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"
48 #define NFSDDBG_FACILITY NFSDDBG_PROC
51 time_t nfsd4_lease
= 90; /* default lease time */
52 time_t nfsd4_grace
= 90;
53 static time_t boot_time
;
55 #define all_ones {{~0,~0},~0}
56 static const stateid_t one_stateid
= {
58 .si_opaque
= all_ones
,
60 static const stateid_t zero_stateid
= {
63 static const stateid_t currentstateid
= {
67 static u64 current_sessionid
= 1;
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
71 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
73 /* forward declarations */
74 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
);
78 /* Currently used for almost all code touching nfsv4 state: */
79 static DEFINE_MUTEX(client_mutex
);
82 * Currently used for the del_recall_lru and file hash table. In an
83 * effort to decrease the scope of the client_mutex, this spinlock may
84 * eventually cover more:
86 static DEFINE_SPINLOCK(recall_lock
);
88 static struct kmem_cache
*openowner_slab
= NULL
;
89 static struct kmem_cache
*lockowner_slab
= NULL
;
90 static struct kmem_cache
*file_slab
= NULL
;
91 static struct kmem_cache
*stateid_slab
= NULL
;
92 static struct kmem_cache
*deleg_slab
= NULL
;
97 mutex_lock(&client_mutex
);
100 static void free_session(struct kref
*);
102 /* Must be called under the client_lock */
103 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
105 kref_put(&ses
->se_ref
, free_session
);
108 static void nfsd4_get_session(struct nfsd4_session
*ses
)
110 kref_get(&ses
->se_ref
);
114 nfs4_unlock_state(void)
116 mutex_unlock(&client_mutex
);
120 opaque_hashval(const void *ptr
, int nbytes
)
122 unsigned char *cptr
= (unsigned char *) ptr
;
132 static struct list_head del_recall_lru
;
134 static void nfsd4_free_file(struct nfs4_file
*f
)
136 kmem_cache_free(file_slab
, f
);
140 put_nfs4_file(struct nfs4_file
*fi
)
142 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
143 list_del(&fi
->fi_hash
);
144 spin_unlock(&recall_lock
);
151 get_nfs4_file(struct nfs4_file
*fi
)
153 atomic_inc(&fi
->fi_ref
);
156 static int num_delegations
;
157 unsigned int max_delegations
;
160 * Open owner state (share locks)
163 /* hash tables for lock and open owners */
164 #define OWNER_HASH_BITS 8
165 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
166 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
168 static unsigned int ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
172 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
174 return ret
& OWNER_HASH_MASK
;
177 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
179 /* hash table for nfs4_file */
180 #define FILE_HASH_BITS 8
181 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
183 static unsigned int file_hashval(struct inode
*ino
)
185 /* XXX: why are we hashing on inode pointer, anyway? */
186 return hash_ptr(ino
, FILE_HASH_BITS
);
189 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
191 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
193 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
194 atomic_inc(&fp
->fi_access
[oflag
]);
197 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
199 if (oflag
== O_RDWR
) {
200 __nfs4_file_get_access(fp
, O_RDONLY
);
201 __nfs4_file_get_access(fp
, O_WRONLY
);
203 __nfs4_file_get_access(fp
, oflag
);
206 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
208 if (fp
->fi_fds
[oflag
]) {
209 fput(fp
->fi_fds
[oflag
]);
210 fp
->fi_fds
[oflag
] = NULL
;
214 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
216 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
217 nfs4_file_put_fd(fp
, oflag
);
219 * It's also safe to get rid of the RDWR open *if*
220 * we no longer have need of the other kind of access
221 * or if we already have the other kind of open:
223 if (fp
->fi_fds
[1-oflag
]
224 || atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
225 nfs4_file_put_fd(fp
, O_RDWR
);
229 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
231 if (oflag
== O_RDWR
) {
232 __nfs4_file_put_access(fp
, O_RDONLY
);
233 __nfs4_file_put_access(fp
, O_WRONLY
);
235 __nfs4_file_put_access(fp
, oflag
);
238 static inline int get_new_stid(struct nfs4_stid
*stid
)
240 static int min_stateid
= 0;
241 struct idr
*stateids
= &stid
->sc_client
->cl_stateids
;
245 error
= idr_get_new_above(stateids
, stid
, min_stateid
, &new_stid
);
247 * Note: the necessary preallocation was done in
248 * nfs4_alloc_stateid(). The idr code caps the number of
249 * preallocations that can exist at a time, but the state lock
250 * prevents anyone from using ours before we get here:
254 * It shouldn't be a problem to reuse an opaque stateid value.
255 * I don't think it is for 4.1. But with 4.0 I worry that, for
256 * example, a stray write retransmission could be accepted by
257 * the server when it should have been rejected. Therefore,
258 * adopt a trick from the sctp code to attempt to maximize the
259 * amount of time until an id is reused, by ensuring they always
260 * "increase" (mod INT_MAX):
263 min_stateid
= new_stid
+1;
264 if (min_stateid
== INT_MAX
)
269 static void init_stid(struct nfs4_stid
*stid
, struct nfs4_client
*cl
, unsigned char type
)
271 stateid_t
*s
= &stid
->sc_stateid
;
274 stid
->sc_type
= type
;
275 stid
->sc_client
= cl
;
276 s
->si_opaque
.so_clid
= cl
->cl_clientid
;
277 new_id
= get_new_stid(stid
);
278 s
->si_opaque
.so_id
= (u32
)new_id
;
279 /* Will be incremented before return to client: */
280 s
->si_generation
= 0;
283 static struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
)
285 struct idr
*stateids
= &cl
->cl_stateids
;
287 if (!idr_pre_get(stateids
, GFP_KERNEL
))
290 * Note: if we fail here (or any time between now and the time
291 * we actually get the new idr), we won't need to undo the idr
292 * preallocation, since the idr code caps the number of
293 * preallocated entries.
295 return kmem_cache_alloc(slab
, GFP_KERNEL
);
298 static struct nfs4_ol_stateid
* nfs4_alloc_stateid(struct nfs4_client
*clp
)
300 return openlockstateid(nfs4_alloc_stid(clp
, stateid_slab
));
303 static struct nfs4_delegation
*
304 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_ol_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
306 struct nfs4_delegation
*dp
;
307 struct nfs4_file
*fp
= stp
->st_file
;
309 dprintk("NFSD alloc_init_deleg\n");
311 * Major work on the lease subsystem (for example, to support
312 * calbacks on stat) will be required before we can support
313 * write delegations properly.
315 if (type
!= NFS4_OPEN_DELEGATE_READ
)
317 if (fp
->fi_had_conflict
)
319 if (num_delegations
> max_delegations
)
321 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
324 init_stid(&dp
->dl_stid
, clp
, NFS4_DELEG_STID
);
326 * delegation seqid's are never incremented. The 4.1 special
327 * meaning of seqid 0 isn't meaningful, really, but let's avoid
328 * 0 anyway just for consistency and use 1:
330 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
332 INIT_LIST_HEAD(&dp
->dl_perfile
);
333 INIT_LIST_HEAD(&dp
->dl_perclnt
);
334 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
338 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
340 atomic_set(&dp
->dl_count
, 1);
341 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
346 nfs4_put_delegation(struct nfs4_delegation
*dp
)
348 if (atomic_dec_and_test(&dp
->dl_count
)) {
349 dprintk("NFSD: freeing dp %p\n",dp
);
350 put_nfs4_file(dp
->dl_file
);
351 kmem_cache_free(deleg_slab
, dp
);
356 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
358 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
359 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
361 fput(fp
->fi_deleg_file
);
362 fp
->fi_deleg_file
= NULL
;
366 static void unhash_stid(struct nfs4_stid
*s
)
368 struct idr
*stateids
= &s
->sc_client
->cl_stateids
;
370 idr_remove(stateids
, s
->sc_stateid
.si_opaque
.so_id
);
373 /* Called under the state lock. */
375 unhash_delegation(struct nfs4_delegation
*dp
)
377 unhash_stid(&dp
->dl_stid
);
378 list_del_init(&dp
->dl_perclnt
);
379 spin_lock(&recall_lock
);
380 list_del_init(&dp
->dl_perfile
);
381 list_del_init(&dp
->dl_recall_lru
);
382 spin_unlock(&recall_lock
);
383 nfs4_put_deleg_lease(dp
->dl_file
);
384 nfs4_put_delegation(dp
);
391 /* client_lock protects the client lru list and session hash table */
392 static DEFINE_SPINLOCK(client_lock
);
394 /* Hash tables for nfs4_clientid state */
395 #define CLIENT_HASH_BITS 4
396 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
397 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
399 static unsigned int clientid_hashval(u32 id
)
401 return id
& CLIENT_HASH_MASK
;
404 static unsigned int clientstr_hashval(const char *name
)
406 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
410 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
411 * used in reboot/reset lease grace period processing
413 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
414 * setclientid_confirmed info.
416 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
419 * client_lru holds client queue ordered by nfs4_client.cl_time
422 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
423 * for last close replay.
425 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
426 static int reclaim_str_hashtbl_size
= 0;
427 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
428 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
429 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
430 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
431 static struct list_head client_lru
;
432 static struct list_head close_lru
;
435 * We store the NONE, READ, WRITE, and BOTH bits separately in the
436 * st_{access,deny}_bmap field of the stateid, in order to track not
437 * only what share bits are currently in force, but also what
438 * combinations of share bits previous opens have used. This allows us
439 * to enforce the recommendation of rfc 3530 14.2.19 that the server
440 * return an error if the client attempt to downgrade to a combination
441 * of share bits not explicable by closing some of its previous opens.
443 * XXX: This enforcement is actually incomplete, since we don't keep
444 * track of access/deny bit combinations; so, e.g., we allow:
446 * OPEN allow read, deny write
447 * OPEN allow both, deny none
448 * DOWNGRADE allow read, deny none
450 * which we should reject.
453 bmap_to_share_mode(unsigned long bmap
) {
455 unsigned int access
= 0;
457 for (i
= 1; i
< 4; i
++) {
458 if (test_bit(i
, &bmap
))
465 test_share(struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
) {
466 unsigned int access
, deny
;
468 access
= bmap_to_share_mode(stp
->st_access_bmap
);
469 deny
= bmap_to_share_mode(stp
->st_deny_bmap
);
470 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
475 /* set share access for a given stateid */
477 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
479 __set_bit(access
, &stp
->st_access_bmap
);
482 /* clear share access for a given stateid */
484 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
486 __clear_bit(access
, &stp
->st_access_bmap
);
489 /* test whether a given stateid has access */
491 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
493 return test_bit(access
, &stp
->st_access_bmap
);
496 /* set share deny for a given stateid */
498 set_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
500 __set_bit(access
, &stp
->st_deny_bmap
);
503 /* clear share deny for a given stateid */
505 clear_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
507 __clear_bit(access
, &stp
->st_deny_bmap
);
510 /* test whether a given stateid is denying specific access */
512 test_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
514 return test_bit(access
, &stp
->st_deny_bmap
);
517 static int nfs4_access_to_omode(u32 access
)
519 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
520 case NFS4_SHARE_ACCESS_READ
:
522 case NFS4_SHARE_ACCESS_WRITE
:
524 case NFS4_SHARE_ACCESS_BOTH
:
530 /* release all access and file references for a given stateid */
532 release_all_access(struct nfs4_ol_stateid
*stp
)
536 for (i
= 1; i
< 4; i
++) {
537 if (test_access(i
, stp
))
538 nfs4_file_put_access(stp
->st_file
,
539 nfs4_access_to_omode(i
));
540 clear_access(i
, stp
);
544 static void unhash_generic_stateid(struct nfs4_ol_stateid
*stp
)
546 list_del(&stp
->st_perfile
);
547 list_del(&stp
->st_perstateowner
);
550 static void close_generic_stateid(struct nfs4_ol_stateid
*stp
)
552 release_all_access(stp
);
553 put_nfs4_file(stp
->st_file
);
557 static void free_generic_stateid(struct nfs4_ol_stateid
*stp
)
559 kmem_cache_free(stateid_slab
, stp
);
562 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
566 unhash_generic_stateid(stp
);
567 unhash_stid(&stp
->st_stid
);
568 file
= find_any_file(stp
->st_file
);
570 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
571 close_generic_stateid(stp
);
572 free_generic_stateid(stp
);
575 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
577 struct nfs4_ol_stateid
*stp
;
579 list_del(&lo
->lo_owner
.so_strhash
);
580 list_del(&lo
->lo_perstateid
);
581 list_del(&lo
->lo_owner_ino_hash
);
582 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
583 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
584 struct nfs4_ol_stateid
, st_perstateowner
);
585 release_lock_stateid(stp
);
589 static void release_lockowner(struct nfs4_lockowner
*lo
)
591 unhash_lockowner(lo
);
592 nfs4_free_lockowner(lo
);
596 release_stateid_lockowners(struct nfs4_ol_stateid
*open_stp
)
598 struct nfs4_lockowner
*lo
;
600 while (!list_empty(&open_stp
->st_lockowners
)) {
601 lo
= list_entry(open_stp
->st_lockowners
.next
,
602 struct nfs4_lockowner
, lo_perstateid
);
603 release_lockowner(lo
);
607 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
)
609 unhash_generic_stateid(stp
);
610 release_stateid_lockowners(stp
);
611 close_generic_stateid(stp
);
614 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
616 unhash_open_stateid(stp
);
617 unhash_stid(&stp
->st_stid
);
618 free_generic_stateid(stp
);
621 static void unhash_openowner(struct nfs4_openowner
*oo
)
623 struct nfs4_ol_stateid
*stp
;
625 list_del(&oo
->oo_owner
.so_strhash
);
626 list_del(&oo
->oo_perclient
);
627 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
628 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
629 struct nfs4_ol_stateid
, st_perstateowner
);
630 release_open_stateid(stp
);
634 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
636 struct nfs4_ol_stateid
*s
= oo
->oo_last_closed_stid
;
639 unhash_stid(&s
->st_stid
);
640 free_generic_stateid(s
);
641 oo
->oo_last_closed_stid
= NULL
;
645 static void release_openowner(struct nfs4_openowner
*oo
)
647 unhash_openowner(oo
);
648 list_del(&oo
->oo_close_lru
);
649 release_last_closed_stateid(oo
);
650 nfs4_free_openowner(oo
);
653 #define SESSION_HASH_SIZE 512
654 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
657 hash_sessionid(struct nfs4_sessionid
*sessionid
)
659 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
661 return sid
->sequence
% SESSION_HASH_SIZE
;
666 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
668 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
669 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
673 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
680 gen_sessionid(struct nfsd4_session
*ses
)
682 struct nfs4_client
*clp
= ses
->se_client
;
683 struct nfsd4_sessionid
*sid
;
685 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
686 sid
->clientid
= clp
->cl_clientid
;
687 sid
->sequence
= current_sessionid
++;
692 * The protocol defines ca_maxresponssize_cached to include the size of
693 * the rpc header, but all we need to cache is the data starting after
694 * the end of the initial SEQUENCE operation--the rest we regenerate
695 * each time. Therefore we can advertise a ca_maxresponssize_cached
696 * value that is the number of bytes in our cache plus a few additional
697 * bytes. In order to stay on the safe side, and not promise more than
698 * we can cache, those additional bytes must be the minimum possible: 24
699 * bytes of rpc header (xid through accept state, with AUTH_NULL
700 * verifier), 12 for the compound header (with zero-length tag), and 44
701 * for the SEQUENCE op response:
703 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
706 free_session_slots(struct nfsd4_session
*ses
)
710 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
711 kfree(ses
->se_slots
[i
]);
715 * We don't actually need to cache the rpc and session headers, so we
716 * can allocate a little less for each slot:
718 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
720 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
723 static int nfsd4_sanitize_slot_size(u32 size
)
725 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
726 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
732 * XXX: If we run out of reserved DRC memory we could (up to a point)
733 * re-negotiate active sessions and reduce their slot usage to make
734 * room for new connections. For now we just fail the create session.
736 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
740 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
742 spin_lock(&nfsd_drc_lock
);
743 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
744 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
745 num
= min_t(int, num
, avail
/ slotsize
);
746 nfsd_drc_mem_used
+= num
* slotsize
;
747 spin_unlock(&nfsd_drc_lock
);
752 static void nfsd4_put_drc_mem(int slotsize
, int num
)
754 spin_lock(&nfsd_drc_lock
);
755 nfsd_drc_mem_used
-= slotsize
* num
;
756 spin_unlock(&nfsd_drc_lock
);
759 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
761 struct nfsd4_session
*new;
764 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
765 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
766 mem
= numslots
* sizeof(struct nfsd4_slot
*);
768 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
771 /* allocate each struct nfsd4_slot and data cache in one piece */
772 for (i
= 0; i
< numslots
; i
++) {
773 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
774 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
775 if (!new->se_slots
[i
])
781 kfree(new->se_slots
[i
]);
786 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
788 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
790 new->maxreqs
= numslots
;
791 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
792 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
793 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
794 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
795 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
798 static void free_conn(struct nfsd4_conn
*c
)
800 svc_xprt_put(c
->cn_xprt
);
804 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
806 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
807 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
809 spin_lock(&clp
->cl_lock
);
810 if (!list_empty(&c
->cn_persession
)) {
811 list_del(&c
->cn_persession
);
814 spin_unlock(&clp
->cl_lock
);
815 nfsd4_probe_callback(clp
);
818 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
820 struct nfsd4_conn
*conn
;
822 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
825 svc_xprt_get(rqstp
->rq_xprt
);
826 conn
->cn_xprt
= rqstp
->rq_xprt
;
827 conn
->cn_flags
= flags
;
828 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
832 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
834 conn
->cn_session
= ses
;
835 list_add(&conn
->cn_persession
, &ses
->se_conns
);
838 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
840 struct nfs4_client
*clp
= ses
->se_client
;
842 spin_lock(&clp
->cl_lock
);
843 __nfsd4_hash_conn(conn
, ses
);
844 spin_unlock(&clp
->cl_lock
);
847 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
849 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
850 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
853 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
855 struct nfsd4_conn
*conn
;
858 conn
= alloc_conn(rqstp
, dir
);
860 return nfserr_jukebox
;
861 nfsd4_hash_conn(conn
, ses
);
862 ret
= nfsd4_register_conn(conn
);
864 /* oops; xprt is already down: */
865 nfsd4_conn_lost(&conn
->cn_xpt_user
);
866 if (ses
->se_client
->cl_cb_state
== NFSD4_CB_DOWN
&&
867 dir
& NFS4_CDFC4_BACK
) {
868 /* callback channel may be back up */
869 nfsd4_probe_callback(ses
->se_client
);
874 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
876 u32 dir
= NFS4_CDFC4_FORE
;
878 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
879 dir
|= NFS4_CDFC4_BACK
;
881 return nfsd4_new_conn(rqstp
, ses
, dir
);
884 /* must be called under client_lock */
885 static void nfsd4_del_conns(struct nfsd4_session
*s
)
887 struct nfs4_client
*clp
= s
->se_client
;
888 struct nfsd4_conn
*c
;
890 spin_lock(&clp
->cl_lock
);
891 while (!list_empty(&s
->se_conns
)) {
892 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
893 list_del_init(&c
->cn_persession
);
894 spin_unlock(&clp
->cl_lock
);
896 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
899 spin_lock(&clp
->cl_lock
);
901 spin_unlock(&clp
->cl_lock
);
904 static void free_session(struct kref
*kref
)
906 struct nfsd4_session
*ses
;
909 lockdep_assert_held(&client_lock
);
910 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
911 nfsd4_del_conns(ses
);
912 spin_lock(&nfsd_drc_lock
);
913 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
914 nfsd_drc_mem_used
-= mem
;
915 spin_unlock(&nfsd_drc_lock
);
916 free_session_slots(ses
);
920 void nfsd4_put_session(struct nfsd4_session
*ses
)
922 spin_lock(&client_lock
);
923 nfsd4_put_session_locked(ses
);
924 spin_unlock(&client_lock
);
927 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
929 struct nfsd4_session
*new;
930 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
931 int numslots
, slotsize
;
936 * Note decreasing slot size below client's request may
937 * make it difficult for client to function correctly, whereas
938 * decreasing the number of slots will (just?) affect
939 * performance. When short on memory we therefore prefer to
940 * decrease number of slots instead of their size.
942 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
943 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
947 new = alloc_session(slotsize
, numslots
);
949 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
952 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
954 new->se_client
= clp
;
957 INIT_LIST_HEAD(&new->se_conns
);
959 new->se_cb_seq_nr
= 1;
960 new->se_flags
= cses
->flags
;
961 new->se_cb_prog
= cses
->callback_prog
;
962 kref_init(&new->se_ref
);
963 idx
= hash_sessionid(&new->se_sessionid
);
964 spin_lock(&client_lock
);
965 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
966 spin_lock(&clp
->cl_lock
);
967 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
968 spin_unlock(&clp
->cl_lock
);
969 spin_unlock(&client_lock
);
971 status
= nfsd4_new_conn_from_crses(rqstp
, new);
972 /* whoops: benny points out, status is ignored! (err, or bogus) */
974 spin_lock(&client_lock
);
975 free_session(&new->se_ref
);
976 spin_unlock(&client_lock
);
979 if (cses
->flags
& SESSION4_BACK_CHAN
) {
980 struct sockaddr
*sa
= svc_addr(rqstp
);
982 * This is a little silly; with sessions there's no real
983 * use for the callback address. Use the peer address
984 * as a reasonable default for now, but consider fixing
985 * the rpc client not to require an address in the
988 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
989 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
991 nfsd4_probe_callback(clp
);
995 /* caller must hold client_lock */
996 static struct nfsd4_session
*
997 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
999 struct nfsd4_session
*elem
;
1002 dump_sessionid(__func__
, sessionid
);
1003 idx
= hash_sessionid(sessionid
);
1004 /* Search in the appropriate list */
1005 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
1006 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1007 NFS4_MAX_SESSIONID_LEN
)) {
1012 dprintk("%s: session not found\n", __func__
);
1016 /* caller must hold client_lock */
1018 unhash_session(struct nfsd4_session
*ses
)
1020 list_del(&ses
->se_hash
);
1021 spin_lock(&ses
->se_client
->cl_lock
);
1022 list_del(&ses
->se_perclnt
);
1023 spin_unlock(&ses
->se_client
->cl_lock
);
1026 /* must be called under the client_lock */
1028 renew_client_locked(struct nfs4_client
*clp
)
1030 if (is_client_expired(clp
)) {
1032 printk("%s: client (clientid %08x/%08x) already expired\n",
1034 clp
->cl_clientid
.cl_boot
,
1035 clp
->cl_clientid
.cl_id
);
1039 dprintk("renewing client (clientid %08x/%08x)\n",
1040 clp
->cl_clientid
.cl_boot
,
1041 clp
->cl_clientid
.cl_id
);
1042 list_move_tail(&clp
->cl_lru
, &client_lru
);
1043 clp
->cl_time
= get_seconds();
1047 renew_client(struct nfs4_client
*clp
)
1049 spin_lock(&client_lock
);
1050 renew_client_locked(clp
);
1051 spin_unlock(&client_lock
);
1054 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1056 STALE_CLIENTID(clientid_t
*clid
)
1058 if (clid
->cl_boot
== boot_time
)
1060 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1061 clid
->cl_boot
, clid
->cl_id
, boot_time
);
1066 * XXX Should we use a slab cache ?
1067 * This type of memory management is somewhat inefficient, but we use it
1068 * anyway since SETCLIENTID is not a common operation.
1070 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1072 struct nfs4_client
*clp
;
1074 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1077 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1078 if (clp
->cl_name
.data
== NULL
) {
1082 clp
->cl_name
.len
= name
.len
;
1087 free_client(struct nfs4_client
*clp
)
1089 lockdep_assert_held(&client_lock
);
1090 while (!list_empty(&clp
->cl_sessions
)) {
1091 struct nfsd4_session
*ses
;
1092 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1094 list_del(&ses
->se_perclnt
);
1095 nfsd4_put_session_locked(ses
);
1097 free_svc_cred(&clp
->cl_cred
);
1098 kfree(clp
->cl_name
.data
);
1103 release_session_client(struct nfsd4_session
*session
)
1105 struct nfs4_client
*clp
= session
->se_client
;
1107 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
1109 if (is_client_expired(clp
)) {
1111 session
->se_client
= NULL
;
1113 renew_client_locked(clp
);
1114 spin_unlock(&client_lock
);
1117 /* must be called under the client_lock */
1119 unhash_client_locked(struct nfs4_client
*clp
)
1121 struct nfsd4_session
*ses
;
1123 mark_client_expired(clp
);
1124 list_del(&clp
->cl_lru
);
1125 spin_lock(&clp
->cl_lock
);
1126 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1127 list_del_init(&ses
->se_hash
);
1128 spin_unlock(&clp
->cl_lock
);
1132 expire_client(struct nfs4_client
*clp
)
1134 struct nfs4_openowner
*oo
;
1135 struct nfs4_delegation
*dp
;
1136 struct list_head reaplist
;
1138 INIT_LIST_HEAD(&reaplist
);
1139 spin_lock(&recall_lock
);
1140 while (!list_empty(&clp
->cl_delegations
)) {
1141 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1142 list_del_init(&dp
->dl_perclnt
);
1143 list_move(&dp
->dl_recall_lru
, &reaplist
);
1145 spin_unlock(&recall_lock
);
1146 while (!list_empty(&reaplist
)) {
1147 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1148 unhash_delegation(dp
);
1150 while (!list_empty(&clp
->cl_openowners
)) {
1151 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1152 release_openowner(oo
);
1154 nfsd4_shutdown_callback(clp
);
1155 if (clp
->cl_cb_conn
.cb_xprt
)
1156 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1157 list_del(&clp
->cl_idhash
);
1158 list_del(&clp
->cl_strhash
);
1159 spin_lock(&client_lock
);
1160 unhash_client_locked(clp
);
1161 if (atomic_read(&clp
->cl_refcount
) == 0)
1163 spin_unlock(&client_lock
);
1166 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1168 memcpy(target
->cl_verifier
.data
, source
->data
,
1169 sizeof(target
->cl_verifier
.data
));
1172 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1174 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1175 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1178 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1180 if (source
->cr_principal
) {
1181 target
->cr_principal
=
1182 kstrdup(source
->cr_principal
, GFP_KERNEL
);
1183 if (target
->cr_principal
== NULL
)
1186 target
->cr_principal
= NULL
;
1187 target
->cr_flavor
= source
->cr_flavor
;
1188 target
->cr_uid
= source
->cr_uid
;
1189 target
->cr_gid
= source
->cr_gid
;
1190 target
->cr_group_info
= source
->cr_group_info
;
1191 get_group_info(target
->cr_group_info
);
1195 static int same_name(const char *n1
, const char *n2
)
1197 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1201 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1203 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1207 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1209 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1212 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
1216 if (g1
->ngroups
!= g2
->ngroups
)
1218 for (i
=0; i
<g1
->ngroups
; i
++)
1219 if (GROUP_AT(g1
, i
) != GROUP_AT(g2
, i
))
1225 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1227 if ((cr1
->cr_flavor
!= cr2
->cr_flavor
)
1228 || (cr1
->cr_uid
!= cr2
->cr_uid
)
1229 || (cr1
->cr_gid
!= cr2
->cr_gid
)
1230 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
1232 if (cr1
->cr_principal
== cr2
->cr_principal
)
1234 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
1236 return 0 == strcmp(cr1
->cr_principal
, cr1
->cr_principal
);
1239 static void gen_clid(struct nfs4_client
*clp
)
1241 static u32 current_clientid
= 1;
1243 clp
->cl_clientid
.cl_boot
= boot_time
;
1244 clp
->cl_clientid
.cl_id
= current_clientid
++;
1247 static void gen_confirm(struct nfs4_client
*clp
)
1252 verf
[0] = (__be32
)get_seconds();
1253 verf
[1] = (__be32
)i
++;
1254 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
1257 static struct nfs4_stid
*find_stateid(struct nfs4_client
*cl
, stateid_t
*t
)
1259 return idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1262 static struct nfs4_stid
*find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1264 struct nfs4_stid
*s
;
1266 s
= find_stateid(cl
, t
);
1269 if (typemask
& s
->sc_type
)
1274 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1275 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1277 struct nfs4_client
*clp
;
1278 struct sockaddr
*sa
= svc_addr(rqstp
);
1281 clp
= alloc_client(name
);
1285 INIT_LIST_HEAD(&clp
->cl_sessions
);
1286 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1288 spin_lock(&client_lock
);
1290 spin_unlock(&client_lock
);
1293 idr_init(&clp
->cl_stateids
);
1294 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1295 atomic_set(&clp
->cl_refcount
, 0);
1296 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1297 INIT_LIST_HEAD(&clp
->cl_idhash
);
1298 INIT_LIST_HEAD(&clp
->cl_strhash
);
1299 INIT_LIST_HEAD(&clp
->cl_openowners
);
1300 INIT_LIST_HEAD(&clp
->cl_delegations
);
1301 INIT_LIST_HEAD(&clp
->cl_lru
);
1302 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1303 spin_lock_init(&clp
->cl_lock
);
1304 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1305 clp
->cl_time
= get_seconds();
1306 clear_bit(0, &clp
->cl_cb_slot_busy
);
1307 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1308 copy_verf(clp
, verf
);
1309 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1311 clp
->cl_cb_session
= NULL
;
1316 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1318 unsigned int idhashval
;
1320 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1321 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1322 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1327 move_to_confirmed(struct nfs4_client
*clp
)
1329 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1330 unsigned int strhashval
;
1332 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1333 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1334 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1335 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1339 static struct nfs4_client
*
1340 find_confirmed_client(clientid_t
*clid
)
1342 struct nfs4_client
*clp
;
1343 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1345 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1346 if (same_clid(&clp
->cl_clientid
, clid
)) {
1354 static struct nfs4_client
*
1355 find_unconfirmed_client(clientid_t
*clid
)
1357 struct nfs4_client
*clp
;
1358 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1360 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1361 if (same_clid(&clp
->cl_clientid
, clid
))
1367 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1369 return clp
->cl_exchange_flags
!= 0;
1372 static struct nfs4_client
*
1373 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1375 struct nfs4_client
*clp
;
1377 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1378 if (same_name(clp
->cl_recdir
, dname
))
1384 static struct nfs4_client
*
1385 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1387 struct nfs4_client
*clp
;
1389 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1390 if (same_name(clp
->cl_recdir
, dname
))
1397 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1399 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1400 struct sockaddr
*sa
= svc_addr(rqstp
);
1401 u32 scopeid
= rpc_get_scope_id(sa
);
1402 unsigned short expected_family
;
1404 /* Currently, we only support tcp and tcp6 for the callback channel */
1405 if (se
->se_callback_netid_len
== 3 &&
1406 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1407 expected_family
= AF_INET
;
1408 else if (se
->se_callback_netid_len
== 4 &&
1409 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1410 expected_family
= AF_INET6
;
1414 conn
->cb_addrlen
= rpc_uaddr2sockaddr(&init_net
, se
->se_callback_addr_val
,
1415 se
->se_callback_addr_len
,
1416 (struct sockaddr
*)&conn
->cb_addr
,
1417 sizeof(conn
->cb_addr
));
1419 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1422 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1423 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1425 conn
->cb_prog
= se
->se_callback_prog
;
1426 conn
->cb_ident
= se
->se_callback_ident
;
1427 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
1430 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1431 conn
->cb_addrlen
= 0;
1432 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1433 "will not receive delegations\n",
1434 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1440 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1443 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1445 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1448 dprintk("--> %s slot %p\n", __func__
, slot
);
1450 slot
->sl_opcnt
= resp
->opcnt
;
1451 slot
->sl_status
= resp
->cstate
.status
;
1453 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
1454 if (nfsd4_not_cached(resp
)) {
1455 slot
->sl_datalen
= 0;
1458 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1459 base
= (char *)resp
->cstate
.datap
-
1460 (char *)resp
->xbuf
->head
[0].iov_base
;
1461 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1463 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1468 * Encode the replay sequence operation from the slot values.
1469 * If cachethis is FALSE encode the uncached rep error on the next
1470 * operation which sets resp->p and increments resp->opcnt for
1471 * nfs4svc_encode_compoundres.
1475 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1476 struct nfsd4_compoundres
*resp
)
1478 struct nfsd4_op
*op
;
1479 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1481 /* Encode the replayed sequence operation */
1482 op
= &args
->ops
[resp
->opcnt
- 1];
1483 nfsd4_encode_operation(resp
, op
);
1485 /* Return nfserr_retry_uncached_rep in next operation. */
1486 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
1487 op
= &args
->ops
[resp
->opcnt
++];
1488 op
->status
= nfserr_retry_uncached_rep
;
1489 nfsd4_encode_operation(resp
, op
);
1495 * The sequence operation is not cached because we can use the slot and
1499 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1500 struct nfsd4_sequence
*seq
)
1502 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1505 dprintk("--> %s slot %p\n", __func__
, slot
);
1507 /* Either returns 0 or nfserr_retry_uncached */
1508 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1509 if (status
== nfserr_retry_uncached_rep
)
1512 /* The sequence operation has been encoded, cstate->datap set. */
1513 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1515 resp
->opcnt
= slot
->sl_opcnt
;
1516 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1517 status
= slot
->sl_status
;
1523 * Set the exchange_id flags returned by the server.
1526 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1528 /* pNFS is not supported */
1529 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1531 /* Referrals are supported, Migration is not. */
1532 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1534 /* set the wire flags to return to client. */
1535 clid
->flags
= new->cl_exchange_flags
;
1538 static bool client_has_state(struct nfs4_client
*clp
)
1541 * Note clp->cl_openowners check isn't quite right: there's no
1542 * need to count owners without stateid's.
1544 * Also note we should probably be using this in 4.0 case too.
1546 return !list_empty(&clp
->cl_openowners
)
1547 || !list_empty(&clp
->cl_delegations
)
1548 || !list_empty(&clp
->cl_sessions
);
1552 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1553 struct nfsd4_compound_state
*cstate
,
1554 struct nfsd4_exchange_id
*exid
)
1556 struct nfs4_client
*unconf
, *conf
, *new;
1558 unsigned int strhashval
;
1559 char dname
[HEXDIR_LEN
];
1560 char addr_str
[INET6_ADDRSTRLEN
];
1561 nfs4_verifier verf
= exid
->verifier
;
1562 struct sockaddr
*sa
= svc_addr(rqstp
);
1563 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
1565 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1566 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1567 "ip_addr=%s flags %x, spa_how %d\n",
1568 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1569 addr_str
, exid
->flags
, exid
->spa_how
);
1571 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
1572 return nfserr_inval
;
1574 /* Currently only support SP4_NONE */
1575 switch (exid
->spa_how
) {
1579 return nfserr_serverfault
;
1581 BUG(); /* checked by xdr code */
1583 return nfserr_serverfault
; /* no excuse :-/ */
1586 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1591 strhashval
= clientstr_hashval(dname
);
1593 /* Cases below refer to rfc 5661 section 18.35.4: */
1595 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1597 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
1598 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
1601 if (!clp_used_exchangeid(conf
)) { /* buggy client */
1602 status
= nfserr_inval
;
1605 if (!creds_match
) { /* case 9 */
1606 status
= nfserr_perm
;
1609 if (!verfs_match
) { /* case 8 */
1610 status
= nfserr_not_same
;
1614 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1618 if (!creds_match
) { /* case 3 */
1619 if (client_has_state(conf
)) {
1620 status
= nfserr_clid_inuse
;
1623 expire_client(conf
);
1626 if (verfs_match
) { /* case 2 */
1627 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1631 /* case 5, client reboot */
1635 if (update
) { /* case 7 */
1636 status
= nfserr_noent
;
1640 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1641 if (unconf
) /* case 4, possible retry or client restart */
1642 expire_client(unconf
);
1644 /* case 1 (normal case) */
1646 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1648 status
= nfserr_jukebox
;
1653 add_to_unconfirmed(new, strhashval
);
1655 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1656 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1658 exid
->seqid
= new->cl_cs_slot
.sl_seqid
+ 1;
1659 nfsd4_set_ex_flags(new, exid
);
1661 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1662 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1666 nfs4_unlock_state();
1671 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1673 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1676 /* The slot is in use, and no response has been sent. */
1678 if (seqid
== slot_seqid
)
1679 return nfserr_jukebox
;
1681 return nfserr_seq_misordered
;
1683 /* Note unsigned 32-bit arithmetic handles wraparound: */
1684 if (likely(seqid
== slot_seqid
+ 1))
1686 if (seqid
== slot_seqid
)
1687 return nfserr_replay_cache
;
1688 return nfserr_seq_misordered
;
1692 * Cache the create session result into the create session single DRC
1693 * slot cache by saving the xdr structure. sl_seqid has been set.
1694 * Do this for solo or embedded create session operations.
1697 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1698 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
1700 slot
->sl_status
= nfserr
;
1701 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1705 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1706 struct nfsd4_clid_slot
*slot
)
1708 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1709 return slot
->sl_status
;
1712 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1713 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1714 1 + /* MIN tag is length with zero, only length */ \
1715 3 + /* version, opcount, opcode */ \
1716 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1717 /* seqid, slotID, slotID, cache */ \
1718 4 ) * sizeof(__be32))
1720 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1721 2 + /* verifier: AUTH_NULL, length 0 */\
1723 1 + /* MIN tag is length with zero, only length */ \
1724 3 + /* opcount, opcode, opstatus*/ \
1725 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1726 /* seqid, slotID, slotID, slotID, status */ \
1727 5 ) * sizeof(__be32))
1729 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1731 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1732 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1736 nfsd4_create_session(struct svc_rqst
*rqstp
,
1737 struct nfsd4_compound_state
*cstate
,
1738 struct nfsd4_create_session
*cr_ses
)
1740 struct sockaddr
*sa
= svc_addr(rqstp
);
1741 struct nfs4_client
*conf
, *unconf
;
1742 struct nfsd4_session
*new;
1743 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1744 bool confirm_me
= false;
1747 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1748 return nfserr_inval
;
1751 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1752 conf
= find_confirmed_client(&cr_ses
->clientid
);
1755 cs_slot
= &conf
->cl_cs_slot
;
1756 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1757 if (status
== nfserr_replay_cache
) {
1758 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1760 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1761 status
= nfserr_seq_misordered
;
1764 } else if (unconf
) {
1765 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1766 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1767 status
= nfserr_clid_inuse
;
1770 cs_slot
= &unconf
->cl_cs_slot
;
1771 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1773 /* an unconfirmed replay returns misordered */
1774 status
= nfserr_seq_misordered
;
1780 status
= nfserr_stale_clientid
;
1785 * XXX: we should probably set this at creation time, and check
1786 * for consistent minorversion use throughout:
1788 conf
->cl_minorversion
= 1;
1790 * We do not support RDMA or persistent sessions
1792 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1793 cr_ses
->flags
&= ~SESSION4_RDMA
;
1795 status
= nfserr_toosmall
;
1796 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1799 status
= nfserr_jukebox
;
1800 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1804 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1805 NFS4_MAX_SESSIONID_LEN
);
1806 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1807 sizeof(struct nfsd4_channel_attrs
));
1808 cs_slot
->sl_seqid
++;
1809 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1811 /* cache solo and embedded create sessions under the state lock */
1812 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1814 unsigned int hash
= clientstr_hashval(unconf
->cl_recdir
);
1815 struct nfs4_client
*old
=
1816 find_confirmed_client_by_str(conf
->cl_recdir
, hash
);
1819 move_to_confirmed(conf
);
1822 nfs4_unlock_state();
1823 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1827 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1829 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1830 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1832 return argp
->opcnt
== resp
->opcnt
;
1835 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1838 case NFS4_CDFC4_FORE
:
1839 case NFS4_CDFC4_BACK
:
1841 case NFS4_CDFC4_FORE_OR_BOTH
:
1842 case NFS4_CDFC4_BACK_OR_BOTH
:
1843 *dir
= NFS4_CDFC4_BOTH
;
1846 return nfserr_inval
;
1849 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1850 struct nfsd4_compound_state
*cstate
,
1851 struct nfsd4_bind_conn_to_session
*bcts
)
1855 if (!nfsd4_last_compound_op(rqstp
))
1856 return nfserr_not_only_op
;
1857 spin_lock(&client_lock
);
1858 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1859 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1860 * client_lock iself: */
1861 if (cstate
->session
) {
1862 nfsd4_get_session(cstate
->session
);
1863 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1865 spin_unlock(&client_lock
);
1866 if (!cstate
->session
)
1867 return nfserr_badsession
;
1869 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1871 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1875 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1879 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1883 nfsd4_destroy_session(struct svc_rqst
*r
,
1884 struct nfsd4_compound_state
*cstate
,
1885 struct nfsd4_destroy_session
*sessionid
)
1887 struct nfsd4_session
*ses
;
1888 __be32 status
= nfserr_badsession
;
1891 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1892 * - Should we return nfserr_back_chan_busy if waiting for
1893 * callbacks on to-be-destroyed session?
1894 * - Do we need to clear any callback info from previous session?
1897 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1898 if (!nfsd4_last_compound_op(r
))
1899 return nfserr_not_only_op
;
1901 dump_sessionid(__func__
, &sessionid
->sessionid
);
1902 spin_lock(&client_lock
);
1903 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1905 spin_unlock(&client_lock
);
1909 unhash_session(ses
);
1910 spin_unlock(&client_lock
);
1913 nfsd4_probe_callback_sync(ses
->se_client
);
1914 nfs4_unlock_state();
1916 spin_lock(&client_lock
);
1917 nfsd4_del_conns(ses
);
1918 nfsd4_put_session_locked(ses
);
1919 spin_unlock(&client_lock
);
1922 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1926 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1928 struct nfsd4_conn
*c
;
1930 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1931 if (c
->cn_xprt
== xpt
) {
1938 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1940 struct nfs4_client
*clp
= ses
->se_client
;
1941 struct nfsd4_conn
*c
;
1944 spin_lock(&clp
->cl_lock
);
1945 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1947 spin_unlock(&clp
->cl_lock
);
1951 __nfsd4_hash_conn(new, ses
);
1952 spin_unlock(&clp
->cl_lock
);
1953 ret
= nfsd4_register_conn(new);
1955 /* oops; xprt is already down: */
1956 nfsd4_conn_lost(&new->cn_xpt_user
);
1960 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1962 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1964 return args
->opcnt
> session
->se_fchannel
.maxops
;
1967 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1968 struct nfsd4_session
*session
)
1970 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1972 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1976 nfsd4_sequence(struct svc_rqst
*rqstp
,
1977 struct nfsd4_compound_state
*cstate
,
1978 struct nfsd4_sequence
*seq
)
1980 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1981 struct nfsd4_session
*session
;
1982 struct nfsd4_slot
*slot
;
1983 struct nfsd4_conn
*conn
;
1986 if (resp
->opcnt
!= 1)
1987 return nfserr_sequence_pos
;
1990 * Will be either used or freed by nfsd4_sequence_check_conn
1993 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1995 return nfserr_jukebox
;
1997 spin_lock(&client_lock
);
1998 status
= nfserr_badsession
;
1999 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
2003 status
= nfserr_too_many_ops
;
2004 if (nfsd4_session_too_many_ops(rqstp
, session
))
2007 status
= nfserr_req_too_big
;
2008 if (nfsd4_request_too_big(rqstp
, session
))
2011 status
= nfserr_badslot
;
2012 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
2015 slot
= session
->se_slots
[seq
->slotid
];
2016 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
2018 /* We do not negotiate the number of slots yet, so set the
2019 * maxslots to the session maxreqs which is used to encode
2020 * sr_highest_slotid and the sr_target_slot id to maxslots */
2021 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
2023 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
2024 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
2025 if (status
== nfserr_replay_cache
) {
2026 status
= nfserr_seq_misordered
;
2027 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
2029 cstate
->slot
= slot
;
2030 cstate
->session
= session
;
2031 /* Return the cached reply status and set cstate->status
2032 * for nfsd4_proc_compound processing */
2033 status
= nfsd4_replay_cache_entry(resp
, seq
);
2034 cstate
->status
= nfserr_replay_cache
;
2040 nfsd4_sequence_check_conn(conn
, session
);
2043 /* Success! bump slot seqid */
2044 slot
->sl_seqid
= seq
->seqid
;
2045 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
2047 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
2049 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
2051 cstate
->slot
= slot
;
2052 cstate
->session
= session
;
2055 /* Hold a session reference until done processing the compound. */
2056 if (cstate
->session
) {
2057 struct nfs4_client
*clp
= session
->se_client
;
2059 nfsd4_get_session(cstate
->session
);
2060 atomic_inc(&clp
->cl_refcount
);
2061 switch (clp
->cl_cb_state
) {
2063 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
2065 case NFSD4_CB_FAULT
:
2066 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
2069 seq
->status_flags
= 0;
2073 spin_unlock(&client_lock
);
2074 dprintk("%s: return %d\n", __func__
, ntohl(status
));
2079 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
2081 struct nfs4_client
*conf
, *unconf
, *clp
;
2085 unconf
= find_unconfirmed_client(&dc
->clientid
);
2086 conf
= find_confirmed_client(&dc
->clientid
);
2091 if (!is_client_expired(conf
) && client_has_state(conf
)) {
2092 status
= nfserr_clientid_busy
;
2096 /* rfc5661 18.50.3 */
2097 if (cstate
->session
&& conf
== cstate
->session
->se_client
) {
2098 status
= nfserr_clientid_busy
;
2104 status
= nfserr_stale_clientid
;
2110 nfs4_unlock_state();
2111 dprintk("%s return %d\n", __func__
, ntohl(status
));
2116 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2120 if (rc
->rca_one_fs
) {
2121 if (!cstate
->current_fh
.fh_dentry
)
2122 return nfserr_nofilehandle
;
2124 * We don't take advantage of the rca_one_fs case.
2125 * That's OK, it's optional, we can safely ignore it.
2131 status
= nfserr_complete_already
;
2132 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
2133 &cstate
->session
->se_client
->cl_flags
))
2136 status
= nfserr_stale_clientid
;
2137 if (is_client_expired(cstate
->session
->se_client
))
2139 * The following error isn't really legal.
2140 * But we only get here if the client just explicitly
2141 * destroyed the client. Surely it no longer cares what
2142 * error it gets back on an operation for the dead
2148 nfsd4_client_record_create(cstate
->session
->se_client
);
2150 nfs4_unlock_state();
2155 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2156 struct nfsd4_setclientid
*setclid
)
2158 struct xdr_netobj clname
= setclid
->se_name
;
2159 nfs4_verifier clverifier
= setclid
->se_verf
;
2160 unsigned int strhashval
;
2161 struct nfs4_client
*conf
, *unconf
, *new;
2163 char dname
[HEXDIR_LEN
];
2165 status
= nfs4_make_rec_clidname(dname
, &clname
);
2169 strhashval
= clientstr_hashval(dname
);
2171 /* Cases below refer to rfc 3530 section 14.2.33: */
2173 conf
= find_confirmed_client_by_str(dname
, strhashval
);
2176 status
= nfserr_clid_inuse
;
2177 if (clp_used_exchangeid(conf
))
2179 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
2180 char addr_str
[INET6_ADDRSTRLEN
];
2181 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
2183 dprintk("NFSD: setclientid: string in use by client "
2184 "at %s\n", addr_str
);
2188 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
2190 expire_client(unconf
);
2191 status
= nfserr_jukebox
;
2192 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2195 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
))
2196 /* case 1: probable callback update */
2197 copy_clid(new, conf
);
2198 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2201 * XXX: we should probably set this at creation time, and check
2202 * for consistent minorversion use throughout:
2204 new->cl_minorversion
= 0;
2205 gen_callback(new, setclid
, rqstp
);
2206 add_to_unconfirmed(new, strhashval
);
2207 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2208 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2209 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2212 nfs4_unlock_state();
2218 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2219 struct nfsd4_compound_state
*cstate
,
2220 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2222 struct nfs4_client
*conf
, *unconf
;
2223 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2224 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2227 if (STALE_CLIENTID(clid
))
2228 return nfserr_stale_clientid
;
2231 conf
= find_confirmed_client(clid
);
2232 unconf
= find_unconfirmed_client(clid
);
2234 * We try hard to give out unique clientid's, so if we get an
2235 * attempt to confirm the same clientid with a different cred,
2236 * there's a bug somewhere. Let's charitably assume it's our
2239 status
= nfserr_serverfault
;
2240 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
2242 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2244 /* cases below refer to rfc 3530 section 14.2.34: */
2245 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
2246 if (conf
&& !unconf
) /* case 2: probable retransmit */
2248 else /* case 4: client hasn't noticed we rebooted yet? */
2249 status
= nfserr_stale_clientid
;
2253 if (conf
) { /* case 1: callback update */
2254 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2255 nfsd4_probe_callback(conf
);
2256 expire_client(unconf
);
2257 } else { /* case 3: normal case; new or rebooted client */
2258 unsigned int hash
= clientstr_hashval(unconf
->cl_recdir
);
2260 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
, hash
);
2262 nfsd4_client_record_remove(conf
);
2263 expire_client(conf
);
2265 move_to_confirmed(unconf
);
2266 nfsd4_probe_callback(unconf
);
2269 nfs4_unlock_state();
2273 static struct nfs4_file
*nfsd4_alloc_file(void)
2275 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2278 /* OPEN Share state helper functions */
2279 static void nfsd4_init_file(struct nfs4_file
*fp
, struct inode
*ino
)
2281 unsigned int hashval
= file_hashval(ino
);
2283 atomic_set(&fp
->fi_ref
, 1);
2284 INIT_LIST_HEAD(&fp
->fi_hash
);
2285 INIT_LIST_HEAD(&fp
->fi_stateids
);
2286 INIT_LIST_HEAD(&fp
->fi_delegations
);
2287 fp
->fi_inode
= igrab(ino
);
2288 fp
->fi_had_conflict
= false;
2289 fp
->fi_lease
= NULL
;
2290 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2291 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2292 spin_lock(&recall_lock
);
2293 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2294 spin_unlock(&recall_lock
);
2298 nfsd4_free_slab(struct kmem_cache
**slab
)
2302 kmem_cache_destroy(*slab
);
2307 nfsd4_free_slabs(void)
2309 nfsd4_free_slab(&openowner_slab
);
2310 nfsd4_free_slab(&lockowner_slab
);
2311 nfsd4_free_slab(&file_slab
);
2312 nfsd4_free_slab(&stateid_slab
);
2313 nfsd4_free_slab(&deleg_slab
);
2317 nfsd4_init_slabs(void)
2319 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2320 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2321 if (openowner_slab
== NULL
)
2323 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2324 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2325 if (lockowner_slab
== NULL
)
2327 file_slab
= kmem_cache_create("nfsd4_files",
2328 sizeof(struct nfs4_file
), 0, 0, NULL
);
2329 if (file_slab
== NULL
)
2331 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2332 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
2333 if (stateid_slab
== NULL
)
2335 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2336 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2337 if (deleg_slab
== NULL
)
2342 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2346 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2348 kfree(oo
->oo_owner
.so_owner
.data
);
2349 kmem_cache_free(openowner_slab
, oo
);
2352 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2354 kfree(lo
->lo_owner
.so_owner
.data
);
2355 kmem_cache_free(lockowner_slab
, lo
);
2358 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2360 rp
->rp_status
= nfserr_serverfault
;
2362 rp
->rp_buf
= rp
->rp_ibuf
;
2365 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2367 struct nfs4_stateowner
*sop
;
2369 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2373 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2374 if (!sop
->so_owner
.data
) {
2375 kmem_cache_free(slab
, sop
);
2378 sop
->so_owner
.len
= owner
->len
;
2380 INIT_LIST_HEAD(&sop
->so_stateids
);
2381 sop
->so_client
= clp
;
2382 init_nfs4_replay(&sop
->so_replay
);
2386 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2388 list_add(&oo
->oo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2389 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2392 static struct nfs4_openowner
*
2393 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2394 struct nfs4_openowner
*oo
;
2396 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2399 oo
->oo_owner
.so_is_open_owner
= 1;
2400 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2401 oo
->oo_flags
= NFS4_OO_NEW
;
2403 oo
->oo_last_closed_stid
= NULL
;
2404 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2405 hash_openowner(oo
, clp
, strhashval
);
2409 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2410 struct nfs4_openowner
*oo
= open
->op_openowner
;
2411 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
2413 init_stid(&stp
->st_stid
, clp
, NFS4_OPEN_STID
);
2414 INIT_LIST_HEAD(&stp
->st_lockowners
);
2415 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2416 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2417 stp
->st_stateowner
= &oo
->oo_owner
;
2420 stp
->st_access_bmap
= 0;
2421 stp
->st_deny_bmap
= 0;
2422 set_access(open
->op_share_access
, stp
);
2423 set_deny(open
->op_share_deny
, stp
);
2424 stp
->st_openstp
= NULL
;
2428 move_to_close_lru(struct nfs4_openowner
*oo
)
2430 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2432 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2433 oo
->oo_time
= get_seconds();
2437 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2440 return (sop
->so_owner
.len
== owner
->len
) &&
2441 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2442 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2445 static struct nfs4_openowner
*
2446 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2448 struct nfs4_stateowner
*so
;
2449 struct nfs4_openowner
*oo
;
2451 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2452 if (!so
->so_is_open_owner
)
2454 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
)) {
2456 renew_client(oo
->oo_owner
.so_client
);
2463 /* search file_hashtbl[] for file */
2464 static struct nfs4_file
*
2465 find_file(struct inode
*ino
)
2467 unsigned int hashval
= file_hashval(ino
);
2468 struct nfs4_file
*fp
;
2470 spin_lock(&recall_lock
);
2471 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2472 if (fp
->fi_inode
== ino
) {
2474 spin_unlock(&recall_lock
);
2478 spin_unlock(&recall_lock
);
2483 * Called to check deny when READ with all zero stateid or
2484 * WRITE with all zero or all one stateid
2487 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2489 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2490 struct nfs4_file
*fp
;
2491 struct nfs4_ol_stateid
*stp
;
2494 dprintk("NFSD: nfs4_share_conflict\n");
2496 fp
= find_file(ino
);
2499 ret
= nfserr_locked
;
2500 /* Search for conflicting share reservations */
2501 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2502 if (test_deny(deny_type
, stp
) ||
2503 test_deny(NFS4_SHARE_DENY_BOTH
, stp
))
2512 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2514 /* We're assuming the state code never drops its reference
2515 * without first removing the lease. Since we're in this lease
2516 * callback (and since the lease code is serialized by the kernel
2517 * lock) we know the server hasn't removed the lease yet, we know
2518 * it's safe to take a reference: */
2519 atomic_inc(&dp
->dl_count
);
2521 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2523 /* only place dl_time is set. protected by lock_flocks*/
2524 dp
->dl_time
= get_seconds();
2526 nfsd4_cb_recall(dp
);
2529 /* Called from break_lease() with lock_flocks() held. */
2530 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2532 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2533 struct nfs4_delegation
*dp
;
2536 /* We assume break_lease is only called once per lease: */
2537 BUG_ON(fp
->fi_had_conflict
);
2539 * We don't want the locks code to timeout the lease for us;
2540 * we'll remove it ourself if a delegation isn't returned
2543 fl
->fl_break_time
= 0;
2545 spin_lock(&recall_lock
);
2546 fp
->fi_had_conflict
= true;
2547 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2548 nfsd_break_one_deleg(dp
);
2549 spin_unlock(&recall_lock
);
2553 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2556 return lease_modify(onlist
, arg
);
2561 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2562 .lm_break
= nfsd_break_deleg_cb
,
2563 .lm_change
= nfsd_change_deleg_cb
,
2566 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2568 if (nfsd4_has_session(cstate
))
2570 if (seqid
== so
->so_seqid
- 1)
2571 return nfserr_replay_me
;
2572 if (seqid
== so
->so_seqid
)
2574 return nfserr_bad_seqid
;
2578 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2579 struct nfsd4_open
*open
)
2581 clientid_t
*clientid
= &open
->op_clientid
;
2582 struct nfs4_client
*clp
= NULL
;
2583 unsigned int strhashval
;
2584 struct nfs4_openowner
*oo
= NULL
;
2587 if (STALE_CLIENTID(&open
->op_clientid
))
2588 return nfserr_stale_clientid
;
2590 * In case we need it later, after we've already created the
2591 * file and don't want to risk a further failure:
2593 open
->op_file
= nfsd4_alloc_file();
2594 if (open
->op_file
== NULL
)
2595 return nfserr_jukebox
;
2597 strhashval
= ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2598 oo
= find_openstateowner_str(strhashval
, open
);
2599 open
->op_openowner
= oo
;
2601 clp
= find_confirmed_client(clientid
);
2603 return nfserr_expired
;
2606 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
2607 /* Replace unconfirmed owners without checking for replay. */
2608 clp
= oo
->oo_owner
.so_client
;
2609 release_openowner(oo
);
2610 open
->op_openowner
= NULL
;
2613 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2616 clp
= oo
->oo_owner
.so_client
;
2619 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2621 return nfserr_jukebox
;
2622 open
->op_openowner
= oo
;
2624 open
->op_stp
= nfs4_alloc_stateid(clp
);
2626 return nfserr_jukebox
;
2630 static inline __be32
2631 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2633 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2634 return nfserr_openmode
;
2639 static int share_access_to_flags(u32 share_access
)
2641 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2644 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
2646 struct nfs4_stid
*ret
;
2648 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
2651 return delegstateid(ret
);
2654 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
2656 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
2657 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
2661 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2662 struct nfs4_delegation
**dp
)
2665 __be32 status
= nfserr_bad_stateid
;
2667 *dp
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
2670 flags
= share_access_to_flags(open
->op_share_access
);
2671 status
= nfs4_check_delegmode(*dp
, flags
);
2675 if (!nfsd4_is_deleg_cur(open
))
2679 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2684 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
**stpp
)
2686 struct nfs4_ol_stateid
*local
;
2687 struct nfs4_openowner
*oo
= open
->op_openowner
;
2689 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2690 /* ignore lock owners */
2691 if (local
->st_stateowner
->so_is_open_owner
== 0)
2693 /* remember if we have seen this open owner */
2694 if (local
->st_stateowner
== &oo
->oo_owner
)
2696 /* check for conflicting share reservations */
2697 if (!test_share(local
, open
))
2698 return nfserr_share_denied
;
2703 static void nfs4_free_stateid(struct nfs4_ol_stateid
*s
)
2705 kmem_cache_free(stateid_slab
, s
);
2708 static inline int nfs4_access_to_access(u32 nfs4_access
)
2712 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2713 flags
|= NFSD_MAY_READ
;
2714 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2715 flags
|= NFSD_MAY_WRITE
;
2719 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2720 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2723 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2724 int access
= nfs4_access_to_access(open
->op_share_access
);
2726 if (!fp
->fi_fds
[oflag
]) {
2727 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2728 &fp
->fi_fds
[oflag
]);
2732 nfs4_file_get_access(fp
, oflag
);
2737 static inline __be32
2738 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2739 struct nfsd4_open
*open
)
2741 struct iattr iattr
= {
2742 .ia_valid
= ATTR_SIZE
,
2745 if (!open
->op_truncate
)
2747 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2748 return nfserr_inval
;
2749 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2753 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
)
2755 u32 op_share_access
= open
->op_share_access
;
2759 new_access
= !test_access(op_share_access
, stp
);
2761 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2765 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2768 int oflag
= nfs4_access_to_omode(op_share_access
);
2769 nfs4_file_put_access(fp
, oflag
);
2773 /* remember the open */
2774 set_access(op_share_access
, stp
);
2775 set_deny(open
->op_share_deny
, stp
);
2782 nfs4_set_claim_prev(struct nfsd4_open
*open
, bool has_session
)
2784 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2787 /* Should we give out recallable state?: */
2788 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2790 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2793 * In the sessions case, since we don't have to establish a
2794 * separate connection for callbacks, we assume it's OK
2795 * until we hear otherwise:
2797 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2800 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2802 struct file_lock
*fl
;
2804 fl
= locks_alloc_lock();
2807 locks_init_lock(fl
);
2808 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2809 fl
->fl_flags
= FL_LEASE
;
2810 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2811 fl
->fl_end
= OFFSET_MAX
;
2812 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2813 fl
->fl_pid
= current
->tgid
;
2817 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2819 struct nfs4_file
*fp
= dp
->dl_file
;
2820 struct file_lock
*fl
;
2823 fl
= nfs4_alloc_init_lease(dp
, flag
);
2826 fl
->fl_file
= find_readable_file(fp
);
2827 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2828 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2830 list_del_init(&dp
->dl_perclnt
);
2831 locks_free_lock(fl
);
2835 fp
->fi_deleg_file
= fl
->fl_file
;
2836 get_file(fp
->fi_deleg_file
);
2837 atomic_set(&fp
->fi_delegees
, 1);
2838 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2842 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2844 struct nfs4_file
*fp
= dp
->dl_file
;
2847 return nfs4_setlease(dp
, flag
);
2848 spin_lock(&recall_lock
);
2849 if (fp
->fi_had_conflict
) {
2850 spin_unlock(&recall_lock
);
2853 atomic_inc(&fp
->fi_delegees
);
2854 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2855 spin_unlock(&recall_lock
);
2856 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2860 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
2862 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2863 if (status
== -EAGAIN
)
2864 open
->op_why_no_deleg
= WND4_CONTENTION
;
2866 open
->op_why_no_deleg
= WND4_RESOURCE
;
2867 switch (open
->op_deleg_want
) {
2868 case NFS4_SHARE_WANT_READ_DELEG
:
2869 case NFS4_SHARE_WANT_WRITE_DELEG
:
2870 case NFS4_SHARE_WANT_ANY_DELEG
:
2872 case NFS4_SHARE_WANT_CANCEL
:
2873 open
->op_why_no_deleg
= WND4_CANCELLED
;
2875 case NFS4_SHARE_WANT_NO_DELEG
:
2876 BUG(); /* not supposed to get here */
2882 * Attempt to hand out a delegation.
2885 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
*stp
)
2887 struct nfs4_delegation
*dp
;
2888 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2890 int status
= 0, flag
= 0;
2892 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2893 flag
= NFS4_OPEN_DELEGATE_NONE
;
2894 open
->op_recall
= 0;
2895 switch (open
->op_claim_type
) {
2896 case NFS4_OPEN_CLAIM_PREVIOUS
:
2898 open
->op_recall
= 1;
2899 flag
= open
->op_delegate_type
;
2900 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2903 case NFS4_OPEN_CLAIM_NULL
:
2904 /* Let's not give out any delegations till everyone's
2905 * had the chance to reclaim theirs.... */
2906 if (locks_in_grace())
2908 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
2910 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2911 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2913 flag
= NFS4_OPEN_DELEGATE_READ
;
2919 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2922 status
= nfs4_set_delegation(dp
, flag
);
2926 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
2928 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2929 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
2931 open
->op_delegate_type
= flag
;
2932 if (flag
== NFS4_OPEN_DELEGATE_NONE
) {
2933 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
2934 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2935 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2937 /* 4.1 client asking for a delegation? */
2938 if (open
->op_deleg_want
)
2939 nfsd4_open_deleg_none_ext(open
, status
);
2943 nfs4_put_delegation(dp
);
2945 flag
= NFS4_OPEN_DELEGATE_NONE
;
2949 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
2950 struct nfs4_delegation
*dp
)
2952 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
2953 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2954 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2955 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
2956 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
2957 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2958 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2959 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
2961 /* Otherwise the client must be confused wanting a delegation
2962 * it already has, therefore we don't return
2963 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2968 * called with nfs4_lock_state() held.
2971 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2973 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2974 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
2975 struct nfs4_file
*fp
= NULL
;
2976 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2977 struct nfs4_ol_stateid
*stp
= NULL
;
2978 struct nfs4_delegation
*dp
= NULL
;
2982 * Lookup file; if found, lookup stateid and check open request,
2983 * and check for delegations in the process of being recalled.
2984 * If not found, create the nfs4_file struct
2986 fp
= find_file(ino
);
2988 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2990 status
= nfs4_check_deleg(cl
, fp
, open
, &dp
);
2994 status
= nfserr_bad_stateid
;
2995 if (nfsd4_is_deleg_cur(open
))
2997 status
= nfserr_jukebox
;
2999 open
->op_file
= NULL
;
3000 nfsd4_init_file(fp
, ino
);
3004 * OPEN the file, or upgrade an existing OPEN.
3005 * If truncate fails, the OPEN fails.
3008 /* Stateid was found, this is an OPEN upgrade */
3009 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
3013 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, open
);
3017 open
->op_stp
= NULL
;
3018 init_open_stateid(stp
, fp
, open
);
3019 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
3021 release_open_stateid(stp
);
3025 update_stateid(&stp
->st_stid
.sc_stateid
);
3026 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3028 if (nfsd4_has_session(&resp
->cstate
)) {
3029 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3031 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
3032 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3033 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
3039 * Attempt to hand out a delegation. No error return, because the
3040 * OPEN succeeds even if we fail.
3042 nfs4_open_delegation(current_fh
, open
, stp
);
3046 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
3047 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3049 /* 4.1 client trying to upgrade/downgrade delegation? */
3050 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
3051 open
->op_deleg_want
)
3052 nfsd4_deleg_xgrade_none_ext(open
, dp
);
3056 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
3057 nfs4_set_claim_prev(open
, nfsd4_has_session(&resp
->cstate
));
3059 * To finish the open response, we just need to set the rflags.
3061 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
3062 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
3063 !nfsd4_has_session(&resp
->cstate
))
3064 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
3069 void nfsd4_cleanup_open_state(struct nfsd4_open
*open
, __be32 status
)
3071 if (open
->op_openowner
) {
3072 struct nfs4_openowner
*oo
= open
->op_openowner
;
3074 if (!list_empty(&oo
->oo_owner
.so_stateids
))
3075 list_del_init(&oo
->oo_close_lru
);
3076 if (oo
->oo_flags
& NFS4_OO_NEW
) {
3078 release_openowner(oo
);
3079 open
->op_openowner
= NULL
;
3081 oo
->oo_flags
&= ~NFS4_OO_NEW
;
3085 nfsd4_free_file(open
->op_file
);
3087 nfs4_free_stateid(open
->op_stp
);
3091 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3094 struct nfs4_client
*clp
;
3098 dprintk("process_renew(%08x/%08x): starting\n",
3099 clid
->cl_boot
, clid
->cl_id
);
3100 status
= nfserr_stale_clientid
;
3101 if (STALE_CLIENTID(clid
))
3103 clp
= find_confirmed_client(clid
);
3104 status
= nfserr_expired
;
3106 /* We assume the client took too long to RENEW. */
3107 dprintk("nfsd4_renew: clientid not found!\n");
3110 status
= nfserr_cb_path_down
;
3111 if (!list_empty(&clp
->cl_delegations
)
3112 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
3116 nfs4_unlock_state();
3120 static struct lock_manager nfsd4_manager
= {
3123 static bool grace_ended
;
3126 nfsd4_end_grace(void)
3128 /* do nothing if grace period already ended */
3132 dprintk("NFSD: end of grace period\n");
3134 nfsd4_record_grace_done(&init_net
, boot_time
);
3135 locks_end_grace(&nfsd4_manager
);
3137 * Now that every NFSv4 client has had the chance to recover and
3138 * to see the (possibly new, possibly shorter) lease time, we
3139 * can safely set the next grace time to the current lease time:
3141 nfsd4_grace
= nfsd4_lease
;
3145 nfs4_laundromat(void)
3147 struct nfs4_client
*clp
;
3148 struct nfs4_openowner
*oo
;
3149 struct nfs4_delegation
*dp
;
3150 struct list_head
*pos
, *next
, reaplist
;
3151 time_t cutoff
= get_seconds() - nfsd4_lease
;
3152 time_t t
, clientid_val
= nfsd4_lease
;
3153 time_t u
, test_val
= nfsd4_lease
;
3157 dprintk("NFSD: laundromat service - starting\n");
3159 INIT_LIST_HEAD(&reaplist
);
3160 spin_lock(&client_lock
);
3161 list_for_each_safe(pos
, next
, &client_lru
) {
3162 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3163 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
3164 t
= clp
->cl_time
- cutoff
;
3165 if (clientid_val
> t
)
3169 if (atomic_read(&clp
->cl_refcount
)) {
3170 dprintk("NFSD: client in use (clientid %08x)\n",
3171 clp
->cl_clientid
.cl_id
);
3174 unhash_client_locked(clp
);
3175 list_add(&clp
->cl_lru
, &reaplist
);
3177 spin_unlock(&client_lock
);
3178 list_for_each_safe(pos
, next
, &reaplist
) {
3179 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3180 dprintk("NFSD: purging unused client (clientid %08x)\n",
3181 clp
->cl_clientid
.cl_id
);
3182 nfsd4_client_record_remove(clp
);
3185 spin_lock(&recall_lock
);
3186 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3187 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3188 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3189 u
= dp
->dl_time
- cutoff
;
3194 list_move(&dp
->dl_recall_lru
, &reaplist
);
3196 spin_unlock(&recall_lock
);
3197 list_for_each_safe(pos
, next
, &reaplist
) {
3198 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3199 unhash_delegation(dp
);
3201 test_val
= nfsd4_lease
;
3202 list_for_each_safe(pos
, next
, &close_lru
) {
3203 oo
= container_of(pos
, struct nfs4_openowner
, oo_close_lru
);
3204 if (time_after((unsigned long)oo
->oo_time
, (unsigned long)cutoff
)) {
3205 u
= oo
->oo_time
- cutoff
;
3210 release_openowner(oo
);
3212 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3213 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3214 nfs4_unlock_state();
3215 return clientid_val
;
3218 static struct workqueue_struct
*laundry_wq
;
3219 static void laundromat_main(struct work_struct
*);
3220 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3223 laundromat_main(struct work_struct
*not_used
)
3227 t
= nfs4_laundromat();
3228 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3229 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3232 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*stp
)
3234 if (fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
)
3235 return nfserr_bad_stateid
;
3240 STALE_STATEID(stateid_t
*stateid
)
3242 if (stateid
->si_opaque
.so_clid
.cl_boot
== boot_time
)
3244 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3245 STATEID_VAL(stateid
));
3250 access_permit_read(struct nfs4_ol_stateid
*stp
)
3252 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
3253 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
3254 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
3258 access_permit_write(struct nfs4_ol_stateid
*stp
)
3260 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
3261 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
3265 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
3267 __be32 status
= nfserr_openmode
;
3269 /* For lock stateid's, we test the parent open, not the lock: */
3270 if (stp
->st_openstp
)
3271 stp
= stp
->st_openstp
;
3272 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
3274 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
3281 static inline __be32
3282 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3284 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3286 else if (locks_in_grace()) {
3287 /* Answer in remaining cases depends on existence of
3288 * conflicting state; so we must wait out the grace period. */
3289 return nfserr_grace
;
3290 } else if (flags
& WR_STATE
)
3291 return nfs4_share_conflict(current_fh
,
3292 NFS4_SHARE_DENY_WRITE
);
3293 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3294 return nfs4_share_conflict(current_fh
,
3295 NFS4_SHARE_DENY_READ
);
3299 * Allow READ/WRITE during grace period on recovered state only for files
3300 * that are not able to provide mandatory locking.
3303 grace_disallows_io(struct inode
*inode
)
3305 return locks_in_grace() && mandatory_lock(inode
);
3308 /* Returns true iff a is later than b: */
3309 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3311 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3314 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3317 * When sessions are used the stateid generation number is ignored
3320 if (has_session
&& in
->si_generation
== 0)
3323 if (in
->si_generation
== ref
->si_generation
)
3326 /* If the client sends us a stateid from the future, it's buggy: */
3327 if (stateid_generation_after(in
, ref
))
3328 return nfserr_bad_stateid
;
3330 * However, we could see a stateid from the past, even from a
3331 * non-buggy client. For example, if the client sends a lock
3332 * while some IO is outstanding, the lock may bump si_generation
3333 * while the IO is still in flight. The client could avoid that
3334 * situation by waiting for responses on all the IO requests,
3335 * but better performance may result in retrying IO that
3336 * receives an old_stateid error if requests are rarely
3337 * reordered in flight:
3339 return nfserr_old_stateid
;
3342 static __be32
nfsd4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
3344 struct nfs4_stid
*s
;
3345 struct nfs4_ol_stateid
*ols
;
3348 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3349 return nfserr_bad_stateid
;
3350 /* Client debugging aid. */
3351 if (!same_clid(&stateid
->si_opaque
.so_clid
, &cl
->cl_clientid
)) {
3352 char addr_str
[INET6_ADDRSTRLEN
];
3353 rpc_ntop((struct sockaddr
*)&cl
->cl_addr
, addr_str
,
3355 pr_warn_ratelimited("NFSD: client %s testing state ID "
3356 "with incorrect client ID\n", addr_str
);
3357 return nfserr_bad_stateid
;
3359 s
= find_stateid(cl
, stateid
);
3361 return nfserr_bad_stateid
;
3362 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3365 if (!(s
->sc_type
& (NFS4_OPEN_STID
| NFS4_LOCK_STID
)))
3367 ols
= openlockstateid(s
);
3368 if (ols
->st_stateowner
->so_is_open_owner
3369 && !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3370 return nfserr_bad_stateid
;
3374 static __be32
nfsd4_lookup_stateid(stateid_t
*stateid
, unsigned char typemask
, struct nfs4_stid
**s
)
3376 struct nfs4_client
*cl
;
3378 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3379 return nfserr_bad_stateid
;
3380 if (STALE_STATEID(stateid
))
3381 return nfserr_stale_stateid
;
3382 cl
= find_confirmed_client(&stateid
->si_opaque
.so_clid
);
3384 return nfserr_expired
;
3385 *s
= find_stateid_by_type(cl
, stateid
, typemask
);
3387 return nfserr_bad_stateid
;
3393 * Checks for stateid operations
3396 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3397 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3399 struct nfs4_stid
*s
;
3400 struct nfs4_ol_stateid
*stp
= NULL
;
3401 struct nfs4_delegation
*dp
= NULL
;
3402 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3403 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3409 if (grace_disallows_io(ino
))
3410 return nfserr_grace
;
3412 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3413 return check_special_stateids(current_fh
, stateid
, flags
);
3415 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
, &s
);
3418 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, nfsd4_has_session(cstate
));
3421 switch (s
->sc_type
) {
3422 case NFS4_DELEG_STID
:
3423 dp
= delegstateid(s
);
3424 status
= nfs4_check_delegmode(dp
, flags
);
3428 *filpp
= dp
->dl_file
->fi_deleg_file
;
3432 case NFS4_OPEN_STID
:
3433 case NFS4_LOCK_STID
:
3434 stp
= openlockstateid(s
);
3435 status
= nfs4_check_fh(current_fh
, stp
);
3438 if (stp
->st_stateowner
->so_is_open_owner
3439 && !(openowner(stp
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3441 status
= nfs4_check_openmode(stp
, flags
);
3445 if (flags
& RD_STATE
)
3446 *filpp
= find_readable_file(stp
->st_file
);
3448 *filpp
= find_writeable_file(stp
->st_file
);
3452 return nfserr_bad_stateid
;
3460 nfsd4_free_lock_stateid(struct nfs4_ol_stateid
*stp
)
3462 if (check_for_locks(stp
->st_file
, lockowner(stp
->st_stateowner
)))
3463 return nfserr_locks_held
;
3464 release_lock_stateid(stp
);
3469 * Test if the stateid is valid
3472 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3473 struct nfsd4_test_stateid
*test_stateid
)
3475 struct nfsd4_test_stateid_id
*stateid
;
3476 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3479 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
3480 stateid
->ts_id_status
=
3481 nfsd4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
3482 nfs4_unlock_state();
3488 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3489 struct nfsd4_free_stateid
*free_stateid
)
3491 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3492 struct nfs4_stid
*s
;
3493 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3494 __be32 ret
= nfserr_bad_stateid
;
3497 s
= find_stateid(cl
, stateid
);
3500 switch (s
->sc_type
) {
3501 case NFS4_DELEG_STID
:
3502 ret
= nfserr_locks_held
;
3504 case NFS4_OPEN_STID
:
3505 case NFS4_LOCK_STID
:
3506 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3509 if (s
->sc_type
== NFS4_LOCK_STID
)
3510 ret
= nfsd4_free_lock_stateid(openlockstateid(s
));
3512 ret
= nfserr_locks_held
;
3515 ret
= nfserr_bad_stateid
;
3518 nfs4_unlock_state();
3525 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3526 RD_STATE
: WR_STATE
;
3529 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
3531 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3532 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
3535 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
3538 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
)
3540 * "Closed" stateid's exist *only* to return
3541 * nfserr_replay_me from the previous step.
3543 return nfserr_bad_stateid
;
3544 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3547 return nfs4_check_fh(current_fh
, stp
);
3551 * Checks for sequence id mutating operations.
3554 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3555 stateid_t
*stateid
, char typemask
,
3556 struct nfs4_ol_stateid
**stpp
)
3559 struct nfs4_stid
*s
;
3561 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3562 seqid
, STATEID_VAL(stateid
));
3565 status
= nfsd4_lookup_stateid(stateid
, typemask
, &s
);
3568 *stpp
= openlockstateid(s
);
3569 cstate
->replay_owner
= (*stpp
)->st_stateowner
;
3571 return nfs4_seqid_op_checks(cstate
, stateid
, seqid
, *stpp
);
3574 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
, stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
)
3577 struct nfs4_openowner
*oo
;
3579 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
3580 NFS4_OPEN_STID
, stpp
);
3583 oo
= openowner((*stpp
)->st_stateowner
);
3584 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
3585 return nfserr_bad_stateid
;
3590 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3591 struct nfsd4_open_confirm
*oc
)
3594 struct nfs4_openowner
*oo
;
3595 struct nfs4_ol_stateid
*stp
;
3597 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3598 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3599 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3601 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3607 status
= nfs4_preprocess_seqid_op(cstate
,
3608 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3609 NFS4_OPEN_STID
, &stp
);
3612 oo
= openowner(stp
->st_stateowner
);
3613 status
= nfserr_bad_stateid
;
3614 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
)
3616 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3617 update_stateid(&stp
->st_stid
.sc_stateid
);
3618 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3619 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3620 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3622 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
3625 if (!cstate
->replay_owner
)
3626 nfs4_unlock_state();
3630 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
3632 if (!test_access(access
, stp
))
3634 nfs4_file_put_access(stp
->st_file
, nfs4_access_to_omode(access
));
3635 clear_access(access
, stp
);
3638 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
3640 switch (to_access
) {
3641 case NFS4_SHARE_ACCESS_READ
:
3642 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
3643 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3645 case NFS4_SHARE_ACCESS_WRITE
:
3646 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
3647 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3649 case NFS4_SHARE_ACCESS_BOTH
:
3657 reset_union_bmap_deny(unsigned long deny
, struct nfs4_ol_stateid
*stp
)
3660 for (i
= 0; i
< 4; i
++) {
3661 if ((i
& deny
) != i
)
3667 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3668 struct nfsd4_compound_state
*cstate
,
3669 struct nfsd4_open_downgrade
*od
)
3672 struct nfs4_ol_stateid
*stp
;
3674 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3675 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3676 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3678 /* We don't yet support WANT bits: */
3679 if (od
->od_deleg_want
)
3680 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
3684 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
3685 &od
->od_stateid
, &stp
);
3688 status
= nfserr_inval
;
3689 if (!test_access(od
->od_share_access
, stp
)) {
3690 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3691 stp
->st_access_bmap
, od
->od_share_access
);
3694 if (!test_deny(od
->od_share_deny
, stp
)) {
3695 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3696 stp
->st_deny_bmap
, od
->od_share_deny
);
3699 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
3701 reset_union_bmap_deny(od
->od_share_deny
, stp
);
3703 update_stateid(&stp
->st_stid
.sc_stateid
);
3704 memcpy(&od
->od_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3707 if (!cstate
->replay_owner
)
3708 nfs4_unlock_state();
3712 void nfsd4_purge_closed_stateid(struct nfs4_stateowner
*so
)
3714 struct nfs4_openowner
*oo
;
3715 struct nfs4_ol_stateid
*s
;
3717 if (!so
->so_is_open_owner
)
3720 s
= oo
->oo_last_closed_stid
;
3723 if (!(oo
->oo_flags
& NFS4_OO_PURGE_CLOSE
)) {
3724 /* Release the last_closed_stid on the next seqid bump: */
3725 oo
->oo_flags
|= NFS4_OO_PURGE_CLOSE
;
3728 oo
->oo_flags
&= ~NFS4_OO_PURGE_CLOSE
;
3729 release_last_closed_stateid(oo
);
3732 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
3734 unhash_open_stateid(s
);
3735 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
3739 * nfs4_unlock_state() called after encode
3742 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3743 struct nfsd4_close
*close
)
3746 struct nfs4_openowner
*oo
;
3747 struct nfs4_ol_stateid
*stp
;
3749 dprintk("NFSD: nfsd4_close on file %.*s\n",
3750 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3751 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3754 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3756 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
3760 oo
= openowner(stp
->st_stateowner
);
3762 update_stateid(&stp
->st_stid
.sc_stateid
);
3763 memcpy(&close
->cl_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3765 nfsd4_close_open_stateid(stp
);
3766 oo
->oo_last_closed_stid
= stp
;
3768 /* place unused nfs4_stateowners on so_close_lru list to be
3769 * released by the laundromat service after the lease period
3770 * to enable us to handle CLOSE replay
3772 if (list_empty(&oo
->oo_owner
.so_stateids
))
3773 move_to_close_lru(oo
);
3775 if (!cstate
->replay_owner
)
3776 nfs4_unlock_state();
3781 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3782 struct nfsd4_delegreturn
*dr
)
3784 struct nfs4_delegation
*dp
;
3785 stateid_t
*stateid
= &dr
->dr_stateid
;
3786 struct nfs4_stid
*s
;
3787 struct inode
*inode
;
3790 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3792 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3795 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
, &s
);
3798 dp
= delegstateid(s
);
3799 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3803 unhash_delegation(dp
);
3805 nfs4_unlock_state();
3811 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3813 #define LOCKOWNER_INO_HASH_BITS 8
3814 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3815 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3818 end_offset(u64 start
, u64 len
)
3823 return end
>= start
? end
: NFS4_MAX_UINT64
;
3826 /* last octet in a range */
3828 last_byte_offset(u64 start
, u64 len
)
3834 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3837 static unsigned int lockowner_ino_hashval(struct inode
*inode
, u32 cl_id
, struct xdr_netobj
*ownername
)
3839 return (file_hashval(inode
) + cl_id
3840 + opaque_hashval(ownername
->data
, ownername
->len
))
3841 & LOCKOWNER_INO_HASH_MASK
;
3844 static struct list_head lockowner_ino_hashtbl
[LOCKOWNER_INO_HASH_SIZE
];
3847 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3848 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3849 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3850 * locking, this prevents us from being completely protocol-compliant. The
3851 * real solution to this problem is to start using unsigned file offsets in
3852 * the VFS, but this is a very deep change!
3855 nfs4_transform_lock_offset(struct file_lock
*lock
)
3857 if (lock
->fl_start
< 0)
3858 lock
->fl_start
= OFFSET_MAX
;
3859 if (lock
->fl_end
< 0)
3860 lock
->fl_end
= OFFSET_MAX
;
3863 /* Hack!: For now, we're defining this just so we can use a pointer to it
3864 * as a unique cookie to identify our (NFSv4's) posix locks. */
3865 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3869 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3871 struct nfs4_lockowner
*lo
;
3873 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3874 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3875 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3876 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3877 if (!deny
->ld_owner
.data
)
3878 /* We just don't care that much */
3880 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3881 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3884 deny
->ld_owner
.len
= 0;
3885 deny
->ld_owner
.data
= NULL
;
3886 deny
->ld_clientid
.cl_boot
= 0;
3887 deny
->ld_clientid
.cl_id
= 0;
3889 deny
->ld_start
= fl
->fl_start
;
3890 deny
->ld_length
= NFS4_MAX_UINT64
;
3891 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3892 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3893 deny
->ld_type
= NFS4_READ_LT
;
3894 if (fl
->fl_type
!= F_RDLCK
)
3895 deny
->ld_type
= NFS4_WRITE_LT
;
3898 static bool same_lockowner_ino(struct nfs4_lockowner
*lo
, struct inode
*inode
, clientid_t
*clid
, struct xdr_netobj
*owner
)
3900 struct nfs4_ol_stateid
*lst
;
3902 if (!same_owner_str(&lo
->lo_owner
, owner
, clid
))
3904 lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3905 struct nfs4_ol_stateid
, st_perstateowner
);
3906 return lst
->st_file
->fi_inode
== inode
;
3909 static struct nfs4_lockowner
*
3910 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3911 struct xdr_netobj
*owner
)
3913 unsigned int hashval
= lockowner_ino_hashval(inode
, clid
->cl_id
, owner
);
3914 struct nfs4_lockowner
*lo
;
3916 list_for_each_entry(lo
, &lockowner_ino_hashtbl
[hashval
], lo_owner_ino_hash
) {
3917 if (same_lockowner_ino(lo
, inode
, clid
, owner
))
3923 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
)
3925 struct inode
*inode
= open_stp
->st_file
->fi_inode
;
3926 unsigned int inohash
= lockowner_ino_hashval(inode
,
3927 clp
->cl_clientid
.cl_id
, &lo
->lo_owner
.so_owner
);
3929 list_add(&lo
->lo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
3930 list_add(&lo
->lo_owner_ino_hash
, &lockowner_ino_hashtbl
[inohash
]);
3931 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3935 * Alloc a lock owner structure.
3936 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3939 * strhashval = ownerstr_hashval
3942 static struct nfs4_lockowner
*
3943 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3944 struct nfs4_lockowner
*lo
;
3946 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
3949 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
3950 lo
->lo_owner
.so_is_open_owner
= 0;
3951 /* It is the openowner seqid that will be incremented in encode in the
3952 * case of new lockowners; so increment the lock seqid manually: */
3953 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3954 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
3958 static struct nfs4_ol_stateid
*
3959 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_ol_stateid
*open_stp
)
3961 struct nfs4_ol_stateid
*stp
;
3962 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
3964 stp
= nfs4_alloc_stateid(clp
);
3967 init_stid(&stp
->st_stid
, clp
, NFS4_LOCK_STID
);
3968 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3969 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
3970 stp
->st_stateowner
= &lo
->lo_owner
;
3973 stp
->st_access_bmap
= 0;
3974 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3975 stp
->st_openstp
= open_stp
;
3980 check_lock_length(u64 offset
, u64 length
)
3982 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3983 LOFF_OVERFLOW(offset
, length
)));
3986 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
3988 struct nfs4_file
*fp
= lock_stp
->st_file
;
3989 int oflag
= nfs4_access_to_omode(access
);
3991 if (test_access(access
, lock_stp
))
3993 nfs4_file_get_access(fp
, oflag
);
3994 set_access(access
, lock_stp
);
3997 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)
3999 struct nfs4_file
*fi
= ost
->st_file
;
4000 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
4001 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
4002 struct nfs4_lockowner
*lo
;
4003 unsigned int strhashval
;
4005 lo
= find_lockowner_str(fi
->fi_inode
, &cl
->cl_clientid
, &lock
->v
.new.owner
);
4007 if (!cstate
->minorversion
)
4008 return nfserr_bad_seqid
;
4009 /* XXX: a lockowner always has exactly one stateid: */
4010 *lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
4011 struct nfs4_ol_stateid
, st_perstateowner
);
4014 strhashval
= ownerstr_hashval(cl
->cl_clientid
.cl_id
,
4015 &lock
->v
.new.owner
);
4016 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
4018 return nfserr_jukebox
;
4019 *lst
= alloc_init_lock_stateid(lo
, fi
, ost
);
4021 release_lockowner(lo
);
4022 return nfserr_jukebox
;
4032 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4033 struct nfsd4_lock
*lock
)
4035 struct nfs4_openowner
*open_sop
= NULL
;
4036 struct nfs4_lockowner
*lock_sop
= NULL
;
4037 struct nfs4_ol_stateid
*lock_stp
;
4038 struct file
*filp
= NULL
;
4039 struct file_lock file_lock
;
4040 struct file_lock conflock
;
4042 bool new_state
= false;
4046 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4047 (long long) lock
->lk_offset
,
4048 (long long) lock
->lk_length
);
4050 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
4051 return nfserr_inval
;
4053 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
4054 S_IFREG
, NFSD_MAY_LOCK
))) {
4055 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4061 if (lock
->lk_is_new
) {
4062 struct nfs4_ol_stateid
*open_stp
= NULL
;
4064 if (nfsd4_has_session(cstate
))
4065 /* See rfc 5661 18.10.3: given clientid is ignored: */
4066 memcpy(&lock
->v
.new.clientid
,
4067 &cstate
->session
->se_client
->cl_clientid
,
4068 sizeof(clientid_t
));
4070 status
= nfserr_stale_clientid
;
4071 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
4074 /* validate and update open stateid and open seqid */
4075 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
4076 lock
->lk_new_open_seqid
,
4077 &lock
->lk_new_open_stateid
,
4081 open_sop
= openowner(open_stp
->st_stateowner
);
4082 status
= nfserr_bad_stateid
;
4083 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
4084 &lock
->v
.new.clientid
))
4086 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
4087 &lock_stp
, &new_state
);
4089 status
= nfs4_preprocess_seqid_op(cstate
,
4090 lock
->lk_old_lock_seqid
,
4091 &lock
->lk_old_lock_stateid
,
4092 NFS4_LOCK_STID
, &lock_stp
);
4095 lock_sop
= lockowner(lock_stp
->st_stateowner
);
4097 lkflg
= setlkflg(lock
->lk_type
);
4098 status
= nfs4_check_openmode(lock_stp
, lkflg
);
4102 status
= nfserr_grace
;
4103 if (locks_in_grace() && !lock
->lk_reclaim
)
4105 status
= nfserr_no_grace
;
4106 if (!locks_in_grace() && lock
->lk_reclaim
)
4109 locks_init_lock(&file_lock
);
4110 switch (lock
->lk_type
) {
4113 filp
= find_readable_file(lock_stp
->st_file
);
4115 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4116 file_lock
.fl_type
= F_RDLCK
;
4119 case NFS4_WRITEW_LT
:
4120 filp
= find_writeable_file(lock_stp
->st_file
);
4122 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4123 file_lock
.fl_type
= F_WRLCK
;
4126 status
= nfserr_inval
;
4130 status
= nfserr_openmode
;
4133 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
4134 file_lock
.fl_pid
= current
->tgid
;
4135 file_lock
.fl_file
= filp
;
4136 file_lock
.fl_flags
= FL_POSIX
;
4137 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4139 file_lock
.fl_start
= lock
->lk_offset
;
4140 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4141 nfs4_transform_lock_offset(&file_lock
);
4144 * Try to lock the file in the VFS.
4145 * Note: locks.c uses the BKL to protect the inode's lock list.
4148 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
4150 case 0: /* success! */
4151 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
4152 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
4156 case (EAGAIN
): /* conflock holds conflicting lock */
4157 status
= nfserr_denied
;
4158 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4159 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
4162 status
= nfserr_deadlock
;
4165 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4166 status
= nfserrno(err
);
4170 if (status
&& new_state
)
4171 release_lockowner(lock_sop
);
4172 if (!cstate
->replay_owner
)
4173 nfs4_unlock_state();
4178 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4179 * so we do a temporary open here just to get an open file to pass to
4180 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4183 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4186 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4188 err
= nfserrno(vfs_test_lock(file
, lock
));
4198 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4199 struct nfsd4_lockt
*lockt
)
4201 struct inode
*inode
;
4202 struct file_lock file_lock
;
4203 struct nfs4_lockowner
*lo
;
4206 if (locks_in_grace())
4207 return nfserr_grace
;
4209 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4210 return nfserr_inval
;
4214 status
= nfserr_stale_clientid
;
4215 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
4218 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4221 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4222 locks_init_lock(&file_lock
);
4223 switch (lockt
->lt_type
) {
4226 file_lock
.fl_type
= F_RDLCK
;
4229 case NFS4_WRITEW_LT
:
4230 file_lock
.fl_type
= F_WRLCK
;
4233 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4234 status
= nfserr_inval
;
4238 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4240 file_lock
.fl_owner
= (fl_owner_t
)lo
;
4241 file_lock
.fl_pid
= current
->tgid
;
4242 file_lock
.fl_flags
= FL_POSIX
;
4244 file_lock
.fl_start
= lockt
->lt_offset
;
4245 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4247 nfs4_transform_lock_offset(&file_lock
);
4249 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4253 if (file_lock
.fl_type
!= F_UNLCK
) {
4254 status
= nfserr_denied
;
4255 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4258 nfs4_unlock_state();
4263 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4264 struct nfsd4_locku
*locku
)
4266 struct nfs4_ol_stateid
*stp
;
4267 struct file
*filp
= NULL
;
4268 struct file_lock file_lock
;
4272 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4273 (long long) locku
->lu_offset
,
4274 (long long) locku
->lu_length
);
4276 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4277 return nfserr_inval
;
4281 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4282 &locku
->lu_stateid
, NFS4_LOCK_STID
, &stp
);
4285 filp
= find_any_file(stp
->st_file
);
4287 status
= nfserr_lock_range
;
4291 locks_init_lock(&file_lock
);
4292 file_lock
.fl_type
= F_UNLCK
;
4293 file_lock
.fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4294 file_lock
.fl_pid
= current
->tgid
;
4295 file_lock
.fl_file
= filp
;
4296 file_lock
.fl_flags
= FL_POSIX
;
4297 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4298 file_lock
.fl_start
= locku
->lu_offset
;
4300 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4301 nfs4_transform_lock_offset(&file_lock
);
4304 * Try to unlock the file in the VFS.
4306 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4308 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4312 * OK, unlock succeeded; the only thing left to do is update the stateid.
4314 update_stateid(&stp
->st_stid
.sc_stateid
);
4315 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4318 if (!cstate
->replay_owner
)
4319 nfs4_unlock_state();
4323 status
= nfserrno(err
);
4329 * 1: locks held by lockowner
4330 * 0: no locks held by lockowner
4333 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4335 struct file_lock
**flpp
;
4336 struct inode
*inode
= filp
->fi_inode
;
4340 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4341 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4352 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4353 struct nfsd4_compound_state
*cstate
,
4354 struct nfsd4_release_lockowner
*rlockowner
)
4356 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4357 struct nfs4_stateowner
*sop
;
4358 struct nfs4_lockowner
*lo
;
4359 struct nfs4_ol_stateid
*stp
;
4360 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4361 struct list_head matches
;
4362 unsigned int hashval
= ownerstr_hashval(clid
->cl_id
, owner
);
4365 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4366 clid
->cl_boot
, clid
->cl_id
);
4368 /* XXX check for lease expiration */
4370 status
= nfserr_stale_clientid
;
4371 if (STALE_CLIENTID(clid
))
4376 status
= nfserr_locks_held
;
4377 INIT_LIST_HEAD(&matches
);
4379 list_for_each_entry(sop
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
4380 if (sop
->so_is_open_owner
)
4382 if (!same_owner_str(sop
, owner
, clid
))
4384 list_for_each_entry(stp
, &sop
->so_stateids
,
4386 lo
= lockowner(sop
);
4387 if (check_for_locks(stp
->st_file
, lo
))
4389 list_add(&lo
->lo_list
, &matches
);
4392 /* Clients probably won't expect us to return with some (but not all)
4393 * of the lockowner state released; so don't release any until all
4394 * have been checked. */
4396 while (!list_empty(&matches
)) {
4397 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4399 /* unhash_stateowner deletes so_perclient only
4400 * for openowners. */
4401 list_del(&lo
->lo_list
);
4402 release_lockowner(lo
);
4405 nfs4_unlock_state();
4409 static inline struct nfs4_client_reclaim
*
4412 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4416 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4418 unsigned int strhashval
= clientstr_hashval(name
);
4419 struct nfs4_client
*clp
;
4421 clp
= find_confirmed_client_by_str(name
, strhashval
);
4424 return test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
4428 * failure => all reset bets are off, nfserr_no_grace...
4431 nfs4_client_to_reclaim(const char *name
)
4433 unsigned int strhashval
;
4434 struct nfs4_client_reclaim
*crp
= NULL
;
4436 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4437 crp
= alloc_reclaim();
4440 strhashval
= clientstr_hashval(name
);
4441 INIT_LIST_HEAD(&crp
->cr_strhash
);
4442 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4443 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4444 reclaim_str_hashtbl_size
++;
4449 nfs4_release_reclaim(void)
4451 struct nfs4_client_reclaim
*crp
= NULL
;
4454 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4455 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4456 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4457 struct nfs4_client_reclaim
, cr_strhash
);
4458 list_del(&crp
->cr_strhash
);
4460 reclaim_str_hashtbl_size
--;
4463 BUG_ON(reclaim_str_hashtbl_size
);
4467 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4468 struct nfs4_client_reclaim
*
4469 nfsd4_find_reclaim_client(struct nfs4_client
*clp
)
4471 unsigned int strhashval
;
4472 struct nfs4_client_reclaim
*crp
= NULL
;
4474 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4475 clp
->cl_name
.len
, clp
->cl_name
.data
,
4478 /* find clp->cl_name in reclaim_str_hashtbl */
4479 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4480 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4481 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4489 * Called from OPEN. Look for clientid in reclaim list.
4492 nfs4_check_open_reclaim(clientid_t
*clid
)
4494 struct nfs4_client
*clp
;
4496 /* find clientid in conf_id_hashtbl */
4497 clp
= find_confirmed_client(clid
);
4499 return nfserr_reclaim_bad
;
4501 return nfsd4_client_record_check(clp
) ? nfserr_reclaim_bad
: nfs_ok
;
4504 #ifdef CONFIG_NFSD_FAULT_INJECTION
4506 void nfsd_forget_clients(u64 num
)
4508 struct nfs4_client
*clp
, *next
;
4512 list_for_each_entry_safe(clp
, next
, &client_lru
, cl_lru
) {
4513 nfsd4_client_record_remove(clp
);
4518 nfs4_unlock_state();
4520 printk(KERN_INFO
"NFSD: Forgot %d clients", count
);
4523 static void release_lockowner_sop(struct nfs4_stateowner
*sop
)
4525 release_lockowner(lockowner(sop
));
4528 static void release_openowner_sop(struct nfs4_stateowner
*sop
)
4530 release_openowner(openowner(sop
));
4533 static int nfsd_release_n_owners(u64 num
, bool is_open_owner
,
4534 void (*release_sop
)(struct nfs4_stateowner
*))
4537 struct nfs4_stateowner
*sop
, *next
;
4539 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4540 list_for_each_entry_safe(sop
, next
, &ownerstr_hashtbl
[i
], so_strhash
) {
4541 if (sop
->so_is_open_owner
!= is_open_owner
)
4551 void nfsd_forget_locks(u64 num
)
4556 count
= nfsd_release_n_owners(num
, false, release_lockowner_sop
);
4557 nfs4_unlock_state();
4559 printk(KERN_INFO
"NFSD: Forgot %d locks", count
);
4562 void nfsd_forget_openowners(u64 num
)
4567 count
= nfsd_release_n_owners(num
, true, release_openowner_sop
);
4568 nfs4_unlock_state();
4570 printk(KERN_INFO
"NFSD: Forgot %d open owners", count
);
4573 int nfsd_process_n_delegations(u64 num
, void (*deleg_func
)(struct nfs4_delegation
*))
4576 struct nfs4_file
*fp
, *fnext
;
4577 struct nfs4_delegation
*dp
, *dnext
;
4579 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4580 list_for_each_entry_safe(fp
, fnext
, &file_hashtbl
[i
], fi_hash
) {
4581 list_for_each_entry_safe(dp
, dnext
, &fp
->fi_delegations
, dl_perfile
) {
4592 void nfsd_forget_delegations(u64 num
)
4597 count
= nfsd_process_n_delegations(num
, unhash_delegation
);
4598 nfs4_unlock_state();
4600 printk(KERN_INFO
"NFSD: Forgot %d delegations", count
);
4603 void nfsd_recall_delegations(u64 num
)
4608 spin_lock(&recall_lock
);
4609 count
= nfsd_process_n_delegations(num
, nfsd_break_one_deleg
);
4610 spin_unlock(&recall_lock
);
4611 nfs4_unlock_state();
4613 printk(KERN_INFO
"NFSD: Recalled %d delegations", count
);
4616 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4618 /* initialization to perform at module load time: */
4621 nfs4_state_init(void)
4625 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4626 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4627 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4628 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4629 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4630 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4632 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4633 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4634 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4635 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4637 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4638 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4640 for (i
= 0; i
< LOCKOWNER_INO_HASH_SIZE
; i
++)
4641 INIT_LIST_HEAD(&lockowner_ino_hashtbl
[i
]);
4642 INIT_LIST_HEAD(&close_lru
);
4643 INIT_LIST_HEAD(&client_lru
);
4644 INIT_LIST_HEAD(&del_recall_lru
);
4645 reclaim_str_hashtbl_size
= 0;
4649 * Since the lifetime of a delegation isn't limited to that of an open, a
4650 * client may quite reasonably hang on to a delegation as long as it has
4651 * the inode cached. This becomes an obvious problem the first time a
4652 * client's inode cache approaches the size of the server's total memory.
4654 * For now we avoid this problem by imposing a hard limit on the number
4655 * of delegations, which varies according to the server's memory size.
4658 set_max_delegations(void)
4661 * Allow at most 4 delegations per megabyte of RAM. Quick
4662 * estimates suggest that in the worst case (where every delegation
4663 * is for a different inode), a delegation could take about 1.5K,
4664 * giving a worst case usage of about 6% of memory.
4666 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4669 /* initialization to perform when the nfsd service is started: */
4672 nfs4_state_start(void)
4677 * FIXME: For now, we hang most of the pernet global stuff off of
4678 * init_net until nfsd is fully containerized. Eventually, we'll
4679 * need to pass a net pointer into this function, take a reference
4680 * to that instead and then do most of the rest of this on a per-net
4684 nfsd4_client_tracking_init(&init_net
);
4685 boot_time
= get_seconds();
4686 locks_start_grace(&nfsd4_manager
);
4687 grace_ended
= false;
4688 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4690 ret
= set_callback_cred();
4695 laundry_wq
= create_singlethread_workqueue("nfsd4");
4696 if (laundry_wq
== NULL
) {
4700 ret
= nfsd4_create_callback_queue();
4702 goto out_free_laundry
;
4703 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4704 set_max_delegations();
4707 destroy_workqueue(laundry_wq
);
4709 nfsd4_client_tracking_exit(&init_net
);
4715 __nfs4_state_shutdown(void)
4718 struct nfs4_client
*clp
= NULL
;
4719 struct nfs4_delegation
*dp
= NULL
;
4720 struct list_head
*pos
, *next
, reaplist
;
4722 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4723 while (!list_empty(&conf_id_hashtbl
[i
])) {
4724 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4727 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4728 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4732 INIT_LIST_HEAD(&reaplist
);
4733 spin_lock(&recall_lock
);
4734 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4735 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4736 list_move(&dp
->dl_recall_lru
, &reaplist
);
4738 spin_unlock(&recall_lock
);
4739 list_for_each_safe(pos
, next
, &reaplist
) {
4740 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4741 unhash_delegation(dp
);
4744 nfsd4_client_tracking_exit(&init_net
);
4749 nfs4_state_shutdown(void)
4751 cancel_delayed_work_sync(&laundromat_work
);
4752 destroy_workqueue(laundry_wq
);
4753 locks_end_grace(&nfsd4_manager
);
4755 __nfs4_state_shutdown();
4756 nfs4_unlock_state();
4757 nfsd4_destroy_callback_queue();
4761 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4763 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
4764 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
4768 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4770 if (cstate
->minorversion
) {
4771 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
4772 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4777 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
4779 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4783 * functions to set current state id
4786 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4788 put_stateid(cstate
, &odp
->od_stateid
);
4792 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
4794 put_stateid(cstate
, &open
->op_stateid
);
4798 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4800 put_stateid(cstate
, &close
->cl_stateid
);
4804 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
4806 put_stateid(cstate
, &lock
->lk_resp_stateid
);
4810 * functions to consume current state id
4814 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4816 get_stateid(cstate
, &odp
->od_stateid
);
4820 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
4822 get_stateid(cstate
, &drp
->dr_stateid
);
4826 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
4828 get_stateid(cstate
, &fsp
->fr_stateid
);
4832 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
4834 get_stateid(cstate
, &setattr
->sa_stateid
);
4838 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4840 get_stateid(cstate
, &close
->cl_stateid
);
4844 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
4846 get_stateid(cstate
, &locku
->lu_stateid
);
4850 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
4852 get_stateid(cstate
, &read
->rd_stateid
);
4856 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
4858 get_stateid(cstate
, &write
->wr_stateid
);