2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
45 #include "current_stateid.h"
47 #define NFSDDBG_FACILITY NFSDDBG_PROC
50 time_t nfsd4_lease
= 90; /* default lease time */
51 time_t nfsd4_grace
= 90;
52 static time_t boot_time
;
54 #define all_ones {{~0,~0},~0}
55 static const stateid_t one_stateid
= {
57 .si_opaque
= all_ones
,
59 static const stateid_t zero_stateid
= {
62 static const stateid_t currentstateid
= {
66 static u64 current_sessionid
= 1;
68 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
69 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
70 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static int check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
);
77 /* Currently used for almost all code touching nfsv4 state: */
78 static DEFINE_MUTEX(client_mutex
);
81 * Currently used for the del_recall_lru and file hash table. In an
82 * effort to decrease the scope of the client_mutex, this spinlock may
83 * eventually cover more:
85 static DEFINE_SPINLOCK(recall_lock
);
87 static struct kmem_cache
*openowner_slab
= NULL
;
88 static struct kmem_cache
*lockowner_slab
= NULL
;
89 static struct kmem_cache
*file_slab
= NULL
;
90 static struct kmem_cache
*stateid_slab
= NULL
;
91 static struct kmem_cache
*deleg_slab
= NULL
;
96 mutex_lock(&client_mutex
);
99 static void free_session(struct kref
*);
101 /* Must be called under the client_lock */
102 static void nfsd4_put_session_locked(struct nfsd4_session
*ses
)
104 kref_put(&ses
->se_ref
, free_session
);
107 static void nfsd4_get_session(struct nfsd4_session
*ses
)
109 kref_get(&ses
->se_ref
);
113 nfs4_unlock_state(void)
115 mutex_unlock(&client_mutex
);
119 opaque_hashval(const void *ptr
, int nbytes
)
121 unsigned char *cptr
= (unsigned char *) ptr
;
131 static struct list_head del_recall_lru
;
133 static void nfsd4_free_file(struct nfs4_file
*f
)
135 kmem_cache_free(file_slab
, f
);
139 put_nfs4_file(struct nfs4_file
*fi
)
141 if (atomic_dec_and_lock(&fi
->fi_ref
, &recall_lock
)) {
142 list_del(&fi
->fi_hash
);
143 spin_unlock(&recall_lock
);
150 get_nfs4_file(struct nfs4_file
*fi
)
152 atomic_inc(&fi
->fi_ref
);
155 static int num_delegations
;
156 unsigned int max_delegations
;
159 * Open owner state (share locks)
162 /* hash tables for lock and open owners */
163 #define OWNER_HASH_BITS 8
164 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
165 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
167 static unsigned int ownerstr_hashval(u32 clientid
, struct xdr_netobj
*ownername
)
171 ret
= opaque_hashval(ownername
->data
, ownername
->len
);
173 return ret
& OWNER_HASH_MASK
;
176 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
178 /* hash table for nfs4_file */
179 #define FILE_HASH_BITS 8
180 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
182 static unsigned int file_hashval(struct inode
*ino
)
184 /* XXX: why are we hashing on inode pointer, anyway? */
185 return hash_ptr(ino
, FILE_HASH_BITS
);
188 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
190 static void __nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
192 BUG_ON(!(fp
->fi_fds
[oflag
] || fp
->fi_fds
[O_RDWR
]));
193 atomic_inc(&fp
->fi_access
[oflag
]);
196 static void nfs4_file_get_access(struct nfs4_file
*fp
, int oflag
)
198 if (oflag
== O_RDWR
) {
199 __nfs4_file_get_access(fp
, O_RDONLY
);
200 __nfs4_file_get_access(fp
, O_WRONLY
);
202 __nfs4_file_get_access(fp
, oflag
);
205 static void nfs4_file_put_fd(struct nfs4_file
*fp
, int oflag
)
207 if (fp
->fi_fds
[oflag
]) {
208 fput(fp
->fi_fds
[oflag
]);
209 fp
->fi_fds
[oflag
] = NULL
;
213 static void __nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
215 if (atomic_dec_and_test(&fp
->fi_access
[oflag
])) {
216 nfs4_file_put_fd(fp
, oflag
);
218 * It's also safe to get rid of the RDWR open *if*
219 * we no longer have need of the other kind of access
220 * or if we already have the other kind of open:
222 if (fp
->fi_fds
[1-oflag
]
223 || atomic_read(&fp
->fi_access
[1 - oflag
]) == 0)
224 nfs4_file_put_fd(fp
, O_RDWR
);
228 static void nfs4_file_put_access(struct nfs4_file
*fp
, int oflag
)
230 if (oflag
== O_RDWR
) {
231 __nfs4_file_put_access(fp
, O_RDONLY
);
232 __nfs4_file_put_access(fp
, O_WRONLY
);
234 __nfs4_file_put_access(fp
, oflag
);
237 static inline int get_new_stid(struct nfs4_stid
*stid
)
239 static int min_stateid
= 0;
240 struct idr
*stateids
= &stid
->sc_client
->cl_stateids
;
244 error
= idr_get_new_above(stateids
, stid
, min_stateid
, &new_stid
);
246 * Note: the necessary preallocation was done in
247 * nfs4_alloc_stateid(). The idr code caps the number of
248 * preallocations that can exist at a time, but the state lock
249 * prevents anyone from using ours before we get here:
253 * It shouldn't be a problem to reuse an opaque stateid value.
254 * I don't think it is for 4.1. But with 4.0 I worry that, for
255 * example, a stray write retransmission could be accepted by
256 * the server when it should have been rejected. Therefore,
257 * adopt a trick from the sctp code to attempt to maximize the
258 * amount of time until an id is reused, by ensuring they always
259 * "increase" (mod INT_MAX):
262 min_stateid
= new_stid
+1;
263 if (min_stateid
== INT_MAX
)
268 static void init_stid(struct nfs4_stid
*stid
, struct nfs4_client
*cl
, unsigned char type
)
270 stateid_t
*s
= &stid
->sc_stateid
;
273 stid
->sc_type
= type
;
274 stid
->sc_client
= cl
;
275 s
->si_opaque
.so_clid
= cl
->cl_clientid
;
276 new_id
= get_new_stid(stid
);
277 s
->si_opaque
.so_id
= (u32
)new_id
;
278 /* Will be incremented before return to client: */
279 s
->si_generation
= 0;
282 static struct nfs4_stid
*nfs4_alloc_stid(struct nfs4_client
*cl
, struct kmem_cache
*slab
)
284 struct idr
*stateids
= &cl
->cl_stateids
;
286 if (!idr_pre_get(stateids
, GFP_KERNEL
))
289 * Note: if we fail here (or any time between now and the time
290 * we actually get the new idr), we won't need to undo the idr
291 * preallocation, since the idr code caps the number of
292 * preallocated entries.
294 return kmem_cache_alloc(slab
, GFP_KERNEL
);
297 static struct nfs4_ol_stateid
* nfs4_alloc_stateid(struct nfs4_client
*clp
)
299 return openlockstateid(nfs4_alloc_stid(clp
, stateid_slab
));
302 static struct nfs4_delegation
*
303 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_ol_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
305 struct nfs4_delegation
*dp
;
306 struct nfs4_file
*fp
= stp
->st_file
;
308 dprintk("NFSD alloc_init_deleg\n");
310 * Major work on the lease subsystem (for example, to support
311 * calbacks on stat) will be required before we can support
312 * write delegations properly.
314 if (type
!= NFS4_OPEN_DELEGATE_READ
)
316 if (fp
->fi_had_conflict
)
318 if (num_delegations
> max_delegations
)
320 dp
= delegstateid(nfs4_alloc_stid(clp
, deleg_slab
));
323 init_stid(&dp
->dl_stid
, clp
, NFS4_DELEG_STID
);
325 * delegation seqid's are never incremented. The 4.1 special
326 * meaning of seqid 0 isn't meaningful, really, but let's avoid
327 * 0 anyway just for consistency and use 1:
329 dp
->dl_stid
.sc_stateid
.si_generation
= 1;
331 INIT_LIST_HEAD(&dp
->dl_perfile
);
332 INIT_LIST_HEAD(&dp
->dl_perclnt
);
333 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
337 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
339 atomic_set(&dp
->dl_count
, 1);
340 INIT_WORK(&dp
->dl_recall
.cb_work
, nfsd4_do_callback_rpc
);
345 nfs4_put_delegation(struct nfs4_delegation
*dp
)
347 if (atomic_dec_and_test(&dp
->dl_count
)) {
348 dprintk("NFSD: freeing dp %p\n",dp
);
349 put_nfs4_file(dp
->dl_file
);
350 kmem_cache_free(deleg_slab
, dp
);
355 static void nfs4_put_deleg_lease(struct nfs4_file
*fp
)
357 if (atomic_dec_and_test(&fp
->fi_delegees
)) {
358 vfs_setlease(fp
->fi_deleg_file
, F_UNLCK
, &fp
->fi_lease
);
360 fput(fp
->fi_deleg_file
);
361 fp
->fi_deleg_file
= NULL
;
365 static void unhash_stid(struct nfs4_stid
*s
)
367 struct idr
*stateids
= &s
->sc_client
->cl_stateids
;
369 idr_remove(stateids
, s
->sc_stateid
.si_opaque
.so_id
);
372 /* Called under the state lock. */
374 unhash_delegation(struct nfs4_delegation
*dp
)
376 unhash_stid(&dp
->dl_stid
);
377 list_del_init(&dp
->dl_perclnt
);
378 spin_lock(&recall_lock
);
379 list_del_init(&dp
->dl_perfile
);
380 list_del_init(&dp
->dl_recall_lru
);
381 spin_unlock(&recall_lock
);
382 nfs4_put_deleg_lease(dp
->dl_file
);
383 nfs4_put_delegation(dp
);
390 /* client_lock protects the client lru list and session hash table */
391 static DEFINE_SPINLOCK(client_lock
);
393 /* Hash tables for nfs4_clientid state */
394 #define CLIENT_HASH_BITS 4
395 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
396 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
398 static unsigned int clientid_hashval(u32 id
)
400 return id
& CLIENT_HASH_MASK
;
403 static unsigned int clientstr_hashval(const char *name
)
405 return opaque_hashval(name
, 8) & CLIENT_HASH_MASK
;
409 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
410 * used in reboot/reset lease grace period processing
412 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
413 * setclientid_confirmed info.
415 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
418 * client_lru holds client queue ordered by nfs4_client.cl_time
421 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
422 * for last close replay.
424 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
425 static int reclaim_str_hashtbl_size
= 0;
426 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
427 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
428 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
429 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
430 static struct list_head client_lru
;
431 static struct list_head close_lru
;
434 * We store the NONE, READ, WRITE, and BOTH bits separately in the
435 * st_{access,deny}_bmap field of the stateid, in order to track not
436 * only what share bits are currently in force, but also what
437 * combinations of share bits previous opens have used. This allows us
438 * to enforce the recommendation of rfc 3530 14.2.19 that the server
439 * return an error if the client attempt to downgrade to a combination
440 * of share bits not explicable by closing some of its previous opens.
442 * XXX: This enforcement is actually incomplete, since we don't keep
443 * track of access/deny bit combinations; so, e.g., we allow:
445 * OPEN allow read, deny write
446 * OPEN allow both, deny none
447 * DOWNGRADE allow read, deny none
449 * which we should reject.
452 bmap_to_share_mode(unsigned long bmap
) {
454 unsigned int access
= 0;
456 for (i
= 1; i
< 4; i
++) {
457 if (test_bit(i
, &bmap
))
464 test_share(struct nfs4_ol_stateid
*stp
, struct nfsd4_open
*open
) {
465 unsigned int access
, deny
;
467 access
= bmap_to_share_mode(stp
->st_access_bmap
);
468 deny
= bmap_to_share_mode(stp
->st_deny_bmap
);
469 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
474 /* set share access for a given stateid */
476 set_access(u32 access
, struct nfs4_ol_stateid
*stp
)
478 __set_bit(access
, &stp
->st_access_bmap
);
481 /* clear share access for a given stateid */
483 clear_access(u32 access
, struct nfs4_ol_stateid
*stp
)
485 __clear_bit(access
, &stp
->st_access_bmap
);
488 /* test whether a given stateid has access */
490 test_access(u32 access
, struct nfs4_ol_stateid
*stp
)
492 return test_bit(access
, &stp
->st_access_bmap
);
495 /* set share deny for a given stateid */
497 set_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
499 __set_bit(access
, &stp
->st_deny_bmap
);
502 /* clear share deny for a given stateid */
504 clear_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
506 __clear_bit(access
, &stp
->st_deny_bmap
);
509 /* test whether a given stateid is denying specific access */
511 test_deny(u32 access
, struct nfs4_ol_stateid
*stp
)
513 return test_bit(access
, &stp
->st_deny_bmap
);
516 static int nfs4_access_to_omode(u32 access
)
518 switch (access
& NFS4_SHARE_ACCESS_BOTH
) {
519 case NFS4_SHARE_ACCESS_READ
:
521 case NFS4_SHARE_ACCESS_WRITE
:
523 case NFS4_SHARE_ACCESS_BOTH
:
529 /* release all access and file references for a given stateid */
531 release_all_access(struct nfs4_ol_stateid
*stp
)
535 for (i
= 1; i
< 4; i
++) {
536 if (test_access(i
, stp
))
537 nfs4_file_put_access(stp
->st_file
,
538 nfs4_access_to_omode(i
));
539 clear_access(i
, stp
);
543 static void unhash_generic_stateid(struct nfs4_ol_stateid
*stp
)
545 list_del(&stp
->st_perfile
);
546 list_del(&stp
->st_perstateowner
);
549 static void close_generic_stateid(struct nfs4_ol_stateid
*stp
)
551 release_all_access(stp
);
552 put_nfs4_file(stp
->st_file
);
556 static void free_generic_stateid(struct nfs4_ol_stateid
*stp
)
558 kmem_cache_free(stateid_slab
, stp
);
561 static void release_lock_stateid(struct nfs4_ol_stateid
*stp
)
565 unhash_generic_stateid(stp
);
566 unhash_stid(&stp
->st_stid
);
567 file
= find_any_file(stp
->st_file
);
569 locks_remove_posix(file
, (fl_owner_t
)lockowner(stp
->st_stateowner
));
570 close_generic_stateid(stp
);
571 free_generic_stateid(stp
);
574 static void unhash_lockowner(struct nfs4_lockowner
*lo
)
576 struct nfs4_ol_stateid
*stp
;
578 list_del(&lo
->lo_owner
.so_strhash
);
579 list_del(&lo
->lo_perstateid
);
580 list_del(&lo
->lo_owner_ino_hash
);
581 while (!list_empty(&lo
->lo_owner
.so_stateids
)) {
582 stp
= list_first_entry(&lo
->lo_owner
.so_stateids
,
583 struct nfs4_ol_stateid
, st_perstateowner
);
584 release_lock_stateid(stp
);
588 static void release_lockowner(struct nfs4_lockowner
*lo
)
590 unhash_lockowner(lo
);
591 nfs4_free_lockowner(lo
);
595 release_stateid_lockowners(struct nfs4_ol_stateid
*open_stp
)
597 struct nfs4_lockowner
*lo
;
599 while (!list_empty(&open_stp
->st_lockowners
)) {
600 lo
= list_entry(open_stp
->st_lockowners
.next
,
601 struct nfs4_lockowner
, lo_perstateid
);
602 release_lockowner(lo
);
606 static void unhash_open_stateid(struct nfs4_ol_stateid
*stp
)
608 unhash_generic_stateid(stp
);
609 release_stateid_lockowners(stp
);
610 close_generic_stateid(stp
);
613 static void release_open_stateid(struct nfs4_ol_stateid
*stp
)
615 unhash_open_stateid(stp
);
616 unhash_stid(&stp
->st_stid
);
617 free_generic_stateid(stp
);
620 static void unhash_openowner(struct nfs4_openowner
*oo
)
622 struct nfs4_ol_stateid
*stp
;
624 list_del(&oo
->oo_owner
.so_strhash
);
625 list_del(&oo
->oo_perclient
);
626 while (!list_empty(&oo
->oo_owner
.so_stateids
)) {
627 stp
= list_first_entry(&oo
->oo_owner
.so_stateids
,
628 struct nfs4_ol_stateid
, st_perstateowner
);
629 release_open_stateid(stp
);
633 static void release_last_closed_stateid(struct nfs4_openowner
*oo
)
635 struct nfs4_ol_stateid
*s
= oo
->oo_last_closed_stid
;
638 unhash_stid(&s
->st_stid
);
639 free_generic_stateid(s
);
640 oo
->oo_last_closed_stid
= NULL
;
644 static void release_openowner(struct nfs4_openowner
*oo
)
646 unhash_openowner(oo
);
647 list_del(&oo
->oo_close_lru
);
648 release_last_closed_stateid(oo
);
649 nfs4_free_openowner(oo
);
652 #define SESSION_HASH_SIZE 512
653 static struct list_head sessionid_hashtbl
[SESSION_HASH_SIZE
];
656 hash_sessionid(struct nfs4_sessionid
*sessionid
)
658 struct nfsd4_sessionid
*sid
= (struct nfsd4_sessionid
*)sessionid
;
660 return sid
->sequence
% SESSION_HASH_SIZE
;
665 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
667 u32
*ptr
= (u32
*)(&sessionid
->data
[0]);
668 dprintk("%s: %u:%u:%u:%u\n", fn
, ptr
[0], ptr
[1], ptr
[2], ptr
[3]);
672 dump_sessionid(const char *fn
, struct nfs4_sessionid
*sessionid
)
679 gen_sessionid(struct nfsd4_session
*ses
)
681 struct nfs4_client
*clp
= ses
->se_client
;
682 struct nfsd4_sessionid
*sid
;
684 sid
= (struct nfsd4_sessionid
*)ses
->se_sessionid
.data
;
685 sid
->clientid
= clp
->cl_clientid
;
686 sid
->sequence
= current_sessionid
++;
691 * The protocol defines ca_maxresponssize_cached to include the size of
692 * the rpc header, but all we need to cache is the data starting after
693 * the end of the initial SEQUENCE operation--the rest we regenerate
694 * each time. Therefore we can advertise a ca_maxresponssize_cached
695 * value that is the number of bytes in our cache plus a few additional
696 * bytes. In order to stay on the safe side, and not promise more than
697 * we can cache, those additional bytes must be the minimum possible: 24
698 * bytes of rpc header (xid through accept state, with AUTH_NULL
699 * verifier), 12 for the compound header (with zero-length tag), and 44
700 * for the SEQUENCE op response:
702 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
705 free_session_slots(struct nfsd4_session
*ses
)
709 for (i
= 0; i
< ses
->se_fchannel
.maxreqs
; i
++)
710 kfree(ses
->se_slots
[i
]);
714 * We don't actually need to cache the rpc and session headers, so we
715 * can allocate a little less for each slot:
717 static inline int slot_bytes(struct nfsd4_channel_attrs
*ca
)
719 return ca
->maxresp_cached
- NFSD_MIN_HDR_SEQ_SZ
;
722 static int nfsd4_sanitize_slot_size(u32 size
)
724 size
-= NFSD_MIN_HDR_SEQ_SZ
; /* We don't cache the rpc header */
725 size
= min_t(u32
, size
, NFSD_SLOT_CACHE_SIZE
);
731 * XXX: If we run out of reserved DRC memory we could (up to a point)
732 * re-negotiate active sessions and reduce their slot usage to make
733 * room for new connections. For now we just fail the create session.
735 static int nfsd4_get_drc_mem(int slotsize
, u32 num
)
739 num
= min_t(u32
, num
, NFSD_MAX_SLOTS_PER_SESSION
);
741 spin_lock(&nfsd_drc_lock
);
742 avail
= min_t(int, NFSD_MAX_MEM_PER_SESSION
,
743 nfsd_drc_max_mem
- nfsd_drc_mem_used
);
744 num
= min_t(int, num
, avail
/ slotsize
);
745 nfsd_drc_mem_used
+= num
* slotsize
;
746 spin_unlock(&nfsd_drc_lock
);
751 static void nfsd4_put_drc_mem(int slotsize
, int num
)
753 spin_lock(&nfsd_drc_lock
);
754 nfsd_drc_mem_used
-= slotsize
* num
;
755 spin_unlock(&nfsd_drc_lock
);
758 static struct nfsd4_session
*alloc_session(int slotsize
, int numslots
)
760 struct nfsd4_session
*new;
763 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION
* sizeof(struct nfsd4_slot
*)
764 + sizeof(struct nfsd4_session
) > PAGE_SIZE
);
765 mem
= numslots
* sizeof(struct nfsd4_slot
*);
767 new = kzalloc(sizeof(*new) + mem
, GFP_KERNEL
);
770 /* allocate each struct nfsd4_slot and data cache in one piece */
771 for (i
= 0; i
< numslots
; i
++) {
772 mem
= sizeof(struct nfsd4_slot
) + slotsize
;
773 new->se_slots
[i
] = kzalloc(mem
, GFP_KERNEL
);
774 if (!new->se_slots
[i
])
780 kfree(new->se_slots
[i
]);
785 static void init_forechannel_attrs(struct nfsd4_channel_attrs
*new, struct nfsd4_channel_attrs
*req
, int numslots
, int slotsize
)
787 u32 maxrpc
= nfsd_serv
->sv_max_mesg
;
789 new->maxreqs
= numslots
;
790 new->maxresp_cached
= min_t(u32
, req
->maxresp_cached
,
791 slotsize
+ NFSD_MIN_HDR_SEQ_SZ
);
792 new->maxreq_sz
= min_t(u32
, req
->maxreq_sz
, maxrpc
);
793 new->maxresp_sz
= min_t(u32
, req
->maxresp_sz
, maxrpc
);
794 new->maxops
= min_t(u32
, req
->maxops
, NFSD_MAX_OPS_PER_COMPOUND
);
797 static void free_conn(struct nfsd4_conn
*c
)
799 svc_xprt_put(c
->cn_xprt
);
803 static void nfsd4_conn_lost(struct svc_xpt_user
*u
)
805 struct nfsd4_conn
*c
= container_of(u
, struct nfsd4_conn
, cn_xpt_user
);
806 struct nfs4_client
*clp
= c
->cn_session
->se_client
;
808 spin_lock(&clp
->cl_lock
);
809 if (!list_empty(&c
->cn_persession
)) {
810 list_del(&c
->cn_persession
);
813 spin_unlock(&clp
->cl_lock
);
814 nfsd4_probe_callback(clp
);
817 static struct nfsd4_conn
*alloc_conn(struct svc_rqst
*rqstp
, u32 flags
)
819 struct nfsd4_conn
*conn
;
821 conn
= kmalloc(sizeof(struct nfsd4_conn
), GFP_KERNEL
);
824 svc_xprt_get(rqstp
->rq_xprt
);
825 conn
->cn_xprt
= rqstp
->rq_xprt
;
826 conn
->cn_flags
= flags
;
827 INIT_LIST_HEAD(&conn
->cn_xpt_user
.list
);
831 static void __nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
833 conn
->cn_session
= ses
;
834 list_add(&conn
->cn_persession
, &ses
->se_conns
);
837 static void nfsd4_hash_conn(struct nfsd4_conn
*conn
, struct nfsd4_session
*ses
)
839 struct nfs4_client
*clp
= ses
->se_client
;
841 spin_lock(&clp
->cl_lock
);
842 __nfsd4_hash_conn(conn
, ses
);
843 spin_unlock(&clp
->cl_lock
);
846 static int nfsd4_register_conn(struct nfsd4_conn
*conn
)
848 conn
->cn_xpt_user
.callback
= nfsd4_conn_lost
;
849 return register_xpt_user(conn
->cn_xprt
, &conn
->cn_xpt_user
);
852 static __be32
nfsd4_new_conn(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
, u32 dir
)
854 struct nfsd4_conn
*conn
;
857 conn
= alloc_conn(rqstp
, dir
);
859 return nfserr_jukebox
;
860 nfsd4_hash_conn(conn
, ses
);
861 ret
= nfsd4_register_conn(conn
);
863 /* oops; xprt is already down: */
864 nfsd4_conn_lost(&conn
->cn_xpt_user
);
868 static __be32
nfsd4_new_conn_from_crses(struct svc_rqst
*rqstp
, struct nfsd4_session
*ses
)
870 u32 dir
= NFS4_CDFC4_FORE
;
872 if (ses
->se_flags
& SESSION4_BACK_CHAN
)
873 dir
|= NFS4_CDFC4_BACK
;
875 return nfsd4_new_conn(rqstp
, ses
, dir
);
878 /* must be called under client_lock */
879 static void nfsd4_del_conns(struct nfsd4_session
*s
)
881 struct nfs4_client
*clp
= s
->se_client
;
882 struct nfsd4_conn
*c
;
884 spin_lock(&clp
->cl_lock
);
885 while (!list_empty(&s
->se_conns
)) {
886 c
= list_first_entry(&s
->se_conns
, struct nfsd4_conn
, cn_persession
);
887 list_del_init(&c
->cn_persession
);
888 spin_unlock(&clp
->cl_lock
);
890 unregister_xpt_user(c
->cn_xprt
, &c
->cn_xpt_user
);
893 spin_lock(&clp
->cl_lock
);
895 spin_unlock(&clp
->cl_lock
);
898 static void free_session(struct kref
*kref
)
900 struct nfsd4_session
*ses
;
903 BUG_ON(!spin_is_locked(&client_lock
));
904 ses
= container_of(kref
, struct nfsd4_session
, se_ref
);
905 nfsd4_del_conns(ses
);
906 spin_lock(&nfsd_drc_lock
);
907 mem
= ses
->se_fchannel
.maxreqs
* slot_bytes(&ses
->se_fchannel
);
908 nfsd_drc_mem_used
-= mem
;
909 spin_unlock(&nfsd_drc_lock
);
910 free_session_slots(ses
);
914 void nfsd4_put_session(struct nfsd4_session
*ses
)
916 spin_lock(&client_lock
);
917 nfsd4_put_session_locked(ses
);
918 spin_unlock(&client_lock
);
921 static struct nfsd4_session
*alloc_init_session(struct svc_rqst
*rqstp
, struct nfs4_client
*clp
, struct nfsd4_create_session
*cses
)
923 struct nfsd4_session
*new;
924 struct nfsd4_channel_attrs
*fchan
= &cses
->fore_channel
;
925 int numslots
, slotsize
;
930 * Note decreasing slot size below client's request may
931 * make it difficult for client to function correctly, whereas
932 * decreasing the number of slots will (just?) affect
933 * performance. When short on memory we therefore prefer to
934 * decrease number of slots instead of their size.
936 slotsize
= nfsd4_sanitize_slot_size(fchan
->maxresp_cached
);
937 numslots
= nfsd4_get_drc_mem(slotsize
, fchan
->maxreqs
);
941 new = alloc_session(slotsize
, numslots
);
943 nfsd4_put_drc_mem(slotsize
, fchan
->maxreqs
);
946 init_forechannel_attrs(&new->se_fchannel
, fchan
, numslots
, slotsize
);
948 new->se_client
= clp
;
951 INIT_LIST_HEAD(&new->se_conns
);
953 new->se_cb_seq_nr
= 1;
954 new->se_flags
= cses
->flags
;
955 new->se_cb_prog
= cses
->callback_prog
;
956 kref_init(&new->se_ref
);
957 idx
= hash_sessionid(&new->se_sessionid
);
958 spin_lock(&client_lock
);
959 list_add(&new->se_hash
, &sessionid_hashtbl
[idx
]);
960 spin_lock(&clp
->cl_lock
);
961 list_add(&new->se_perclnt
, &clp
->cl_sessions
);
962 spin_unlock(&clp
->cl_lock
);
963 spin_unlock(&client_lock
);
965 status
= nfsd4_new_conn_from_crses(rqstp
, new);
966 /* whoops: benny points out, status is ignored! (err, or bogus) */
968 spin_lock(&client_lock
);
969 free_session(&new->se_ref
);
970 spin_unlock(&client_lock
);
973 if (cses
->flags
& SESSION4_BACK_CHAN
) {
974 struct sockaddr
*sa
= svc_addr(rqstp
);
976 * This is a little silly; with sessions there's no real
977 * use for the callback address. Use the peer address
978 * as a reasonable default for now, but consider fixing
979 * the rpc client not to require an address in the
982 rpc_copy_addr((struct sockaddr
*)&clp
->cl_cb_conn
.cb_addr
, sa
);
983 clp
->cl_cb_conn
.cb_addrlen
= svc_addr_len(sa
);
985 nfsd4_probe_callback(clp
);
989 /* caller must hold client_lock */
990 static struct nfsd4_session
*
991 find_in_sessionid_hashtbl(struct nfs4_sessionid
*sessionid
)
993 struct nfsd4_session
*elem
;
996 dump_sessionid(__func__
, sessionid
);
997 idx
= hash_sessionid(sessionid
);
998 /* Search in the appropriate list */
999 list_for_each_entry(elem
, &sessionid_hashtbl
[idx
], se_hash
) {
1000 if (!memcmp(elem
->se_sessionid
.data
, sessionid
->data
,
1001 NFS4_MAX_SESSIONID_LEN
)) {
1006 dprintk("%s: session not found\n", __func__
);
1010 /* caller must hold client_lock */
1012 unhash_session(struct nfsd4_session
*ses
)
1014 list_del(&ses
->se_hash
);
1015 spin_lock(&ses
->se_client
->cl_lock
);
1016 list_del(&ses
->se_perclnt
);
1017 spin_unlock(&ses
->se_client
->cl_lock
);
1020 /* must be called under the client_lock */
1022 renew_client_locked(struct nfs4_client
*clp
)
1024 if (is_client_expired(clp
)) {
1025 dprintk("%s: client (clientid %08x/%08x) already expired\n",
1027 clp
->cl_clientid
.cl_boot
,
1028 clp
->cl_clientid
.cl_id
);
1032 dprintk("renewing client (clientid %08x/%08x)\n",
1033 clp
->cl_clientid
.cl_boot
,
1034 clp
->cl_clientid
.cl_id
);
1035 list_move_tail(&clp
->cl_lru
, &client_lru
);
1036 clp
->cl_time
= get_seconds();
1040 renew_client(struct nfs4_client
*clp
)
1042 spin_lock(&client_lock
);
1043 renew_client_locked(clp
);
1044 spin_unlock(&client_lock
);
1047 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1049 STALE_CLIENTID(clientid_t
*clid
)
1051 if (clid
->cl_boot
== boot_time
)
1053 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1054 clid
->cl_boot
, clid
->cl_id
, boot_time
);
1059 * XXX Should we use a slab cache ?
1060 * This type of memory management is somewhat inefficient, but we use it
1061 * anyway since SETCLIENTID is not a common operation.
1063 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
1065 struct nfs4_client
*clp
;
1067 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
1070 clp
->cl_name
.data
= kmemdup(name
.data
, name
.len
, GFP_KERNEL
);
1071 if (clp
->cl_name
.data
== NULL
) {
1075 clp
->cl_name
.len
= name
.len
;
1080 free_client(struct nfs4_client
*clp
)
1082 BUG_ON(!spin_is_locked(&client_lock
));
1083 while (!list_empty(&clp
->cl_sessions
)) {
1084 struct nfsd4_session
*ses
;
1085 ses
= list_entry(clp
->cl_sessions
.next
, struct nfsd4_session
,
1087 list_del(&ses
->se_perclnt
);
1088 nfsd4_put_session_locked(ses
);
1090 if (clp
->cl_cred
.cr_group_info
)
1091 put_group_info(clp
->cl_cred
.cr_group_info
);
1092 kfree(clp
->cl_principal
);
1093 kfree(clp
->cl_name
.data
);
1098 release_session_client(struct nfsd4_session
*session
)
1100 struct nfs4_client
*clp
= session
->se_client
;
1102 if (!atomic_dec_and_lock(&clp
->cl_refcount
, &client_lock
))
1104 if (is_client_expired(clp
)) {
1106 session
->se_client
= NULL
;
1108 renew_client_locked(clp
);
1109 spin_unlock(&client_lock
);
1112 /* must be called under the client_lock */
1114 unhash_client_locked(struct nfs4_client
*clp
)
1116 struct nfsd4_session
*ses
;
1118 mark_client_expired(clp
);
1119 list_del(&clp
->cl_lru
);
1120 spin_lock(&clp
->cl_lock
);
1121 list_for_each_entry(ses
, &clp
->cl_sessions
, se_perclnt
)
1122 list_del_init(&ses
->se_hash
);
1123 spin_unlock(&clp
->cl_lock
);
1127 expire_client(struct nfs4_client
*clp
)
1129 struct nfs4_openowner
*oo
;
1130 struct nfs4_delegation
*dp
;
1131 struct list_head reaplist
;
1133 INIT_LIST_HEAD(&reaplist
);
1134 spin_lock(&recall_lock
);
1135 while (!list_empty(&clp
->cl_delegations
)) {
1136 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
1137 list_del_init(&dp
->dl_perclnt
);
1138 list_move(&dp
->dl_recall_lru
, &reaplist
);
1140 spin_unlock(&recall_lock
);
1141 while (!list_empty(&reaplist
)) {
1142 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
1143 unhash_delegation(dp
);
1145 while (!list_empty(&clp
->cl_openowners
)) {
1146 oo
= list_entry(clp
->cl_openowners
.next
, struct nfs4_openowner
, oo_perclient
);
1147 release_openowner(oo
);
1149 nfsd4_shutdown_callback(clp
);
1150 if (clp
->cl_cb_conn
.cb_xprt
)
1151 svc_xprt_put(clp
->cl_cb_conn
.cb_xprt
);
1152 list_del(&clp
->cl_idhash
);
1153 list_del(&clp
->cl_strhash
);
1154 spin_lock(&client_lock
);
1155 unhash_client_locked(clp
);
1156 if (atomic_read(&clp
->cl_refcount
) == 0)
1158 spin_unlock(&client_lock
);
1161 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
1163 memcpy(target
->cl_verifier
.data
, source
->data
,
1164 sizeof(target
->cl_verifier
.data
));
1167 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
1169 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
1170 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
1173 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
1175 target
->cr_uid
= source
->cr_uid
;
1176 target
->cr_gid
= source
->cr_gid
;
1177 target
->cr_group_info
= source
->cr_group_info
;
1178 get_group_info(target
->cr_group_info
);
1181 static int same_name(const char *n1
, const char *n2
)
1183 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
1187 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
1189 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
1193 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
1195 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
1198 /* XXX what about NGROUP */
1200 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
1202 return cr1
->cr_uid
== cr2
->cr_uid
;
1205 static void gen_clid(struct nfs4_client
*clp
)
1207 static u32 current_clientid
= 1;
1209 clp
->cl_clientid
.cl_boot
= boot_time
;
1210 clp
->cl_clientid
.cl_id
= current_clientid
++;
1213 static void gen_confirm(struct nfs4_client
*clp
)
1218 verf
[0] = (__be32
)get_seconds();
1219 verf
[1] = (__be32
)i
++;
1220 memcpy(clp
->cl_confirm
.data
, verf
, sizeof(clp
->cl_confirm
.data
));
1223 static struct nfs4_stid
*find_stateid(struct nfs4_client
*cl
, stateid_t
*t
)
1225 return idr_find(&cl
->cl_stateids
, t
->si_opaque
.so_id
);
1228 static struct nfs4_stid
*find_stateid_by_type(struct nfs4_client
*cl
, stateid_t
*t
, char typemask
)
1230 struct nfs4_stid
*s
;
1232 s
= find_stateid(cl
, t
);
1235 if (typemask
& s
->sc_type
)
1240 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
,
1241 struct svc_rqst
*rqstp
, nfs4_verifier
*verf
)
1243 struct nfs4_client
*clp
;
1244 struct sockaddr
*sa
= svc_addr(rqstp
);
1247 clp
= alloc_client(name
);
1251 INIT_LIST_HEAD(&clp
->cl_sessions
);
1253 princ
= svc_gss_principal(rqstp
);
1255 clp
->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
1256 if (clp
->cl_principal
== NULL
) {
1257 spin_lock(&client_lock
);
1259 spin_unlock(&client_lock
);
1264 idr_init(&clp
->cl_stateids
);
1265 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
1266 atomic_set(&clp
->cl_refcount
, 0);
1267 clp
->cl_cb_state
= NFSD4_CB_UNKNOWN
;
1268 INIT_LIST_HEAD(&clp
->cl_idhash
);
1269 INIT_LIST_HEAD(&clp
->cl_strhash
);
1270 INIT_LIST_HEAD(&clp
->cl_openowners
);
1271 INIT_LIST_HEAD(&clp
->cl_delegations
);
1272 INIT_LIST_HEAD(&clp
->cl_lru
);
1273 INIT_LIST_HEAD(&clp
->cl_callbacks
);
1274 spin_lock_init(&clp
->cl_lock
);
1275 INIT_WORK(&clp
->cl_cb_null
.cb_work
, nfsd4_do_callback_rpc
);
1276 clp
->cl_time
= get_seconds();
1277 clear_bit(0, &clp
->cl_cb_slot_busy
);
1278 rpc_init_wait_queue(&clp
->cl_cb_waitq
, "Backchannel slot table");
1279 copy_verf(clp
, verf
);
1280 rpc_copy_addr((struct sockaddr
*) &clp
->cl_addr
, sa
);
1281 clp
->cl_flavor
= rqstp
->rq_flavor
;
1282 copy_cred(&clp
->cl_cred
, &rqstp
->rq_cred
);
1284 clp
->cl_cb_session
= NULL
;
1289 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
1291 unsigned int idhashval
;
1293 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
1294 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1295 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
1300 move_to_confirmed(struct nfs4_client
*clp
)
1302 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
1303 unsigned int strhashval
;
1305 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
1306 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
1307 strhashval
= clientstr_hashval(clp
->cl_recdir
);
1308 list_move(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
1312 static struct nfs4_client
*
1313 find_confirmed_client(clientid_t
*clid
)
1315 struct nfs4_client
*clp
;
1316 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1318 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
1319 if (same_clid(&clp
->cl_clientid
, clid
)) {
1327 static struct nfs4_client
*
1328 find_unconfirmed_client(clientid_t
*clid
)
1330 struct nfs4_client
*clp
;
1331 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
1333 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
1334 if (same_clid(&clp
->cl_clientid
, clid
))
1340 static bool clp_used_exchangeid(struct nfs4_client
*clp
)
1342 return clp
->cl_exchange_flags
!= 0;
1345 static struct nfs4_client
*
1346 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
1348 struct nfs4_client
*clp
;
1350 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
1351 if (same_name(clp
->cl_recdir
, dname
))
1357 static struct nfs4_client
*
1358 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
1360 struct nfs4_client
*clp
;
1362 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
1363 if (same_name(clp
->cl_recdir
, dname
))
1370 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
, struct svc_rqst
*rqstp
)
1372 struct nfs4_cb_conn
*conn
= &clp
->cl_cb_conn
;
1373 struct sockaddr
*sa
= svc_addr(rqstp
);
1374 u32 scopeid
= rpc_get_scope_id(sa
);
1375 unsigned short expected_family
;
1377 /* Currently, we only support tcp and tcp6 for the callback channel */
1378 if (se
->se_callback_netid_len
== 3 &&
1379 !memcmp(se
->se_callback_netid_val
, "tcp", 3))
1380 expected_family
= AF_INET
;
1381 else if (se
->se_callback_netid_len
== 4 &&
1382 !memcmp(se
->se_callback_netid_val
, "tcp6", 4))
1383 expected_family
= AF_INET6
;
1387 conn
->cb_addrlen
= rpc_uaddr2sockaddr(&init_net
, se
->se_callback_addr_val
,
1388 se
->se_callback_addr_len
,
1389 (struct sockaddr
*)&conn
->cb_addr
,
1390 sizeof(conn
->cb_addr
));
1392 if (!conn
->cb_addrlen
|| conn
->cb_addr
.ss_family
!= expected_family
)
1395 if (conn
->cb_addr
.ss_family
== AF_INET6
)
1396 ((struct sockaddr_in6
*)&conn
->cb_addr
)->sin6_scope_id
= scopeid
;
1398 conn
->cb_prog
= se
->se_callback_prog
;
1399 conn
->cb_ident
= se
->se_callback_ident
;
1400 memcpy(&conn
->cb_saddr
, &rqstp
->rq_daddr
, rqstp
->rq_daddrlen
);
1403 conn
->cb_addr
.ss_family
= AF_UNSPEC
;
1404 conn
->cb_addrlen
= 0;
1405 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
1406 "will not receive delegations\n",
1407 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1413 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1416 nfsd4_store_cache_entry(struct nfsd4_compoundres
*resp
)
1418 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1421 dprintk("--> %s slot %p\n", __func__
, slot
);
1423 slot
->sl_opcnt
= resp
->opcnt
;
1424 slot
->sl_status
= resp
->cstate
.status
;
1426 slot
->sl_flags
|= NFSD4_SLOT_INITIALIZED
;
1427 if (nfsd4_not_cached(resp
)) {
1428 slot
->sl_datalen
= 0;
1431 slot
->sl_datalen
= (char *)resp
->p
- (char *)resp
->cstate
.datap
;
1432 base
= (char *)resp
->cstate
.datap
-
1433 (char *)resp
->xbuf
->head
[0].iov_base
;
1434 if (read_bytes_from_xdr_buf(resp
->xbuf
, base
, slot
->sl_data
,
1436 WARN("%s: sessions DRC could not cache compound\n", __func__
);
1441 * Encode the replay sequence operation from the slot values.
1442 * If cachethis is FALSE encode the uncached rep error on the next
1443 * operation which sets resp->p and increments resp->opcnt for
1444 * nfs4svc_encode_compoundres.
1448 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs
*args
,
1449 struct nfsd4_compoundres
*resp
)
1451 struct nfsd4_op
*op
;
1452 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1454 /* Encode the replayed sequence operation */
1455 op
= &args
->ops
[resp
->opcnt
- 1];
1456 nfsd4_encode_operation(resp
, op
);
1458 /* Return nfserr_retry_uncached_rep in next operation. */
1459 if (args
->opcnt
> 1 && !(slot
->sl_flags
& NFSD4_SLOT_CACHETHIS
)) {
1460 op
= &args
->ops
[resp
->opcnt
++];
1461 op
->status
= nfserr_retry_uncached_rep
;
1462 nfsd4_encode_operation(resp
, op
);
1468 * The sequence operation is not cached because we can use the slot and
1472 nfsd4_replay_cache_entry(struct nfsd4_compoundres
*resp
,
1473 struct nfsd4_sequence
*seq
)
1475 struct nfsd4_slot
*slot
= resp
->cstate
.slot
;
1478 dprintk("--> %s slot %p\n", __func__
, slot
);
1480 /* Either returns 0 or nfserr_retry_uncached */
1481 status
= nfsd4_enc_sequence_replay(resp
->rqstp
->rq_argp
, resp
);
1482 if (status
== nfserr_retry_uncached_rep
)
1485 /* The sequence operation has been encoded, cstate->datap set. */
1486 memcpy(resp
->cstate
.datap
, slot
->sl_data
, slot
->sl_datalen
);
1488 resp
->opcnt
= slot
->sl_opcnt
;
1489 resp
->p
= resp
->cstate
.datap
+ XDR_QUADLEN(slot
->sl_datalen
);
1490 status
= slot
->sl_status
;
1496 * Set the exchange_id flags returned by the server.
1499 nfsd4_set_ex_flags(struct nfs4_client
*new, struct nfsd4_exchange_id
*clid
)
1501 /* pNFS is not supported */
1502 new->cl_exchange_flags
|= EXCHGID4_FLAG_USE_NON_PNFS
;
1504 /* Referrals are supported, Migration is not. */
1505 new->cl_exchange_flags
|= EXCHGID4_FLAG_SUPP_MOVED_REFER
;
1507 /* set the wire flags to return to client. */
1508 clid
->flags
= new->cl_exchange_flags
;
1512 nfsd4_exchange_id(struct svc_rqst
*rqstp
,
1513 struct nfsd4_compound_state
*cstate
,
1514 struct nfsd4_exchange_id
*exid
)
1516 struct nfs4_client
*unconf
, *conf
, *new;
1518 unsigned int strhashval
;
1519 char dname
[HEXDIR_LEN
];
1520 char addr_str
[INET6_ADDRSTRLEN
];
1521 nfs4_verifier verf
= exid
->verifier
;
1522 struct sockaddr
*sa
= svc_addr(rqstp
);
1524 rpc_ntop(sa
, addr_str
, sizeof(addr_str
));
1525 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1526 "ip_addr=%s flags %x, spa_how %d\n",
1527 __func__
, rqstp
, exid
, exid
->clname
.len
, exid
->clname
.data
,
1528 addr_str
, exid
->flags
, exid
->spa_how
);
1530 if (exid
->flags
& ~EXCHGID4_FLAG_MASK_A
)
1531 return nfserr_inval
;
1533 /* Currently only support SP4_NONE */
1534 switch (exid
->spa_how
) {
1538 return nfserr_serverfault
;
1540 BUG(); /* checked by xdr code */
1542 return nfserr_serverfault
; /* no excuse :-/ */
1545 status
= nfs4_make_rec_clidname(dname
, &exid
->clname
);
1550 strhashval
= clientstr_hashval(dname
);
1553 conf
= find_confirmed_client_by_str(dname
, strhashval
);
1555 if (!clp_used_exchangeid(conf
)) {
1556 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1557 status
= nfserr_inval
; /* buggy client */
1561 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
1562 /* 18.35.4 case 9 */
1563 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
)
1564 status
= nfserr_perm
;
1566 status
= nfserr_clid_inuse
;
1569 if (!same_verf(&verf
, &conf
->cl_verifier
)) {
1570 /* 18.35.4 case 8 */
1571 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1572 status
= nfserr_not_same
;
1575 /* Client reboot: destroy old state */
1576 expire_client(conf
);
1579 if (!clp_used_exchangeid(conf
)) {
1580 status
= nfserr_inval
;
1584 * Set bit when the owner id and verifier map to an already
1585 * confirmed client id (18.35.3).
1587 exid
->flags
|= EXCHGID4_FLAG_CONFIRMED_R
;
1590 * Falling into 18.35.4 case 2, possible router replay.
1591 * Leave confirmed record intact and return same result.
1597 /* 18.35.4 case 7 */
1598 if (exid
->flags
& EXCHGID4_FLAG_UPD_CONFIRMED_REC_A
) {
1599 status
= nfserr_noent
;
1603 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
1606 * Possible retry or client restart. Per 18.35.4 case 4,
1607 * a new unconfirmed record should be generated regardless
1608 * of whether any properties have changed.
1610 expire_client(unconf
);
1615 new = create_client(exid
->clname
, dname
, rqstp
, &verf
);
1617 status
= nfserr_jukebox
;
1622 add_to_unconfirmed(new, strhashval
);
1624 exid
->clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
1625 exid
->clientid
.cl_id
= new->cl_clientid
.cl_id
;
1628 nfsd4_set_ex_flags(new, exid
);
1630 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1631 new->cl_cs_slot
.sl_seqid
, new->cl_exchange_flags
);
1635 nfs4_unlock_state();
1640 check_slot_seqid(u32 seqid
, u32 slot_seqid
, int slot_inuse
)
1642 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__
, seqid
,
1645 /* The slot is in use, and no response has been sent. */
1647 if (seqid
== slot_seqid
)
1648 return nfserr_jukebox
;
1650 return nfserr_seq_misordered
;
1652 /* Note unsigned 32-bit arithmetic handles wraparound: */
1653 if (likely(seqid
== slot_seqid
+ 1))
1655 if (seqid
== slot_seqid
)
1656 return nfserr_replay_cache
;
1657 return nfserr_seq_misordered
;
1661 * Cache the create session result into the create session single DRC
1662 * slot cache by saving the xdr structure. sl_seqid has been set.
1663 * Do this for solo or embedded create session operations.
1666 nfsd4_cache_create_session(struct nfsd4_create_session
*cr_ses
,
1667 struct nfsd4_clid_slot
*slot
, __be32 nfserr
)
1669 slot
->sl_status
= nfserr
;
1670 memcpy(&slot
->sl_cr_ses
, cr_ses
, sizeof(*cr_ses
));
1674 nfsd4_replay_create_session(struct nfsd4_create_session
*cr_ses
,
1675 struct nfsd4_clid_slot
*slot
)
1677 memcpy(cr_ses
, &slot
->sl_cr_ses
, sizeof(*cr_ses
));
1678 return slot
->sl_status
;
1681 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1682 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1683 1 + /* MIN tag is length with zero, only length */ \
1684 3 + /* version, opcount, opcode */ \
1685 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1686 /* seqid, slotID, slotID, cache */ \
1687 4 ) * sizeof(__be32))
1689 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1690 2 + /* verifier: AUTH_NULL, length 0 */\
1692 1 + /* MIN tag is length with zero, only length */ \
1693 3 + /* opcount, opcode, opstatus*/ \
1694 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1695 /* seqid, slotID, slotID, slotID, status */ \
1696 5 ) * sizeof(__be32))
1698 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel
)
1700 return fchannel
.maxreq_sz
< NFSD_MIN_REQ_HDR_SEQ_SZ
1701 || fchannel
.maxresp_sz
< NFSD_MIN_RESP_HDR_SEQ_SZ
;
1705 nfsd4_create_session(struct svc_rqst
*rqstp
,
1706 struct nfsd4_compound_state
*cstate
,
1707 struct nfsd4_create_session
*cr_ses
)
1709 struct sockaddr
*sa
= svc_addr(rqstp
);
1710 struct nfs4_client
*conf
, *unconf
;
1711 struct nfsd4_session
*new;
1712 struct nfsd4_clid_slot
*cs_slot
= NULL
;
1713 bool confirm_me
= false;
1716 if (cr_ses
->flags
& ~SESSION4_FLAG_MASK_A
)
1717 return nfserr_inval
;
1720 unconf
= find_unconfirmed_client(&cr_ses
->clientid
);
1721 conf
= find_confirmed_client(&cr_ses
->clientid
);
1724 cs_slot
= &conf
->cl_cs_slot
;
1725 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1726 if (status
== nfserr_replay_cache
) {
1727 dprintk("Got a create_session replay! seqid= %d\n",
1729 /* Return the cached reply status */
1730 status
= nfsd4_replay_create_session(cr_ses
, cs_slot
);
1732 } else if (cr_ses
->seqid
!= cs_slot
->sl_seqid
+ 1) {
1733 status
= nfserr_seq_misordered
;
1734 dprintk("Sequence misordered!\n");
1735 dprintk("Expected seqid= %d but got seqid= %d\n",
1736 cs_slot
->sl_seqid
, cr_ses
->seqid
);
1739 } else if (unconf
) {
1740 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
) ||
1741 !rpc_cmp_addr(sa
, (struct sockaddr
*) &unconf
->cl_addr
)) {
1742 status
= nfserr_clid_inuse
;
1746 cs_slot
= &unconf
->cl_cs_slot
;
1747 status
= check_slot_seqid(cr_ses
->seqid
, cs_slot
->sl_seqid
, 0);
1749 /* an unconfirmed replay returns misordered */
1750 status
= nfserr_seq_misordered
;
1757 status
= nfserr_stale_clientid
;
1762 * XXX: we should probably set this at creation time, and check
1763 * for consistent minorversion use throughout:
1765 conf
->cl_minorversion
= 1;
1767 * We do not support RDMA or persistent sessions
1769 cr_ses
->flags
&= ~SESSION4_PERSIST
;
1770 cr_ses
->flags
&= ~SESSION4_RDMA
;
1772 status
= nfserr_toosmall
;
1773 if (check_forechannel_attrs(cr_ses
->fore_channel
))
1776 status
= nfserr_jukebox
;
1777 new = alloc_init_session(rqstp
, conf
, cr_ses
);
1781 memcpy(cr_ses
->sessionid
.data
, new->se_sessionid
.data
,
1782 NFS4_MAX_SESSIONID_LEN
);
1783 memcpy(&cr_ses
->fore_channel
, &new->se_fchannel
,
1784 sizeof(struct nfsd4_channel_attrs
));
1785 cs_slot
->sl_seqid
++;
1786 cr_ses
->seqid
= cs_slot
->sl_seqid
;
1788 /* cache solo and embedded create sessions under the state lock */
1789 nfsd4_cache_create_session(cr_ses
, cs_slot
, status
);
1791 move_to_confirmed(conf
);
1793 nfs4_unlock_state();
1794 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1798 static bool nfsd4_last_compound_op(struct svc_rqst
*rqstp
)
1800 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1801 struct nfsd4_compoundargs
*argp
= rqstp
->rq_argp
;
1803 return argp
->opcnt
== resp
->opcnt
;
1806 static __be32
nfsd4_map_bcts_dir(u32
*dir
)
1809 case NFS4_CDFC4_FORE
:
1810 case NFS4_CDFC4_BACK
:
1812 case NFS4_CDFC4_FORE_OR_BOTH
:
1813 case NFS4_CDFC4_BACK_OR_BOTH
:
1814 *dir
= NFS4_CDFC4_BOTH
;
1817 return nfserr_inval
;
1820 __be32
nfsd4_bind_conn_to_session(struct svc_rqst
*rqstp
,
1821 struct nfsd4_compound_state
*cstate
,
1822 struct nfsd4_bind_conn_to_session
*bcts
)
1826 if (!nfsd4_last_compound_op(rqstp
))
1827 return nfserr_not_only_op
;
1828 spin_lock(&client_lock
);
1829 cstate
->session
= find_in_sessionid_hashtbl(&bcts
->sessionid
);
1830 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1831 * client_lock iself: */
1832 if (cstate
->session
) {
1833 nfsd4_get_session(cstate
->session
);
1834 atomic_inc(&cstate
->session
->se_client
->cl_refcount
);
1836 spin_unlock(&client_lock
);
1837 if (!cstate
->session
)
1838 return nfserr_badsession
;
1840 status
= nfsd4_map_bcts_dir(&bcts
->dir
);
1842 nfsd4_new_conn(rqstp
, cstate
->session
, bcts
->dir
);
1846 static bool nfsd4_compound_in_session(struct nfsd4_session
*session
, struct nfs4_sessionid
*sid
)
1850 return !memcmp(sid
, &session
->se_sessionid
, sizeof(*sid
));
1854 nfsd4_destroy_session(struct svc_rqst
*r
,
1855 struct nfsd4_compound_state
*cstate
,
1856 struct nfsd4_destroy_session
*sessionid
)
1858 struct nfsd4_session
*ses
;
1859 __be32 status
= nfserr_badsession
;
1862 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1863 * - Should we return nfserr_back_chan_busy if waiting for
1864 * callbacks on to-be-destroyed session?
1865 * - Do we need to clear any callback info from previous session?
1868 if (nfsd4_compound_in_session(cstate
->session
, &sessionid
->sessionid
)) {
1869 if (!nfsd4_last_compound_op(r
))
1870 return nfserr_not_only_op
;
1872 dump_sessionid(__func__
, &sessionid
->sessionid
);
1873 spin_lock(&client_lock
);
1874 ses
= find_in_sessionid_hashtbl(&sessionid
->sessionid
);
1876 spin_unlock(&client_lock
);
1880 unhash_session(ses
);
1881 spin_unlock(&client_lock
);
1884 nfsd4_probe_callback_sync(ses
->se_client
);
1885 nfs4_unlock_state();
1887 spin_lock(&client_lock
);
1888 nfsd4_del_conns(ses
);
1889 nfsd4_put_session_locked(ses
);
1890 spin_unlock(&client_lock
);
1893 dprintk("%s returns %d\n", __func__
, ntohl(status
));
1897 static struct nfsd4_conn
*__nfsd4_find_conn(struct svc_xprt
*xpt
, struct nfsd4_session
*s
)
1899 struct nfsd4_conn
*c
;
1901 list_for_each_entry(c
, &s
->se_conns
, cn_persession
) {
1902 if (c
->cn_xprt
== xpt
) {
1909 static void nfsd4_sequence_check_conn(struct nfsd4_conn
*new, struct nfsd4_session
*ses
)
1911 struct nfs4_client
*clp
= ses
->se_client
;
1912 struct nfsd4_conn
*c
;
1915 spin_lock(&clp
->cl_lock
);
1916 c
= __nfsd4_find_conn(new->cn_xprt
, ses
);
1918 spin_unlock(&clp
->cl_lock
);
1922 __nfsd4_hash_conn(new, ses
);
1923 spin_unlock(&clp
->cl_lock
);
1924 ret
= nfsd4_register_conn(new);
1926 /* oops; xprt is already down: */
1927 nfsd4_conn_lost(&new->cn_xpt_user
);
1931 static bool nfsd4_session_too_many_ops(struct svc_rqst
*rqstp
, struct nfsd4_session
*session
)
1933 struct nfsd4_compoundargs
*args
= rqstp
->rq_argp
;
1935 return args
->opcnt
> session
->se_fchannel
.maxops
;
1938 static bool nfsd4_request_too_big(struct svc_rqst
*rqstp
,
1939 struct nfsd4_session
*session
)
1941 struct xdr_buf
*xb
= &rqstp
->rq_arg
;
1943 return xb
->len
> session
->se_fchannel
.maxreq_sz
;
1947 nfsd4_sequence(struct svc_rqst
*rqstp
,
1948 struct nfsd4_compound_state
*cstate
,
1949 struct nfsd4_sequence
*seq
)
1951 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
1952 struct nfsd4_session
*session
;
1953 struct nfsd4_slot
*slot
;
1954 struct nfsd4_conn
*conn
;
1957 if (resp
->opcnt
!= 1)
1958 return nfserr_sequence_pos
;
1961 * Will be either used or freed by nfsd4_sequence_check_conn
1964 conn
= alloc_conn(rqstp
, NFS4_CDFC4_FORE
);
1966 return nfserr_jukebox
;
1968 spin_lock(&client_lock
);
1969 status
= nfserr_badsession
;
1970 session
= find_in_sessionid_hashtbl(&seq
->sessionid
);
1974 status
= nfserr_too_many_ops
;
1975 if (nfsd4_session_too_many_ops(rqstp
, session
))
1978 status
= nfserr_req_too_big
;
1979 if (nfsd4_request_too_big(rqstp
, session
))
1982 status
= nfserr_badslot
;
1983 if (seq
->slotid
>= session
->se_fchannel
.maxreqs
)
1986 slot
= session
->se_slots
[seq
->slotid
];
1987 dprintk("%s: slotid %d\n", __func__
, seq
->slotid
);
1989 /* We do not negotiate the number of slots yet, so set the
1990 * maxslots to the session maxreqs which is used to encode
1991 * sr_highest_slotid and the sr_target_slot id to maxslots */
1992 seq
->maxslots
= session
->se_fchannel
.maxreqs
;
1994 status
= check_slot_seqid(seq
->seqid
, slot
->sl_seqid
,
1995 slot
->sl_flags
& NFSD4_SLOT_INUSE
);
1996 if (status
== nfserr_replay_cache
) {
1997 status
= nfserr_seq_misordered
;
1998 if (!(slot
->sl_flags
& NFSD4_SLOT_INITIALIZED
))
2000 cstate
->slot
= slot
;
2001 cstate
->session
= session
;
2002 /* Return the cached reply status and set cstate->status
2003 * for nfsd4_proc_compound processing */
2004 status
= nfsd4_replay_cache_entry(resp
, seq
);
2005 cstate
->status
= nfserr_replay_cache
;
2011 nfsd4_sequence_check_conn(conn
, session
);
2014 /* Success! bump slot seqid */
2015 slot
->sl_seqid
= seq
->seqid
;
2016 slot
->sl_flags
|= NFSD4_SLOT_INUSE
;
2018 slot
->sl_flags
|= NFSD4_SLOT_CACHETHIS
;
2020 slot
->sl_flags
&= ~NFSD4_SLOT_CACHETHIS
;
2022 cstate
->slot
= slot
;
2023 cstate
->session
= session
;
2026 /* Hold a session reference until done processing the compound. */
2027 if (cstate
->session
) {
2028 struct nfs4_client
*clp
= session
->se_client
;
2030 nfsd4_get_session(cstate
->session
);
2031 atomic_inc(&clp
->cl_refcount
);
2032 switch (clp
->cl_cb_state
) {
2034 seq
->status_flags
= SEQ4_STATUS_CB_PATH_DOWN
;
2036 case NFSD4_CB_FAULT
:
2037 seq
->status_flags
= SEQ4_STATUS_BACKCHANNEL_FAULT
;
2040 seq
->status_flags
= 0;
2044 spin_unlock(&client_lock
);
2045 dprintk("%s: return %d\n", __func__
, ntohl(status
));
2049 static inline bool has_resources(struct nfs4_client
*clp
)
2051 return !list_empty(&clp
->cl_openowners
)
2052 || !list_empty(&clp
->cl_delegations
)
2053 || !list_empty(&clp
->cl_sessions
);
2057 nfsd4_destroy_clientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_destroy_clientid
*dc
)
2059 struct nfs4_client
*conf
, *unconf
, *clp
;
2063 unconf
= find_unconfirmed_client(&dc
->clientid
);
2064 conf
= find_confirmed_client(&dc
->clientid
);
2069 if (!is_client_expired(conf
) && has_resources(conf
)) {
2070 status
= nfserr_clientid_busy
;
2074 /* rfc5661 18.50.3 */
2075 if (cstate
->session
&& conf
== cstate
->session
->se_client
) {
2076 status
= nfserr_clientid_busy
;
2082 status
= nfserr_stale_clientid
;
2088 nfs4_unlock_state();
2089 dprintk("%s return %d\n", __func__
, ntohl(status
));
2094 nfsd4_reclaim_complete(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
, struct nfsd4_reclaim_complete
*rc
)
2098 if (rc
->rca_one_fs
) {
2099 if (!cstate
->current_fh
.fh_dentry
)
2100 return nfserr_nofilehandle
;
2102 * We don't take advantage of the rca_one_fs case.
2103 * That's OK, it's optional, we can safely ignore it.
2109 status
= nfserr_complete_already
;
2110 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE
,
2111 &cstate
->session
->se_client
->cl_flags
))
2114 status
= nfserr_stale_clientid
;
2115 if (is_client_expired(cstate
->session
->se_client
))
2117 * The following error isn't really legal.
2118 * But we only get here if the client just explicitly
2119 * destroyed the client. Surely it no longer cares what
2120 * error it gets back on an operation for the dead
2126 nfsd4_client_record_create(cstate
->session
->se_client
);
2128 nfs4_unlock_state();
2133 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2134 struct nfsd4_setclientid
*setclid
)
2136 struct xdr_netobj clname
= setclid
->se_name
;
2137 nfs4_verifier clverifier
= setclid
->se_verf
;
2138 unsigned int strhashval
;
2139 struct nfs4_client
*conf
, *unconf
, *new;
2141 char dname
[HEXDIR_LEN
];
2143 status
= nfs4_make_rec_clidname(dname
, &clname
);
2148 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2149 * We get here on a DRC miss.
2152 strhashval
= clientstr_hashval(dname
);
2155 conf
= find_confirmed_client_by_str(dname
, strhashval
);
2157 /* RFC 3530 14.2.33 CASE 0: */
2158 status
= nfserr_clid_inuse
;
2159 if (clp_used_exchangeid(conf
))
2161 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)) {
2162 char addr_str
[INET6_ADDRSTRLEN
];
2163 rpc_ntop((struct sockaddr
*) &conf
->cl_addr
, addr_str
,
2165 dprintk("NFSD: setclientid: string in use by client "
2166 "at %s\n", addr_str
);
2171 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2172 * has a description of SETCLIENTID request processing consisting
2173 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2175 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
2176 status
= nfserr_jukebox
;
2179 * RFC 3530 14.2.33 CASE 4:
2180 * placed first, because it is the normal case
2183 expire_client(unconf
);
2184 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2188 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
2190 * RFC 3530 14.2.33 CASE 1:
2191 * probable callback update
2194 /* Note this is removing unconfirmed {*x***},
2195 * which is stronger than RFC recommended {vxc**}.
2196 * This has the advantage that there is at most
2197 * one {*x***} in either list at any time.
2199 expire_client(unconf
);
2201 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2204 copy_clid(new, conf
);
2205 } else if (!unconf
) {
2207 * RFC 3530 14.2.33 CASE 2:
2208 * probable client reboot; state will be removed if
2211 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2217 * RFC 3530 14.2.33 CASE 3:
2218 * probable client reboot; state will be removed if
2221 expire_client(unconf
);
2222 new = create_client(clname
, dname
, rqstp
, &clverifier
);
2228 * XXX: we should probably set this at creation time, and check
2229 * for consistent minorversion use throughout:
2231 new->cl_minorversion
= 0;
2232 gen_callback(new, setclid
, rqstp
);
2233 add_to_unconfirmed(new, strhashval
);
2234 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
2235 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
2236 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
2239 nfs4_unlock_state();
2245 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2246 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2247 * bullets, labeled as CASE1 - CASE4 below.
2250 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
2251 struct nfsd4_compound_state
*cstate
,
2252 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
2254 struct nfs4_client
*conf
, *unconf
;
2255 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
2256 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
2259 if (STALE_CLIENTID(clid
))
2260 return nfserr_stale_clientid
;
2262 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2263 * We get here on a DRC miss.
2268 conf
= find_confirmed_client(clid
);
2269 unconf
= find_unconfirmed_client(clid
);
2272 * section 14.2.34 of RFC 3530 has a description of
2273 * SETCLIENTID_CONFIRM request processing consisting
2274 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2276 status
= nfserr_clid_inuse
;
2277 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
2279 * RFC 3530 14.2.34 CASE 1:
2282 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
2283 status
= nfserr_clid_inuse
;
2285 nfsd4_change_callback(conf
, &unconf
->cl_cb_conn
);
2286 nfsd4_probe_callback(conf
);
2287 expire_client(unconf
);
2290 } else if (conf
&& !unconf
) {
2292 * RFC 3530 14.2.34 CASE 2:
2293 * probable retransmitted request; play it safe and
2296 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
2297 status
= nfserr_clid_inuse
;
2300 } else if (!conf
&& unconf
2301 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
2303 * RFC 3530 14.2.34 CASE 3:
2304 * Normal case; new or rebooted client:
2306 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
2307 status
= nfserr_clid_inuse
;
2310 clientstr_hashval(unconf
->cl_recdir
);
2311 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
2314 nfsd4_client_record_remove(conf
);
2315 expire_client(conf
);
2317 move_to_confirmed(unconf
);
2319 nfsd4_probe_callback(conf
);
2322 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
2323 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
2326 * RFC 3530 14.2.34 CASE 4:
2327 * Client probably hasn't noticed that we rebooted yet.
2329 status
= nfserr_stale_clientid
;
2332 nfs4_unlock_state();
2336 static struct nfs4_file
*nfsd4_alloc_file(void)
2338 return kmem_cache_alloc(file_slab
, GFP_KERNEL
);
2341 /* OPEN Share state helper functions */
2342 static void nfsd4_init_file(struct nfs4_file
*fp
, struct inode
*ino
)
2344 unsigned int hashval
= file_hashval(ino
);
2346 atomic_set(&fp
->fi_ref
, 1);
2347 INIT_LIST_HEAD(&fp
->fi_hash
);
2348 INIT_LIST_HEAD(&fp
->fi_stateids
);
2349 INIT_LIST_HEAD(&fp
->fi_delegations
);
2350 fp
->fi_inode
= igrab(ino
);
2351 fp
->fi_had_conflict
= false;
2352 fp
->fi_lease
= NULL
;
2353 memset(fp
->fi_fds
, 0, sizeof(fp
->fi_fds
));
2354 memset(fp
->fi_access
, 0, sizeof(fp
->fi_access
));
2355 spin_lock(&recall_lock
);
2356 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
2357 spin_unlock(&recall_lock
);
2361 nfsd4_free_slab(struct kmem_cache
**slab
)
2365 kmem_cache_destroy(*slab
);
2370 nfsd4_free_slabs(void)
2372 nfsd4_free_slab(&openowner_slab
);
2373 nfsd4_free_slab(&lockowner_slab
);
2374 nfsd4_free_slab(&file_slab
);
2375 nfsd4_free_slab(&stateid_slab
);
2376 nfsd4_free_slab(&deleg_slab
);
2380 nfsd4_init_slabs(void)
2382 openowner_slab
= kmem_cache_create("nfsd4_openowners",
2383 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2384 if (openowner_slab
== NULL
)
2386 lockowner_slab
= kmem_cache_create("nfsd4_lockowners",
2387 sizeof(struct nfs4_openowner
), 0, 0, NULL
);
2388 if (lockowner_slab
== NULL
)
2390 file_slab
= kmem_cache_create("nfsd4_files",
2391 sizeof(struct nfs4_file
), 0, 0, NULL
);
2392 if (file_slab
== NULL
)
2394 stateid_slab
= kmem_cache_create("nfsd4_stateids",
2395 sizeof(struct nfs4_ol_stateid
), 0, 0, NULL
);
2396 if (stateid_slab
== NULL
)
2398 deleg_slab
= kmem_cache_create("nfsd4_delegations",
2399 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
2400 if (deleg_slab
== NULL
)
2405 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2409 void nfs4_free_openowner(struct nfs4_openowner
*oo
)
2411 kfree(oo
->oo_owner
.so_owner
.data
);
2412 kmem_cache_free(openowner_slab
, oo
);
2415 void nfs4_free_lockowner(struct nfs4_lockowner
*lo
)
2417 kfree(lo
->lo_owner
.so_owner
.data
);
2418 kmem_cache_free(lockowner_slab
, lo
);
2421 static void init_nfs4_replay(struct nfs4_replay
*rp
)
2423 rp
->rp_status
= nfserr_serverfault
;
2425 rp
->rp_buf
= rp
->rp_ibuf
;
2428 static inline void *alloc_stateowner(struct kmem_cache
*slab
, struct xdr_netobj
*owner
, struct nfs4_client
*clp
)
2430 struct nfs4_stateowner
*sop
;
2432 sop
= kmem_cache_alloc(slab
, GFP_KERNEL
);
2436 sop
->so_owner
.data
= kmemdup(owner
->data
, owner
->len
, GFP_KERNEL
);
2437 if (!sop
->so_owner
.data
) {
2438 kmem_cache_free(slab
, sop
);
2441 sop
->so_owner
.len
= owner
->len
;
2443 INIT_LIST_HEAD(&sop
->so_stateids
);
2444 sop
->so_client
= clp
;
2445 init_nfs4_replay(&sop
->so_replay
);
2449 static void hash_openowner(struct nfs4_openowner
*oo
, struct nfs4_client
*clp
, unsigned int strhashval
)
2451 list_add(&oo
->oo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
2452 list_add(&oo
->oo_perclient
, &clp
->cl_openowners
);
2455 static struct nfs4_openowner
*
2456 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
2457 struct nfs4_openowner
*oo
;
2459 oo
= alloc_stateowner(openowner_slab
, &open
->op_owner
, clp
);
2462 oo
->oo_owner
.so_is_open_owner
= 1;
2463 oo
->oo_owner
.so_seqid
= open
->op_seqid
;
2464 oo
->oo_flags
= NFS4_OO_NEW
;
2466 oo
->oo_last_closed_stid
= NULL
;
2467 INIT_LIST_HEAD(&oo
->oo_close_lru
);
2468 hash_openowner(oo
, clp
, strhashval
);
2472 static void init_open_stateid(struct nfs4_ol_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
2473 struct nfs4_openowner
*oo
= open
->op_openowner
;
2474 struct nfs4_client
*clp
= oo
->oo_owner
.so_client
;
2476 init_stid(&stp
->st_stid
, clp
, NFS4_OPEN_STID
);
2477 INIT_LIST_HEAD(&stp
->st_lockowners
);
2478 list_add(&stp
->st_perstateowner
, &oo
->oo_owner
.so_stateids
);
2479 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2480 stp
->st_stateowner
= &oo
->oo_owner
;
2483 stp
->st_access_bmap
= 0;
2484 stp
->st_deny_bmap
= 0;
2485 set_access(open
->op_share_access
, stp
);
2486 set_deny(open
->op_share_deny
, stp
);
2487 stp
->st_openstp
= NULL
;
2491 move_to_close_lru(struct nfs4_openowner
*oo
)
2493 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo
);
2495 list_move_tail(&oo
->oo_close_lru
, &close_lru
);
2496 oo
->oo_time
= get_seconds();
2500 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
2503 return (sop
->so_owner
.len
== owner
->len
) &&
2504 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
2505 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
2508 static struct nfs4_openowner
*
2509 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
2511 struct nfs4_stateowner
*so
;
2512 struct nfs4_openowner
*oo
;
2514 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
2515 if (!so
->so_is_open_owner
)
2517 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
)) {
2519 renew_client(oo
->oo_owner
.so_client
);
2526 /* search file_hashtbl[] for file */
2527 static struct nfs4_file
*
2528 find_file(struct inode
*ino
)
2530 unsigned int hashval
= file_hashval(ino
);
2531 struct nfs4_file
*fp
;
2533 spin_lock(&recall_lock
);
2534 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
2535 if (fp
->fi_inode
== ino
) {
2537 spin_unlock(&recall_lock
);
2541 spin_unlock(&recall_lock
);
2546 * Called to check deny when READ with all zero stateid or
2547 * WRITE with all zero or all one stateid
2550 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
2552 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2553 struct nfs4_file
*fp
;
2554 struct nfs4_ol_stateid
*stp
;
2557 dprintk("NFSD: nfs4_share_conflict\n");
2559 fp
= find_file(ino
);
2562 ret
= nfserr_locked
;
2563 /* Search for conflicting share reservations */
2564 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
2565 if (test_deny(deny_type
, stp
) ||
2566 test_deny(NFS4_SHARE_DENY_BOTH
, stp
))
2575 static void nfsd_break_one_deleg(struct nfs4_delegation
*dp
)
2577 /* We're assuming the state code never drops its reference
2578 * without first removing the lease. Since we're in this lease
2579 * callback (and since the lease code is serialized by the kernel
2580 * lock) we know the server hasn't removed the lease yet, we know
2581 * it's safe to take a reference: */
2582 atomic_inc(&dp
->dl_count
);
2584 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
2586 /* only place dl_time is set. protected by lock_flocks*/
2587 dp
->dl_time
= get_seconds();
2589 nfsd4_cb_recall(dp
);
2592 /* Called from break_lease() with lock_flocks() held. */
2593 static void nfsd_break_deleg_cb(struct file_lock
*fl
)
2595 struct nfs4_file
*fp
= (struct nfs4_file
*)fl
->fl_owner
;
2596 struct nfs4_delegation
*dp
;
2599 /* We assume break_lease is only called once per lease: */
2600 BUG_ON(fp
->fi_had_conflict
);
2602 * We don't want the locks code to timeout the lease for us;
2603 * we'll remove it ourself if a delegation isn't returned
2606 fl
->fl_break_time
= 0;
2608 spin_lock(&recall_lock
);
2609 fp
->fi_had_conflict
= true;
2610 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
)
2611 nfsd_break_one_deleg(dp
);
2612 spin_unlock(&recall_lock
);
2616 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
2619 return lease_modify(onlist
, arg
);
2624 static const struct lock_manager_operations nfsd_lease_mng_ops
= {
2625 .lm_break
= nfsd_break_deleg_cb
,
2626 .lm_change
= nfsd_change_deleg_cb
,
2629 static __be32
nfsd4_check_seqid(struct nfsd4_compound_state
*cstate
, struct nfs4_stateowner
*so
, u32 seqid
)
2631 if (nfsd4_has_session(cstate
))
2633 if (seqid
== so
->so_seqid
- 1)
2634 return nfserr_replay_me
;
2635 if (seqid
== so
->so_seqid
)
2637 return nfserr_bad_seqid
;
2641 nfsd4_process_open1(struct nfsd4_compound_state
*cstate
,
2642 struct nfsd4_open
*open
)
2644 clientid_t
*clientid
= &open
->op_clientid
;
2645 struct nfs4_client
*clp
= NULL
;
2646 unsigned int strhashval
;
2647 struct nfs4_openowner
*oo
= NULL
;
2650 if (STALE_CLIENTID(&open
->op_clientid
))
2651 return nfserr_stale_clientid
;
2653 * In case we need it later, after we've already created the
2654 * file and don't want to risk a further failure:
2656 open
->op_file
= nfsd4_alloc_file();
2657 if (open
->op_file
== NULL
)
2658 return nfserr_jukebox
;
2660 strhashval
= ownerstr_hashval(clientid
->cl_id
, &open
->op_owner
);
2661 oo
= find_openstateowner_str(strhashval
, open
);
2662 open
->op_openowner
= oo
;
2664 clp
= find_confirmed_client(clientid
);
2666 return nfserr_expired
;
2669 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
)) {
2670 /* Replace unconfirmed owners without checking for replay. */
2671 clp
= oo
->oo_owner
.so_client
;
2672 release_openowner(oo
);
2673 open
->op_openowner
= NULL
;
2676 status
= nfsd4_check_seqid(cstate
, &oo
->oo_owner
, open
->op_seqid
);
2679 clp
= oo
->oo_owner
.so_client
;
2682 oo
= alloc_init_open_stateowner(strhashval
, clp
, open
);
2684 return nfserr_jukebox
;
2685 open
->op_openowner
= oo
;
2687 open
->op_stp
= nfs4_alloc_stateid(clp
);
2689 return nfserr_jukebox
;
2693 static inline __be32
2694 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
2696 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
2697 return nfserr_openmode
;
2702 static int share_access_to_flags(u32 share_access
)
2704 return share_access
== NFS4_SHARE_ACCESS_READ
? RD_STATE
: WR_STATE
;
2707 static struct nfs4_delegation
*find_deleg_stateid(struct nfs4_client
*cl
, stateid_t
*s
)
2709 struct nfs4_stid
*ret
;
2711 ret
= find_stateid_by_type(cl
, s
, NFS4_DELEG_STID
);
2714 return delegstateid(ret
);
2717 static bool nfsd4_is_deleg_cur(struct nfsd4_open
*open
)
2719 return open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
||
2720 open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEG_CUR_FH
;
2724 nfs4_check_deleg(struct nfs4_client
*cl
, struct nfs4_file
*fp
, struct nfsd4_open
*open
,
2725 struct nfs4_delegation
**dp
)
2728 __be32 status
= nfserr_bad_stateid
;
2730 *dp
= find_deleg_stateid(cl
, &open
->op_delegate_stateid
);
2733 flags
= share_access_to_flags(open
->op_share_access
);
2734 status
= nfs4_check_delegmode(*dp
, flags
);
2738 if (!nfsd4_is_deleg_cur(open
))
2742 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2747 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
**stpp
)
2749 struct nfs4_ol_stateid
*local
;
2750 struct nfs4_openowner
*oo
= open
->op_openowner
;
2752 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
2753 /* ignore lock owners */
2754 if (local
->st_stateowner
->so_is_open_owner
== 0)
2756 /* remember if we have seen this open owner */
2757 if (local
->st_stateowner
== &oo
->oo_owner
)
2759 /* check for conflicting share reservations */
2760 if (!test_share(local
, open
))
2761 return nfserr_share_denied
;
2766 static void nfs4_free_stateid(struct nfs4_ol_stateid
*s
)
2768 kmem_cache_free(stateid_slab
, s
);
2771 static inline int nfs4_access_to_access(u32 nfs4_access
)
2775 if (nfs4_access
& NFS4_SHARE_ACCESS_READ
)
2776 flags
|= NFSD_MAY_READ
;
2777 if (nfs4_access
& NFS4_SHARE_ACCESS_WRITE
)
2778 flags
|= NFSD_MAY_WRITE
;
2782 static __be32
nfs4_get_vfs_file(struct svc_rqst
*rqstp
, struct nfs4_file
*fp
,
2783 struct svc_fh
*cur_fh
, struct nfsd4_open
*open
)
2786 int oflag
= nfs4_access_to_omode(open
->op_share_access
);
2787 int access
= nfs4_access_to_access(open
->op_share_access
);
2789 if (!fp
->fi_fds
[oflag
]) {
2790 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, access
,
2791 &fp
->fi_fds
[oflag
]);
2795 nfs4_file_get_access(fp
, oflag
);
2800 static inline __be32
2801 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
2802 struct nfsd4_open
*open
)
2804 struct iattr iattr
= {
2805 .ia_valid
= ATTR_SIZE
,
2808 if (!open
->op_truncate
)
2810 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
2811 return nfserr_inval
;
2812 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
2816 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
)
2818 u32 op_share_access
= open
->op_share_access
;
2822 new_access
= !test_access(op_share_access
, stp
);
2824 status
= nfs4_get_vfs_file(rqstp
, fp
, cur_fh
, open
);
2828 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
2831 int oflag
= nfs4_access_to_omode(op_share_access
);
2832 nfs4_file_put_access(fp
, oflag
);
2836 /* remember the open */
2837 set_access(op_share_access
, stp
);
2838 set_deny(open
->op_share_deny
, stp
);
2845 nfs4_set_claim_prev(struct nfsd4_open
*open
, bool has_session
)
2847 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
2850 /* Should we give out recallable state?: */
2851 static bool nfsd4_cb_channel_good(struct nfs4_client
*clp
)
2853 if (clp
->cl_cb_state
== NFSD4_CB_UP
)
2856 * In the sessions case, since we don't have to establish a
2857 * separate connection for callbacks, we assume it's OK
2858 * until we hear otherwise:
2860 return clp
->cl_minorversion
&& clp
->cl_cb_state
== NFSD4_CB_UNKNOWN
;
2863 static struct file_lock
*nfs4_alloc_init_lease(struct nfs4_delegation
*dp
, int flag
)
2865 struct file_lock
*fl
;
2867 fl
= locks_alloc_lock();
2870 locks_init_lock(fl
);
2871 fl
->fl_lmops
= &nfsd_lease_mng_ops
;
2872 fl
->fl_flags
= FL_LEASE
;
2873 fl
->fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
2874 fl
->fl_end
= OFFSET_MAX
;
2875 fl
->fl_owner
= (fl_owner_t
)(dp
->dl_file
);
2876 fl
->fl_pid
= current
->tgid
;
2880 static int nfs4_setlease(struct nfs4_delegation
*dp
, int flag
)
2882 struct nfs4_file
*fp
= dp
->dl_file
;
2883 struct file_lock
*fl
;
2886 fl
= nfs4_alloc_init_lease(dp
, flag
);
2889 fl
->fl_file
= find_readable_file(fp
);
2890 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2891 status
= vfs_setlease(fl
->fl_file
, fl
->fl_type
, &fl
);
2893 list_del_init(&dp
->dl_perclnt
);
2894 locks_free_lock(fl
);
2898 fp
->fi_deleg_file
= fl
->fl_file
;
2899 get_file(fp
->fi_deleg_file
);
2900 atomic_set(&fp
->fi_delegees
, 1);
2901 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2905 static int nfs4_set_delegation(struct nfs4_delegation
*dp
, int flag
)
2907 struct nfs4_file
*fp
= dp
->dl_file
;
2910 return nfs4_setlease(dp
, flag
);
2911 spin_lock(&recall_lock
);
2912 if (fp
->fi_had_conflict
) {
2913 spin_unlock(&recall_lock
);
2916 atomic_inc(&fp
->fi_delegees
);
2917 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
2918 spin_unlock(&recall_lock
);
2919 list_add(&dp
->dl_perclnt
, &dp
->dl_stid
.sc_client
->cl_delegations
);
2923 static void nfsd4_open_deleg_none_ext(struct nfsd4_open
*open
, int status
)
2925 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
2926 if (status
== -EAGAIN
)
2927 open
->op_why_no_deleg
= WND4_CONTENTION
;
2929 open
->op_why_no_deleg
= WND4_RESOURCE
;
2930 switch (open
->op_deleg_want
) {
2931 case NFS4_SHARE_WANT_READ_DELEG
:
2932 case NFS4_SHARE_WANT_WRITE_DELEG
:
2933 case NFS4_SHARE_WANT_ANY_DELEG
:
2935 case NFS4_SHARE_WANT_CANCEL
:
2936 open
->op_why_no_deleg
= WND4_CANCELLED
;
2938 case NFS4_SHARE_WANT_NO_DELEG
:
2939 BUG(); /* not supposed to get here */
2945 * Attempt to hand out a delegation.
2948 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_ol_stateid
*stp
)
2950 struct nfs4_delegation
*dp
;
2951 struct nfs4_openowner
*oo
= container_of(stp
->st_stateowner
, struct nfs4_openowner
, oo_owner
);
2953 int status
= 0, flag
= 0;
2955 cb_up
= nfsd4_cb_channel_good(oo
->oo_owner
.so_client
);
2956 flag
= NFS4_OPEN_DELEGATE_NONE
;
2957 open
->op_recall
= 0;
2958 switch (open
->op_claim_type
) {
2959 case NFS4_OPEN_CLAIM_PREVIOUS
:
2961 open
->op_recall
= 1;
2962 flag
= open
->op_delegate_type
;
2963 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
2966 case NFS4_OPEN_CLAIM_NULL
:
2967 /* Let's not give out any delegations till everyone's
2968 * had the chance to reclaim theirs.... */
2969 if (locks_in_grace())
2971 if (!cb_up
|| !(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
2973 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
2974 flag
= NFS4_OPEN_DELEGATE_WRITE
;
2976 flag
= NFS4_OPEN_DELEGATE_READ
;
2982 dp
= alloc_init_deleg(oo
->oo_owner
.so_client
, stp
, fh
, flag
);
2985 status
= nfs4_set_delegation(dp
, flag
);
2989 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stid
.sc_stateid
, sizeof(dp
->dl_stid
.sc_stateid
));
2991 dprintk("NFSD: delegation stateid=" STATEID_FMT
"\n",
2992 STATEID_VAL(&dp
->dl_stid
.sc_stateid
));
2994 open
->op_delegate_type
= flag
;
2995 if (flag
== NFS4_OPEN_DELEGATE_NONE
) {
2996 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
&&
2997 open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
2998 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3000 /* 4.1 client asking for a delegation? */
3001 if (open
->op_deleg_want
)
3002 nfsd4_open_deleg_none_ext(open
, status
);
3006 nfs4_put_delegation(dp
);
3008 flag
= NFS4_OPEN_DELEGATE_NONE
;
3012 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open
*open
,
3013 struct nfs4_delegation
*dp
)
3015 if (open
->op_deleg_want
== NFS4_SHARE_WANT_READ_DELEG
&&
3016 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
3017 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3018 open
->op_why_no_deleg
= WND4_NOT_SUPP_DOWNGRADE
;
3019 } else if (open
->op_deleg_want
== NFS4_SHARE_WANT_WRITE_DELEG
&&
3020 dp
->dl_type
== NFS4_OPEN_DELEGATE_WRITE
) {
3021 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3022 open
->op_why_no_deleg
= WND4_NOT_SUPP_UPGRADE
;
3024 /* Otherwise the client must be confused wanting a delegation
3025 * it already has, therefore we don't return
3026 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3031 * called with nfs4_lock_state() held.
3034 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
3036 struct nfsd4_compoundres
*resp
= rqstp
->rq_resp
;
3037 struct nfs4_client
*cl
= open
->op_openowner
->oo_owner
.so_client
;
3038 struct nfs4_file
*fp
= NULL
;
3039 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3040 struct nfs4_ol_stateid
*stp
= NULL
;
3041 struct nfs4_delegation
*dp
= NULL
;
3045 * Lookup file; if found, lookup stateid and check open request,
3046 * and check for delegations in the process of being recalled.
3047 * If not found, create the nfs4_file struct
3049 fp
= find_file(ino
);
3051 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
3053 status
= nfs4_check_deleg(cl
, fp
, open
, &dp
);
3057 status
= nfserr_bad_stateid
;
3058 if (nfsd4_is_deleg_cur(open
))
3060 status
= nfserr_jukebox
;
3062 open
->op_file
= NULL
;
3063 nfsd4_init_file(fp
, ino
);
3067 * OPEN the file, or upgrade an existing OPEN.
3068 * If truncate fails, the OPEN fails.
3071 /* Stateid was found, this is an OPEN upgrade */
3072 status
= nfs4_upgrade_open(rqstp
, fp
, current_fh
, stp
, open
);
3076 status
= nfs4_get_vfs_file(rqstp
, fp
, current_fh
, open
);
3080 open
->op_stp
= NULL
;
3081 init_open_stateid(stp
, fp
, open
);
3082 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
3084 release_open_stateid(stp
);
3088 update_stateid(&stp
->st_stid
.sc_stateid
);
3089 memcpy(&open
->op_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3091 if (nfsd4_has_session(&resp
->cstate
)) {
3092 open
->op_openowner
->oo_flags
|= NFS4_OO_CONFIRMED
;
3094 if (open
->op_deleg_want
& NFS4_SHARE_WANT_NO_DELEG
) {
3095 open
->op_delegate_type
= NFS4_OPEN_DELEGATE_NONE_EXT
;
3096 open
->op_why_no_deleg
= WND4_NOT_WANTED
;
3102 * Attempt to hand out a delegation. No error return, because the
3103 * OPEN succeeds even if we fail.
3105 nfs4_open_delegation(current_fh
, open
, stp
);
3109 dprintk("%s: stateid=" STATEID_FMT
"\n", __func__
,
3110 STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3112 /* 4.1 client trying to upgrade/downgrade delegation? */
3113 if (open
->op_delegate_type
== NFS4_OPEN_DELEGATE_NONE
&& dp
&&
3114 open
->op_deleg_want
)
3115 nfsd4_deleg_xgrade_none_ext(open
, dp
);
3119 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
3120 nfs4_set_claim_prev(open
, nfsd4_has_session(&resp
->cstate
));
3122 * To finish the open response, we just need to set the rflags.
3124 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
3125 if (!(open
->op_openowner
->oo_flags
& NFS4_OO_CONFIRMED
) &&
3126 !nfsd4_has_session(&resp
->cstate
))
3127 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
3132 void nfsd4_cleanup_open_state(struct nfsd4_open
*open
, __be32 status
)
3134 if (open
->op_openowner
) {
3135 struct nfs4_openowner
*oo
= open
->op_openowner
;
3137 if (!list_empty(&oo
->oo_owner
.so_stateids
))
3138 list_del_init(&oo
->oo_close_lru
);
3139 if (oo
->oo_flags
& NFS4_OO_NEW
) {
3141 release_openowner(oo
);
3142 open
->op_openowner
= NULL
;
3144 oo
->oo_flags
&= ~NFS4_OO_NEW
;
3148 nfsd4_free_file(open
->op_file
);
3150 nfs4_free_stateid(open
->op_stp
);
3154 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3157 struct nfs4_client
*clp
;
3161 dprintk("process_renew(%08x/%08x): starting\n",
3162 clid
->cl_boot
, clid
->cl_id
);
3163 status
= nfserr_stale_clientid
;
3164 if (STALE_CLIENTID(clid
))
3166 clp
= find_confirmed_client(clid
);
3167 status
= nfserr_expired
;
3169 /* We assume the client took too long to RENEW. */
3170 dprintk("nfsd4_renew: clientid not found!\n");
3173 status
= nfserr_cb_path_down
;
3174 if (!list_empty(&clp
->cl_delegations
)
3175 && clp
->cl_cb_state
!= NFSD4_CB_UP
)
3179 nfs4_unlock_state();
3183 static struct lock_manager nfsd4_manager
= {
3186 static bool grace_ended
;
3189 nfsd4_end_grace(void)
3191 /* do nothing if grace period already ended */
3195 dprintk("NFSD: end of grace period\n");
3197 nfsd4_record_grace_done(&init_net
, boot_time
);
3198 locks_end_grace(&nfsd4_manager
);
3200 * Now that every NFSv4 client has had the chance to recover and
3201 * to see the (possibly new, possibly shorter) lease time, we
3202 * can safely set the next grace time to the current lease time:
3204 nfsd4_grace
= nfsd4_lease
;
3208 nfs4_laundromat(void)
3210 struct nfs4_client
*clp
;
3211 struct nfs4_openowner
*oo
;
3212 struct nfs4_delegation
*dp
;
3213 struct list_head
*pos
, *next
, reaplist
;
3214 time_t cutoff
= get_seconds() - nfsd4_lease
;
3215 time_t t
, clientid_val
= nfsd4_lease
;
3216 time_t u
, test_val
= nfsd4_lease
;
3220 dprintk("NFSD: laundromat service - starting\n");
3222 INIT_LIST_HEAD(&reaplist
);
3223 spin_lock(&client_lock
);
3224 list_for_each_safe(pos
, next
, &client_lru
) {
3225 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3226 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
3227 t
= clp
->cl_time
- cutoff
;
3228 if (clientid_val
> t
)
3232 if (atomic_read(&clp
->cl_refcount
)) {
3233 dprintk("NFSD: client in use (clientid %08x)\n",
3234 clp
->cl_clientid
.cl_id
);
3237 unhash_client_locked(clp
);
3238 list_add(&clp
->cl_lru
, &reaplist
);
3240 spin_unlock(&client_lock
);
3241 list_for_each_safe(pos
, next
, &reaplist
) {
3242 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
3243 dprintk("NFSD: purging unused client (clientid %08x)\n",
3244 clp
->cl_clientid
.cl_id
);
3245 nfsd4_client_record_remove(clp
);
3248 spin_lock(&recall_lock
);
3249 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3250 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3251 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
3252 u
= dp
->dl_time
- cutoff
;
3257 list_move(&dp
->dl_recall_lru
, &reaplist
);
3259 spin_unlock(&recall_lock
);
3260 list_for_each_safe(pos
, next
, &reaplist
) {
3261 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3262 unhash_delegation(dp
);
3264 test_val
= nfsd4_lease
;
3265 list_for_each_safe(pos
, next
, &close_lru
) {
3266 oo
= container_of(pos
, struct nfs4_openowner
, oo_close_lru
);
3267 if (time_after((unsigned long)oo
->oo_time
, (unsigned long)cutoff
)) {
3268 u
= oo
->oo_time
- cutoff
;
3273 release_openowner(oo
);
3275 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
3276 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
3277 nfs4_unlock_state();
3278 return clientid_val
;
3281 static struct workqueue_struct
*laundry_wq
;
3282 static void laundromat_main(struct work_struct
*);
3283 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
3286 laundromat_main(struct work_struct
*not_used
)
3290 t
= nfs4_laundromat();
3291 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
3292 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
3295 static inline __be32
nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_ol_stateid
*stp
)
3297 if (fhp
->fh_dentry
->d_inode
!= stp
->st_file
->fi_inode
)
3298 return nfserr_bad_stateid
;
3303 STALE_STATEID(stateid_t
*stateid
)
3305 if (stateid
->si_opaque
.so_clid
.cl_boot
== boot_time
)
3307 dprintk("NFSD: stale stateid " STATEID_FMT
"!\n",
3308 STATEID_VAL(stateid
));
3313 access_permit_read(struct nfs4_ol_stateid
*stp
)
3315 return test_access(NFS4_SHARE_ACCESS_READ
, stp
) ||
3316 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
) ||
3317 test_access(NFS4_SHARE_ACCESS_WRITE
, stp
);
3321 access_permit_write(struct nfs4_ol_stateid
*stp
)
3323 return test_access(NFS4_SHARE_ACCESS_WRITE
, stp
) ||
3324 test_access(NFS4_SHARE_ACCESS_BOTH
, stp
);
3328 __be32
nfs4_check_openmode(struct nfs4_ol_stateid
*stp
, int flags
)
3330 __be32 status
= nfserr_openmode
;
3332 /* For lock stateid's, we test the parent open, not the lock: */
3333 if (stp
->st_openstp
)
3334 stp
= stp
->st_openstp
;
3335 if ((flags
& WR_STATE
) && !access_permit_write(stp
))
3337 if ((flags
& RD_STATE
) && !access_permit_read(stp
))
3344 static inline __be32
3345 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
3347 if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
3349 else if (locks_in_grace()) {
3350 /* Answer in remaining cases depends on existence of
3351 * conflicting state; so we must wait out the grace period. */
3352 return nfserr_grace
;
3353 } else if (flags
& WR_STATE
)
3354 return nfs4_share_conflict(current_fh
,
3355 NFS4_SHARE_DENY_WRITE
);
3356 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3357 return nfs4_share_conflict(current_fh
,
3358 NFS4_SHARE_DENY_READ
);
3362 * Allow READ/WRITE during grace period on recovered state only for files
3363 * that are not able to provide mandatory locking.
3366 grace_disallows_io(struct inode
*inode
)
3368 return locks_in_grace() && mandatory_lock(inode
);
3371 /* Returns true iff a is later than b: */
3372 static bool stateid_generation_after(stateid_t
*a
, stateid_t
*b
)
3374 return (s32
)a
->si_generation
- (s32
)b
->si_generation
> 0;
3377 static __be32
check_stateid_generation(stateid_t
*in
, stateid_t
*ref
, bool has_session
)
3380 * When sessions are used the stateid generation number is ignored
3383 if (has_session
&& in
->si_generation
== 0)
3386 if (in
->si_generation
== ref
->si_generation
)
3389 /* If the client sends us a stateid from the future, it's buggy: */
3390 if (stateid_generation_after(in
, ref
))
3391 return nfserr_bad_stateid
;
3393 * However, we could see a stateid from the past, even from a
3394 * non-buggy client. For example, if the client sends a lock
3395 * while some IO is outstanding, the lock may bump si_generation
3396 * while the IO is still in flight. The client could avoid that
3397 * situation by waiting for responses on all the IO requests,
3398 * but better performance may result in retrying IO that
3399 * receives an old_stateid error if requests are rarely
3400 * reordered in flight:
3402 return nfserr_old_stateid
;
3405 __be32
nfs4_validate_stateid(struct nfs4_client
*cl
, stateid_t
*stateid
)
3407 struct nfs4_stid
*s
;
3408 struct nfs4_ol_stateid
*ols
;
3411 if (STALE_STATEID(stateid
))
3412 return nfserr_stale_stateid
;
3414 s
= find_stateid(cl
, stateid
);
3416 return nfserr_stale_stateid
;
3417 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3420 if (!(s
->sc_type
& (NFS4_OPEN_STID
| NFS4_LOCK_STID
)))
3422 ols
= openlockstateid(s
);
3423 if (ols
->st_stateowner
->so_is_open_owner
3424 && !(openowner(ols
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3425 return nfserr_bad_stateid
;
3429 static __be32
nfsd4_lookup_stateid(stateid_t
*stateid
, unsigned char typemask
, struct nfs4_stid
**s
)
3431 struct nfs4_client
*cl
;
3433 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3434 return nfserr_bad_stateid
;
3435 if (STALE_STATEID(stateid
))
3436 return nfserr_stale_stateid
;
3437 cl
= find_confirmed_client(&stateid
->si_opaque
.so_clid
);
3439 return nfserr_expired
;
3440 *s
= find_stateid_by_type(cl
, stateid
, typemask
);
3442 return nfserr_bad_stateid
;
3448 * Checks for stateid operations
3451 nfs4_preprocess_stateid_op(struct nfsd4_compound_state
*cstate
,
3452 stateid_t
*stateid
, int flags
, struct file
**filpp
)
3454 struct nfs4_stid
*s
;
3455 struct nfs4_ol_stateid
*stp
= NULL
;
3456 struct nfs4_delegation
*dp
= NULL
;
3457 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3458 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
3464 if (grace_disallows_io(ino
))
3465 return nfserr_grace
;
3467 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
3468 return check_special_stateids(current_fh
, stateid
, flags
);
3470 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
|NFS4_OPEN_STID
|NFS4_LOCK_STID
, &s
);
3473 status
= check_stateid_generation(stateid
, &s
->sc_stateid
, nfsd4_has_session(cstate
));
3476 switch (s
->sc_type
) {
3477 case NFS4_DELEG_STID
:
3478 dp
= delegstateid(s
);
3479 status
= nfs4_check_delegmode(dp
, flags
);
3483 *filpp
= dp
->dl_file
->fi_deleg_file
;
3487 case NFS4_OPEN_STID
:
3488 case NFS4_LOCK_STID
:
3489 stp
= openlockstateid(s
);
3490 status
= nfs4_check_fh(current_fh
, stp
);
3493 if (stp
->st_stateowner
->so_is_open_owner
3494 && !(openowner(stp
->st_stateowner
)->oo_flags
& NFS4_OO_CONFIRMED
))
3496 status
= nfs4_check_openmode(stp
, flags
);
3500 if (flags
& RD_STATE
)
3501 *filpp
= find_readable_file(stp
->st_file
);
3503 *filpp
= find_writeable_file(stp
->st_file
);
3507 return nfserr_bad_stateid
;
3515 nfsd4_free_lock_stateid(struct nfs4_ol_stateid
*stp
)
3517 if (check_for_locks(stp
->st_file
, lockowner(stp
->st_stateowner
)))
3518 return nfserr_locks_held
;
3519 release_lock_stateid(stp
);
3524 * Test if the stateid is valid
3527 nfsd4_test_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3528 struct nfsd4_test_stateid
*test_stateid
)
3530 struct nfsd4_test_stateid_id
*stateid
;
3531 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3534 list_for_each_entry(stateid
, &test_stateid
->ts_stateid_list
, ts_id_list
)
3535 stateid
->ts_id_status
= nfs4_validate_stateid(cl
, &stateid
->ts_id_stateid
);
3536 nfs4_unlock_state();
3542 nfsd4_free_stateid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3543 struct nfsd4_free_stateid
*free_stateid
)
3545 stateid_t
*stateid
= &free_stateid
->fr_stateid
;
3546 struct nfs4_stid
*s
;
3547 struct nfs4_client
*cl
= cstate
->session
->se_client
;
3548 __be32 ret
= nfserr_bad_stateid
;
3551 s
= find_stateid(cl
, stateid
);
3554 switch (s
->sc_type
) {
3555 case NFS4_DELEG_STID
:
3556 ret
= nfserr_locks_held
;
3558 case NFS4_OPEN_STID
:
3559 case NFS4_LOCK_STID
:
3560 ret
= check_stateid_generation(stateid
, &s
->sc_stateid
, 1);
3563 if (s
->sc_type
== NFS4_LOCK_STID
)
3564 ret
= nfsd4_free_lock_stateid(openlockstateid(s
));
3566 ret
= nfserr_locks_held
;
3569 ret
= nfserr_bad_stateid
;
3572 nfs4_unlock_state();
3579 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
3580 RD_STATE
: WR_STATE
;
3583 static __be32
nfs4_seqid_op_checks(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
, u32 seqid
, struct nfs4_ol_stateid
*stp
)
3585 struct svc_fh
*current_fh
= &cstate
->current_fh
;
3586 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
3589 status
= nfsd4_check_seqid(cstate
, sop
, seqid
);
3592 if (stp
->st_stid
.sc_type
== NFS4_CLOSED_STID
)
3594 * "Closed" stateid's exist *only* to return
3595 * nfserr_replay_me from the previous step.
3597 return nfserr_bad_stateid
;
3598 status
= check_stateid_generation(stateid
, &stp
->st_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3601 return nfs4_check_fh(current_fh
, stp
);
3605 * Checks for sequence id mutating operations.
3608 nfs4_preprocess_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
,
3609 stateid_t
*stateid
, char typemask
,
3610 struct nfs4_ol_stateid
**stpp
)
3613 struct nfs4_stid
*s
;
3615 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT
"\n", __func__
,
3616 seqid
, STATEID_VAL(stateid
));
3619 status
= nfsd4_lookup_stateid(stateid
, typemask
, &s
);
3622 *stpp
= openlockstateid(s
);
3623 cstate
->replay_owner
= (*stpp
)->st_stateowner
;
3625 return nfs4_seqid_op_checks(cstate
, stateid
, seqid
, *stpp
);
3628 static __be32
nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state
*cstate
, u32 seqid
, stateid_t
*stateid
, struct nfs4_ol_stateid
**stpp
)
3631 struct nfs4_openowner
*oo
;
3633 status
= nfs4_preprocess_seqid_op(cstate
, seqid
, stateid
,
3634 NFS4_OPEN_STID
, stpp
);
3637 oo
= openowner((*stpp
)->st_stateowner
);
3638 if (!(oo
->oo_flags
& NFS4_OO_CONFIRMED
))
3639 return nfserr_bad_stateid
;
3644 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3645 struct nfsd4_open_confirm
*oc
)
3648 struct nfs4_openowner
*oo
;
3649 struct nfs4_ol_stateid
*stp
;
3651 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3652 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3653 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3655 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
3661 status
= nfs4_preprocess_seqid_op(cstate
,
3662 oc
->oc_seqid
, &oc
->oc_req_stateid
,
3663 NFS4_OPEN_STID
, &stp
);
3666 oo
= openowner(stp
->st_stateowner
);
3667 status
= nfserr_bad_stateid
;
3668 if (oo
->oo_flags
& NFS4_OO_CONFIRMED
)
3670 oo
->oo_flags
|= NFS4_OO_CONFIRMED
;
3671 update_stateid(&stp
->st_stid
.sc_stateid
);
3672 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3673 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT
"\n",
3674 __func__
, oc
->oc_seqid
, STATEID_VAL(&stp
->st_stid
.sc_stateid
));
3676 nfsd4_client_record_create(oo
->oo_owner
.so_client
);
3679 if (!cstate
->replay_owner
)
3680 nfs4_unlock_state();
3684 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid
*stp
, u32 access
)
3686 if (!test_access(access
, stp
))
3688 nfs4_file_put_access(stp
->st_file
, nfs4_access_to_omode(access
));
3689 clear_access(access
, stp
);
3692 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid
*stp
, u32 to_access
)
3694 switch (to_access
) {
3695 case NFS4_SHARE_ACCESS_READ
:
3696 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_WRITE
);
3697 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3699 case NFS4_SHARE_ACCESS_WRITE
:
3700 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_READ
);
3701 nfs4_stateid_downgrade_bit(stp
, NFS4_SHARE_ACCESS_BOTH
);
3703 case NFS4_SHARE_ACCESS_BOTH
:
3711 reset_union_bmap_deny(unsigned long deny
, struct nfs4_ol_stateid
*stp
)
3714 for (i
= 0; i
< 4; i
++) {
3715 if ((i
& deny
) != i
)
3721 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
3722 struct nfsd4_compound_state
*cstate
,
3723 struct nfsd4_open_downgrade
*od
)
3726 struct nfs4_ol_stateid
*stp
;
3728 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3729 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3730 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3732 /* We don't yet support WANT bits: */
3733 if (od
->od_deleg_want
)
3734 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__
,
3738 status
= nfs4_preprocess_confirmed_seqid_op(cstate
, od
->od_seqid
,
3739 &od
->od_stateid
, &stp
);
3742 status
= nfserr_inval
;
3743 if (!test_access(od
->od_share_access
, stp
)) {
3744 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3745 stp
->st_access_bmap
, od
->od_share_access
);
3748 if (!test_deny(od
->od_share_deny
, stp
)) {
3749 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3750 stp
->st_deny_bmap
, od
->od_share_deny
);
3753 nfs4_stateid_downgrade(stp
, od
->od_share_access
);
3755 reset_union_bmap_deny(od
->od_share_deny
, stp
);
3757 update_stateid(&stp
->st_stid
.sc_stateid
);
3758 memcpy(&od
->od_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3761 if (!cstate
->replay_owner
)
3762 nfs4_unlock_state();
3766 void nfsd4_purge_closed_stateid(struct nfs4_stateowner
*so
)
3768 struct nfs4_openowner
*oo
;
3769 struct nfs4_ol_stateid
*s
;
3771 if (!so
->so_is_open_owner
)
3774 s
= oo
->oo_last_closed_stid
;
3777 if (!(oo
->oo_flags
& NFS4_OO_PURGE_CLOSE
)) {
3778 /* Release the last_closed_stid on the next seqid bump: */
3779 oo
->oo_flags
|= NFS4_OO_PURGE_CLOSE
;
3782 oo
->oo_flags
&= ~NFS4_OO_PURGE_CLOSE
;
3783 release_last_closed_stateid(oo
);
3786 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid
*s
)
3788 unhash_open_stateid(s
);
3789 s
->st_stid
.sc_type
= NFS4_CLOSED_STID
;
3793 * nfs4_unlock_state() called after encode
3796 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3797 struct nfsd4_close
*close
)
3800 struct nfs4_openowner
*oo
;
3801 struct nfs4_ol_stateid
*stp
;
3803 dprintk("NFSD: nfsd4_close on file %.*s\n",
3804 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
3805 cstate
->current_fh
.fh_dentry
->d_name
.name
);
3808 status
= nfs4_preprocess_seqid_op(cstate
, close
->cl_seqid
,
3810 NFS4_OPEN_STID
|NFS4_CLOSED_STID
,
3814 oo
= openowner(stp
->st_stateowner
);
3816 update_stateid(&stp
->st_stid
.sc_stateid
);
3817 memcpy(&close
->cl_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
3819 nfsd4_close_open_stateid(stp
);
3820 oo
->oo_last_closed_stid
= stp
;
3822 /* place unused nfs4_stateowners on so_close_lru list to be
3823 * released by the laundromat service after the lease period
3824 * to enable us to handle CLOSE replay
3826 if (list_empty(&oo
->oo_owner
.so_stateids
))
3827 move_to_close_lru(oo
);
3829 if (!cstate
->replay_owner
)
3830 nfs4_unlock_state();
3835 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
3836 struct nfsd4_delegreturn
*dr
)
3838 struct nfs4_delegation
*dp
;
3839 stateid_t
*stateid
= &dr
->dr_stateid
;
3840 struct nfs4_stid
*s
;
3841 struct inode
*inode
;
3844 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
3846 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
3849 status
= nfsd4_lookup_stateid(stateid
, NFS4_DELEG_STID
, &s
);
3852 dp
= delegstateid(s
);
3853 status
= check_stateid_generation(stateid
, &dp
->dl_stid
.sc_stateid
, nfsd4_has_session(cstate
));
3857 unhash_delegation(dp
);
3859 nfs4_unlock_state();
3865 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3867 #define LOCKOWNER_INO_HASH_BITS 8
3868 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3869 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3872 end_offset(u64 start
, u64 len
)
3877 return end
>= start
? end
: NFS4_MAX_UINT64
;
3880 /* last octet in a range */
3882 last_byte_offset(u64 start
, u64 len
)
3888 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
3891 static unsigned int lockowner_ino_hashval(struct inode
*inode
, u32 cl_id
, struct xdr_netobj
*ownername
)
3893 return (file_hashval(inode
) + cl_id
3894 + opaque_hashval(ownername
->data
, ownername
->len
))
3895 & LOCKOWNER_INO_HASH_MASK
;
3898 static struct list_head lockowner_ino_hashtbl
[LOCKOWNER_INO_HASH_SIZE
];
3901 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3902 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3903 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3904 * locking, this prevents us from being completely protocol-compliant. The
3905 * real solution to this problem is to start using unsigned file offsets in
3906 * the VFS, but this is a very deep change!
3909 nfs4_transform_lock_offset(struct file_lock
*lock
)
3911 if (lock
->fl_start
< 0)
3912 lock
->fl_start
= OFFSET_MAX
;
3913 if (lock
->fl_end
< 0)
3914 lock
->fl_end
= OFFSET_MAX
;
3917 /* Hack!: For now, we're defining this just so we can use a pointer to it
3918 * as a unique cookie to identify our (NFSv4's) posix locks. */
3919 static const struct lock_manager_operations nfsd_posix_mng_ops
= {
3923 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
3925 struct nfs4_lockowner
*lo
;
3927 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
3928 lo
= (struct nfs4_lockowner
*) fl
->fl_owner
;
3929 deny
->ld_owner
.data
= kmemdup(lo
->lo_owner
.so_owner
.data
,
3930 lo
->lo_owner
.so_owner
.len
, GFP_KERNEL
);
3931 if (!deny
->ld_owner
.data
)
3932 /* We just don't care that much */
3934 deny
->ld_owner
.len
= lo
->lo_owner
.so_owner
.len
;
3935 deny
->ld_clientid
= lo
->lo_owner
.so_client
->cl_clientid
;
3938 deny
->ld_owner
.len
= 0;
3939 deny
->ld_owner
.data
= NULL
;
3940 deny
->ld_clientid
.cl_boot
= 0;
3941 deny
->ld_clientid
.cl_id
= 0;
3943 deny
->ld_start
= fl
->fl_start
;
3944 deny
->ld_length
= NFS4_MAX_UINT64
;
3945 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
3946 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
3947 deny
->ld_type
= NFS4_READ_LT
;
3948 if (fl
->fl_type
!= F_RDLCK
)
3949 deny
->ld_type
= NFS4_WRITE_LT
;
3952 static bool same_lockowner_ino(struct nfs4_lockowner
*lo
, struct inode
*inode
, clientid_t
*clid
, struct xdr_netobj
*owner
)
3954 struct nfs4_ol_stateid
*lst
;
3956 if (!same_owner_str(&lo
->lo_owner
, owner
, clid
))
3958 lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
3959 struct nfs4_ol_stateid
, st_perstateowner
);
3960 return lst
->st_file
->fi_inode
== inode
;
3963 static struct nfs4_lockowner
*
3964 find_lockowner_str(struct inode
*inode
, clientid_t
*clid
,
3965 struct xdr_netobj
*owner
)
3967 unsigned int hashval
= lockowner_ino_hashval(inode
, clid
->cl_id
, owner
);
3968 struct nfs4_lockowner
*lo
;
3970 list_for_each_entry(lo
, &lockowner_ino_hashtbl
[hashval
], lo_owner_ino_hash
) {
3971 if (same_lockowner_ino(lo
, inode
, clid
, owner
))
3977 static void hash_lockowner(struct nfs4_lockowner
*lo
, unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
)
3979 struct inode
*inode
= open_stp
->st_file
->fi_inode
;
3980 unsigned int inohash
= lockowner_ino_hashval(inode
,
3981 clp
->cl_clientid
.cl_id
, &lo
->lo_owner
.so_owner
);
3983 list_add(&lo
->lo_owner
.so_strhash
, &ownerstr_hashtbl
[strhashval
]);
3984 list_add(&lo
->lo_owner_ino_hash
, &lockowner_ino_hashtbl
[inohash
]);
3985 list_add(&lo
->lo_perstateid
, &open_stp
->st_lockowners
);
3989 * Alloc a lock owner structure.
3990 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3993 * strhashval = ownerstr_hashval
3996 static struct nfs4_lockowner
*
3997 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_ol_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
3998 struct nfs4_lockowner
*lo
;
4000 lo
= alloc_stateowner(lockowner_slab
, &lock
->lk_new_owner
, clp
);
4003 INIT_LIST_HEAD(&lo
->lo_owner
.so_stateids
);
4004 lo
->lo_owner
.so_is_open_owner
= 0;
4005 /* It is the openowner seqid that will be incremented in encode in the
4006 * case of new lockowners; so increment the lock seqid manually: */
4007 lo
->lo_owner
.so_seqid
= lock
->lk_new_lock_seqid
+ 1;
4008 hash_lockowner(lo
, strhashval
, clp
, open_stp
);
4012 static struct nfs4_ol_stateid
*
4013 alloc_init_lock_stateid(struct nfs4_lockowner
*lo
, struct nfs4_file
*fp
, struct nfs4_ol_stateid
*open_stp
)
4015 struct nfs4_ol_stateid
*stp
;
4016 struct nfs4_client
*clp
= lo
->lo_owner
.so_client
;
4018 stp
= nfs4_alloc_stateid(clp
);
4021 init_stid(&stp
->st_stid
, clp
, NFS4_LOCK_STID
);
4022 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
4023 list_add(&stp
->st_perstateowner
, &lo
->lo_owner
.so_stateids
);
4024 stp
->st_stateowner
= &lo
->lo_owner
;
4027 stp
->st_access_bmap
= 0;
4028 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
4029 stp
->st_openstp
= open_stp
;
4034 check_lock_length(u64 offset
, u64 length
)
4036 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
4037 LOFF_OVERFLOW(offset
, length
)));
4040 static void get_lock_access(struct nfs4_ol_stateid
*lock_stp
, u32 access
)
4042 struct nfs4_file
*fp
= lock_stp
->st_file
;
4043 int oflag
= nfs4_access_to_omode(access
);
4045 if (test_access(access
, lock_stp
))
4047 nfs4_file_get_access(fp
, oflag
);
4048 set_access(access
, lock_stp
);
4051 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)
4053 struct nfs4_file
*fi
= ost
->st_file
;
4054 struct nfs4_openowner
*oo
= openowner(ost
->st_stateowner
);
4055 struct nfs4_client
*cl
= oo
->oo_owner
.so_client
;
4056 struct nfs4_lockowner
*lo
;
4057 unsigned int strhashval
;
4059 lo
= find_lockowner_str(fi
->fi_inode
, &cl
->cl_clientid
, &lock
->v
.new.owner
);
4061 if (!cstate
->minorversion
)
4062 return nfserr_bad_seqid
;
4063 /* XXX: a lockowner always has exactly one stateid: */
4064 *lst
= list_first_entry(&lo
->lo_owner
.so_stateids
,
4065 struct nfs4_ol_stateid
, st_perstateowner
);
4068 strhashval
= ownerstr_hashval(cl
->cl_clientid
.cl_id
,
4069 &lock
->v
.new.owner
);
4070 lo
= alloc_init_lock_stateowner(strhashval
, cl
, ost
, lock
);
4072 return nfserr_jukebox
;
4073 *lst
= alloc_init_lock_stateid(lo
, fi
, ost
);
4075 release_lockowner(lo
);
4076 return nfserr_jukebox
;
4086 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4087 struct nfsd4_lock
*lock
)
4089 struct nfs4_openowner
*open_sop
= NULL
;
4090 struct nfs4_lockowner
*lock_sop
= NULL
;
4091 struct nfs4_ol_stateid
*lock_stp
;
4092 struct file
*filp
= NULL
;
4093 struct file_lock file_lock
;
4094 struct file_lock conflock
;
4096 bool new_state
= false;
4100 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4101 (long long) lock
->lk_offset
,
4102 (long long) lock
->lk_length
);
4104 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
4105 return nfserr_inval
;
4107 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
4108 S_IFREG
, NFSD_MAY_LOCK
))) {
4109 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4115 if (lock
->lk_is_new
) {
4117 * Client indicates that this is a new lockowner.
4118 * Use open owner and open stateid to create lock owner and
4121 struct nfs4_ol_stateid
*open_stp
= NULL
;
4123 if (nfsd4_has_session(cstate
))
4124 /* See rfc 5661 18.10.3: given clientid is ignored: */
4125 memcpy(&lock
->v
.new.clientid
,
4126 &cstate
->session
->se_client
->cl_clientid
,
4127 sizeof(clientid_t
));
4129 status
= nfserr_stale_clientid
;
4130 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
4133 /* validate and update open stateid and open seqid */
4134 status
= nfs4_preprocess_confirmed_seqid_op(cstate
,
4135 lock
->lk_new_open_seqid
,
4136 &lock
->lk_new_open_stateid
,
4140 open_sop
= openowner(open_stp
->st_stateowner
);
4141 status
= nfserr_bad_stateid
;
4142 if (!same_clid(&open_sop
->oo_owner
.so_client
->cl_clientid
,
4143 &lock
->v
.new.clientid
))
4145 status
= lookup_or_create_lock_state(cstate
, open_stp
, lock
,
4146 &lock_stp
, &new_state
);
4150 /* lock (lock owner + lock stateid) already exists */
4151 status
= nfs4_preprocess_seqid_op(cstate
,
4152 lock
->lk_old_lock_seqid
,
4153 &lock
->lk_old_lock_stateid
,
4154 NFS4_LOCK_STID
, &lock_stp
);
4158 lock_sop
= lockowner(lock_stp
->st_stateowner
);
4160 lkflg
= setlkflg(lock
->lk_type
);
4161 status
= nfs4_check_openmode(lock_stp
, lkflg
);
4165 status
= nfserr_grace
;
4166 if (locks_in_grace() && !lock
->lk_reclaim
)
4168 status
= nfserr_no_grace
;
4169 if (!locks_in_grace() && lock
->lk_reclaim
)
4172 locks_init_lock(&file_lock
);
4173 switch (lock
->lk_type
) {
4176 filp
= find_readable_file(lock_stp
->st_file
);
4178 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_READ
);
4179 file_lock
.fl_type
= F_RDLCK
;
4182 case NFS4_WRITEW_LT
:
4183 filp
= find_writeable_file(lock_stp
->st_file
);
4185 get_lock_access(lock_stp
, NFS4_SHARE_ACCESS_WRITE
);
4186 file_lock
.fl_type
= F_WRLCK
;
4189 status
= nfserr_inval
;
4193 status
= nfserr_openmode
;
4196 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
4197 file_lock
.fl_pid
= current
->tgid
;
4198 file_lock
.fl_file
= filp
;
4199 file_lock
.fl_flags
= FL_POSIX
;
4200 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4202 file_lock
.fl_start
= lock
->lk_offset
;
4203 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
4204 nfs4_transform_lock_offset(&file_lock
);
4207 * Try to lock the file in the VFS.
4208 * Note: locks.c uses the BKL to protect the inode's lock list.
4211 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, &conflock
);
4213 case 0: /* success! */
4214 update_stateid(&lock_stp
->st_stid
.sc_stateid
);
4215 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stid
.sc_stateid
,
4219 case (EAGAIN
): /* conflock holds conflicting lock */
4220 status
= nfserr_denied
;
4221 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4222 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
4225 status
= nfserr_deadlock
;
4228 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
4229 status
= nfserrno(err
);
4233 if (status
&& new_state
)
4234 release_lockowner(lock_sop
);
4235 if (!cstate
->replay_owner
)
4236 nfs4_unlock_state();
4241 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4242 * so we do a temporary open here just to get an open file to pass to
4243 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4246 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
4251 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
4254 err
= vfs_test_lock(file
, lock
);
4263 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4264 struct nfsd4_lockt
*lockt
)
4266 struct inode
*inode
;
4267 struct file_lock file_lock
;
4268 struct nfs4_lockowner
*lo
;
4272 if (locks_in_grace())
4273 return nfserr_grace
;
4275 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
4276 return nfserr_inval
;
4280 status
= nfserr_stale_clientid
;
4281 if (!nfsd4_has_session(cstate
) && STALE_CLIENTID(&lockt
->lt_clientid
))
4284 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
4287 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
4288 locks_init_lock(&file_lock
);
4289 switch (lockt
->lt_type
) {
4292 file_lock
.fl_type
= F_RDLCK
;
4295 case NFS4_WRITEW_LT
:
4296 file_lock
.fl_type
= F_WRLCK
;
4299 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4300 status
= nfserr_inval
;
4304 lo
= find_lockowner_str(inode
, &lockt
->lt_clientid
, &lockt
->lt_owner
);
4306 file_lock
.fl_owner
= (fl_owner_t
)lo
;
4307 file_lock
.fl_pid
= current
->tgid
;
4308 file_lock
.fl_flags
= FL_POSIX
;
4310 file_lock
.fl_start
= lockt
->lt_offset
;
4311 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
4313 nfs4_transform_lock_offset(&file_lock
);
4316 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
4318 status
= nfserrno(error
);
4321 if (file_lock
.fl_type
!= F_UNLCK
) {
4322 status
= nfserr_denied
;
4323 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
4326 nfs4_unlock_state();
4331 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
4332 struct nfsd4_locku
*locku
)
4334 struct nfs4_ol_stateid
*stp
;
4335 struct file
*filp
= NULL
;
4336 struct file_lock file_lock
;
4340 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4341 (long long) locku
->lu_offset
,
4342 (long long) locku
->lu_length
);
4344 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
4345 return nfserr_inval
;
4349 status
= nfs4_preprocess_seqid_op(cstate
, locku
->lu_seqid
,
4350 &locku
->lu_stateid
, NFS4_LOCK_STID
, &stp
);
4353 filp
= find_any_file(stp
->st_file
);
4355 status
= nfserr_lock_range
;
4359 locks_init_lock(&file_lock
);
4360 file_lock
.fl_type
= F_UNLCK
;
4361 file_lock
.fl_owner
= (fl_owner_t
)lockowner(stp
->st_stateowner
);
4362 file_lock
.fl_pid
= current
->tgid
;
4363 file_lock
.fl_file
= filp
;
4364 file_lock
.fl_flags
= FL_POSIX
;
4365 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
4366 file_lock
.fl_start
= locku
->lu_offset
;
4368 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
4369 nfs4_transform_lock_offset(&file_lock
);
4372 * Try to unlock the file in the VFS.
4374 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
4376 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4380 * OK, unlock succeeded; the only thing left to do is update the stateid.
4382 update_stateid(&stp
->st_stid
.sc_stateid
);
4383 memcpy(&locku
->lu_stateid
, &stp
->st_stid
.sc_stateid
, sizeof(stateid_t
));
4386 if (!cstate
->replay_owner
)
4387 nfs4_unlock_state();
4391 status
= nfserrno(err
);
4397 * 1: locks held by lockowner
4398 * 0: no locks held by lockowner
4401 check_for_locks(struct nfs4_file
*filp
, struct nfs4_lockowner
*lowner
)
4403 struct file_lock
**flpp
;
4404 struct inode
*inode
= filp
->fi_inode
;
4408 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
4409 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
4420 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
4421 struct nfsd4_compound_state
*cstate
,
4422 struct nfsd4_release_lockowner
*rlockowner
)
4424 clientid_t
*clid
= &rlockowner
->rl_clientid
;
4425 struct nfs4_stateowner
*sop
;
4426 struct nfs4_lockowner
*lo
;
4427 struct nfs4_ol_stateid
*stp
;
4428 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
4429 struct list_head matches
;
4430 unsigned int hashval
= ownerstr_hashval(clid
->cl_id
, owner
);
4433 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4434 clid
->cl_boot
, clid
->cl_id
);
4436 /* XXX check for lease expiration */
4438 status
= nfserr_stale_clientid
;
4439 if (STALE_CLIENTID(clid
))
4444 status
= nfserr_locks_held
;
4445 INIT_LIST_HEAD(&matches
);
4447 list_for_each_entry(sop
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
4448 if (sop
->so_is_open_owner
)
4450 if (!same_owner_str(sop
, owner
, clid
))
4452 list_for_each_entry(stp
, &sop
->so_stateids
,
4454 lo
= lockowner(sop
);
4455 if (check_for_locks(stp
->st_file
, lo
))
4457 list_add(&lo
->lo_list
, &matches
);
4460 /* Clients probably won't expect us to return with some (but not all)
4461 * of the lockowner state released; so don't release any until all
4462 * have been checked. */
4464 while (!list_empty(&matches
)) {
4465 lo
= list_entry(matches
.next
, struct nfs4_lockowner
,
4467 /* unhash_stateowner deletes so_perclient only
4468 * for openowners. */
4469 list_del(&lo
->lo_list
);
4470 release_lockowner(lo
);
4473 nfs4_unlock_state();
4477 static inline struct nfs4_client_reclaim
*
4480 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
4484 nfs4_has_reclaimed_state(const char *name
, bool use_exchange_id
)
4486 unsigned int strhashval
= clientstr_hashval(name
);
4487 struct nfs4_client
*clp
;
4489 clp
= find_confirmed_client_by_str(name
, strhashval
);
4492 return test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
4496 * failure => all reset bets are off, nfserr_no_grace...
4499 nfs4_client_to_reclaim(const char *name
)
4501 unsigned int strhashval
;
4502 struct nfs4_client_reclaim
*crp
= NULL
;
4504 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
4505 crp
= alloc_reclaim();
4508 strhashval
= clientstr_hashval(name
);
4509 INIT_LIST_HEAD(&crp
->cr_strhash
);
4510 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
4511 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
4512 reclaim_str_hashtbl_size
++;
4517 nfs4_release_reclaim(void)
4519 struct nfs4_client_reclaim
*crp
= NULL
;
4522 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4523 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
4524 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
4525 struct nfs4_client_reclaim
, cr_strhash
);
4526 list_del(&crp
->cr_strhash
);
4528 reclaim_str_hashtbl_size
--;
4531 BUG_ON(reclaim_str_hashtbl_size
);
4535 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4536 struct nfs4_client_reclaim
*
4537 nfsd4_find_reclaim_client(struct nfs4_client
*clp
)
4539 unsigned int strhashval
;
4540 struct nfs4_client_reclaim
*crp
= NULL
;
4542 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4543 clp
->cl_name
.len
, clp
->cl_name
.data
,
4546 /* find clp->cl_name in reclaim_str_hashtbl */
4547 strhashval
= clientstr_hashval(clp
->cl_recdir
);
4548 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
4549 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
4557 * Called from OPEN. Look for clientid in reclaim list.
4560 nfs4_check_open_reclaim(clientid_t
*clid
)
4562 struct nfs4_client
*clp
;
4564 /* find clientid in conf_id_hashtbl */
4565 clp
= find_confirmed_client(clid
);
4567 return nfserr_reclaim_bad
;
4569 return nfsd4_client_record_check(clp
) ? nfserr_reclaim_bad
: nfs_ok
;
4572 #ifdef CONFIG_NFSD_FAULT_INJECTION
4574 void nfsd_forget_clients(u64 num
)
4576 struct nfs4_client
*clp
, *next
;
4580 list_for_each_entry_safe(clp
, next
, &client_lru
, cl_lru
) {
4581 nfsd4_client_record_remove(clp
);
4586 nfs4_unlock_state();
4588 printk(KERN_INFO
"NFSD: Forgot %d clients", count
);
4591 static void release_lockowner_sop(struct nfs4_stateowner
*sop
)
4593 release_lockowner(lockowner(sop
));
4596 static void release_openowner_sop(struct nfs4_stateowner
*sop
)
4598 release_openowner(openowner(sop
));
4601 static int nfsd_release_n_owners(u64 num
, bool is_open_owner
,
4602 void (*release_sop
)(struct nfs4_stateowner
*))
4605 struct nfs4_stateowner
*sop
, *next
;
4607 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4608 list_for_each_entry_safe(sop
, next
, &ownerstr_hashtbl
[i
], so_strhash
) {
4609 if (sop
->so_is_open_owner
!= is_open_owner
)
4619 void nfsd_forget_locks(u64 num
)
4624 count
= nfsd_release_n_owners(num
, false, release_lockowner_sop
);
4625 nfs4_unlock_state();
4627 printk(KERN_INFO
"NFSD: Forgot %d locks", count
);
4630 void nfsd_forget_openowners(u64 num
)
4635 count
= nfsd_release_n_owners(num
, true, release_openowner_sop
);
4636 nfs4_unlock_state();
4638 printk(KERN_INFO
"NFSD: Forgot %d open owners", count
);
4641 int nfsd_process_n_delegations(u64 num
, void (*deleg_func
)(struct nfs4_delegation
*))
4644 struct nfs4_file
*fp
, *fnext
;
4645 struct nfs4_delegation
*dp
, *dnext
;
4647 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4648 list_for_each_entry_safe(fp
, fnext
, &file_hashtbl
[i
], fi_hash
) {
4649 list_for_each_entry_safe(dp
, dnext
, &fp
->fi_delegations
, dl_perfile
) {
4660 void nfsd_forget_delegations(u64 num
)
4665 count
= nfsd_process_n_delegations(num
, unhash_delegation
);
4666 nfs4_unlock_state();
4668 printk(KERN_INFO
"NFSD: Forgot %d delegations", count
);
4671 void nfsd_recall_delegations(u64 num
)
4676 spin_lock(&recall_lock
);
4677 count
= nfsd_process_n_delegations(num
, nfsd_break_one_deleg
);
4678 spin_unlock(&recall_lock
);
4679 nfs4_unlock_state();
4681 printk(KERN_INFO
"NFSD: Recalled %d delegations", count
);
4684 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4686 /* initialization to perform at module load time: */
4689 nfs4_state_init(void)
4693 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4694 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
4695 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
4696 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
4697 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
4698 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
4700 for (i
= 0; i
< SESSION_HASH_SIZE
; i
++)
4701 INIT_LIST_HEAD(&sessionid_hashtbl
[i
]);
4702 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
4703 INIT_LIST_HEAD(&file_hashtbl
[i
]);
4705 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
4706 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
4708 for (i
= 0; i
< LOCKOWNER_INO_HASH_SIZE
; i
++)
4709 INIT_LIST_HEAD(&lockowner_ino_hashtbl
[i
]);
4710 INIT_LIST_HEAD(&close_lru
);
4711 INIT_LIST_HEAD(&client_lru
);
4712 INIT_LIST_HEAD(&del_recall_lru
);
4713 reclaim_str_hashtbl_size
= 0;
4717 * Since the lifetime of a delegation isn't limited to that of an open, a
4718 * client may quite reasonably hang on to a delegation as long as it has
4719 * the inode cached. This becomes an obvious problem the first time a
4720 * client's inode cache approaches the size of the server's total memory.
4722 * For now we avoid this problem by imposing a hard limit on the number
4723 * of delegations, which varies according to the server's memory size.
4726 set_max_delegations(void)
4729 * Allow at most 4 delegations per megabyte of RAM. Quick
4730 * estimates suggest that in the worst case (where every delegation
4731 * is for a different inode), a delegation could take about 1.5K,
4732 * giving a worst case usage of about 6% of memory.
4734 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
4737 /* initialization to perform when the nfsd service is started: */
4740 nfs4_state_start(void)
4745 * FIXME: For now, we hang most of the pernet global stuff off of
4746 * init_net until nfsd is fully containerized. Eventually, we'll
4747 * need to pass a net pointer into this function, take a reference
4748 * to that instead and then do most of the rest of this on a per-net
4752 nfsd4_client_tracking_init(&init_net
);
4753 boot_time
= get_seconds();
4754 locks_start_grace(&nfsd4_manager
);
4755 grace_ended
= false;
4756 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
4758 ret
= set_callback_cred();
4763 laundry_wq
= create_singlethread_workqueue("nfsd4");
4764 if (laundry_wq
== NULL
) {
4768 ret
= nfsd4_create_callback_queue();
4770 goto out_free_laundry
;
4771 queue_delayed_work(laundry_wq
, &laundromat_work
, nfsd4_grace
* HZ
);
4772 set_max_delegations();
4775 destroy_workqueue(laundry_wq
);
4777 nfsd4_client_tracking_exit(&init_net
);
4783 __nfs4_state_shutdown(void)
4786 struct nfs4_client
*clp
= NULL
;
4787 struct nfs4_delegation
*dp
= NULL
;
4788 struct list_head
*pos
, *next
, reaplist
;
4790 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
4791 while (!list_empty(&conf_id_hashtbl
[i
])) {
4792 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
4795 while (!list_empty(&unconf_str_hashtbl
[i
])) {
4796 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
4800 INIT_LIST_HEAD(&reaplist
);
4801 spin_lock(&recall_lock
);
4802 list_for_each_safe(pos
, next
, &del_recall_lru
) {
4803 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4804 list_move(&dp
->dl_recall_lru
, &reaplist
);
4806 spin_unlock(&recall_lock
);
4807 list_for_each_safe(pos
, next
, &reaplist
) {
4808 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
4809 unhash_delegation(dp
);
4812 nfsd4_client_tracking_exit(&init_net
);
4817 nfs4_state_shutdown(void)
4819 cancel_delayed_work_sync(&laundromat_work
);
4820 destroy_workqueue(laundry_wq
);
4821 locks_end_grace(&nfsd4_manager
);
4823 __nfs4_state_shutdown();
4824 nfs4_unlock_state();
4825 nfsd4_destroy_callback_queue();
4829 get_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4831 if (HAS_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
) && CURRENT_STATEID(stateid
))
4832 memcpy(stateid
, &cstate
->current_stateid
, sizeof(stateid_t
));
4836 put_stateid(struct nfsd4_compound_state
*cstate
, stateid_t
*stateid
)
4838 if (cstate
->minorversion
) {
4839 memcpy(&cstate
->current_stateid
, stateid
, sizeof(stateid_t
));
4840 SET_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4845 clear_current_stateid(struct nfsd4_compound_state
*cstate
)
4847 CLEAR_STATE_ID(cstate
, CURRENT_STATE_ID_FLAG
);
4851 * functions to set current state id
4854 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4856 put_stateid(cstate
, &odp
->od_stateid
);
4860 nfsd4_set_openstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open
*open
)
4862 put_stateid(cstate
, &open
->op_stateid
);
4866 nfsd4_set_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4868 put_stateid(cstate
, &close
->cl_stateid
);
4872 nfsd4_set_lockstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_lock
*lock
)
4874 put_stateid(cstate
, &lock
->lk_resp_stateid
);
4878 * functions to consume current state id
4882 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_open_downgrade
*odp
)
4884 get_stateid(cstate
, &odp
->od_stateid
);
4888 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_delegreturn
*drp
)
4890 get_stateid(cstate
, &drp
->dr_stateid
);
4894 nfsd4_get_freestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_free_stateid
*fsp
)
4896 get_stateid(cstate
, &fsp
->fr_stateid
);
4900 nfsd4_get_setattrstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_setattr
*setattr
)
4902 get_stateid(cstate
, &setattr
->sa_stateid
);
4906 nfsd4_get_closestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_close
*close
)
4908 get_stateid(cstate
, &close
->cl_stateid
);
4912 nfsd4_get_lockustateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_locku
*locku
)
4914 get_stateid(cstate
, &locku
->lu_stateid
);
4918 nfsd4_get_readstateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_read
*read
)
4920 get_stateid(cstate
, &read
->rd_stateid
);
4924 nfsd4_get_writestateid(struct nfsd4_compound_state
*cstate
, struct nfsd4_write
*write
)
4926 get_stateid(cstate
, &write
->wr_stateid
);