2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/clnt.h>
46 #include "current_stateid.h"
47 #include "fault_inject.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 time_t nfsd4_lease
= 90; /* default lease time */
55 time_t nfsd4_grace
= 90;
56 static time_t boot_time
;
58 #define all_ones {{~0,~0},~0}
59 static const stateid_t one_stateid
= {
61 .si_opaque
= all_ones
,
63 static const stateid_t zero_stateid
= {
66 static const stateid_t currentstateid
= {
70 static u64 current_sessionid
= 1;
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
74 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
76 /* forward declarations */
77 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
);
81 /* Currently used for almost all code touching nfsv4 state: */
82 static DEFINE_MUTEX(client_mutex
);
85 * Currently used for the del_recall_lru and file hash table. In an
86 * effort to decrease the scope of the client_mutex, this spinlock may
87 * eventually cover more:
89 static DEFINE_SPINLOCK(recall_lock
);
91 static struct kmem_cache
*openowner_slab
= NULL
;
92 static struct kmem_cache
*lockowner_slab
= NULL
;
93 static struct kmem_cache
*file_slab
= NULL
;
94 static struct kmem_cache
*stateid_slab
= NULL
;
95 static struct kmem_cache
*deleg_slab
= NULL
;
100 mutex_lock(&client_mutex
);
103 static void free_session(struct kref
*);
105 /* Must be called under the client_lock */
106 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
108 kref_put(&ses
->se_ref
, free_session
);
111 static void nfsd4_get_session(struct nfsd4_session
*ses
)
113 kref_get(&ses
->se_ref
);
117 nfs4_unlock_state(void)
119 mutex_unlock(&client_mutex
);
123 opaque_hashval(const void *ptr
, int nbytes
)
125 unsigned char *cptr
= (unsigned char *) ptr
;
135 static struct list_head del_recall_lru
;
137 static void nfsd4_free_file(struct nfs4_file
*f
)
139 kmem_cache_free(file_slab
, f
);
143 put_nfs4_file(struct nfs4_file
*fi
)
145 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
146 list_del(&fi
->fi_hash
);
147 spin_unlock(&recall_lock
);
154 get_nfs4_file(struct nfs4_file
*fi
)
156 atomic_inc(&fi
->fi_ref
);
159 static int num_delegations
;
160 unsigned int max_delegations
;
163 * Open owner state (share locks)
166 /* hash tables for lock and open owners */
167 #define OWNER_HASH_BITS 8
168 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
169 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
171 static unsigned int ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
175 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
177 return ret
& OWNER_HASH_MASK
;
180 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
182 /* hash table for nfs4_file */
183 #define FILE_HASH_BITS 8
184 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
186 static unsigned int file_hashval(struct inode
*ino
)
188 /* XXX: why are we hashing on inode pointer, anyway? */
189 return hash_ptr(ino
, FILE_HASH_BITS
);
192 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
194 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
196 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
197 atomic_inc(&fp
->fi_access
[oflag
]);
200 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
202 if (oflag
== O_RDWR
) {
203 __nfs4_file_get_access(fp
, O_RDONLY
);
204 __nfs4_file_get_access(fp
, O_WRONLY
);
206 __nfs4_file_get_access(fp
, oflag
);
209 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
211 if (fp
->fi_fds
[oflag
]) {
212 fput(fp
->fi_fds
[oflag
]);
213 fp
->fi_fds
[oflag
] = NULL
;
217 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
219 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
220 nfs4_file_put_fd(fp
, oflag
);
222 * It's also safe to get rid of the RDWR open *if*
223 * we no longer have need of the other kind of access
224 * or if we already have the other kind of open:
226 if (fp
->fi_fds
[1-oflag
]
227 || atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
228 nfs4_file_put_fd(fp
, O_RDWR
);
232 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
234 if (oflag
== O_RDWR
) {
235 __nfs4_file_put_access(fp
, O_RDONLY
);
236 __nfs4_file_put_access(fp
, O_WRONLY
);
238 __nfs4_file_put_access(fp
, oflag
);
241 static inline int get_new_stid(struct nfs4_stid
*stid
)
243 static int min_stateid
= 0;
244 struct idr
*stateids
= &stid
->sc_client
->cl_stateids
;
248 error
= idr_get_new_above(stateids
, stid
, min_stateid
, &new_stid
);
250 * Note: the necessary preallocation was done in
251 * nfs4_alloc_stateid(). The idr code caps the number of
252 * preallocations that can exist at a time, but the state lock
253 * prevents anyone from using ours before we get here:
257 * It shouldn't be a problem to reuse an opaque stateid value.
258 * I don't think it is for 4.1. But with 4.0 I worry that, for
259 * example, a stray write retransmission could be accepted by
260 * the server when it should have been rejected. Therefore,
261 * adopt a trick from the sctp code to attempt to maximize the
262 * amount of time until an id is reused, by ensuring they always
263 * "increase" (mod INT_MAX):
266 min_stateid
= new_stid
+1;
267 if (min_stateid
== INT_MAX
)
272 static void init_stid(struct nfs4_stid
*stid
, struct nfs4_client
*cl
, unsigned char type
)
274 stateid_t
*s
= &stid
->sc_stateid
;
277 stid
->sc_type
= type
;
278 stid
->sc_client
= cl
;
279 s
->si_opaque
.so_clid
= cl
->cl_clientid
;
280 new_id
= get_new_stid(stid
);
281 s
->si_opaque
.so_id
= (u32
)new_id
;
282 /* Will be incremented before return to client: */
283 s
->si_generation
= 0;
286 static struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
)
288 struct idr
*stateids
= &cl
->cl_stateids
;
290 if (!idr_pre_get(stateids
, GFP_KERNEL
))
293 * Note: if we fail here (or any time between now and the time
294 * we actually get the new idr), we won't need to undo the idr
295 * preallocation, since the idr code caps the number of
296 * preallocated entries.
298 return kmem_cache_alloc(slab
, GFP_KERNEL
);
301 static struct nfs4_ol_stateid
* nfs4_alloc_stateid(struct nfs4_client
*clp
)
303 return openlockstateid(nfs4_alloc_stid(clp
, stateid_slab
));
306 static struct nfs4_delegation
*
307 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_ol_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
309 struct nfs4_delegation
*dp
;
310 struct nfs4_file
*fp
= stp
->st_file
;
312 dprintk("NFSD alloc_init_deleg\n");
314 * Major work on the lease subsystem (for example, to support
315 * calbacks on stat) will be required before we can support
316 * write delegations properly.
318 if (type
!= NFS4_OPEN_DELEGATE_READ
)
320 if (fp
->fi_had_conflict
)
322 if (num_delegations
> max_delegations
)
324 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
327 init_stid(&dp
->dl_stid
, clp
, NFS4_DELEG_STID
);
329 * delegation seqid's are never incremented. The 4.1 special
330 * meaning of seqid 0 isn't meaningful, really, but let's avoid
331 * 0 anyway just for consistency and use 1:
333 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
335 INIT_LIST_HEAD(&dp
->dl_perfile
);
336 INIT_LIST_HEAD(&dp
->dl_perclnt
);
337 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
341 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
343 atomic_set(&dp
->dl_count
, 1);
344 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
349 nfs4_put_delegation(struct nfs4_delegation
*dp
)
351 if (atomic_dec_and_test(&dp
->dl_count
)) {
352 dprintk("NFSD: freeing dp %p\n",dp
);
353 put_nfs4_file(dp
->dl_file
);
354 kmem_cache_free(deleg_slab
, dp
);
359 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
361 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
362 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
364 fput(fp
->fi_deleg_file
);
365 fp
->fi_deleg_file
= NULL
;
369 static void unhash_stid(struct nfs4_stid
*s
)
371 struct idr
*stateids
= &s
->sc_client
->cl_stateids
;
373 idr_remove(stateids
, s
->sc_stateid
.si_opaque
.so_id
);
376 /* Called under the state lock. */
378 unhash_delegation(struct nfs4_delegation
*dp
)
380 unhash_stid(&dp
->dl_stid
);
381 list_del_init(&dp
->dl_perclnt
);
382 spin_lock(&recall_lock
);
383 list_del_init(&dp
->dl_perfile
);
384 list_del_init(&dp
->dl_recall_lru
);
385 spin_unlock(&recall_lock
);
386 nfs4_put_deleg_lease(dp
->dl_file
);
387 nfs4_put_delegation(dp
);
394 /* client_lock protects the client lru list and session hash table */
395 static DEFINE_SPINLOCK(client_lock
);
397 /* Hash tables for nfs4_clientid state */
398 #define CLIENT_HASH_BITS 4
399 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
400 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
402 static unsigned int clientid_hashval(u32 id
)
404 return id
& CLIENT_HASH_MASK
;
407 static unsigned int clientstr_hashval(const char *name
)
409 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
413 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
414 * used in reboot/reset lease grace period processing
416 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
417 * setclientid_confirmed info.
419 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
422 * client_lru holds client queue ordered by nfs4_client.cl_time
425 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
426 * for last close replay.
428 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
429 static int reclaim_str_hashtbl_size
= 0;
430 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
431 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
432 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
433 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
434 static struct list_head client_lru
;
435 static struct list_head close_lru
;
438 * We store the NONE, READ, WRITE, and BOTH bits separately in the
439 * st_{access,deny}_bmap field of the stateid, in order to track not
440 * only what share bits are currently in force, but also what
441 * combinations of share bits previous opens have used. This allows us
442 * to enforce the recommendation of rfc 3530 14.2.19 that the server
443 * return an error if the client attempt to downgrade to a combination
444 * of share bits not explicable by closing some of its previous opens.
446 * XXX: This enforcement is actually incomplete, since we don't keep
447 * track of access/deny bit combinations; so, e.g., we allow:
449 * OPEN allow read, deny write
450 * OPEN allow both, deny none
451 * DOWNGRADE allow read, deny none
453 * which we should reject.
456 bmap_to_share_mode(unsigned long bmap
) {
458 unsigned int access
= 0;
460 for (i
= 1; i
< 4; i
++) {
461 if (test_bit(i
, &bmap
))
468 test_share(struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
) {
469 unsigned int access
, deny
;
471 access
= bmap_to_share_mode(stp
->st_access_bmap
);
472 deny
= bmap_to_share_mode(stp
->st_deny_bmap
);
473 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
478 /* set share access for a given stateid */
480 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
482 __set_bit(access
, &stp
->st_access_bmap
);
485 /* clear share access for a given stateid */
487 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
489 __clear_bit(access
, &stp
->st_access_bmap
);
492 /* test whether a given stateid has access */
494 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
496 return test_bit(access
, &stp
->st_access_bmap
);
499 /* set share deny for a given stateid */
501 set_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
503 __set_bit(access
, &stp
->st_deny_bmap
);
506 /* clear share deny for a given stateid */
508 clear_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
510 __clear_bit(access
, &stp
->st_deny_bmap
);
513 /* test whether a given stateid is denying specific access */
515 test_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
517 return test_bit(access
, &stp
->st_deny_bmap
);
520 static int nfs4_access_to_omode(u32 access
)
522 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
523 case NFS4_SHARE_ACCESS_READ
:
525 case NFS4_SHARE_ACCESS_WRITE
:
527 case NFS4_SHARE_ACCESS_BOTH
:
533 /* release all access and file references for a given stateid */
535 release_all_access(struct nfs4_ol_stateid
*stp
)
539 for (i
= 1; i
< 4; i
++) {
540 if (test_access(i
, stp
))
541 nfs4_file_put_access(stp
->st_file
,
542 nfs4_access_to_omode(i
));
543 clear_access(i
, stp
);
547 static void unhash_generic_stateid(struct nfs4_ol_stateid
*stp
)
549 list_del(&stp
->st_perfile
);
550 list_del(&stp
->st_perstateowner
);
553 static void close_generic_stateid(struct nfs4_ol_stateid
*stp
)
555 release_all_access(stp
);
556 put_nfs4_file(stp
->st_file
);
560 static void free_generic_stateid(struct nfs4_ol_stateid
*stp
)
562 kmem_cache_free(stateid_slab
, stp
);
565 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
569 unhash_generic_stateid(stp
);
570 unhash_stid(&stp
->st_stid
);
571 file
= find_any_file(stp
->st_file
);
573 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
574 close_generic_stateid(stp
);
575 free_generic_stateid(stp
);
578 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
580 struct nfs4_ol_stateid
*stp
;
582 list_del(&lo
->lo_owner
.so_strhash
);
583 list_del(&lo
->lo_perstateid
);
584 list_del(&lo
->lo_owner_ino_hash
);
585 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
586 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
587 struct nfs4_ol_stateid
, st_perstateowner
);
588 release_lock_stateid(stp
);
592 static void release_lockowner(struct nfs4_lockowner
*lo
)
594 unhash_lockowner(lo
);
595 nfs4_free_lockowner(lo
);
599 release_stateid_lockowners(struct nfs4_ol_stateid
*open_stp
)
601 struct nfs4_lockowner
*lo
;
603 while (!list_empty(&open_stp
->st_lockowners
)) {
604 lo
= list_entry(open_stp
->st_lockowners
.next
,
605 struct nfs4_lockowner
, lo_perstateid
);
606 release_lockowner(lo
);
610 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
)
612 unhash_generic_stateid(stp
);
613 release_stateid_lockowners(stp
);
614 close_generic_stateid(stp
);
617 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
619 unhash_open_stateid(stp
);
620 unhash_stid(&stp
->st_stid
);
621 free_generic_stateid(stp
);
624 static void unhash_openowner(struct nfs4_openowner
*oo
)
626 struct nfs4_ol_stateid
*stp
;
628 list_del(&oo
->oo_owner
.so_strhash
);
629 list_del(&oo
->oo_perclient
);
630 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
631 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
632 struct nfs4_ol_stateid
, st_perstateowner
);
633 release_open_stateid(stp
);
637 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
639 struct nfs4_ol_stateid
*s
= oo
->oo_last_closed_stid
;
642 unhash_stid(&s
->st_stid
);
643 free_generic_stateid(s
);
644 oo
->oo_last_closed_stid
= NULL
;
648 static void release_openowner(struct nfs4_openowner
*oo
)
650 unhash_openowner(oo
);
651 list_del(&oo
->oo_close_lru
);
652 release_last_closed_stateid(oo
);
653 nfs4_free_openowner(oo
);
656 #define SESSION_HASH_SIZE 512
657 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
660 hash_sessionid(struct nfs4_sessionid
*sessionid
)
662 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
664 return sid
->sequence
% SESSION_HASH_SIZE
;
669 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
671 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
672 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
676 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
683 gen_sessionid(struct nfsd4_session
*ses
)
685 struct nfs4_client
*clp
= ses
->se_client
;
686 struct nfsd4_sessionid
*sid
;
688 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
689 sid
->clientid
= clp
->cl_clientid
;
690 sid
->sequence
= current_sessionid
++;
695 * The protocol defines ca_maxresponssize_cached to include the size of
696 * the rpc header, but all we need to cache is the data starting after
697 * the end of the initial SEQUENCE operation--the rest we regenerate
698 * each time. Therefore we can advertise a ca_maxresponssize_cached
699 * value that is the number of bytes in our cache plus a few additional
700 * bytes. In order to stay on the safe side, and not promise more than
701 * we can cache, those additional bytes must be the minimum possible: 24
702 * bytes of rpc header (xid through accept state, with AUTH_NULL
703 * verifier), 12 for the compound header (with zero-length tag), and 44
704 * for the SEQUENCE op response:
706 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
709 free_session_slots(struct nfsd4_session
*ses
)
713 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
714 kfree(ses
->se_slots
[i
]);
718 * We don't actually need to cache the rpc and session headers, so we
719 * can allocate a little less for each slot:
721 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
723 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
726 static int nfsd4_sanitize_slot_size(u32 size
)
728 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
729 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
735 * XXX: If we run out of reserved DRC memory we could (up to a point)
736 * re-negotiate active sessions and reduce their slot usage to make
737 * room for new connections. For now we just fail the create session.
739 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
743 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
745 spin_lock(&nfsd_drc_lock
);
746 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
747 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
748 num
= min_t(int, num
, avail
/ slotsize
);
749 nfsd_drc_mem_used
+= num
* slotsize
;
750 spin_unlock(&nfsd_drc_lock
);
755 static void nfsd4_put_drc_mem(int slotsize
, int num
)
757 spin_lock(&nfsd_drc_lock
);
758 nfsd_drc_mem_used
-= slotsize
* num
;
759 spin_unlock(&nfsd_drc_lock
);
762 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
764 struct nfsd4_session
*new;
767 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
768 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
769 mem
= numslots
* sizeof(struct nfsd4_slot
*);
771 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
774 /* allocate each struct nfsd4_slot and data cache in one piece */
775 for (i
= 0; i
< numslots
; i
++) {
776 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
777 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
778 if (!new->se_slots
[i
])
784 kfree(new->se_slots
[i
]);
789 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
791 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
793 new->maxreqs
= numslots
;
794 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
795 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
796 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
797 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
798 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
801 static void free_conn(struct nfsd4_conn
*c
)
803 svc_xprt_put(c
->cn_xprt
);
807 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
809 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
810 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
812 spin_lock(&clp
->cl_lock
);
813 if (!list_empty(&c
->cn_persession
)) {
814 list_del(&c
->cn_persession
);
817 spin_unlock(&clp
->cl_lock
);
818 nfsd4_probe_callback(clp
);
821 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
823 struct nfsd4_conn
*conn
;
825 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
828 svc_xprt_get(rqstp
->rq_xprt
);
829 conn
->cn_xprt
= rqstp
->rq_xprt
;
830 conn
->cn_flags
= flags
;
831 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
835 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
837 conn
->cn_session
= ses
;
838 list_add(&conn
->cn_persession
, &ses
->se_conns
);
841 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
843 struct nfs4_client
*clp
= ses
->se_client
;
845 spin_lock(&clp
->cl_lock
);
846 __nfsd4_hash_conn(conn
, ses
);
847 spin_unlock(&clp
->cl_lock
);
850 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
852 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
853 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
856 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
858 struct nfsd4_conn
*conn
;
861 conn
= alloc_conn(rqstp
, dir
);
863 return nfserr_jukebox
;
864 nfsd4_hash_conn(conn
, ses
);
865 ret
= nfsd4_register_conn(conn
);
867 /* oops; xprt is already down: */
868 nfsd4_conn_lost(&conn
->cn_xpt_user
);
869 if (ses
->se_client
->cl_cb_state
== NFSD4_CB_DOWN
&&
870 dir
& NFS4_CDFC4_BACK
) {
871 /* callback channel may be back up */
872 nfsd4_probe_callback(ses
->se_client
);
877 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
879 u32 dir
= NFS4_CDFC4_FORE
;
881 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
882 dir
|= NFS4_CDFC4_BACK
;
884 return nfsd4_new_conn(rqstp
, ses
, dir
);
887 /* must be called under client_lock */
888 static void nfsd4_del_conns(struct nfsd4_session
*s
)
890 struct nfs4_client
*clp
= s
->se_client
;
891 struct nfsd4_conn
*c
;
893 spin_lock(&clp
->cl_lock
);
894 while (!list_empty(&s
->se_conns
)) {
895 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
896 list_del_init(&c
->cn_persession
);
897 spin_unlock(&clp
->cl_lock
);
899 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
902 spin_lock(&clp
->cl_lock
);
904 spin_unlock(&clp
->cl_lock
);
907 static void free_session(struct kref
*kref
)
909 struct nfsd4_session
*ses
;
912 lockdep_assert_held(&client_lock
);
913 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
914 nfsd4_del_conns(ses
);
915 spin_lock(&nfsd_drc_lock
);
916 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
917 nfsd_drc_mem_used
-= mem
;
918 spin_unlock(&nfsd_drc_lock
);
919 free_session_slots(ses
);
923 void nfsd4_put_session(struct nfsd4_session
*ses
)
925 spin_lock(&client_lock
);
926 nfsd4_put_session_locked(ses
);
927 spin_unlock(&client_lock
);
930 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
932 struct nfsd4_session
*new;
933 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
934 int numslots
, slotsize
;
939 * Note decreasing slot size below client's request may
940 * make it difficult for client to function correctly, whereas
941 * decreasing the number of slots will (just?) affect
942 * performance. When short on memory we therefore prefer to
943 * decrease number of slots instead of their size.
945 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
946 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
950 new = alloc_session(slotsize
, numslots
);
952 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
955 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
957 new->se_client
= clp
;
960 INIT_LIST_HEAD(&new->se_conns
);
962 new->se_cb_seq_nr
= 1;
963 new->se_flags
= cses
->flags
;
964 new->se_cb_prog
= cses
->callback_prog
;
965 kref_init(&new->se_ref
);
966 idx
= hash_sessionid(&new->se_sessionid
);
967 spin_lock(&client_lock
);
968 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
969 spin_lock(&clp
->cl_lock
);
970 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
971 spin_unlock(&clp
->cl_lock
);
972 spin_unlock(&client_lock
);
974 status
= nfsd4_new_conn_from_crses(rqstp
, new);
975 /* whoops: benny points out, status is ignored! (err, or bogus) */
977 spin_lock(&client_lock
);
978 free_session(&new->se_ref
);
979 spin_unlock(&client_lock
);
982 if (cses
->flags
& SESSION4_BACK_CHAN
) {
983 struct sockaddr
*sa
= svc_addr(rqstp
);
985 * This is a little silly; with sessions there's no real
986 * use for the callback address. Use the peer address
987 * as a reasonable default for now, but consider fixing
988 * the rpc client not to require an address in the
991 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
992 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
994 nfsd4_probe_callback(clp
);
998 /* caller must hold client_lock */
999 static struct nfsd4_session
*
1000 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
1002 struct nfsd4_session
*elem
;
1005 dump_sessionid(__func__
, sessionid
);
1006 idx
= hash_sessionid(sessionid
);
1007 /* Search in the appropriate list */
1008 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
1009 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1010 NFS4_MAX_SESSIONID_LEN
)) {
1015 dprintk("%s: session not found\n", __func__
);
1019 /* caller must hold client_lock */
1021 unhash_session(struct nfsd4_session
*ses
)
1023 list_del(&ses
->se_hash
);
1024 spin_lock(&ses
->se_client
->cl_lock
);
1025 list_del(&ses
->se_perclnt
);
1026 spin_unlock(&ses
->se_client
->cl_lock
);
1029 /* must be called under the client_lock */
1031 renew_client_locked(struct nfs4_client
*clp
)
1033 if (is_client_expired(clp
)) {
1035 printk("%s: client (clientid %08x/%08x) already expired\n",
1037 clp
->cl_clientid
.cl_boot
,
1038 clp
->cl_clientid
.cl_id
);
1042 dprintk("renewing client (clientid %08x/%08x)\n",
1043 clp
->cl_clientid
.cl_boot
,
1044 clp
->cl_clientid
.cl_id
);
1045 list_move_tail(&clp
->cl_lru
, &client_lru
);
1046 clp
->cl_time
= get_seconds();
1050 renew_client(struct nfs4_client
*clp
)
1052 spin_lock(&client_lock
);
1053 renew_client_locked(clp
);
1054 spin_unlock(&client_lock
);
1057 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1059 STALE_CLIENTID(clientid_t
*clid
)
1061 if (clid
->cl_boot
== boot_time
)
1063 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1064 clid
->cl_boot
, clid
->cl_id
, boot_time
);
1069 * XXX Should we use a slab cache ?
1070 * This type of memory management is somewhat inefficient, but we use it
1071 * anyway since SETCLIENTID is not a common operation.
1073 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1075 struct nfs4_client
*clp
;
1077 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1080 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1081 if (clp
->cl_name
.data
== NULL
) {
1085 clp
->cl_name
.len
= name
.len
;
1090 free_client(struct nfs4_client
*clp
)
1092 lockdep_assert_held(&client_lock
);
1093 while (!list_empty(&clp
->cl_sessions
)) {
1094 struct nfsd4_session
*ses
;
1095 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1097 list_del(&ses
->se_perclnt
);
1098 nfsd4_put_session_locked(ses
);
1100 free_svc_cred(&clp
->cl_cred
);
1101 kfree(clp
->cl_name
.data
);
1106 release_session_client(struct nfsd4_session
*session
)
1108 struct nfs4_client
*clp
= session
->se_client
;
1110 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
1112 if (is_client_expired(clp
)) {
1114 session
->se_client
= NULL
;
1116 renew_client_locked(clp
);
1117 spin_unlock(&client_lock
);
1120 /* must be called under the client_lock */
1122 unhash_client_locked(struct nfs4_client
*clp
)
1124 struct nfsd4_session
*ses
;
1126 mark_client_expired(clp
);
1127 list_del(&clp
->cl_lru
);
1128 spin_lock(&clp
->cl_lock
);
1129 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1130 list_del_init(&ses
->se_hash
);
1131 spin_unlock(&clp
->cl_lock
);
1135 expire_client(struct nfs4_client
*clp
)
1137 struct nfs4_openowner
*oo
;
1138 struct nfs4_delegation
*dp
;
1139 struct list_head reaplist
;
1141 INIT_LIST_HEAD(&reaplist
);
1142 spin_lock(&recall_lock
);
1143 while (!list_empty(&clp
->cl_delegations
)) {
1144 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1145 list_del_init(&dp
->dl_perclnt
);
1146 list_move(&dp
->dl_recall_lru
, &reaplist
);
1148 spin_unlock(&recall_lock
);
1149 while (!list_empty(&reaplist
)) {
1150 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1151 unhash_delegation(dp
);
1153 while (!list_empty(&clp
->cl_openowners
)) {
1154 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1155 release_openowner(oo
);
1157 nfsd4_shutdown_callback(clp
);
1158 if (clp
->cl_cb_conn
.cb_xprt
)
1159 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1160 list_del(&clp
->cl_idhash
);
1161 list_del(&clp
->cl_strhash
);
1162 spin_lock(&client_lock
);
1163 unhash_client_locked(clp
);
1164 if (atomic_read(&clp
->cl_refcount
) == 0)
1166 spin_unlock(&client_lock
);
1169 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1171 memcpy(target
->cl_verifier
.data
, source
->data
,
1172 sizeof(target
->cl_verifier
.data
));
1175 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1177 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1178 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1181 static int copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1183 if (source
->cr_principal
) {
1184 target
->cr_principal
=
1185 kstrdup(source
->cr_principal
, GFP_KERNEL
);
1186 if (target
->cr_principal
== NULL
)
1189 target
->cr_principal
= NULL
;
1190 target
->cr_flavor
= source
->cr_flavor
;
1191 target
->cr_uid
= source
->cr_uid
;
1192 target
->cr_gid
= source
->cr_gid
;
1193 target
->cr_group_info
= source
->cr_group_info
;
1194 get_group_info(target
->cr_group_info
);
1198 static int same_name(const char *n1
, const char *n2
)
1200 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1204 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1206 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1210 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1212 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1215 static bool groups_equal(struct group_info
*g1
, struct group_info
*g2
)
1219 if (g1
->ngroups
!= g2
->ngroups
)
1221 for (i
=0; i
<g1
->ngroups
; i
++)
1222 if (GROUP_AT(g1
, i
) != GROUP_AT(g2
, i
))
1228 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1230 if ((cr1
->cr_flavor
!= cr2
->cr_flavor
)
1231 || (cr1
->cr_uid
!= cr2
->cr_uid
)
1232 || (cr1
->cr_gid
!= cr2
->cr_gid
)
1233 || !groups_equal(cr1
->cr_group_info
, cr2
->cr_group_info
))
1235 if (cr1
->cr_principal
== cr2
->cr_principal
)
1237 if (!cr1
->cr_principal
|| !cr2
->cr_principal
)
1239 return 0 == strcmp(cr1
->cr_principal
, cr2
->cr_principal
);
1242 static void gen_clid(struct nfs4_client
*clp
)
1244 static u32 current_clientid
= 1;
1246 clp
->cl_clientid
.cl_boot
= boot_time
;
1247 clp
->cl_clientid
.cl_id
= current_clientid
++;
1250 static void gen_confirm(struct nfs4_client
*clp
)
1255 verf
[0] = (__be32
)get_seconds();
1256 verf
[1] = (__be32
)i
++;
1257 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
1260 static struct nfs4_stid
*find_stateid(struct nfs4_client
*cl
, stateid_t
*t
)
1262 return idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1265 static struct nfs4_stid
*find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1267 struct nfs4_stid
*s
;
1269 s
= find_stateid(cl
, t
);
1272 if (typemask
& s
->sc_type
)
1277 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1278 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1280 struct nfs4_client
*clp
;
1281 struct sockaddr
*sa
= svc_addr(rqstp
);
1284 clp
= alloc_client(name
);
1288 INIT_LIST_HEAD(&clp
->cl_sessions
);
1289 ret
= copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1291 spin_lock(&client_lock
);
1293 spin_unlock(&client_lock
);
1296 idr_init(&clp
->cl_stateids
);
1297 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1298 atomic_set(&clp
->cl_refcount
, 0);
1299 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1300 INIT_LIST_HEAD(&clp
->cl_idhash
);
1301 INIT_LIST_HEAD(&clp
->cl_strhash
);
1302 INIT_LIST_HEAD(&clp
->cl_openowners
);
1303 INIT_LIST_HEAD(&clp
->cl_delegations
);
1304 INIT_LIST_HEAD(&clp
->cl_lru
);
1305 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1306 spin_lock_init(&clp
->cl_lock
);
1307 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1308 clp
->cl_time
= get_seconds();
1309 clear_bit(0, &clp
->cl_cb_slot_busy
);
1310 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1311 copy_verf(clp
, verf
);
1312 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1314 clp
->cl_cb_session
= NULL
;
1319 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1321 unsigned int idhashval
;
1323 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1324 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1325 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1330 move_to_confirmed(struct nfs4_client
*clp
)
1332 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1333 unsigned int strhashval
;
1335 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1336 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1337 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1338 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1342 static struct nfs4_client
*
1343 find_confirmed_client(clientid_t
*clid
)
1345 struct nfs4_client
*clp
;
1346 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1348 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1349 if (same_clid(&clp
->cl_clientid
, clid
)) {
1357 static struct nfs4_client
*
1358 find_unconfirmed_client(clientid_t
*clid
)
1360 struct nfs4_client
*clp
;
1361 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1363 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1364 if (same_clid(&clp
->cl_clientid
, clid
))
1370 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1372 return clp
->cl_exchange_flags
!= 0;
1375 static struct nfs4_client
*
1376 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1378 struct nfs4_client
*clp
;
1380 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1381 if (same_name(clp
->cl_recdir
, dname
))
1387 static struct nfs4_client
*
1388 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1390 struct nfs4_client
*clp
;
1392 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1393 if (same_name(clp
->cl_recdir
, dname
))
1400 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1402 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1403 struct sockaddr
*sa
= svc_addr(rqstp
);
1404 u32 scopeid
= rpc_get_scope_id(sa
);
1405 unsigned short expected_family
;
1407 /* Currently, we only support tcp and tcp6 for the callback channel */
1408 if (se
->se_callback_netid_len
== 3 &&
1409 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1410 expected_family
= AF_INET
;
1411 else if (se
->se_callback_netid_len
== 4 &&
1412 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1413 expected_family
= AF_INET6
;
1417 conn
->cb_addrlen
= rpc_uaddr2sockaddr(&init_net
, se
->se_callback_addr_val
,
1418 se
->se_callback_addr_len
,
1419 (struct sockaddr
*)&conn
->cb_addr
,
1420 sizeof(conn
->cb_addr
));
1422 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1425 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1426 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1428 conn
->cb_prog
= se
->se_callback_prog
;
1429 conn
->cb_ident
= se
->se_callback_ident
;
1430 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
1433 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1434 conn
->cb_addrlen
= 0;
1435 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1436 "will not receive delegations\n",
1437 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1443 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1446 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1448 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1451 dprintk("--> %s slot %p\n", __func__
, slot
);
1453 slot
->sl_opcnt
= resp
->opcnt
;
1454 slot
->sl_status
= resp
->cstate
.status
;
1456 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
1457 if (nfsd4_not_cached(resp
)) {
1458 slot
->sl_datalen
= 0;
1461 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1462 base
= (char *)resp
->cstate
.datap
-
1463 (char *)resp
->xbuf
->head
[0].iov_base
;
1464 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1466 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1471 * Encode the replay sequence operation from the slot values.
1472 * If cachethis is FALSE encode the uncached rep error on the next
1473 * operation which sets resp->p and increments resp->opcnt for
1474 * nfs4svc_encode_compoundres.
1478 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1479 struct nfsd4_compoundres
*resp
)
1481 struct nfsd4_op
*op
;
1482 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1484 /* Encode the replayed sequence operation */
1485 op
= &args
->ops
[resp
->opcnt
- 1];
1486 nfsd4_encode_operation(resp
, op
);
1488 /* Return nfserr_retry_uncached_rep in next operation. */
1489 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
1490 op
= &args
->ops
[resp
->opcnt
++];
1491 op
->status
= nfserr_retry_uncached_rep
;
1492 nfsd4_encode_operation(resp
, op
);
1498 * The sequence operation is not cached because we can use the slot and
1502 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1503 struct nfsd4_sequence
*seq
)
1505 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1508 dprintk("--> %s slot %p\n", __func__
, slot
);
1510 /* Either returns 0 or nfserr_retry_uncached */
1511 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1512 if (status
== nfserr_retry_uncached_rep
)
1515 /* The sequence operation has been encoded, cstate->datap set. */
1516 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1518 resp
->opcnt
= slot
->sl_opcnt
;
1519 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1520 status
= slot
->sl_status
;
1526 * Set the exchange_id flags returned by the server.
1529 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1531 /* pNFS is not supported */
1532 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1534 /* Referrals are supported, Migration is not. */
1535 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1537 /* set the wire flags to return to client. */
1538 clid
->flags
= new->cl_exchange_flags
;
1541 static bool client_has_state(struct nfs4_client
*clp
)
1544 * Note clp->cl_openowners check isn't quite right: there's no
1545 * need to count owners without stateid's.
1547 * Also note we should probably be using this in 4.0 case too.
1549 return !list_empty(&clp
->cl_openowners
)
1550 || !list_empty(&clp
->cl_delegations
)
1551 || !list_empty(&clp
->cl_sessions
);
1555 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1556 struct nfsd4_compound_state
*cstate
,
1557 struct nfsd4_exchange_id
*exid
)
1559 struct nfs4_client
*unconf
, *conf
, *new;
1561 unsigned int strhashval
;
1562 char dname
[HEXDIR_LEN
];
1563 char addr_str
[INET6_ADDRSTRLEN
];
1564 nfs4_verifier verf
= exid
->verifier
;
1565 struct sockaddr
*sa
= svc_addr(rqstp
);
1566 bool update
= exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
;
1568 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1569 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1570 "ip_addr=%s flags %x, spa_how %d\n",
1571 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1572 addr_str
, exid
->flags
, exid
->spa_how
);
1574 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
1575 return nfserr_inval
;
1577 /* Currently only support SP4_NONE */
1578 switch (exid
->spa_how
) {
1582 return nfserr_serverfault
;
1584 BUG(); /* checked by xdr code */
1586 return nfserr_serverfault
; /* no excuse :-/ */
1589 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1594 strhashval
= clientstr_hashval(dname
);
1596 /* Cases below refer to rfc 5661 section 18.35.4: */
1598 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1600 bool creds_match
= same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
);
1601 bool verfs_match
= same_verf(&verf
, &conf
->cl_verifier
);
1604 if (!clp_used_exchangeid(conf
)) { /* buggy client */
1605 status
= nfserr_inval
;
1608 if (!creds_match
) { /* case 9 */
1609 status
= nfserr_perm
;
1612 if (!verfs_match
) { /* case 8 */
1613 status
= nfserr_not_same
;
1617 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1621 if (!creds_match
) { /* case 3 */
1622 if (client_has_state(conf
)) {
1623 status
= nfserr_clid_inuse
;
1626 expire_client(conf
);
1629 if (verfs_match
) { /* case 2 */
1630 conf
->cl_exchange_flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1634 /* case 5, client reboot */
1638 if (update
) { /* case 7 */
1639 status
= nfserr_noent
;
1643 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1644 if (unconf
) /* case 4, possible retry or client restart */
1645 expire_client(unconf
);
1647 /* case 1 (normal case) */
1649 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1651 status
= nfserr_jukebox
;
1656 add_to_unconfirmed(new, strhashval
);
1658 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1659 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1661 exid
->seqid
= new->cl_cs_slot
.sl_seqid
+ 1;
1662 nfsd4_set_ex_flags(new, exid
);
1664 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1665 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1669 nfs4_unlock_state();
1674 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1676 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1679 /* The slot is in use, and no response has been sent. */
1681 if (seqid
== slot_seqid
)
1682 return nfserr_jukebox
;
1684 return nfserr_seq_misordered
;
1686 /* Note unsigned 32-bit arithmetic handles wraparound: */
1687 if (likely(seqid
== slot_seqid
+ 1))
1689 if (seqid
== slot_seqid
)
1690 return nfserr_replay_cache
;
1691 return nfserr_seq_misordered
;
1695 * Cache the create session result into the create session single DRC
1696 * slot cache by saving the xdr structure. sl_seqid has been set.
1697 * Do this for solo or embedded create session operations.
1700 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1701 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
1703 slot
->sl_status
= nfserr
;
1704 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1708 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1709 struct nfsd4_clid_slot
*slot
)
1711 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1712 return slot
->sl_status
;
1715 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1716 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1717 1 + /* MIN tag is length with zero, only length */ \
1718 3 + /* version, opcount, opcode */ \
1719 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1720 /* seqid, slotID, slotID, cache */ \
1721 4 ) * sizeof(__be32))
1723 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1724 2 + /* verifier: AUTH_NULL, length 0 */\
1726 1 + /* MIN tag is length with zero, only length */ \
1727 3 + /* opcount, opcode, opstatus*/ \
1728 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1729 /* seqid, slotID, slotID, slotID, status */ \
1730 5 ) * sizeof(__be32))
1732 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1734 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1735 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1739 nfsd4_create_session(struct svc_rqst
*rqstp
,
1740 struct nfsd4_compound_state
*cstate
,
1741 struct nfsd4_create_session
*cr_ses
)
1743 struct sockaddr
*sa
= svc_addr(rqstp
);
1744 struct nfs4_client
*conf
, *unconf
;
1745 struct nfsd4_session
*new;
1746 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1747 bool confirm_me
= false;
1750 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1751 return nfserr_inval
;
1754 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1755 conf
= find_confirmed_client(&cr_ses
->clientid
);
1758 cs_slot
= &conf
->cl_cs_slot
;
1759 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1760 if (status
== nfserr_replay_cache
) {
1761 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1763 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1764 status
= nfserr_seq_misordered
;
1767 } else if (unconf
) {
1768 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1769 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1770 status
= nfserr_clid_inuse
;
1773 cs_slot
= &unconf
->cl_cs_slot
;
1774 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1776 /* an unconfirmed replay returns misordered */
1777 status
= nfserr_seq_misordered
;
1783 status
= nfserr_stale_clientid
;
1788 * XXX: we should probably set this at creation time, and check
1789 * for consistent minorversion use throughout:
1791 conf
->cl_minorversion
= 1;
1793 * We do not support RDMA or persistent sessions
1795 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1796 cr_ses
->flags
&= ~SESSION4_RDMA
;
1798 status
= nfserr_toosmall
;
1799 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1802 status
= nfserr_jukebox
;
1803 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1807 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1808 NFS4_MAX_SESSIONID_LEN
);
1809 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1810 sizeof(struct nfsd4_channel_attrs
));
1811 cs_slot
->sl_seqid
++;
1812 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1814 /* cache solo and embedded create sessions under the state lock */
1815 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1817 unsigned int hash
= clientstr_hashval(unconf
->cl_recdir
);
1818 struct nfs4_client
*old
=
1819 find_confirmed_client_by_str(conf
->cl_recdir
, hash
);
1822 move_to_confirmed(conf
);
1825 nfs4_unlock_state();
1826 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1830 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1832 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1833 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1835 return argp
->opcnt
== resp
->opcnt
;
1838 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1841 case NFS4_CDFC4_FORE
:
1842 case NFS4_CDFC4_BACK
:
1844 case NFS4_CDFC4_FORE_OR_BOTH
:
1845 case NFS4_CDFC4_BACK_OR_BOTH
:
1846 *dir
= NFS4_CDFC4_BOTH
;
1849 return nfserr_inval
;
1852 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1853 struct nfsd4_compound_state
*cstate
,
1854 struct nfsd4_bind_conn_to_session
*bcts
)
1858 if (!nfsd4_last_compound_op(rqstp
))
1859 return nfserr_not_only_op
;
1860 spin_lock(&client_lock
);
1861 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1862 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1863 * client_lock iself: */
1864 if (cstate
->session
) {
1865 nfsd4_get_session(cstate
->session
);
1866 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1868 spin_unlock(&client_lock
);
1869 if (!cstate
->session
)
1870 return nfserr_badsession
;
1872 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1874 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1878 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1882 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1886 nfsd4_destroy_session(struct svc_rqst
*r
,
1887 struct nfsd4_compound_state
*cstate
,
1888 struct nfsd4_destroy_session
*sessionid
)
1890 struct nfsd4_session
*ses
;
1891 __be32 status
= nfserr_badsession
;
1894 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1895 * - Should we return nfserr_back_chan_busy if waiting for
1896 * callbacks on to-be-destroyed session?
1897 * - Do we need to clear any callback info from previous session?
1900 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1901 if (!nfsd4_last_compound_op(r
))
1902 return nfserr_not_only_op
;
1904 dump_sessionid(__func__
, &sessionid
->sessionid
);
1905 spin_lock(&client_lock
);
1906 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1908 spin_unlock(&client_lock
);
1912 unhash_session(ses
);
1913 spin_unlock(&client_lock
);
1916 nfsd4_probe_callback_sync(ses
->se_client
);
1917 nfs4_unlock_state();
1919 spin_lock(&client_lock
);
1920 nfsd4_del_conns(ses
);
1921 nfsd4_put_session_locked(ses
);
1922 spin_unlock(&client_lock
);
1925 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1929 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1931 struct nfsd4_conn
*c
;
1933 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1934 if (c
->cn_xprt
== xpt
) {
1941 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1943 struct nfs4_client
*clp
= ses
->se_client
;
1944 struct nfsd4_conn
*c
;
1947 spin_lock(&clp
->cl_lock
);
1948 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1950 spin_unlock(&clp
->cl_lock
);
1954 __nfsd4_hash_conn(new, ses
);
1955 spin_unlock(&clp
->cl_lock
);
1956 ret
= nfsd4_register_conn(new);
1958 /* oops; xprt is already down: */
1959 nfsd4_conn_lost(&new->cn_xpt_user
);
1963 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1965 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1967 return args
->opcnt
> session
->se_fchannel
.maxops
;
1970 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1971 struct nfsd4_session
*session
)
1973 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1975 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1979 nfsd4_sequence(struct svc_rqst
*rqstp
,
1980 struct nfsd4_compound_state
*cstate
,
1981 struct nfsd4_sequence
*seq
)
1983 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1984 struct nfsd4_session
*session
;
1985 struct nfsd4_slot
*slot
;
1986 struct nfsd4_conn
*conn
;
1989 if (resp
->opcnt
!= 1)
1990 return nfserr_sequence_pos
;
1993 * Will be either used or freed by nfsd4_sequence_check_conn
1996 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1998 return nfserr_jukebox
;
2000 spin_lock(&client_lock
);
2001 status
= nfserr_badsession
;
2002 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
2006 status
= nfserr_too_many_ops
;
2007 if (nfsd4_session_too_many_ops(rqstp
, session
))
2010 status
= nfserr_req_too_big
;
2011 if (nfsd4_request_too_big(rqstp
, session
))
2014 status
= nfserr_badslot
;
2015 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
2018 slot
= session
->se_slots
[seq
->slotid
];
2019 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
2021 /* We do not negotiate the number of slots yet, so set the
2022 * maxslots to the session maxreqs which is used to encode
2023 * sr_highest_slotid and the sr_target_slot id to maxslots */
2024 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
2026 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
2027 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
2028 if (status
== nfserr_replay_cache
) {
2029 status
= nfserr_seq_misordered
;
2030 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
2032 cstate
->slot
= slot
;
2033 cstate
->session
= session
;
2034 /* Return the cached reply status and set cstate->status
2035 * for nfsd4_proc_compound processing */
2036 status
= nfsd4_replay_cache_entry(resp
, seq
);
2037 cstate
->status
= nfserr_replay_cache
;
2043 nfsd4_sequence_check_conn(conn
, session
);
2046 /* Success! bump slot seqid */
2047 slot
->sl_seqid
= seq
->seqid
;
2048 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
2050 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
2052 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
2054 cstate
->slot
= slot
;
2055 cstate
->session
= session
;
2058 /* Hold a session reference until done processing the compound. */
2059 if (cstate
->session
) {
2060 struct nfs4_client
*clp
= session
->se_client
;
2062 nfsd4_get_session(cstate
->session
);
2063 atomic_inc(&clp
->cl_refcount
);
2064 switch (clp
->cl_cb_state
) {
2066 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
2068 case NFSD4_CB_FAULT
:
2069 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
2072 seq
->status_flags
= 0;
2076 spin_unlock(&client_lock
);
2077 dprintk("%s: return %d\n", __func__
, ntohl(status
));
2082 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
2084 struct nfs4_client
*conf
, *unconf
, *clp
;
2088 unconf
= find_unconfirmed_client(&dc
->clientid
);
2089 conf
= find_confirmed_client(&dc
->clientid
);
2094 if (!is_client_expired(conf
) && client_has_state(conf
)) {
2095 status
= nfserr_clientid_busy
;
2099 /* rfc5661 18.50.3 */
2100 if (cstate
->session
&& conf
== cstate
->session
->se_client
) {
2101 status
= nfserr_clientid_busy
;
2107 status
= nfserr_stale_clientid
;
2113 nfs4_unlock_state();
2114 dprintk("%s return %d\n", __func__
, ntohl(status
));
2119 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2123 if (rc
->rca_one_fs
) {
2124 if (!cstate
->current_fh
.fh_dentry
)
2125 return nfserr_nofilehandle
;
2127 * We don't take advantage of the rca_one_fs case.
2128 * That's OK, it's optional, we can safely ignore it.
2134 status
= nfserr_complete_already
;
2135 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
2136 &cstate
->session
->se_client
->cl_flags
))
2139 status
= nfserr_stale_clientid
;
2140 if (is_client_expired(cstate
->session
->se_client
))
2142 * The following error isn't really legal.
2143 * But we only get here if the client just explicitly
2144 * destroyed the client. Surely it no longer cares what
2145 * error it gets back on an operation for the dead
2151 nfsd4_client_record_create(cstate
->session
->se_client
);
2153 nfs4_unlock_state();
2158 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2159 struct nfsd4_setclientid
*setclid
)
2161 struct xdr_netobj clname
= setclid
->se_name
;
2162 nfs4_verifier clverifier
= setclid
->se_verf
;
2163 unsigned int strhashval
;
2164 struct nfs4_client
*conf
, *unconf
, *new;
2166 char dname
[HEXDIR_LEN
];
2168 status
= nfs4_make_rec_clidname(dname
, &clname
);
2172 strhashval
= clientstr_hashval(dname
);
2174 /* Cases below refer to rfc 3530 section 14.2.33: */
2176 conf
= find_confirmed_client_by_str(dname
, strhashval
);
2179 status
= nfserr_clid_inuse
;
2180 if (clp_used_exchangeid(conf
))
2182 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
2183 char addr_str
[INET6_ADDRSTRLEN
];
2184 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
2186 dprintk("NFSD: setclientid: string in use by client "
2187 "at %s\n", addr_str
);
2191 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
2193 expire_client(unconf
);
2194 status
= nfserr_jukebox
;
2195 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2198 if (conf
&& same_verf(&conf
->cl_verifier
, &clverifier
))
2199 /* case 1: probable callback update */
2200 copy_clid(new, conf
);
2201 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2204 * XXX: we should probably set this at creation time, and check
2205 * for consistent minorversion use throughout:
2207 new->cl_minorversion
= 0;
2208 gen_callback(new, setclid
, rqstp
);
2209 add_to_unconfirmed(new, strhashval
);
2210 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2211 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2212 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2215 nfs4_unlock_state();
2221 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2222 struct nfsd4_compound_state
*cstate
,
2223 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2225 struct nfs4_client
*conf
, *unconf
;
2226 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2227 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2230 if (STALE_CLIENTID(clid
))
2231 return nfserr_stale_clientid
;
2234 conf
= find_confirmed_client(clid
);
2235 unconf
= find_unconfirmed_client(clid
);
2237 * We try hard to give out unique clientid's, so if we get an
2238 * attempt to confirm the same clientid with a different cred,
2239 * there's a bug somewhere. Let's charitably assume it's our
2242 status
= nfserr_serverfault
;
2243 if (unconf
&& !same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
))
2245 if (conf
&& !same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2247 /* cases below refer to rfc 3530 section 14.2.34: */
2248 if (!unconf
|| !same_verf(&confirm
, &unconf
->cl_confirm
)) {
2249 if (conf
&& !unconf
) /* case 2: probable retransmit */
2251 else /* case 4: client hasn't noticed we rebooted yet? */
2252 status
= nfserr_stale_clientid
;
2256 if (conf
) { /* case 1: callback update */
2257 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2258 nfsd4_probe_callback(conf
);
2259 expire_client(unconf
);
2260 } else { /* case 3: normal case; new or rebooted client */
2261 unsigned int hash
= clientstr_hashval(unconf
->cl_recdir
);
2263 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
, hash
);
2265 nfsd4_client_record_remove(conf
);
2266 expire_client(conf
);
2268 move_to_confirmed(unconf
);
2269 nfsd4_probe_callback(unconf
);
2272 nfs4_unlock_state();
2276 static struct nfs4_file
*nfsd4_alloc_file(void)
2278 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2281 /* OPEN Share state helper functions */
2282 static void nfsd4_init_file(struct nfs4_file
*fp
, struct inode
*ino
)
2284 unsigned int hashval
= file_hashval(ino
);
2286 atomic_set(&fp
->fi_ref
, 1);
2287 INIT_LIST_HEAD(&fp
->fi_hash
);
2288 INIT_LIST_HEAD(&fp
->fi_stateids
);
2289 INIT_LIST_HEAD(&fp
->fi_delegations
);
2290 fp
->fi_inode
= igrab(ino
);
2291 fp
->fi_had_conflict
= false;
2292 fp
->fi_lease
= NULL
;
2293 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2294 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2295 spin_lock(&recall_lock
);
2296 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2297 spin_unlock(&recall_lock
);
2301 nfsd4_free_slab(struct kmem_cache
**slab
)
2305 kmem_cache_destroy(*slab
);
2310 nfsd4_free_slabs(void)
2312 nfsd4_free_slab(&openowner_slab
);
2313 nfsd4_free_slab(&lockowner_slab
);
2314 nfsd4_free_slab(&file_slab
);
2315 nfsd4_free_slab(&stateid_slab
);
2316 nfsd4_free_slab(&deleg_slab
);
2320 nfsd4_init_slabs(void)
2322 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2323 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2324 if (openowner_slab
== NULL
)
2326 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2327 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2328 if (lockowner_slab
== NULL
)
2330 file_slab
= kmem_cache_create("nfsd4_files",
2331 sizeof(struct nfs4_file
), 0, 0, NULL
);
2332 if (file_slab
== NULL
)
2334 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2335 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
2336 if (stateid_slab
== NULL
)
2338 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2339 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2340 if (deleg_slab
== NULL
)
2345 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2349 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2351 kfree(oo
->oo_owner
.so_owner
.data
);
2352 kmem_cache_free(openowner_slab
, oo
);
2355 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2357 kfree(lo
->lo_owner
.so_owner
.data
);
2358 kmem_cache_free(lockowner_slab
, lo
);
2361 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2363 rp
->rp_status
= nfserr_serverfault
;
2365 rp
->rp_buf
= rp
->rp_ibuf
;
2368 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2370 struct nfs4_stateowner
*sop
;
2372 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2376 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2377 if (!sop
->so_owner
.data
) {
2378 kmem_cache_free(slab
, sop
);
2381 sop
->so_owner
.len
= owner
->len
;
2383 INIT_LIST_HEAD(&sop
->so_stateids
);
2384 sop
->so_client
= clp
;
2385 init_nfs4_replay(&sop
->so_replay
);
2389 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2391 list_add(&oo
->oo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2392 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2395 static struct nfs4_openowner
*
2396 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2397 struct nfs4_openowner
*oo
;
2399 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2402 oo
->oo_owner
.so_is_open_owner
= 1;
2403 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2404 oo
->oo_flags
= NFS4_OO_NEW
;
2406 oo
->oo_last_closed_stid
= NULL
;
2407 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2408 hash_openowner(oo
, clp
, strhashval
);
2412 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2413 struct nfs4_openowner
*oo
= open
->op_openowner
;
2414 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
2416 init_stid(&stp
->st_stid
, clp
, NFS4_OPEN_STID
);
2417 INIT_LIST_HEAD(&stp
->st_lockowners
);
2418 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2419 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2420 stp
->st_stateowner
= &oo
->oo_owner
;
2423 stp
->st_access_bmap
= 0;
2424 stp
->st_deny_bmap
= 0;
2425 set_access(open
->op_share_access
, stp
);
2426 set_deny(open
->op_share_deny
, stp
);
2427 stp
->st_openstp
= NULL
;
2431 move_to_close_lru(struct nfs4_openowner
*oo
)
2433 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2435 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2436 oo
->oo_time
= get_seconds();
2440 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2443 return (sop
->so_owner
.len
== owner
->len
) &&
2444 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2445 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2448 static struct nfs4_openowner
*
2449 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2451 struct nfs4_stateowner
*so
;
2452 struct nfs4_openowner
*oo
;
2454 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2455 if (!so
->so_is_open_owner
)
2457 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
)) {
2459 renew_client(oo
->oo_owner
.so_client
);
2466 /* search file_hashtbl[] for file */
2467 static struct nfs4_file
*
2468 find_file(struct inode
*ino
)
2470 unsigned int hashval
= file_hashval(ino
);
2471 struct nfs4_file
*fp
;
2473 spin_lock(&recall_lock
);
2474 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2475 if (fp
->fi_inode
== ino
) {
2477 spin_unlock(&recall_lock
);
2481 spin_unlock(&recall_lock
);
2486 * Called to check deny when READ with all zero stateid or
2487 * WRITE with all zero or all one stateid
2490 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2492 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2493 struct nfs4_file
*fp
;
2494 struct nfs4_ol_stateid
*stp
;
2497 dprintk("NFSD: nfs4_share_conflict\n");
2499 fp
= find_file(ino
);
2502 ret
= nfserr_locked
;
2503 /* Search for conflicting share reservations */
2504 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2505 if (test_deny(deny_type
, stp
) ||
2506 test_deny(NFS4_SHARE_DENY_BOTH
, stp
))
2515 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2517 /* We're assuming the state code never drops its reference
2518 * without first removing the lease. Since we're in this lease
2519 * callback (and since the lease code is serialized by the kernel
2520 * lock) we know the server hasn't removed the lease yet, we know
2521 * it's safe to take a reference: */
2522 atomic_inc(&dp
->dl_count
);
2524 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2526 /* only place dl_time is set. protected by lock_flocks*/
2527 dp
->dl_time
= get_seconds();
2529 nfsd4_cb_recall(dp
);
2532 /* Called from break_lease() with lock_flocks() held. */
2533 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2535 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2536 struct nfs4_delegation
*dp
;
2539 /* We assume break_lease is only called once per lease: */
2540 BUG_ON(fp
->fi_had_conflict
);
2542 * We don't want the locks code to timeout the lease for us;
2543 * we'll remove it ourself if a delegation isn't returned
2546 fl
->fl_break_time
= 0;
2548 spin_lock(&recall_lock
);
2549 fp
->fi_had_conflict
= true;
2550 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2551 nfsd_break_one_deleg(dp
);
2552 spin_unlock(&recall_lock
);
2556 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2559 return lease_modify(onlist
, arg
);
2564 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2565 .lm_break
= nfsd_break_deleg_cb
,
2566 .lm_change
= nfsd_change_deleg_cb
,
2569 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2571 if (nfsd4_has_session(cstate
))
2573 if (seqid
== so
->so_seqid
- 1)
2574 return nfserr_replay_me
;
2575 if (seqid
== so
->so_seqid
)
2577 return nfserr_bad_seqid
;
2581 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2582 struct nfsd4_open
*open
)
2584 clientid_t
*clientid
= &open
->op_clientid
;
2585 struct nfs4_client
*clp
= NULL
;
2586 unsigned int strhashval
;
2587 struct nfs4_openowner
*oo
= NULL
;
2590 if (STALE_CLIENTID(&open
->op_clientid
))
2591 return nfserr_stale_clientid
;
2593 * In case we need it later, after we've already created the
2594 * file and don't want to risk a further failure:
2596 open
->op_file
= nfsd4_alloc_file();
2597 if (open
->op_file
== NULL
)
2598 return nfserr_jukebox
;
2600 strhashval
= ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2601 oo
= find_openstateowner_str(strhashval
, open
);
2602 open
->op_openowner
= oo
;
2604 clp
= find_confirmed_client(clientid
);
2606 return nfserr_expired
;
2609 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
2610 /* Replace unconfirmed owners without checking for replay. */
2611 clp
= oo
->oo_owner
.so_client
;
2612 release_openowner(oo
);
2613 open
->op_openowner
= NULL
;
2616 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2619 clp
= oo
->oo_owner
.so_client
;
2622 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2624 return nfserr_jukebox
;
2625 open
->op_openowner
= oo
;
2627 open
->op_stp
= nfs4_alloc_stateid(clp
);
2629 return nfserr_jukebox
;
2633 static inline __be32
2634 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2636 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2637 return nfserr_openmode
;
2642 static int share_access_to_flags(u32 share_access
)
2644 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2647 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
2649 struct nfs4_stid
*ret
;
2651 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
2654 return delegstateid(ret
);
2657 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
2659 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
2660 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
2664 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2665 struct nfs4_delegation
**dp
)
2668 __be32 status
= nfserr_bad_stateid
;
2670 *dp
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
2673 flags
= share_access_to_flags(open
->op_share_access
);
2674 status
= nfs4_check_delegmode(*dp
, flags
);
2678 if (!nfsd4_is_deleg_cur(open
))
2682 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2687 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
**stpp
)
2689 struct nfs4_ol_stateid
*local
;
2690 struct nfs4_openowner
*oo
= open
->op_openowner
;
2692 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2693 /* ignore lock owners */
2694 if (local
->st_stateowner
->so_is_open_owner
== 0)
2696 /* remember if we have seen this open owner */
2697 if (local
->st_stateowner
== &oo
->oo_owner
)
2699 /* check for conflicting share reservations */
2700 if (!test_share(local
, open
))
2701 return nfserr_share_denied
;
2706 static void nfs4_free_stateid(struct nfs4_ol_stateid
*s
)
2708 kmem_cache_free(stateid_slab
, s
);
2711 static inline int nfs4_access_to_access(u32 nfs4_access
)
2715 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2716 flags
|= NFSD_MAY_READ
;
2717 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2718 flags
|= NFSD_MAY_WRITE
;
2722 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2723 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2726 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2727 int access
= nfs4_access_to_access(open
->op_share_access
);
2729 if (!fp
->fi_fds
[oflag
]) {
2730 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2731 &fp
->fi_fds
[oflag
]);
2735 nfs4_file_get_access(fp
, oflag
);
2740 static inline __be32
2741 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2742 struct nfsd4_open
*open
)
2744 struct iattr iattr
= {
2745 .ia_valid
= ATTR_SIZE
,
2748 if (!open
->op_truncate
)
2750 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2751 return nfserr_inval
;
2752 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2756 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
)
2758 u32 op_share_access
= open
->op_share_access
;
2762 new_access
= !test_access(op_share_access
, stp
);
2764 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2768 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2771 int oflag
= nfs4_access_to_omode(op_share_access
);
2772 nfs4_file_put_access(fp
, oflag
);
2776 /* remember the open */
2777 set_access(op_share_access
, stp
);
2778 set_deny(open
->op_share_deny
, stp
);
2785 nfs4_set_claim_prev(struct nfsd4_open
*open
, bool has_session
)
2787 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2790 /* Should we give out recallable state?: */
2791 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2793 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2796 * In the sessions case, since we don't have to establish a
2797 * separate connection for callbacks, we assume it's OK
2798 * until we hear otherwise:
2800 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2803 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2805 struct file_lock
*fl
;
2807 fl
= locks_alloc_lock();
2810 locks_init_lock(fl
);
2811 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2812 fl
->fl_flags
= FL_LEASE
;
2813 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2814 fl
->fl_end
= OFFSET_MAX
;
2815 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2816 fl
->fl_pid
= current
->tgid
;
2820 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2822 struct nfs4_file
*fp
= dp
->dl_file
;
2823 struct file_lock
*fl
;
2826 fl
= nfs4_alloc_init_lease(dp
, flag
);
2829 fl
->fl_file
= find_readable_file(fp
);
2830 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2831 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2833 list_del_init(&dp
->dl_perclnt
);
2834 locks_free_lock(fl
);
2838 fp
->fi_deleg_file
= fl
->fl_file
;
2839 get_file(fp
->fi_deleg_file
);
2840 atomic_set(&fp
->fi_delegees
, 1);
2841 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2845 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2847 struct nfs4_file
*fp
= dp
->dl_file
;
2850 return nfs4_setlease(dp
, flag
);
2851 spin_lock(&recall_lock
);
2852 if (fp
->fi_had_conflict
) {
2853 spin_unlock(&recall_lock
);
2856 atomic_inc(&fp
->fi_delegees
);
2857 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2858 spin_unlock(&recall_lock
);
2859 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2863 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
2865 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2866 if (status
== -EAGAIN
)
2867 open
->op_why_no_deleg
= WND4_CONTENTION
;
2869 open
->op_why_no_deleg
= WND4_RESOURCE
;
2870 switch (open
->op_deleg_want
) {
2871 case NFS4_SHARE_WANT_READ_DELEG
:
2872 case NFS4_SHARE_WANT_WRITE_DELEG
:
2873 case NFS4_SHARE_WANT_ANY_DELEG
:
2875 case NFS4_SHARE_WANT_CANCEL
:
2876 open
->op_why_no_deleg
= WND4_CANCELLED
;
2878 case NFS4_SHARE_WANT_NO_DELEG
:
2879 BUG(); /* not supposed to get here */
2885 * Attempt to hand out a delegation.
2888 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
*stp
)
2890 struct nfs4_delegation
*dp
;
2891 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2893 int status
= 0, flag
= 0;
2895 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2896 flag
= NFS4_OPEN_DELEGATE_NONE
;
2897 open
->op_recall
= 0;
2898 switch (open
->op_claim_type
) {
2899 case NFS4_OPEN_CLAIM_PREVIOUS
:
2901 open
->op_recall
= 1;
2902 flag
= open
->op_delegate_type
;
2903 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2906 case NFS4_OPEN_CLAIM_NULL
:
2907 /* Let's not give out any delegations till everyone's
2908 * had the chance to reclaim theirs.... */
2909 if (locks_in_grace())
2911 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
2913 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2914 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2916 flag
= NFS4_OPEN_DELEGATE_READ
;
2922 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2925 status
= nfs4_set_delegation(dp
, flag
);
2929 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
2931 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2932 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
2934 open
->op_delegate_type
= flag
;
2935 if (flag
== NFS4_OPEN_DELEGATE_NONE
) {
2936 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
2937 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2938 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2940 /* 4.1 client asking for a delegation? */
2941 if (open
->op_deleg_want
)
2942 nfsd4_open_deleg_none_ext(open
, status
);
2946 nfs4_put_delegation(dp
);
2948 flag
= NFS4_OPEN_DELEGATE_NONE
;
2952 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
2953 struct nfs4_delegation
*dp
)
2955 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
2956 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2957 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2958 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
2959 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
2960 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
2961 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2962 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
2964 /* Otherwise the client must be confused wanting a delegation
2965 * it already has, therefore we don't return
2966 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2971 * called with nfs4_lock_state() held.
2974 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
2976 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
2977 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
2978 struct nfs4_file
*fp
= NULL
;
2979 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2980 struct nfs4_ol_stateid
*stp
= NULL
;
2981 struct nfs4_delegation
*dp
= NULL
;
2985 * Lookup file; if found, lookup stateid and check open request,
2986 * and check for delegations in the process of being recalled.
2987 * If not found, create the nfs4_file struct
2989 fp
= find_file(ino
);
2991 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
2993 status
= nfs4_check_deleg(cl
, fp
, open
, &dp
);
2997 status
= nfserr_bad_stateid
;
2998 if (nfsd4_is_deleg_cur(open
))
3000 status
= nfserr_jukebox
;
3002 open
->op_file
= NULL
;
3003 nfsd4_init_file(fp
, ino
);
3007 * OPEN the file, or upgrade an existing OPEN.
3008 * If truncate fails, the OPEN fails.
3011 /* Stateid was found, this is an OPEN upgrade */
3012 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
3016 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, open
);
3019 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
3023 open
->op_stp
= NULL
;
3024 init_open_stateid(stp
, fp
, open
);
3026 update_stateid(&stp
->st_stid
.sc_stateid
);
3027 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3029 if (nfsd4_has_session(&resp
->cstate
)) {
3030 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3032 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
3033 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3034 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
3040 * Attempt to hand out a delegation. No error return, because the
3041 * OPEN succeeds even if we fail.
3043 nfs4_open_delegation(current_fh
, open
, stp
);
3047 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
3048 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3050 /* 4.1 client trying to upgrade/downgrade delegation? */
3051 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
3052 open
->op_deleg_want
)
3053 nfsd4_deleg_xgrade_none_ext(open
, dp
);
3057 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
3058 nfs4_set_claim_prev(open
, nfsd4_has_session(&resp
->cstate
));
3060 * To finish the open response, we just need to set the rflags.
3062 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
3063 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
3064 !nfsd4_has_session(&resp
->cstate
))
3065 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
3070 void nfsd4_cleanup_open_state(struct nfsd4_open
*open
, __be32 status
)
3072 if (open
->op_openowner
) {
3073 struct nfs4_openowner
*oo
= open
->op_openowner
;
3075 if (!list_empty(&oo
->oo_owner
.so_stateids
))
3076 list_del_init(&oo
->oo_close_lru
);
3077 if (oo
->oo_flags
& NFS4_OO_NEW
) {
3079 release_openowner(oo
);
3080 open
->op_openowner
= NULL
;
3082 oo
->oo_flags
&= ~NFS4_OO_NEW
;
3086 nfsd4_free_file(open
->op_file
);
3088 nfs4_free_stateid(open
->op_stp
);
3092 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3095 struct nfs4_client
*clp
;
3099 dprintk("process_renew(%08x/%08x): starting\n",
3100 clid
->cl_boot
, clid
->cl_id
);
3101 status
= nfserr_stale_clientid
;
3102 if (STALE_CLIENTID(clid
))
3104 clp
= find_confirmed_client(clid
);
3105 status
= nfserr_expired
;
3107 /* We assume the client took too long to RENEW. */
3108 dprintk("nfsd4_renew: clientid not found!\n");
3111 status
= nfserr_cb_path_down
;
3112 if (!list_empty(&clp
->cl_delegations
)
3113 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
3117 nfs4_unlock_state();
3121 static bool grace_ended
;
3124 nfsd4_end_grace(struct net
*net
)
3126 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
3128 /* do nothing if grace period already ended */
3132 dprintk("NFSD: end of grace period\n");
3134 nfsd4_record_grace_done(net
, boot_time
);
3135 locks_end_grace(&nn
->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");
3158 nfsd4_end_grace(&init_net
);
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 if (list_empty(&oo
->oo_owner
.so_stateids
)) {
3769 if (cstate
->minorversion
) {
3770 release_openowner(oo
);
3771 cstate
->replay_owner
= NULL
;
3774 * In the 4.0 case we need to keep the owners around a
3775 * little while to handle CLOSE replay.
3777 if (list_empty(&oo
->oo_owner
.so_stateids
))
3778 move_to_close_lru(oo
);
3782 if (!cstate
->replay_owner
)
3783 nfs4_unlock_state();
3788 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3789 struct nfsd4_delegreturn
*dr
)
3791 struct nfs4_delegation
*dp
;
3792 stateid_t
*stateid
= &dr
->dr_stateid
;
3793 struct nfs4_stid
*s
;
3794 struct inode
*inode
;
3797 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3799 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3802 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
, &s
);
3805 dp
= delegstateid(s
);
3806 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3810 unhash_delegation(dp
);
3812 nfs4_unlock_state();
3818 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3820 #define LOCKOWNER_INO_HASH_BITS 8
3821 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3822 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3825 end_offset(u64 start
, u64 len
)
3830 return end
>= start
? end
: NFS4_MAX_UINT64
;
3833 /* last octet in a range */
3835 last_byte_offset(u64 start
, u64 len
)
3841 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3844 static unsigned int lockowner_ino_hashval(struct inode
*inode
, u32 cl_id
, struct xdr_netobj
*ownername
)
3846 return (file_hashval(inode
) + cl_id
3847 + opaque_hashval(ownername
->data
, ownername
->len
))
3848 & LOCKOWNER_INO_HASH_MASK
;
3851 static struct list_head lockowner_ino_hashtbl
[LOCKOWNER_INO_HASH_SIZE
];
3854 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3855 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3856 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3857 * locking, this prevents us from being completely protocol-compliant. The
3858 * real solution to this problem is to start using unsigned file offsets in
3859 * the VFS, but this is a very deep change!
3862 nfs4_transform_lock_offset(struct file_lock
*lock
)
3864 if (lock
->fl_start
< 0)
3865 lock
->fl_start
= OFFSET_MAX
;
3866 if (lock
->fl_end
< 0)
3867 lock
->fl_end
= OFFSET_MAX
;
3870 /* Hack!: For now, we're defining this just so we can use a pointer to it
3871 * as a unique cookie to identify our (NFSv4's) posix locks. */
3872 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3876 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3878 struct nfs4_lockowner
*lo
;
3880 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3881 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3882 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3883 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3884 if (!deny
->ld_owner
.data
)
3885 /* We just don't care that much */
3887 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3888 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3891 deny
->ld_owner
.len
= 0;
3892 deny
->ld_owner
.data
= NULL
;
3893 deny
->ld_clientid
.cl_boot
= 0;
3894 deny
->ld_clientid
.cl_id
= 0;
3896 deny
->ld_start
= fl
->fl_start
;
3897 deny
->ld_length
= NFS4_MAX_UINT64
;
3898 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3899 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3900 deny
->ld_type
= NFS4_READ_LT
;
3901 if (fl
->fl_type
!= F_RDLCK
)
3902 deny
->ld_type
= NFS4_WRITE_LT
;
3905 static bool same_lockowner_ino(struct nfs4_lockowner
*lo
, struct inode
*inode
, clientid_t
*clid
, struct xdr_netobj
*owner
)
3907 struct nfs4_ol_stateid
*lst
;
3909 if (!same_owner_str(&lo
->lo_owner
, owner
, clid
))
3911 lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3912 struct nfs4_ol_stateid
, st_perstateowner
);
3913 return lst
->st_file
->fi_inode
== inode
;
3916 static struct nfs4_lockowner
*
3917 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3918 struct xdr_netobj
*owner
)
3920 unsigned int hashval
= lockowner_ino_hashval(inode
, clid
->cl_id
, owner
);
3921 struct nfs4_lockowner
*lo
;
3923 list_for_each_entry(lo
, &lockowner_ino_hashtbl
[hashval
], lo_owner_ino_hash
) {
3924 if (same_lockowner_ino(lo
, inode
, clid
, owner
))
3930 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
)
3932 struct inode
*inode
= open_stp
->st_file
->fi_inode
;
3933 unsigned int inohash
= lockowner_ino_hashval(inode
,
3934 clp
->cl_clientid
.cl_id
, &lo
->lo_owner
.so_owner
);
3936 list_add(&lo
->lo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
3937 list_add(&lo
->lo_owner_ino_hash
, &lockowner_ino_hashtbl
[inohash
]);
3938 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3942 * Alloc a lock owner structure.
3943 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3946 * strhashval = ownerstr_hashval
3949 static struct nfs4_lockowner
*
3950 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3951 struct nfs4_lockowner
*lo
;
3953 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
3956 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
3957 lo
->lo_owner
.so_is_open_owner
= 0;
3958 /* It is the openowner seqid that will be incremented in encode in the
3959 * case of new lockowners; so increment the lock seqid manually: */
3960 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
3961 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
3965 static struct nfs4_ol_stateid
*
3966 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_ol_stateid
*open_stp
)
3968 struct nfs4_ol_stateid
*stp
;
3969 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
3971 stp
= nfs4_alloc_stateid(clp
);
3974 init_stid(&stp
->st_stid
, clp
, NFS4_LOCK_STID
);
3975 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
3976 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
3977 stp
->st_stateowner
= &lo
->lo_owner
;
3980 stp
->st_access_bmap
= 0;
3981 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
3982 stp
->st_openstp
= open_stp
;
3987 check_lock_length(u64 offset
, u64 length
)
3989 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
3990 LOFF_OVERFLOW(offset
, length
)));
3993 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
3995 struct nfs4_file
*fp
= lock_stp
->st_file
;
3996 int oflag
= nfs4_access_to_omode(access
);
3998 if (test_access(access
, lock_stp
))
4000 nfs4_file_get_access(fp
, oflag
);
4001 set_access(access
, lock_stp
);
4004 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)
4006 struct nfs4_file
*fi
= ost
->st_file
;
4007 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
4008 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
4009 struct nfs4_lockowner
*lo
;
4010 unsigned int strhashval
;
4012 lo
= find_lockowner_str(fi
->fi_inode
, &cl
->cl_clientid
, &lock
->v
.new.owner
);
4014 if (!cstate
->minorversion
)
4015 return nfserr_bad_seqid
;
4016 /* XXX: a lockowner always has exactly one stateid: */
4017 *lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
4018 struct nfs4_ol_stateid
, st_perstateowner
);
4021 strhashval
= ownerstr_hashval(cl
->cl_clientid
.cl_id
,
4022 &lock
->v
.new.owner
);
4023 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
4025 return nfserr_jukebox
;
4026 *lst
= alloc_init_lock_stateid(lo
, fi
, ost
);
4028 release_lockowner(lo
);
4029 return nfserr_jukebox
;
4039 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4040 struct nfsd4_lock
*lock
)
4042 struct nfs4_openowner
*open_sop
= NULL
;
4043 struct nfs4_lockowner
*lock_sop
= NULL
;
4044 struct nfs4_ol_stateid
*lock_stp
;
4045 struct file
*filp
= NULL
;
4046 struct file_lock file_lock
;
4047 struct file_lock conflock
;
4049 bool new_state
= false;
4053 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4054 (long long) lock
->lk_offset
,
4055 (long long) lock
->lk_length
);
4057 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
4058 return nfserr_inval
;
4060 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
4061 S_IFREG
, NFSD_MAY_LOCK
))) {
4062 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4068 if (lock
->lk_is_new
) {
4069 struct nfs4_ol_stateid
*open_stp
= NULL
;
4071 if (nfsd4_has_session(cstate
))
4072 /* See rfc 5661 18.10.3: given clientid is ignored: */
4073 memcpy(&lock
->v
.new.clientid
,
4074 &cstate
->session
->se_client
->cl_clientid
,
4075 sizeof(clientid_t
));
4077 status
= nfserr_stale_clientid
;
4078 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
4081 /* validate and update open stateid and open seqid */
4082 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
4083 lock
->lk_new_open_seqid
,
4084 &lock
->lk_new_open_stateid
,
4088 open_sop
= openowner(open_stp
->st_stateowner
);
4089 status
= nfserr_bad_stateid
;
4090 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
4091 &lock
->v
.new.clientid
))
4093 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
4094 &lock_stp
, &new_state
);
4096 status
= nfs4_preprocess_seqid_op(cstate
,
4097 lock
->lk_old_lock_seqid
,
4098 &lock
->lk_old_lock_stateid
,
4099 NFS4_LOCK_STID
, &lock_stp
);
4102 lock_sop
= lockowner(lock_stp
->st_stateowner
);
4104 lkflg
= setlkflg(lock
->lk_type
);
4105 status
= nfs4_check_openmode(lock_stp
, lkflg
);
4109 status
= nfserr_grace
;
4110 if (locks_in_grace() && !lock
->lk_reclaim
)
4112 status
= nfserr_no_grace
;
4113 if (!locks_in_grace() && lock
->lk_reclaim
)
4116 locks_init_lock(&file_lock
);
4117 switch (lock
->lk_type
) {
4120 filp
= find_readable_file(lock_stp
->st_file
);
4122 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4123 file_lock
.fl_type
= F_RDLCK
;
4126 case NFS4_WRITEW_LT
:
4127 filp
= find_writeable_file(lock_stp
->st_file
);
4129 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4130 file_lock
.fl_type
= F_WRLCK
;
4133 status
= nfserr_inval
;
4137 status
= nfserr_openmode
;
4140 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
4141 file_lock
.fl_pid
= current
->tgid
;
4142 file_lock
.fl_file
= filp
;
4143 file_lock
.fl_flags
= FL_POSIX
;
4144 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4146 file_lock
.fl_start
= lock
->lk_offset
;
4147 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4148 nfs4_transform_lock_offset(&file_lock
);
4151 * Try to lock the file in the VFS.
4152 * Note: locks.c uses the BKL to protect the inode's lock list.
4155 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
4157 case 0: /* success! */
4158 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
4159 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
4163 case (EAGAIN
): /* conflock holds conflicting lock */
4164 status
= nfserr_denied
;
4165 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4166 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
4169 status
= nfserr_deadlock
;
4172 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4173 status
= nfserrno(err
);
4177 if (status
&& new_state
)
4178 release_lockowner(lock_sop
);
4179 if (!cstate
->replay_owner
)
4180 nfs4_unlock_state();
4185 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4186 * so we do a temporary open here just to get an open file to pass to
4187 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4190 static __be32
nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4193 __be32 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4195 err
= nfserrno(vfs_test_lock(file
, lock
));
4205 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4206 struct nfsd4_lockt
*lockt
)
4208 struct inode
*inode
;
4209 struct file_lock file_lock
;
4210 struct nfs4_lockowner
*lo
;
4213 if (locks_in_grace())
4214 return nfserr_grace
;
4216 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4217 return nfserr_inval
;
4221 status
= nfserr_stale_clientid
;
4222 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
4225 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4228 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4229 locks_init_lock(&file_lock
);
4230 switch (lockt
->lt_type
) {
4233 file_lock
.fl_type
= F_RDLCK
;
4236 case NFS4_WRITEW_LT
:
4237 file_lock
.fl_type
= F_WRLCK
;
4240 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4241 status
= nfserr_inval
;
4245 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4247 file_lock
.fl_owner
= (fl_owner_t
)lo
;
4248 file_lock
.fl_pid
= current
->tgid
;
4249 file_lock
.fl_flags
= FL_POSIX
;
4251 file_lock
.fl_start
= lockt
->lt_offset
;
4252 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4254 nfs4_transform_lock_offset(&file_lock
);
4256 status
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4260 if (file_lock
.fl_type
!= F_UNLCK
) {
4261 status
= nfserr_denied
;
4262 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4265 nfs4_unlock_state();
4270 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4271 struct nfsd4_locku
*locku
)
4273 struct nfs4_ol_stateid
*stp
;
4274 struct file
*filp
= NULL
;
4275 struct file_lock file_lock
;
4279 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4280 (long long) locku
->lu_offset
,
4281 (long long) locku
->lu_length
);
4283 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4284 return nfserr_inval
;
4288 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4289 &locku
->lu_stateid
, NFS4_LOCK_STID
, &stp
);
4292 filp
= find_any_file(stp
->st_file
);
4294 status
= nfserr_lock_range
;
4298 locks_init_lock(&file_lock
);
4299 file_lock
.fl_type
= F_UNLCK
;
4300 file_lock
.fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4301 file_lock
.fl_pid
= current
->tgid
;
4302 file_lock
.fl_file
= filp
;
4303 file_lock
.fl_flags
= FL_POSIX
;
4304 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4305 file_lock
.fl_start
= locku
->lu_offset
;
4307 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4308 nfs4_transform_lock_offset(&file_lock
);
4311 * Try to unlock the file in the VFS.
4313 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4315 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4319 * OK, unlock succeeded; the only thing left to do is update the stateid.
4321 update_stateid(&stp
->st_stid
.sc_stateid
);
4322 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4325 if (!cstate
->replay_owner
)
4326 nfs4_unlock_state();
4330 status
= nfserrno(err
);
4336 * 1: locks held by lockowner
4337 * 0: no locks held by lockowner
4340 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4342 struct file_lock
**flpp
;
4343 struct inode
*inode
= filp
->fi_inode
;
4347 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4348 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4359 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4360 struct nfsd4_compound_state
*cstate
,
4361 struct nfsd4_release_lockowner
*rlockowner
)
4363 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4364 struct nfs4_stateowner
*sop
;
4365 struct nfs4_lockowner
*lo
;
4366 struct nfs4_ol_stateid
*stp
;
4367 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4368 struct list_head matches
;
4369 unsigned int hashval
= ownerstr_hashval(clid
->cl_id
, owner
);
4372 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4373 clid
->cl_boot
, clid
->cl_id
);
4375 /* XXX check for lease expiration */
4377 status
= nfserr_stale_clientid
;
4378 if (STALE_CLIENTID(clid
))
4383 status
= nfserr_locks_held
;
4384 INIT_LIST_HEAD(&matches
);
4386 list_for_each_entry(sop
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
4387 if (sop
->so_is_open_owner
)
4389 if (!same_owner_str(sop
, owner
, clid
))
4391 list_for_each_entry(stp
, &sop
->so_stateids
,
4393 lo
= lockowner(sop
);
4394 if (check_for_locks(stp
->st_file
, lo
))
4396 list_add(&lo
->lo_list
, &matches
);
4399 /* Clients probably won't expect us to return with some (but not all)
4400 * of the lockowner state released; so don't release any until all
4401 * have been checked. */
4403 while (!list_empty(&matches
)) {
4404 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4406 /* unhash_stateowner deletes so_perclient only
4407 * for openowners. */
4408 list_del(&lo
->lo_list
);
4409 release_lockowner(lo
);
4412 nfs4_unlock_state();
4416 static inline struct nfs4_client_reclaim
*
4419 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4423 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4425 unsigned int strhashval
= clientstr_hashval(name
);
4426 struct nfs4_client
*clp
;
4428 clp
= find_confirmed_client_by_str(name
, strhashval
);
4431 return test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
4435 * failure => all reset bets are off, nfserr_no_grace...
4438 nfs4_client_to_reclaim(const char *name
)
4440 unsigned int strhashval
;
4441 struct nfs4_client_reclaim
*crp
= NULL
;
4443 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4444 crp
= alloc_reclaim();
4447 strhashval
= clientstr_hashval(name
);
4448 INIT_LIST_HEAD(&crp
->cr_strhash
);
4449 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4450 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4451 reclaim_str_hashtbl_size
++;
4456 nfs4_release_reclaim(void)
4458 struct nfs4_client_reclaim
*crp
= NULL
;
4461 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4462 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4463 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4464 struct nfs4_client_reclaim
, cr_strhash
);
4465 list_del(&crp
->cr_strhash
);
4467 reclaim_str_hashtbl_size
--;
4470 BUG_ON(reclaim_str_hashtbl_size
);
4474 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4475 struct nfs4_client_reclaim
*
4476 nfsd4_find_reclaim_client(struct nfs4_client
*clp
)
4478 unsigned int strhashval
;
4479 struct nfs4_client_reclaim
*crp
= NULL
;
4481 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4482 clp
->cl_name
.len
, clp
->cl_name
.data
,
4485 /* find clp->cl_name in reclaim_str_hashtbl */
4486 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4487 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4488 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4496 * Called from OPEN. Look for clientid in reclaim list.
4499 nfs4_check_open_reclaim(clientid_t
*clid
)
4501 struct nfs4_client
*clp
;
4503 /* find clientid in conf_id_hashtbl */
4504 clp
= find_confirmed_client(clid
);
4506 return nfserr_reclaim_bad
;
4508 return nfsd4_client_record_check(clp
) ? nfserr_reclaim_bad
: nfs_ok
;
4511 #ifdef CONFIG_NFSD_FAULT_INJECTION
4513 void nfsd_forget_clients(u64 num
)
4515 struct nfs4_client
*clp
, *next
;
4519 list_for_each_entry_safe(clp
, next
, &client_lru
, cl_lru
) {
4520 nfsd4_client_record_remove(clp
);
4525 nfs4_unlock_state();
4527 printk(KERN_INFO
"NFSD: Forgot %d clients", count
);
4530 static void release_lockowner_sop(struct nfs4_stateowner
*sop
)
4532 release_lockowner(lockowner(sop
));
4535 static void release_openowner_sop(struct nfs4_stateowner
*sop
)
4537 release_openowner(openowner(sop
));
4540 static int nfsd_release_n_owners(u64 num
, bool is_open_owner
,
4541 void (*release_sop
)(struct nfs4_stateowner
*))
4544 struct nfs4_stateowner
*sop
, *next
;
4546 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4547 list_for_each_entry_safe(sop
, next
, &ownerstr_hashtbl
[i
], so_strhash
) {
4548 if (sop
->so_is_open_owner
!= is_open_owner
)
4558 void nfsd_forget_locks(u64 num
)
4563 count
= nfsd_release_n_owners(num
, false, release_lockowner_sop
);
4564 nfs4_unlock_state();
4566 printk(KERN_INFO
"NFSD: Forgot %d locks", count
);
4569 void nfsd_forget_openowners(u64 num
)
4574 count
= nfsd_release_n_owners(num
, true, release_openowner_sop
);
4575 nfs4_unlock_state();
4577 printk(KERN_INFO
"NFSD: Forgot %d open owners", count
);
4580 int nfsd_process_n_delegations(u64 num
, struct list_head
*list
)
4583 struct nfs4_file
*fp
, *fnext
;
4584 struct nfs4_delegation
*dp
, *dnext
;
4586 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4587 list_for_each_entry_safe(fp
, fnext
, &file_hashtbl
[i
], fi_hash
) {
4588 list_for_each_entry_safe(dp
, dnext
, &fp
->fi_delegations
, dl_perfile
) {
4589 list_move(&dp
->dl_recall_lru
, list
);
4599 void nfsd_forget_delegations(u64 num
)
4603 struct nfs4_delegation
*dp
, *dnext
;
4605 spin_lock(&recall_lock
);
4606 count
= nfsd_process_n_delegations(num
, &victims
);
4607 spin_unlock(&recall_lock
);
4610 list_for_each_entry_safe(dp
, dnext
, &victims
, dl_recall_lru
)
4611 unhash_delegation(dp
);
4612 nfs4_unlock_state();
4614 printk(KERN_INFO
"NFSD: Forgot %d delegations", count
);
4617 void nfsd_recall_delegations(u64 num
)
4621 struct nfs4_delegation
*dp
, *dnext
;
4623 spin_lock(&recall_lock
);
4624 count
= nfsd_process_n_delegations(num
, &victims
);
4625 list_for_each_entry_safe(dp
, dnext
, &victims
, dl_recall_lru
) {
4626 list_del(&dp
->dl_recall_lru
);
4627 nfsd_break_one_deleg(dp
);
4629 spin_unlock(&recall_lock
);
4631 printk(KERN_INFO
"NFSD: Recalled %d delegations", count
);
4634 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4636 /* initialization to perform at module load time: */
4639 nfs4_state_init(void)
4643 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4644 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4645 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4646 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4647 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4648 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4650 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4651 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4652 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4653 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4655 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4656 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4658 for (i
= 0; i
< LOCKOWNER_INO_HASH_SIZE
; i
++)
4659 INIT_LIST_HEAD(&lockowner_ino_hashtbl
[i
]);
4660 INIT_LIST_HEAD(&close_lru
);
4661 INIT_LIST_HEAD(&client_lru
);
4662 INIT_LIST_HEAD(&del_recall_lru
);
4663 reclaim_str_hashtbl_size
= 0;
4667 * Since the lifetime of a delegation isn't limited to that of an open, a
4668 * client may quite reasonably hang on to a delegation as long as it has
4669 * the inode cached. This becomes an obvious problem the first time a
4670 * client's inode cache approaches the size of the server's total memory.
4672 * For now we avoid this problem by imposing a hard limit on the number
4673 * of delegations, which varies according to the server's memory size.
4676 set_max_delegations(void)
4679 * Allow at most 4 delegations per megabyte of RAM. Quick
4680 * estimates suggest that in the worst case (where every delegation
4681 * is for a different inode), a delegation could take about 1.5K,
4682 * giving a worst case usage of about 6% of memory.
4684 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4687 /* initialization to perform when the nfsd service is started: */
4690 nfs4_state_start(void)
4692 struct net
*net
= &init_net
;
4693 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4697 * FIXME: For now, we hang most of the pernet global stuff off of
4698 * init_net until nfsd is fully containerized. Eventually, we'll
4699 * need to pass a net pointer into this function, take a reference
4700 * to that instead and then do most of the rest of this on a per-net
4704 nfsd4_client_tracking_init(net
);
4705 boot_time
= get_seconds();
4706 locks_start_grace(&nn
->nfsd4_manager
);
4707 grace_ended
= false;
4708 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4710 ret
= set_callback_cred();
4715 laundry_wq
= create_singlethread_workqueue("nfsd4");
4716 if (laundry_wq
== NULL
) {
4720 ret
= nfsd4_create_callback_queue();
4722 goto out_free_laundry
;
4723 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4724 set_max_delegations();
4727 destroy_workqueue(laundry_wq
);
4729 nfsd4_client_tracking_exit(net
);
4735 __nfs4_state_shutdown(void)
4738 struct nfs4_client
*clp
= NULL
;
4739 struct nfs4_delegation
*dp
= NULL
;
4740 struct list_head
*pos
, *next
, reaplist
;
4742 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4743 while (!list_empty(&conf_id_hashtbl
[i
])) {
4744 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4747 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4748 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4752 INIT_LIST_HEAD(&reaplist
);
4753 spin_lock(&recall_lock
);
4754 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4755 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4756 list_move(&dp
->dl_recall_lru
, &reaplist
);
4758 spin_unlock(&recall_lock
);
4759 list_for_each_safe(pos
, next
, &reaplist
) {
4760 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4761 unhash_delegation(dp
);
4764 nfsd4_client_tracking_exit(&init_net
);
4769 nfs4_state_shutdown(void)
4771 struct net
*net
= &init_net
;
4772 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
4774 cancel_delayed_work_sync(&laundromat_work
);
4775 destroy_workqueue(laundry_wq
);
4776 locks_end_grace(&nn
->nfsd4_manager
);
4778 __nfs4_state_shutdown();
4779 nfs4_unlock_state();
4780 nfsd4_destroy_callback_queue();
4784 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4786 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
4787 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
4791 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4793 if (cstate
->minorversion
) {
4794 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
4795 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4800 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
4802 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4806 * functions to set current state id
4809 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4811 put_stateid(cstate
, &odp
->od_stateid
);
4815 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
4817 put_stateid(cstate
, &open
->op_stateid
);
4821 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4823 put_stateid(cstate
, &close
->cl_stateid
);
4827 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
4829 put_stateid(cstate
, &lock
->lk_resp_stateid
);
4833 * functions to consume current state id
4837 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4839 get_stateid(cstate
, &odp
->od_stateid
);
4843 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
4845 get_stateid(cstate
, &drp
->dr_stateid
);
4849 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
4851 get_stateid(cstate
, &fsp
->fr_stateid
);
4855 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
4857 get_stateid(cstate
, &setattr
->sa_stateid
);
4861 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4863 get_stateid(cstate
, &close
->cl_stateid
);
4867 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
4869 get_stateid(cstate
, &locku
->lu_stateid
);
4873 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
4875 get_stateid(cstate
, &read
->rd_stateid
);
4879 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
4881 get_stateid(cstate
, &write
->wr_stateid
);