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3693 /* We don't yet support WANT bits: */
3694 if (od
->od_deleg_want
)
3695 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
3699 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
3700 &od
->od_stateid
, &stp
);
3703 status
= nfserr_inval
;
3704 if (!test_access(od
->od_share_access
, stp
)) {
3705 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3706 stp
->st_access_bmap
, od
->od_share_access
);
3709 if (!test_deny(od
->od_share_deny
, stp
)) {
3710 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3711 stp
->st_deny_bmap
, od
->od_share_deny
);
3714 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
3716 reset_union_bmap_deny(od
->od_share_deny
, stp
);
3718 update_stateid(&stp
->st_stid
.sc_stateid
);
3719 memcpy(&od
->od_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3722 if (!cstate
->replay_owner
)
3723 nfs4_unlock_state();
3727 void nfsd4_purge_closed_stateid(struct nfs4_stateowner
*so
)
3729 struct nfs4_openowner
*oo
;
3730 struct nfs4_ol_stateid
*s
;
3732 if (!so
->so_is_open_owner
)
3735 s
= oo
->oo_last_closed_stid
;
3738 if (!(oo
->oo_flags
& NFS4_OO_PURGE_CLOSE
)) {
3739 /* Release the last_closed_stid on the next seqid bump: */
3740 oo
->oo_flags
|= NFS4_OO_PURGE_CLOSE
;
3743 oo
->oo_flags
&= ~NFS4_OO_PURGE_CLOSE
;
3744 release_last_closed_stateid(oo
);
3747 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
3749 unhash_open_stateid(s
);
3750 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
3754 * nfs4_unlock_state() called after encode
3757 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3758 struct nfsd4_close
*close
)
3761 struct nfs4_openowner
*oo
;
3762 struct nfs4_ol_stateid
*stp
;
3764 dprintk("NFSD: nfsd4_close on file %.*s\n",
3765 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3766 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3769 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3771 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
3775 oo
= openowner(stp
->st_stateowner
);
3777 update_stateid(&stp
->st_stid
.sc_stateid
);
3778 memcpy(&close
->cl_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3780 nfsd4_close_open_stateid(stp
);
3781 release_last_closed_stateid(oo
);
3782 oo
->oo_last_closed_stid
= stp
;
3784 if (list_empty(&oo
->oo_owner
.so_stateids
)) {
3785 if (cstate
->minorversion
) {
3786 release_openowner(oo
);
3787 cstate
->replay_owner
= NULL
;
3790 * In the 4.0 case we need to keep the owners around a
3791 * little while to handle CLOSE replay.
3793 if (list_empty(&oo
->oo_owner
.so_stateids
))
3794 move_to_close_lru(oo
);
3798 if (!cstate
->replay_owner
)
3799 nfs4_unlock_state();
3804 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3805 struct nfsd4_delegreturn
*dr
)
3807 struct nfs4_delegation
*dp
;
3808 stateid_t
*stateid
= &dr
->dr_stateid
;
3809 struct nfs4_stid
*s
;
3810 struct inode
*inode
;
3813 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3815 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3818 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
, &s
, cstate
->minorversion
);
3821 dp
= delegstateid(s
);
3822 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3826 unhash_delegation(dp
);
3828 nfs4_unlock_state();
3834 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3836 #define LOCKOWNER_INO_HASH_BITS 8
3837 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3838 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3841 end_offset(u64 start
, u64 len
)
3846 return end
>= start
? end
: NFS4_MAX_UINT64
;
3849 /* last octet in a range */
3851 last_byte_offset(u64 start
, u64 len
)
3857 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3860 static unsigned int lockowner_ino_hashval(struct inode
*inode
, u32 cl_id
, struct xdr_netobj
*ownername
)
3862 return (file_hashval(inode
) + cl_id
3863 + opaque_hashval(ownername
->data
, ownername
->len
))
3864 & LOCKOWNER_INO_HASH_MASK
;
3867 static struct list_head lockowner_ino_hashtbl
[LOCKOWNER_INO_HASH_SIZE
];
3870 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3871 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3872 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3873 * locking, this prevents us from being completely protocol-compliant. The
3874 * real solution to this problem is to start using unsigned file offsets in
3875 * the VFS, but this is a very deep change!
3878 nfs4_transform_lock_offset(struct file_lock
*lock
)
3880 if (lock
->fl_start
< 0)
3881 lock
->fl_start
= OFFSET_MAX
;
3882 if (lock
->fl_end
< 0)
3883 lock
->fl_end
= OFFSET_MAX
;
3886 /* Hack!: For now, we're defining this just so we can use a pointer to it
3887 * as a unique cookie to identify our (NFSv4's) posix locks. */
3888 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3892 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3894 struct nfs4_lockowner
*lo
;
3896 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3897 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3898 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3899 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3900 if (!deny
->ld_owner
.data
)
3901 /* We just don't care that much */
3903 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3904 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3907 deny
->ld_owner
.len
= 0;
3908 deny
->ld_owner
.data
= NULL
;
3909 deny
->ld_clientid
.cl_boot
= 0;
3910 deny
->ld_clientid
.cl_id
= 0;
3912 deny
->ld_start
= fl
->fl_start
;
3913 deny
->ld_length
= NFS4_MAX_UINT64
;
3914 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3915 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3916 deny
->ld_type
= NFS4_READ_LT
;
3917 if (fl
->fl_type
!= F_RDLCK
)
3918 deny
->ld_type
= NFS4_WRITE_LT
;
3921 static bool same_lockowner_ino(struct nfs4_lockowner
*lo
, struct inode
*inode
, clientid_t
*clid
, struct xdr_netobj
*owner
)
3923 struct nfs4_ol_stateid
*lst
;
3925 if (!same_owner_str(&lo
->lo_owner
, owner
, clid
))
3927 lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3928 struct nfs4_ol_stateid
, st_perstateowner
);
3929 return lst
->st_file
->fi_inode
== inode
;
3932 static struct nfs4_lockowner
*
3933 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3934 struct xdr_netobj
*owner
)
3936 unsigned int hashval
= lockowner_ino_hashval(inode
, clid
->cl_id
, owner
);
3937 struct nfs4_lockowner
*lo
;
3939 list_for_each_entry(lo
, &lockowner_ino_hashtbl
[hashval
], lo_owner_ino_hash
) {
3940 if (same_lockowner_ino(lo
, inode
, clid
, owner
))
3946 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
)
3948 struct inode
*inode
= open_stp
->st_file
->fi_inode
;
3949 unsigned int inohash
= lockowner_ino_hashval(inode
,
3950 clp
->cl_clientid
.cl_id
, &lo
->lo_owner
.so_owner
);
3952 list_add(&lo
->lo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
3953 list_add(&lo
->lo_owner_ino_hash
, &lockowner_ino_hashtbl
[inohash
]);
3954 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3958 * Alloc a lock owner structure.
3959 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3962 * strhashval = ownerstr_hashval
3965 static struct nfs4_lockowner
*
3966 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3967 struct nfs4_lockowner
*lo
;
3969 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
3972 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
3973 lo
->lo_owner
.so_is_open_owner
= 0;
3974 /* It is the openowner seqid that will be incremented in encode in the
3975 * case of new lockowners; so increment the lock seqid manually: */
3976 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3977 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
3981 static struct nfs4_ol_stateid
*
3982 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_ol_stateid
*open_stp
)
3984 struct nfs4_ol_stateid
*stp
;
3985 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
3987 stp
= nfs4_alloc_stateid(clp
);
3990 init_stid(&stp
->st_stid
, clp
, NFS4_LOCK_STID
);
3991 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3992 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
3993 stp
->st_stateowner
= &lo
->lo_owner
;
3996 stp
->st_access_bmap
= 0;
3997 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3998 stp
->st_openstp
= open_stp
;
4003 check_lock_length(u64 offset
, u64 length
)
4005 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
4006 LOFF_OVERFLOW(offset
, length
)));
4009 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
4011 struct nfs4_file
*fp
= lock_stp
->st_file
;
4012 int oflag
= nfs4_access_to_omode(access
);
4014 if (test_access(access
, lock_stp
))
4016 nfs4_file_get_access(fp
, oflag
);
4017 set_access(access
, lock_stp
);
4020 static __be32
lookup_or_create_lock_state(struct nfsd4_compound_state
*cstate
, struct nfs4_ol_stateid
*ost
, struct nfsd4_lock
*lock
, struct nfs4_ol_stateid
**lst
, bool *new)
4022 struct nfs4_file
*fi
= ost
->st_file
;
4023 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
4024 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
4025 struct nfs4_lockowner
*lo
;
4026 unsigned int strhashval
;
4028 lo
= find_lockowner_str(fi
->fi_inode
, &cl
->cl_clientid
, &lock
->v
.new.owner
);
4030 if (!cstate
->minorversion
)
4031 return nfserr_bad_seqid
;
4032 /* XXX: a lockowner always has exactly one stateid: */
4033 *lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
4034 struct nfs4_ol_stateid
, st_perstateowner
);
4037 strhashval
= ownerstr_hashval(cl
->cl_clientid
.cl_id
,
4038 &lock
->v
.new.owner
);
4039 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
4041 return nfserr_jukebox
;
4042 *lst
= alloc_init_lock_stateid(lo
, fi
, ost
);
4044 release_lockowner(lo
);
4045 return nfserr_jukebox
;
4055 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4056 struct nfsd4_lock
*lock
)
4058 struct nfs4_openowner
*open_sop
= NULL
;
4059 struct nfs4_lockowner
*lock_sop
= NULL
;
4060 struct nfs4_ol_stateid
*lock_stp
;
4061 struct file
*filp
= NULL
;
4062 struct file_lock
*file_lock
= NULL
;
4063 struct file_lock
*conflock
= NULL
;
4065 bool new_state
= false;
4068 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
4070 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4071 (long long) lock
->lk_offset
,
4072 (long long) lock
->lk_length
);
4074 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
4075 return nfserr_inval
;
4077 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
4078 S_IFREG
, NFSD_MAY_LOCK
))) {
4079 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4085 if (lock
->lk_is_new
) {
4086 struct nfs4_ol_stateid
*open_stp
= NULL
;
4088 if (nfsd4_has_session(cstate
))
4089 /* See rfc 5661 18.10.3: given clientid is ignored: */
4090 memcpy(&lock
->v
.new.clientid
,
4091 &cstate
->session
->se_client
->cl_clientid
,
4092 sizeof(clientid_t
));
4094 status
= nfserr_stale_clientid
;
4095 if (STALE_CLIENTID(&lock
->lk_new_clientid
, nn
))
4098 /* validate and update open stateid and open seqid */
4099 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
4100 lock
->lk_new_open_seqid
,
4101 &lock
->lk_new_open_stateid
,
4105 open_sop
= openowner(open_stp
->st_stateowner
);
4106 status
= nfserr_bad_stateid
;
4107 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
4108 &lock
->v
.new.clientid
))
4110 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
4111 &lock_stp
, &new_state
);
4113 status
= nfs4_preprocess_seqid_op(cstate
,
4114 lock
->lk_old_lock_seqid
,
4115 &lock
->lk_old_lock_stateid
,
4116 NFS4_LOCK_STID
, &lock_stp
);
4119 lock_sop
= lockowner(lock_stp
->st_stateowner
);
4121 lkflg
= setlkflg(lock
->lk_type
);
4122 status
= nfs4_check_openmode(lock_stp
, lkflg
);
4126 status
= nfserr_grace
;
4127 if (locks_in_grace(SVC_NET(rqstp
)) && !lock
->lk_reclaim
)
4129 status
= nfserr_no_grace
;
4130 if (!locks_in_grace(SVC_NET(rqstp
)) && lock
->lk_reclaim
)
4133 file_lock
= locks_alloc_lock();
4135 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
4136 status
= nfserr_jukebox
;
4140 locks_init_lock(file_lock
);
4141 switch (lock
->lk_type
) {
4144 filp
= find_readable_file(lock_stp
->st_file
);
4146 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4147 file_lock
->fl_type
= F_RDLCK
;
4150 case NFS4_WRITEW_LT
:
4151 filp
= find_writeable_file(lock_stp
->st_file
);
4153 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4154 file_lock
->fl_type
= F_WRLCK
;
4157 status
= nfserr_inval
;
4161 status
= nfserr_openmode
;
4164 file_lock
->fl_owner
= (fl_owner_t
)lock_sop
;
4165 file_lock
->fl_pid
= current
->tgid
;
4166 file_lock
->fl_file
= filp
;
4167 file_lock
->fl_flags
= FL_POSIX
;
4168 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
4169 file_lock
->fl_start
= lock
->lk_offset
;
4170 file_lock
->fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4171 nfs4_transform_lock_offset(file_lock
);
4173 conflock
= locks_alloc_lock();
4175 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
4176 status
= nfserr_jukebox
;
4180 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, conflock
);
4182 case 0: /* success! */
4183 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
4184 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
4188 case (EAGAIN
): /* conflock holds conflicting lock */
4189 status
= nfserr_denied
;
4190 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4191 nfs4_set_lock_denied(conflock
, &lock
->lk_denied
);
4194 status
= nfserr_deadlock
;
4197 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4198 status
= nfserrno(err
);
4202 if (status
&& new_state
)
4203 release_lockowner(lock_sop
);
4204 if (!cstate
->replay_owner
)
4205 nfs4_unlock_state();
4207 locks_free_lock(file_lock
);
4209 locks_free_lock(conflock
);
4214 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4215 * so we do a temporary open here just to get an open file to pass to
4216 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4219 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4222 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4224 err
= nfserrno(vfs_test_lock(file
, lock
));
4234 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4235 struct nfsd4_lockt
*lockt
)
4237 struct inode
*inode
;
4238 struct file_lock
*file_lock
= NULL
;
4239 struct nfs4_lockowner
*lo
;
4241 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
4243 if (locks_in_grace(SVC_NET(rqstp
)))
4244 return nfserr_grace
;
4246 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4247 return nfserr_inval
;
4251 status
= nfserr_stale_clientid
;
4252 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
, nn
))
4255 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4258 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4259 file_lock
= locks_alloc_lock();
4261 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
4262 status
= nfserr_jukebox
;
4265 locks_init_lock(file_lock
);
4266 switch (lockt
->lt_type
) {
4269 file_lock
->fl_type
= F_RDLCK
;
4272 case NFS4_WRITEW_LT
:
4273 file_lock
->fl_type
= F_WRLCK
;
4276 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4277 status
= nfserr_inval
;
4281 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4283 file_lock
->fl_owner
= (fl_owner_t
)lo
;
4284 file_lock
->fl_pid
= current
->tgid
;
4285 file_lock
->fl_flags
= FL_POSIX
;
4287 file_lock
->fl_start
= lockt
->lt_offset
;
4288 file_lock
->fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4290 nfs4_transform_lock_offset(file_lock
);
4292 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, file_lock
);
4296 if (file_lock
->fl_type
!= F_UNLCK
) {
4297 status
= nfserr_denied
;
4298 nfs4_set_lock_denied(file_lock
, &lockt
->lt_denied
);
4301 nfs4_unlock_state();
4303 locks_free_lock(file_lock
);
4308 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4309 struct nfsd4_locku
*locku
)
4311 struct nfs4_ol_stateid
*stp
;
4312 struct file
*filp
= NULL
;
4313 struct file_lock
*file_lock
= NULL
;
4317 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4318 (long long) locku
->lu_offset
,
4319 (long long) locku
->lu_length
);
4321 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4322 return nfserr_inval
;
4326 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4327 &locku
->lu_stateid
, NFS4_LOCK_STID
, &stp
);
4330 filp
= find_any_file(stp
->st_file
);
4332 status
= nfserr_lock_range
;
4335 file_lock
= locks_alloc_lock();
4337 dprintk("NFSD: %s: unable to allocate lock!\n", __func__
);
4338 status
= nfserr_jukebox
;
4341 locks_init_lock(file_lock
);
4342 file_lock
->fl_type
= F_UNLCK
;
4343 file_lock
->fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4344 file_lock
->fl_pid
= current
->tgid
;
4345 file_lock
->fl_file
= filp
;
4346 file_lock
->fl_flags
= FL_POSIX
;
4347 file_lock
->fl_lmops
= &nfsd_posix_mng_ops
;
4348 file_lock
->fl_start
= locku
->lu_offset
;
4350 file_lock
->fl_end
= last_byte_offset(locku
->lu_offset
,
4352 nfs4_transform_lock_offset(file_lock
);
4355 * Try to unlock the file in the VFS.
4357 err
= vfs_lock_file(filp
, F_SETLK
, file_lock
, NULL
);
4359 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4363 * OK, unlock succeeded; the only thing left to do is update the stateid.
4365 update_stateid(&stp
->st_stid
.sc_stateid
);
4366 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4369 if (!cstate
->replay_owner
)
4370 nfs4_unlock_state();
4372 locks_free_lock(file_lock
);
4376 status
= nfserrno(err
);
4382 * 1: locks held by lockowner
4383 * 0: no locks held by lockowner
4386 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4388 struct file_lock
**flpp
;
4389 struct inode
*inode
= filp
->fi_inode
;
4393 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4394 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4405 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4406 struct nfsd4_compound_state
*cstate
,
4407 struct nfsd4_release_lockowner
*rlockowner
)
4409 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4410 struct nfs4_stateowner
*sop
;
4411 struct nfs4_lockowner
*lo
;
4412 struct nfs4_ol_stateid
*stp
;
4413 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4414 struct list_head matches
;
4415 unsigned int hashval
= ownerstr_hashval(clid
->cl_id
, owner
);
4417 struct nfsd_net
*nn
= net_generic(&init_net
, nfsd_net_id
);
4419 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4420 clid
->cl_boot
, clid
->cl_id
);
4422 /* XXX check for lease expiration */
4424 status
= nfserr_stale_clientid
;
4425 if (STALE_CLIENTID(clid
, nn
))
4430 status
= nfserr_locks_held
;
4431 INIT_LIST_HEAD(&matches
);
4433 list_for_each_entry(sop
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
4434 if (sop
->so_is_open_owner
)
4436 if (!same_owner_str(sop
, owner
, clid
))
4438 list_for_each_entry(stp
, &sop
->so_stateids
,
4440 lo
= lockowner(sop
);
4441 if (check_for_locks(stp
->st_file
, lo
))
4443 list_add(&lo
->lo_list
, &matches
);
4446 /* Clients probably won't expect us to return with some (but not all)
4447 * of the lockowner state released; so don't release any until all
4448 * have been checked. */
4450 while (!list_empty(&matches
)) {
4451 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4453 /* unhash_stateowner deletes so_perclient only
4454 * for openowners. */
4455 list_del(&lo
->lo_list
);
4456 release_lockowner(lo
);
4459 nfs4_unlock_state();
4463 static inline struct nfs4_client_reclaim
*
4466 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4470 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4472 unsigned int strhashval
= clientstr_hashval(name
);
4473 struct nfs4_client
*clp
;
4475 clp
= find_confirmed_client_by_str(name
, strhashval
);
4478 return test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
4482 * failure => all reset bets are off, nfserr_no_grace...
4485 nfs4_client_to_reclaim(const char *name
)
4487 unsigned int strhashval
;
4488 struct nfs4_client_reclaim
*crp
= NULL
;
4490 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4491 crp
= alloc_reclaim();
4494 strhashval
= clientstr_hashval(name
);
4495 INIT_LIST_HEAD(&crp
->cr_strhash
);
4496 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4497 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4498 reclaim_str_hashtbl_size
++;
4503 nfs4_release_reclaim(void)
4505 struct nfs4_client_reclaim
*crp
= NULL
;
4508 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4509 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4510 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4511 struct nfs4_client_reclaim
, cr_strhash
);
4512 list_del(&crp
->cr_strhash
);
4514 reclaim_str_hashtbl_size
--;
4517 BUG_ON(reclaim_str_hashtbl_size
);
4521 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4522 struct nfs4_client_reclaim
*
4523 nfsd4_find_reclaim_client(struct nfs4_client
*clp
)
4525 unsigned int strhashval
;
4526 struct nfs4_client_reclaim
*crp
= NULL
;
4528 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4529 clp
->cl_name
.len
, clp
->cl_name
.data
,
4532 /* find clp->cl_name in reclaim_str_hashtbl */
4533 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4534 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4535 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4543 * Called from OPEN. Look for clientid in reclaim list.
4546 nfs4_check_open_reclaim(clientid_t
*clid
, bool sessions
)
4548 struct nfs4_client
*clp
;
4550 /* find clientid in conf_id_hashtbl */
4551 clp
= find_confirmed_client(clid
, sessions
);
4553 return nfserr_reclaim_bad
;
4555 return nfsd4_client_record_check(clp
) ? nfserr_reclaim_bad
: nfs_ok
;
4558 #ifdef CONFIG_NFSD_FAULT_INJECTION
4560 void nfsd_forget_clients(u64 num
)
4562 struct nfs4_client
*clp
, *next
;
4566 list_for_each_entry_safe(clp
, next
, &client_lru
, cl_lru
) {
4567 nfsd4_client_record_remove(clp
);
4572 nfs4_unlock_state();
4574 printk(KERN_INFO
"NFSD: Forgot %d clients", count
);
4577 static void release_lockowner_sop(struct nfs4_stateowner
*sop
)
4579 release_lockowner(lockowner(sop
));
4582 static void release_openowner_sop(struct nfs4_stateowner
*sop
)
4584 release_openowner(openowner(sop
));
4587 static int nfsd_release_n_owners(u64 num
, bool is_open_owner
,
4588 void (*release_sop
)(struct nfs4_stateowner
*))
4591 struct nfs4_stateowner
*sop
, *next
;
4593 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4594 list_for_each_entry_safe(sop
, next
, &ownerstr_hashtbl
[i
], so_strhash
) {
4595 if (sop
->so_is_open_owner
!= is_open_owner
)
4605 void nfsd_forget_locks(u64 num
)
4610 count
= nfsd_release_n_owners(num
, false, release_lockowner_sop
);
4611 nfs4_unlock_state();
4613 printk(KERN_INFO
"NFSD: Forgot %d locks", count
);
4616 void nfsd_forget_openowners(u64 num
)
4621 count
= nfsd_release_n_owners(num
, true, release_openowner_sop
);
4622 nfs4_unlock_state();
4624 printk(KERN_INFO
"NFSD: Forgot %d open owners", count
);
4627 static int nfsd_process_n_delegations(u64 num
, struct list_head
*list
)
4630 struct nfs4_file
*fp
, *fnext
;
4631 struct nfs4_delegation
*dp
, *dnext
;
4633 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4634 list_for_each_entry_safe(fp
, fnext
, &file_hashtbl
[i
], fi_hash
) {
4635 list_for_each_entry_safe(dp
, dnext
, &fp
->fi_delegations
, dl_perfile
) {
4636 list_move(&dp
->dl_recall_lru
, list
);
4646 void nfsd_forget_delegations(u64 num
)
4650 struct nfs4_delegation
*dp
, *dnext
;
4652 spin_lock(&recall_lock
);
4653 count
= nfsd_process_n_delegations(num
, &victims
);
4654 spin_unlock(&recall_lock
);
4657 list_for_each_entry_safe(dp
, dnext
, &victims
, dl_recall_lru
)
4658 unhash_delegation(dp
);
4659 nfs4_unlock_state();
4661 printk(KERN_INFO
"NFSD: Forgot %d delegations", count
);
4664 void nfsd_recall_delegations(u64 num
)
4668 struct nfs4_delegation
*dp
, *dnext
;
4670 spin_lock(&recall_lock
);
4671 count
= nfsd_process_n_delegations(num
, &victims
);
4672 list_for_each_entry_safe(dp
, dnext
, &victims
, dl_recall_lru
) {
4673 list_del(&dp
->dl_recall_lru
);
4674 nfsd_break_one_deleg(dp
);
4676 spin_unlock(&recall_lock
);
4678 printk(KERN_INFO
"NFSD: Recalled %d delegations", count
);
4681 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4683 /* initialization to perform at module load time: */
4686 nfs4_state_init(void)
4690 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4691 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4692 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4693 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4694 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4695 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4697 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4698 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4699 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4700 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4702 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4703 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4705 for (i
= 0; i
< LOCKOWNER_INO_HASH_SIZE
; i
++)
4706 INIT_LIST_HEAD(&lockowner_ino_hashtbl
[i
]);
4707 INIT_LIST_HEAD(&close_lru
);
4708 INIT_LIST_HEAD(&client_lru
);
4709 INIT_LIST_HEAD(&del_recall_lru
);
4710 reclaim_str_hashtbl_size
= 0;
4714 * Since the lifetime of a delegation isn't limited to that of an open, a
4715 * client may quite reasonably hang on to a delegation as long as it has
4716 * the inode cached. This becomes an obvious problem the first time a
4717 * client's inode cache approaches the size of the server's total memory.
4719 * For now we avoid this problem by imposing a hard limit on the number
4720 * of delegations, which varies according to the server's memory size.
4723 set_max_delegations(void)
4726 * Allow at most 4 delegations per megabyte of RAM. Quick
4727 * estimates suggest that in the worst case (where every delegation
4728 * is for a different inode), a delegation could take about 1.5K,
4729 * giving a worst case usage of about 6% of memory.
4731 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4734 /* initialization to perform when the nfsd service is started: */
4737 nfs4_state_start(void)
4739 struct net
*net
= &init_net
;
4740 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4744 * FIXME: For now, we hang most of the pernet global stuff off of
4745 * init_net until nfsd is fully containerized. Eventually, we'll
4746 * need to pass a net pointer into this function, take a reference
4747 * to that instead and then do most of the rest of this on a per-net
4751 nfsd4_client_tracking_init(net
);
4752 nn
->boot_time
= get_seconds();
4753 locks_start_grace(net
, &nn
->nfsd4_manager
);
4754 nn
->grace_ended
= false;
4755 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4757 ret
= set_callback_cred();
4762 laundry_wq
= create_singlethread_workqueue("nfsd4");
4763 if (laundry_wq
== NULL
) {
4767 ret
= nfsd4_create_callback_queue();
4769 goto out_free_laundry
;
4770 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4771 set_max_delegations();
4774 destroy_workqueue(laundry_wq
);
4776 nfsd4_client_tracking_exit(net
);
4782 __nfs4_state_shutdown(void)
4785 struct nfs4_client
*clp
= NULL
;
4786 struct nfs4_delegation
*dp
= NULL
;
4787 struct list_head
*pos
, *next
, reaplist
;
4789 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4790 while (!list_empty(&conf_id_hashtbl
[i
])) {
4791 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4794 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4795 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4799 INIT_LIST_HEAD(&reaplist
);
4800 spin_lock(&recall_lock
);
4801 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4802 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4803 list_move(&dp
->dl_recall_lru
, &reaplist
);
4805 spin_unlock(&recall_lock
);
4806 list_for_each_safe(pos
, next
, &reaplist
) {
4807 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4808 unhash_delegation(dp
);
4811 nfsd4_client_tracking_exit(&init_net
);
4816 nfs4_state_shutdown(void)
4818 struct net
*net
= &init_net
;
4819 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4821 cancel_delayed_work_sync(&laundromat_work
);
4822 destroy_workqueue(laundry_wq
);
4823 locks_end_grace(&nn
->nfsd4_manager
);
4825 __nfs4_state_shutdown();
4826 nfs4_unlock_state();
4827 nfsd4_destroy_callback_queue();
4831 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4833 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
4834 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
4838 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4840 if (cstate
->minorversion
) {
4841 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
4842 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4847 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
4849 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4853 * functions to set current state id
4856 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4858 put_stateid(cstate
, &odp
->od_stateid
);
4862 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
4864 put_stateid(cstate
, &open
->op_stateid
);
4868 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4870 put_stateid(cstate
, &close
->cl_stateid
);
4874 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
4876 put_stateid(cstate
, &lock
->lk_resp_stateid
);
4880 * functions to consume current state id
4884 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4886 get_stateid(cstate
, &odp
->od_stateid
);
4890 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
4892 get_stateid(cstate
, &drp
->dr_stateid
);
4896 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
4898 get_stateid(cstate
, &fsp
->fr_stateid
);
4902 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
4904 get_stateid(cstate
, &setattr
->sa_stateid
);
4908 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4910 get_stateid(cstate
, &close
->cl_stateid
);
4914 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
4916 get_stateid(cstate
, &locku
->lu_stateid
);
4920 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
4922 get_stateid(cstate
, &read
->rd_stateid
);
4926 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
4928 get_stateid(cstate
, &write
->wr_stateid
);