2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/hash.h>
48 #include "current_stateid.h"
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
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(state_lock);
87 static struct kmem_cache *openowner_slab;
88 static struct kmem_cache *lockowner_slab;
89 static struct kmem_cache *file_slab;
90 static struct kmem_cache *stateid_slab;
91 static struct kmem_cache *deleg_slab;
96 mutex_lock(&client_mutex);
99 static void free_session(struct nfsd4_session *);
101 static bool is_session_dead(struct nfsd4_session *ses)
103 return ses->se_flags & NFS4_SESSION_DEAD;
106 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
108 if (atomic_read(&ses->se_ref) > ref_held_by_me)
109 return nfserr_jukebox;
110 ses->se_flags |= NFS4_SESSION_DEAD;
115 nfs4_unlock_state(void)
117 mutex_unlock(&client_mutex);
120 static bool is_client_expired(struct nfs4_client *clp)
122 return clp->cl_time == 0;
125 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
127 if (atomic_read(&clp->cl_refcount))
128 return nfserr_jukebox;
133 static __be32 mark_client_expired(struct nfs4_client *clp)
135 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
138 spin_lock(&nn->client_lock);
139 ret = mark_client_expired_locked(clp);
140 spin_unlock(&nn->client_lock);
144 static __be32 get_client_locked(struct nfs4_client *clp)
146 if (is_client_expired(clp))
147 return nfserr_expired;
148 atomic_inc(&clp->cl_refcount);
152 /* must be called under the client_lock */
154 renew_client_locked(struct nfs4_client *clp)
156 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
158 if (is_client_expired(clp)) {
160 printk("%s: client (clientid %08x/%08x) already expired\n",
162 clp->cl_clientid.cl_boot,
163 clp->cl_clientid.cl_id);
167 dprintk("renewing client (clientid %08x/%08x)\n",
168 clp->cl_clientid.cl_boot,
169 clp->cl_clientid.cl_id);
170 list_move_tail(&clp->cl_lru, &nn->client_lru);
171 clp->cl_time = get_seconds();
175 renew_client(struct nfs4_client *clp)
177 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
179 spin_lock(&nn->client_lock);
180 renew_client_locked(clp);
181 spin_unlock(&nn->client_lock);
184 static void put_client_renew_locked(struct nfs4_client *clp)
186 if (!atomic_dec_and_test(&clp->cl_refcount))
188 if (!is_client_expired(clp))
189 renew_client_locked(clp);
192 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
196 if (is_session_dead(ses))
197 return nfserr_badsession;
198 status = get_client_locked(ses->se_client);
201 atomic_inc(&ses->se_ref);
205 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
207 struct nfs4_client *clp = ses->se_client;
209 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
211 put_client_renew_locked(clp);
214 static void nfsd4_put_session(struct nfsd4_session *ses)
216 struct nfs4_client *clp = ses->se_client;
217 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
219 spin_lock(&nn->client_lock);
220 nfsd4_put_session_locked(ses);
221 spin_unlock(&nn->client_lock);
226 opaque_hashval(const void *ptr, int nbytes)
228 unsigned char *cptr = (unsigned char *) ptr;
238 static void nfsd4_free_file(struct nfs4_file *f)
240 kmem_cache_free(file_slab, f);
244 put_nfs4_file(struct nfs4_file *fi)
246 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
247 hlist_del(&fi->fi_hash);
248 spin_unlock(&state_lock);
255 get_nfs4_file(struct nfs4_file *fi)
257 atomic_inc(&fi->fi_ref);
260 static int num_delegations;
261 unsigned long max_delegations;
264 * Open owner state (share locks)
267 /* hash tables for lock and open owners */
268 #define OWNER_HASH_BITS 8
269 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
270 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
272 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
276 ret = opaque_hashval(ownername->data, ownername->len);
278 return ret & OWNER_HASH_MASK;
281 /* hash table for nfs4_file */
282 #define FILE_HASH_BITS 8
283 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
285 static unsigned int file_hashval(struct inode *ino)
287 /* XXX: why are we hashing on inode pointer, anyway? */
288 return hash_ptr(ino, FILE_HASH_BITS);
291 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
293 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
295 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
296 atomic_inc(&fp->fi_access[oflag]);
299 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
301 if (oflag == O_RDWR) {
302 __nfs4_file_get_access(fp, O_RDONLY);
303 __nfs4_file_get_access(fp, O_WRONLY);
305 __nfs4_file_get_access(fp, oflag);
308 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
310 if (fp->fi_fds[oflag]) {
311 fput(fp->fi_fds[oflag]);
312 fp->fi_fds[oflag] = NULL;
316 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
318 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
319 nfs4_file_put_fd(fp, oflag);
320 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
321 nfs4_file_put_fd(fp, O_RDWR);
325 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
327 if (oflag == O_RDWR) {
328 __nfs4_file_put_access(fp, O_RDONLY);
329 __nfs4_file_put_access(fp, O_WRONLY);
331 __nfs4_file_put_access(fp, oflag);
334 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
337 struct idr *stateids = &cl->cl_stateids;
338 struct nfs4_stid *stid;
341 stid = kmem_cache_alloc(slab, GFP_KERNEL);
345 new_id = idr_alloc_cyclic(stateids, stid, 0, 0, GFP_KERNEL);
348 stid->sc_client = cl;
350 stid->sc_stateid.si_opaque.so_id = new_id;
351 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
352 /* Will be incremented before return to client: */
353 stid->sc_stateid.si_generation = 0;
356 * It shouldn't be a problem to reuse an opaque stateid value.
357 * I don't think it is for 4.1. But with 4.0 I worry that, for
358 * example, a stray write retransmission could be accepted by
359 * the server when it should have been rejected. Therefore,
360 * adopt a trick from the sctp code to attempt to maximize the
361 * amount of time until an id is reused, by ensuring they always
362 * "increase" (mod INT_MAX):
366 kmem_cache_free(slab, stid);
370 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
372 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
376 * When we recall a delegation, we should be careful not to hand it
377 * out again straight away.
378 * To ensure this we keep a pair of bloom filters ('new' and 'old')
379 * in which the filehandles of recalled delegations are "stored".
380 * If a filehandle appear in either filter, a delegation is blocked.
381 * When a delegation is recalled, the filehandle is stored in the "new"
383 * Every 30 seconds we swap the filters and clear the "new" one,
384 * unless both are empty of course.
386 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
387 * low 3 bytes as hash-table indices.
389 * 'state_lock', which is always held when block_delegations() is called,
390 * is used to manage concurrent access. Testing does not need the lock
391 * except when swapping the two filters.
393 static struct bloom_pair {
394 int entries, old_entries;
396 int new; /* index into 'set' */
397 DECLARE_BITMAP(set[2], 256);
398 } blocked_delegations;
400 static int delegation_blocked(struct knfsd_fh *fh)
403 struct bloom_pair *bd = &blocked_delegations;
405 if (bd->entries == 0)
407 if (seconds_since_boot() - bd->swap_time > 30) {
408 spin_lock(&state_lock);
409 if (seconds_since_boot() - bd->swap_time > 30) {
410 bd->entries -= bd->old_entries;
411 bd->old_entries = bd->entries;
412 memset(bd->set[bd->new], 0,
415 bd->swap_time = seconds_since_boot();
417 spin_unlock(&state_lock);
419 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
420 if (test_bit(hash&255, bd->set[0]) &&
421 test_bit((hash>>8)&255, bd->set[0]) &&
422 test_bit((hash>>16)&255, bd->set[0]))
425 if (test_bit(hash&255, bd->set[1]) &&
426 test_bit((hash>>8)&255, bd->set[1]) &&
427 test_bit((hash>>16)&255, bd->set[1]))
433 static void block_delegations(struct knfsd_fh *fh)
436 struct bloom_pair *bd = &blocked_delegations;
438 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
440 __set_bit(hash&255, bd->set[bd->new]);
441 __set_bit((hash>>8)&255, bd->set[bd->new]);
442 __set_bit((hash>>16)&255, bd->set[bd->new]);
443 if (bd->entries == 0)
444 bd->swap_time = seconds_since_boot();
448 static struct nfs4_delegation *
449 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh)
451 struct nfs4_delegation *dp;
453 dprintk("NFSD alloc_init_deleg\n");
454 if (num_delegations > max_delegations)
456 if (delegation_blocked(¤t_fh->fh_handle))
458 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
462 * delegation seqid's are never incremented. The 4.1 special
463 * meaning of seqid 0 isn't meaningful, really, but let's avoid
464 * 0 anyway just for consistency and use 1:
466 dp->dl_stid.sc_stateid.si_generation = 1;
468 INIT_LIST_HEAD(&dp->dl_perfile);
469 INIT_LIST_HEAD(&dp->dl_perclnt);
470 INIT_LIST_HEAD(&dp->dl_recall_lru);
472 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
473 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
475 atomic_set(&dp->dl_count, 1);
476 nfsd4_init_callback(&dp->dl_recall);
480 static void remove_stid(struct nfs4_stid *s)
482 struct idr *stateids = &s->sc_client->cl_stateids;
484 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
487 static void nfs4_free_stid(struct kmem_cache *slab, struct nfs4_stid *s)
489 kmem_cache_free(slab, s);
493 nfs4_put_delegation(struct nfs4_delegation *dp)
495 if (atomic_dec_and_test(&dp->dl_count)) {
496 nfs4_free_stid(deleg_slab, &dp->dl_stid);
501 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
505 if (atomic_dec_and_test(&fp->fi_delegees)) {
506 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
508 fput(fp->fi_deleg_file);
509 fp->fi_deleg_file = NULL;
513 static void unhash_stid(struct nfs4_stid *s)
519 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
521 lockdep_assert_held(&state_lock);
523 dp->dl_stid.sc_type = NFS4_DELEG_STID;
524 list_add(&dp->dl_perfile, &fp->fi_delegations);
525 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
528 /* Called under the state lock. */
530 unhash_delegation(struct nfs4_delegation *dp)
532 spin_lock(&state_lock);
533 list_del_init(&dp->dl_perclnt);
534 list_del_init(&dp->dl_perfile);
535 list_del_init(&dp->dl_recall_lru);
536 spin_unlock(&state_lock);
538 nfs4_put_deleg_lease(dp->dl_file);
539 put_nfs4_file(dp->dl_file);
546 static void destroy_revoked_delegation(struct nfs4_delegation *dp)
548 list_del_init(&dp->dl_recall_lru);
549 remove_stid(&dp->dl_stid);
550 nfs4_put_delegation(dp);
553 static void destroy_delegation(struct nfs4_delegation *dp)
555 unhash_delegation(dp);
556 remove_stid(&dp->dl_stid);
557 nfs4_put_delegation(dp);
560 static void revoke_delegation(struct nfs4_delegation *dp)
562 struct nfs4_client *clp = dp->dl_stid.sc_client;
564 if (clp->cl_minorversion == 0)
565 destroy_delegation(dp);
567 unhash_delegation(dp);
568 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
569 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
577 static unsigned int clientid_hashval(u32 id)
579 return id & CLIENT_HASH_MASK;
582 static unsigned int clientstr_hashval(const char *name)
584 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
588 * We store the NONE, READ, WRITE, and BOTH bits separately in the
589 * st_{access,deny}_bmap field of the stateid, in order to track not
590 * only what share bits are currently in force, but also what
591 * combinations of share bits previous opens have used. This allows us
592 * to enforce the recommendation of rfc 3530 14.2.19 that the server
593 * return an error if the client attempt to downgrade to a combination
594 * of share bits not explicable by closing some of its previous opens.
596 * XXX: This enforcement is actually incomplete, since we don't keep
597 * track of access/deny bit combinations; so, e.g., we allow:
599 * OPEN allow read, deny write
600 * OPEN allow both, deny none
601 * DOWNGRADE allow read, deny none
603 * which we should reject.
606 bmap_to_share_mode(unsigned long bmap) {
608 unsigned int access = 0;
610 for (i = 1; i < 4; i++) {
611 if (test_bit(i, &bmap))
618 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
619 unsigned int access, deny;
621 access = bmap_to_share_mode(stp->st_access_bmap);
622 deny = bmap_to_share_mode(stp->st_deny_bmap);
623 if ((access & open->op_share_deny) || (deny & open->op_share_access))
628 /* set share access for a given stateid */
630 set_access(u32 access, struct nfs4_ol_stateid *stp)
632 __set_bit(access, &stp->st_access_bmap);
635 /* clear share access for a given stateid */
637 clear_access(u32 access, struct nfs4_ol_stateid *stp)
639 __clear_bit(access, &stp->st_access_bmap);
642 /* test whether a given stateid has access */
644 test_access(u32 access, struct nfs4_ol_stateid *stp)
646 return test_bit(access, &stp->st_access_bmap);
649 /* set share deny for a given stateid */
651 set_deny(u32 access, struct nfs4_ol_stateid *stp)
653 __set_bit(access, &stp->st_deny_bmap);
656 /* clear share deny for a given stateid */
658 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
660 __clear_bit(access, &stp->st_deny_bmap);
663 /* test whether a given stateid is denying specific access */
665 test_deny(u32 access, struct nfs4_ol_stateid *stp)
667 return test_bit(access, &stp->st_deny_bmap);
670 static int nfs4_access_to_omode(u32 access)
672 switch (access & NFS4_SHARE_ACCESS_BOTH) {
673 case NFS4_SHARE_ACCESS_READ:
675 case NFS4_SHARE_ACCESS_WRITE:
677 case NFS4_SHARE_ACCESS_BOTH:
684 /* release all access and file references for a given stateid */
686 release_all_access(struct nfs4_ol_stateid *stp)
690 for (i = 1; i < 4; i++) {
691 if (test_access(i, stp))
692 nfs4_file_put_access(stp->st_file,
693 nfs4_access_to_omode(i));
694 clear_access(i, stp);
698 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
700 list_del(&stp->st_perfile);
701 list_del(&stp->st_perstateowner);
704 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
706 release_all_access(stp);
707 put_nfs4_file(stp->st_file);
711 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
713 remove_stid(&stp->st_stid);
714 nfs4_free_stid(stateid_slab, &stp->st_stid);
717 static void __release_lock_stateid(struct nfs4_ol_stateid *stp)
721 list_del(&stp->st_locks);
722 unhash_generic_stateid(stp);
723 unhash_stid(&stp->st_stid);
724 file = find_any_file(stp->st_file);
726 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
727 close_generic_stateid(stp);
728 free_generic_stateid(stp);
731 static void unhash_lockowner(struct nfs4_lockowner *lo)
733 struct nfs4_ol_stateid *stp;
735 list_del(&lo->lo_owner.so_strhash);
736 while (!list_empty(&lo->lo_owner.so_stateids)) {
737 stp = list_first_entry(&lo->lo_owner.so_stateids,
738 struct nfs4_ol_stateid, st_perstateowner);
739 __release_lock_stateid(stp);
743 static void nfs4_free_lockowner(struct nfs4_lockowner *lo)
745 kfree(lo->lo_owner.so_owner.data);
746 kmem_cache_free(lockowner_slab, lo);
749 static void release_lockowner(struct nfs4_lockowner *lo)
751 unhash_lockowner(lo);
752 nfs4_free_lockowner(lo);
755 static void release_lockowner_if_empty(struct nfs4_lockowner *lo)
757 if (list_empty(&lo->lo_owner.so_stateids))
758 release_lockowner(lo);
761 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
763 struct nfs4_lockowner *lo;
765 lo = lockowner(stp->st_stateowner);
766 __release_lock_stateid(stp);
767 release_lockowner_if_empty(lo);
770 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp)
772 struct nfs4_ol_stateid *stp;
774 while (!list_empty(&open_stp->st_locks)) {
775 stp = list_entry(open_stp->st_locks.next,
776 struct nfs4_ol_stateid, st_locks);
777 release_lock_stateid(stp);
781 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
783 unhash_generic_stateid(stp);
784 release_open_stateid_locks(stp);
785 close_generic_stateid(stp);
788 static void release_open_stateid(struct nfs4_ol_stateid *stp)
790 unhash_open_stateid(stp);
791 free_generic_stateid(stp);
794 static void unhash_openowner(struct nfs4_openowner *oo)
796 struct nfs4_ol_stateid *stp;
798 list_del(&oo->oo_owner.so_strhash);
799 list_del(&oo->oo_perclient);
800 while (!list_empty(&oo->oo_owner.so_stateids)) {
801 stp = list_first_entry(&oo->oo_owner.so_stateids,
802 struct nfs4_ol_stateid, st_perstateowner);
803 release_open_stateid(stp);
807 static void release_last_closed_stateid(struct nfs4_openowner *oo)
809 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
812 free_generic_stateid(s);
813 oo->oo_last_closed_stid = NULL;
817 static void nfs4_free_openowner(struct nfs4_openowner *oo)
819 kfree(oo->oo_owner.so_owner.data);
820 kmem_cache_free(openowner_slab, oo);
823 static void release_openowner(struct nfs4_openowner *oo)
825 unhash_openowner(oo);
826 list_del(&oo->oo_close_lru);
827 release_last_closed_stateid(oo);
828 nfs4_free_openowner(oo);
832 hash_sessionid(struct nfs4_sessionid *sessionid)
834 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
836 return sid->sequence % SESSION_HASH_SIZE;
841 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
843 u32 *ptr = (u32 *)(&sessionid->data[0]);
844 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
848 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
854 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
855 * won't be used for replay.
857 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
859 struct nfs4_stateowner *so = cstate->replay_owner;
861 if (nfserr == nfserr_replay_me)
864 if (!seqid_mutating_err(ntohl(nfserr))) {
865 cstate->replay_owner = NULL;
870 if (so->so_is_open_owner)
871 release_last_closed_stateid(openowner(so));
877 gen_sessionid(struct nfsd4_session *ses)
879 struct nfs4_client *clp = ses->se_client;
880 struct nfsd4_sessionid *sid;
882 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
883 sid->clientid = clp->cl_clientid;
884 sid->sequence = current_sessionid++;
889 * The protocol defines ca_maxresponssize_cached to include the size of
890 * the rpc header, but all we need to cache is the data starting after
891 * the end of the initial SEQUENCE operation--the rest we regenerate
892 * each time. Therefore we can advertise a ca_maxresponssize_cached
893 * value that is the number of bytes in our cache plus a few additional
894 * bytes. In order to stay on the safe side, and not promise more than
895 * we can cache, those additional bytes must be the minimum possible: 24
896 * bytes of rpc header (xid through accept state, with AUTH_NULL
897 * verifier), 12 for the compound header (with zero-length tag), and 44
898 * for the SEQUENCE op response:
900 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
903 free_session_slots(struct nfsd4_session *ses)
907 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
908 kfree(ses->se_slots[i]);
912 * We don't actually need to cache the rpc and session headers, so we
913 * can allocate a little less for each slot:
915 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
919 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
922 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
923 return size + sizeof(struct nfsd4_slot);
927 * XXX: If we run out of reserved DRC memory we could (up to a point)
928 * re-negotiate active sessions and reduce their slot usage to make
929 * room for new connections. For now we just fail the create session.
931 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
933 u32 slotsize = slot_bytes(ca);
934 u32 num = ca->maxreqs;
937 spin_lock(&nfsd_drc_lock);
938 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
939 nfsd_drc_max_mem - nfsd_drc_mem_used);
940 num = min_t(int, num, avail / slotsize);
941 nfsd_drc_mem_used += num * slotsize;
942 spin_unlock(&nfsd_drc_lock);
947 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
949 int slotsize = slot_bytes(ca);
951 spin_lock(&nfsd_drc_lock);
952 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
953 spin_unlock(&nfsd_drc_lock);
956 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
957 struct nfsd4_channel_attrs *battrs)
959 int numslots = fattrs->maxreqs;
960 int slotsize = slot_bytes(fattrs);
961 struct nfsd4_session *new;
964 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
965 + sizeof(struct nfsd4_session) > PAGE_SIZE);
966 mem = numslots * sizeof(struct nfsd4_slot *);
968 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
971 /* allocate each struct nfsd4_slot and data cache in one piece */
972 for (i = 0; i < numslots; i++) {
973 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
974 if (!new->se_slots[i])
978 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
979 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
984 kfree(new->se_slots[i]);
989 static void free_conn(struct nfsd4_conn *c)
991 svc_xprt_put(c->cn_xprt);
995 static void nfsd4_conn_lost(struct svc_xpt_user *u)
997 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
998 struct nfs4_client *clp = c->cn_session->se_client;
1000 spin_lock(&clp->cl_lock);
1001 if (!list_empty(&c->cn_persession)) {
1002 list_del(&c->cn_persession);
1005 nfsd4_probe_callback(clp);
1006 spin_unlock(&clp->cl_lock);
1009 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1011 struct nfsd4_conn *conn;
1013 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1016 svc_xprt_get(rqstp->rq_xprt);
1017 conn->cn_xprt = rqstp->rq_xprt;
1018 conn->cn_flags = flags;
1019 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1023 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1025 conn->cn_session = ses;
1026 list_add(&conn->cn_persession, &ses->se_conns);
1029 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1031 struct nfs4_client *clp = ses->se_client;
1033 spin_lock(&clp->cl_lock);
1034 __nfsd4_hash_conn(conn, ses);
1035 spin_unlock(&clp->cl_lock);
1038 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1040 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1041 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1044 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1048 nfsd4_hash_conn(conn, ses);
1049 ret = nfsd4_register_conn(conn);
1051 /* oops; xprt is already down: */
1052 nfsd4_conn_lost(&conn->cn_xpt_user);
1053 if (conn->cn_flags & NFS4_CDFC4_BACK) {
1054 /* callback channel may be back up */
1055 nfsd4_probe_callback(ses->se_client);
1059 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1061 u32 dir = NFS4_CDFC4_FORE;
1063 if (cses->flags & SESSION4_BACK_CHAN)
1064 dir |= NFS4_CDFC4_BACK;
1065 return alloc_conn(rqstp, dir);
1068 /* must be called under client_lock */
1069 static void nfsd4_del_conns(struct nfsd4_session *s)
1071 struct nfs4_client *clp = s->se_client;
1072 struct nfsd4_conn *c;
1074 spin_lock(&clp->cl_lock);
1075 while (!list_empty(&s->se_conns)) {
1076 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1077 list_del_init(&c->cn_persession);
1078 spin_unlock(&clp->cl_lock);
1080 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1083 spin_lock(&clp->cl_lock);
1085 spin_unlock(&clp->cl_lock);
1088 static void __free_session(struct nfsd4_session *ses)
1090 free_session_slots(ses);
1094 static void free_session(struct nfsd4_session *ses)
1096 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
1098 lockdep_assert_held(&nn->client_lock);
1099 nfsd4_del_conns(ses);
1100 nfsd4_put_drc_mem(&ses->se_fchannel);
1101 __free_session(ses);
1104 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1107 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1109 new->se_client = clp;
1112 INIT_LIST_HEAD(&new->se_conns);
1114 new->se_cb_seq_nr = 1;
1115 new->se_flags = cses->flags;
1116 new->se_cb_prog = cses->callback_prog;
1117 new->se_cb_sec = cses->cb_sec;
1118 atomic_set(&new->se_ref, 0);
1119 idx = hash_sessionid(&new->se_sessionid);
1120 spin_lock(&nn->client_lock);
1121 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1122 spin_lock(&clp->cl_lock);
1123 list_add(&new->se_perclnt, &clp->cl_sessions);
1124 spin_unlock(&clp->cl_lock);
1125 spin_unlock(&nn->client_lock);
1127 if (cses->flags & SESSION4_BACK_CHAN) {
1128 struct sockaddr *sa = svc_addr(rqstp);
1130 * This is a little silly; with sessions there's no real
1131 * use for the callback address. Use the peer address
1132 * as a reasonable default for now, but consider fixing
1133 * the rpc client not to require an address in the
1136 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1137 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1141 /* caller must hold client_lock */
1142 static struct nfsd4_session *
1143 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1145 struct nfsd4_session *elem;
1147 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1149 dump_sessionid(__func__, sessionid);
1150 idx = hash_sessionid(sessionid);
1151 /* Search in the appropriate list */
1152 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1153 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1154 NFS4_MAX_SESSIONID_LEN)) {
1159 dprintk("%s: session not found\n", __func__);
1163 static struct nfsd4_session *
1164 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1167 struct nfsd4_session *session;
1168 __be32 status = nfserr_badsession;
1170 session = __find_in_sessionid_hashtbl(sessionid, net);
1173 status = nfsd4_get_session_locked(session);
1181 /* caller must hold client_lock */
1183 unhash_session(struct nfsd4_session *ses)
1185 list_del(&ses->se_hash);
1186 spin_lock(&ses->se_client->cl_lock);
1187 list_del(&ses->se_perclnt);
1188 spin_unlock(&ses->se_client->cl_lock);
1191 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1193 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1195 if (clid->cl_boot == nn->boot_time)
1197 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1198 clid->cl_boot, clid->cl_id, nn->boot_time);
1203 * XXX Should we use a slab cache ?
1204 * This type of memory management is somewhat inefficient, but we use it
1205 * anyway since SETCLIENTID is not a common operation.
1207 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1209 struct nfs4_client *clp;
1211 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1214 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1215 if (clp->cl_name.data == NULL) {
1219 clp->cl_name.len = name.len;
1220 INIT_LIST_HEAD(&clp->cl_sessions);
1221 idr_init(&clp->cl_stateids);
1222 atomic_set(&clp->cl_refcount, 0);
1223 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1224 INIT_LIST_HEAD(&clp->cl_idhash);
1225 INIT_LIST_HEAD(&clp->cl_openowners);
1226 INIT_LIST_HEAD(&clp->cl_delegations);
1227 INIT_LIST_HEAD(&clp->cl_lru);
1228 INIT_LIST_HEAD(&clp->cl_callbacks);
1229 INIT_LIST_HEAD(&clp->cl_revoked);
1230 spin_lock_init(&clp->cl_lock);
1231 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1236 free_client(struct nfs4_client *clp)
1238 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1240 lockdep_assert_held(&nn->client_lock);
1241 while (!list_empty(&clp->cl_sessions)) {
1242 struct nfsd4_session *ses;
1243 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1245 list_del(&ses->se_perclnt);
1246 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1249 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1250 free_svc_cred(&clp->cl_cred);
1251 kfree(clp->cl_name.data);
1252 idr_destroy(&clp->cl_stateids);
1256 /* must be called under the client_lock */
1258 unhash_client_locked(struct nfs4_client *clp)
1260 struct nfsd4_session *ses;
1262 list_del(&clp->cl_lru);
1263 spin_lock(&clp->cl_lock);
1264 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1265 list_del_init(&ses->se_hash);
1266 spin_unlock(&clp->cl_lock);
1270 destroy_client(struct nfs4_client *clp)
1272 struct nfs4_openowner *oo;
1273 struct nfs4_delegation *dp;
1274 struct list_head reaplist;
1275 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1277 INIT_LIST_HEAD(&reaplist);
1278 spin_lock(&state_lock);
1279 while (!list_empty(&clp->cl_delegations)) {
1280 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1281 list_del_init(&dp->dl_perclnt);
1282 list_move(&dp->dl_recall_lru, &reaplist);
1284 spin_unlock(&state_lock);
1285 while (!list_empty(&reaplist)) {
1286 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1287 destroy_delegation(dp);
1289 list_splice_init(&clp->cl_revoked, &reaplist);
1290 while (!list_empty(&reaplist)) {
1291 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1292 destroy_revoked_delegation(dp);
1294 while (!list_empty(&clp->cl_openowners)) {
1295 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1296 release_openowner(oo);
1298 nfsd4_shutdown_callback(clp);
1299 if (clp->cl_cb_conn.cb_xprt)
1300 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1301 list_del(&clp->cl_idhash);
1302 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1303 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1305 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1306 spin_lock(&nn->client_lock);
1307 unhash_client_locked(clp);
1308 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1310 spin_unlock(&nn->client_lock);
1313 static void expire_client(struct nfs4_client *clp)
1315 nfsd4_client_record_remove(clp);
1316 destroy_client(clp);
1319 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1321 memcpy(target->cl_verifier.data, source->data,
1322 sizeof(target->cl_verifier.data));
1325 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1327 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1328 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1331 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1333 if (source->cr_principal) {
1334 target->cr_principal =
1335 kstrdup(source->cr_principal, GFP_KERNEL);
1336 if (target->cr_principal == NULL)
1339 target->cr_principal = NULL;
1340 target->cr_flavor = source->cr_flavor;
1341 target->cr_uid = source->cr_uid;
1342 target->cr_gid = source->cr_gid;
1343 target->cr_group_info = source->cr_group_info;
1344 get_group_info(target->cr_group_info);
1345 target->cr_gss_mech = source->cr_gss_mech;
1346 if (source->cr_gss_mech)
1347 gss_mech_get(source->cr_gss_mech);
1352 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1356 res = o1->len - o2->len;
1359 return (long long)memcmp(o1->data, o2->data, o1->len);
1362 static int same_name(const char *n1, const char *n2)
1364 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1368 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1370 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1374 same_clid(clientid_t *cl1, clientid_t *cl2)
1376 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1379 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1383 if (g1->ngroups != g2->ngroups)
1385 for (i=0; i<g1->ngroups; i++)
1386 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1392 * RFC 3530 language requires clid_inuse be returned when the
1393 * "principal" associated with a requests differs from that previously
1394 * used. We use uid, gid's, and gss principal string as our best
1395 * approximation. We also don't want to allow non-gss use of a client
1396 * established using gss: in theory cr_principal should catch that
1397 * change, but in practice cr_principal can be null even in the gss case
1398 * since gssd doesn't always pass down a principal string.
1400 static bool is_gss_cred(struct svc_cred *cr)
1402 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1403 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1408 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1410 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1411 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1412 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1413 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1415 if (cr1->cr_principal == cr2->cr_principal)
1417 if (!cr1->cr_principal || !cr2->cr_principal)
1419 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1422 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1424 struct svc_cred *cr = &rqstp->rq_cred;
1427 if (!cr->cr_gss_mech)
1429 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1430 return service == RPC_GSS_SVC_INTEGRITY ||
1431 service == RPC_GSS_SVC_PRIVACY;
1434 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1436 struct svc_cred *cr = &rqstp->rq_cred;
1438 if (!cl->cl_mach_cred)
1440 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1442 if (!svc_rqst_integrity_protected(rqstp))
1444 if (!cr->cr_principal)
1446 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1449 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1451 static u32 current_clientid = 1;
1453 clp->cl_clientid.cl_boot = nn->boot_time;
1454 clp->cl_clientid.cl_id = current_clientid++;
1457 static void gen_confirm(struct nfs4_client *clp)
1463 * This is opaque to client, so no need to byte-swap. Use
1464 * __force to keep sparse happy
1466 verf[0] = (__force __be32)get_seconds();
1467 verf[1] = (__force __be32)i++;
1468 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1471 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1473 struct nfs4_stid *ret;
1475 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1476 if (!ret || !ret->sc_type)
1481 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1483 struct nfs4_stid *s;
1485 s = find_stateid(cl, t);
1488 if (typemask & s->sc_type)
1493 static struct nfs4_client *create_client(struct xdr_netobj name,
1494 struct svc_rqst *rqstp, nfs4_verifier *verf)
1496 struct nfs4_client *clp;
1497 struct sockaddr *sa = svc_addr(rqstp);
1499 struct net *net = SVC_NET(rqstp);
1500 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1502 clp = alloc_client(name);
1506 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1508 spin_lock(&nn->client_lock);
1510 spin_unlock(&nn->client_lock);
1513 nfsd4_init_callback(&clp->cl_cb_null);
1514 clp->cl_time = get_seconds();
1515 clear_bit(0, &clp->cl_cb_slot_busy);
1516 copy_verf(clp, verf);
1517 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1519 clp->cl_cb_session = NULL;
1525 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1527 struct rb_node **new = &(root->rb_node), *parent = NULL;
1528 struct nfs4_client *clp;
1531 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1534 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1535 new = &((*new)->rb_left);
1537 new = &((*new)->rb_right);
1540 rb_link_node(&new_clp->cl_namenode, parent, new);
1541 rb_insert_color(&new_clp->cl_namenode, root);
1544 static struct nfs4_client *
1545 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1548 struct rb_node *node = root->rb_node;
1549 struct nfs4_client *clp;
1552 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1553 cmp = compare_blob(&clp->cl_name, name);
1555 node = node->rb_left;
1557 node = node->rb_right;
1565 add_to_unconfirmed(struct nfs4_client *clp)
1567 unsigned int idhashval;
1568 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1570 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1571 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1572 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1573 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1578 move_to_confirmed(struct nfs4_client *clp)
1580 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1581 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1583 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1584 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1585 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1586 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1587 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1591 static struct nfs4_client *
1592 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1594 struct nfs4_client *clp;
1595 unsigned int idhashval = clientid_hashval(clid->cl_id);
1597 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1598 if (same_clid(&clp->cl_clientid, clid)) {
1599 if ((bool)clp->cl_minorversion != sessions)
1608 static struct nfs4_client *
1609 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1611 struct list_head *tbl = nn->conf_id_hashtbl;
1613 return find_client_in_id_table(tbl, clid, sessions);
1616 static struct nfs4_client *
1617 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1619 struct list_head *tbl = nn->unconf_id_hashtbl;
1621 return find_client_in_id_table(tbl, clid, sessions);
1624 static bool clp_used_exchangeid(struct nfs4_client *clp)
1626 return clp->cl_exchange_flags != 0;
1629 static struct nfs4_client *
1630 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1632 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1635 static struct nfs4_client *
1636 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1638 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1642 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1644 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1645 struct sockaddr *sa = svc_addr(rqstp);
1646 u32 scopeid = rpc_get_scope_id(sa);
1647 unsigned short expected_family;
1649 /* Currently, we only support tcp and tcp6 for the callback channel */
1650 if (se->se_callback_netid_len == 3 &&
1651 !memcmp(se->se_callback_netid_val, "tcp", 3))
1652 expected_family = AF_INET;
1653 else if (se->se_callback_netid_len == 4 &&
1654 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1655 expected_family = AF_INET6;
1659 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1660 se->se_callback_addr_len,
1661 (struct sockaddr *)&conn->cb_addr,
1662 sizeof(conn->cb_addr));
1664 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1667 if (conn->cb_addr.ss_family == AF_INET6)
1668 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1670 conn->cb_prog = se->se_callback_prog;
1671 conn->cb_ident = se->se_callback_ident;
1672 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1675 conn->cb_addr.ss_family = AF_UNSPEC;
1676 conn->cb_addrlen = 0;
1677 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1678 "will not receive delegations\n",
1679 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1685 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
1688 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1690 struct xdr_buf *buf = resp->xdr.buf;
1691 struct nfsd4_slot *slot = resp->cstate.slot;
1694 dprintk("--> %s slot %p\n", __func__, slot);
1696 slot->sl_opcnt = resp->opcnt;
1697 slot->sl_status = resp->cstate.status;
1699 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1700 if (nfsd4_not_cached(resp)) {
1701 slot->sl_datalen = 0;
1704 base = resp->cstate.data_offset;
1705 slot->sl_datalen = buf->len - base;
1706 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
1707 WARN("%s: sessions DRC could not cache compound\n", __func__);
1712 * Encode the replay sequence operation from the slot values.
1713 * If cachethis is FALSE encode the uncached rep error on the next
1714 * operation which sets resp->p and increments resp->opcnt for
1715 * nfs4svc_encode_compoundres.
1719 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1720 struct nfsd4_compoundres *resp)
1722 struct nfsd4_op *op;
1723 struct nfsd4_slot *slot = resp->cstate.slot;
1725 /* Encode the replayed sequence operation */
1726 op = &args->ops[resp->opcnt - 1];
1727 nfsd4_encode_operation(resp, op);
1729 /* Return nfserr_retry_uncached_rep in next operation. */
1730 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1731 op = &args->ops[resp->opcnt++];
1732 op->status = nfserr_retry_uncached_rep;
1733 nfsd4_encode_operation(resp, op);
1739 * The sequence operation is not cached because we can use the slot and
1743 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1744 struct nfsd4_sequence *seq)
1746 struct nfsd4_slot *slot = resp->cstate.slot;
1747 struct xdr_stream *xdr = &resp->xdr;
1751 dprintk("--> %s slot %p\n", __func__, slot);
1753 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1757 p = xdr_reserve_space(xdr, slot->sl_datalen);
1760 return nfserr_serverfault;
1762 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
1763 xdr_commit_encode(xdr);
1765 resp->opcnt = slot->sl_opcnt;
1766 return slot->sl_status;
1770 * Set the exchange_id flags returned by the server.
1773 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1775 /* pNFS is not supported */
1776 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1778 /* Referrals are supported, Migration is not. */
1779 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1781 /* set the wire flags to return to client. */
1782 clid->flags = new->cl_exchange_flags;
1785 static bool client_has_state(struct nfs4_client *clp)
1788 * Note clp->cl_openowners check isn't quite right: there's no
1789 * need to count owners without stateid's.
1791 * Also note we should probably be using this in 4.0 case too.
1793 return !list_empty(&clp->cl_openowners)
1794 || !list_empty(&clp->cl_delegations)
1795 || !list_empty(&clp->cl_sessions);
1799 nfsd4_exchange_id(struct svc_rqst *rqstp,
1800 struct nfsd4_compound_state *cstate,
1801 struct nfsd4_exchange_id *exid)
1803 struct nfs4_client *unconf, *conf, *new;
1805 char addr_str[INET6_ADDRSTRLEN];
1806 nfs4_verifier verf = exid->verifier;
1807 struct sockaddr *sa = svc_addr(rqstp);
1808 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1809 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1811 rpc_ntop(sa, addr_str, sizeof(addr_str));
1812 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1813 "ip_addr=%s flags %x, spa_how %d\n",
1814 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1815 addr_str, exid->flags, exid->spa_how);
1817 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1818 return nfserr_inval;
1820 switch (exid->spa_how) {
1822 if (!svc_rqst_integrity_protected(rqstp))
1823 return nfserr_inval;
1826 default: /* checked by xdr code */
1829 return nfserr_encr_alg_unsupp;
1832 /* Cases below refer to rfc 5661 section 18.35.4: */
1834 conf = find_confirmed_client_by_name(&exid->clname, nn);
1836 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1837 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1840 if (!clp_used_exchangeid(conf)) { /* buggy client */
1841 status = nfserr_inval;
1844 if (!mach_creds_match(conf, rqstp)) {
1845 status = nfserr_wrong_cred;
1848 if (!creds_match) { /* case 9 */
1849 status = nfserr_perm;
1852 if (!verfs_match) { /* case 8 */
1853 status = nfserr_not_same;
1857 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1861 if (!creds_match) { /* case 3 */
1862 if (client_has_state(conf)) {
1863 status = nfserr_clid_inuse;
1866 expire_client(conf);
1869 if (verfs_match) { /* case 2 */
1870 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1874 /* case 5, client reboot */
1878 if (update) { /* case 7 */
1879 status = nfserr_noent;
1883 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1884 if (unconf) /* case 4, possible retry or client restart */
1885 expire_client(unconf);
1887 /* case 1 (normal case) */
1889 new = create_client(exid->clname, rqstp, &verf);
1891 status = nfserr_jukebox;
1894 new->cl_minorversion = cstate->minorversion;
1895 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
1898 add_to_unconfirmed(new);
1900 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1901 exid->clientid.cl_id = new->cl_clientid.cl_id;
1903 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1904 nfsd4_set_ex_flags(new, exid);
1906 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1907 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1911 nfs4_unlock_state();
1916 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1918 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1921 /* The slot is in use, and no response has been sent. */
1923 if (seqid == slot_seqid)
1924 return nfserr_jukebox;
1926 return nfserr_seq_misordered;
1928 /* Note unsigned 32-bit arithmetic handles wraparound: */
1929 if (likely(seqid == slot_seqid + 1))
1931 if (seqid == slot_seqid)
1932 return nfserr_replay_cache;
1933 return nfserr_seq_misordered;
1937 * Cache the create session result into the create session single DRC
1938 * slot cache by saving the xdr structure. sl_seqid has been set.
1939 * Do this for solo or embedded create session operations.
1942 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1943 struct nfsd4_clid_slot *slot, __be32 nfserr)
1945 slot->sl_status = nfserr;
1946 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1950 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1951 struct nfsd4_clid_slot *slot)
1953 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1954 return slot->sl_status;
1957 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1958 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1959 1 + /* MIN tag is length with zero, only length */ \
1960 3 + /* version, opcount, opcode */ \
1961 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1962 /* seqid, slotID, slotID, cache */ \
1963 4 ) * sizeof(__be32))
1965 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1966 2 + /* verifier: AUTH_NULL, length 0 */\
1968 1 + /* MIN tag is length with zero, only length */ \
1969 3 + /* opcount, opcode, opstatus*/ \
1970 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1971 /* seqid, slotID, slotID, slotID, status */ \
1972 5 ) * sizeof(__be32))
1974 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1976 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
1978 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
1979 return nfserr_toosmall;
1980 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
1981 return nfserr_toosmall;
1982 ca->headerpadsz = 0;
1983 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
1984 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
1985 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
1986 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
1987 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
1988 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
1990 * Note decreasing slot size below client's request may make it
1991 * difficult for client to function correctly, whereas
1992 * decreasing the number of slots will (just?) affect
1993 * performance. When short on memory we therefore prefer to
1994 * decrease number of slots instead of their size. Clients that
1995 * request larger slots than they need will get poor results:
1997 ca->maxreqs = nfsd4_get_drc_mem(ca);
1999 return nfserr_jukebox;
2004 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
2005 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2006 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
2007 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2009 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2011 ca->headerpadsz = 0;
2014 * These RPC_MAX_HEADER macros are overkill, especially since we
2015 * don't even do gss on the backchannel yet. But this is still
2016 * less than 1k. Tighten up this estimate in the unlikely event
2017 * it turns out to be a problem for some client:
2019 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2020 return nfserr_toosmall;
2021 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2022 return nfserr_toosmall;
2023 ca->maxresp_cached = 0;
2025 return nfserr_toosmall;
2030 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2032 switch (cbs->flavor) {
2038 * GSS case: the spec doesn't allow us to return this
2039 * error. But it also doesn't allow us not to support
2041 * I'd rather this fail hard than return some error the
2042 * client might think it can already handle:
2044 return nfserr_encr_alg_unsupp;
2049 nfsd4_create_session(struct svc_rqst *rqstp,
2050 struct nfsd4_compound_state *cstate,
2051 struct nfsd4_create_session *cr_ses)
2053 struct sockaddr *sa = svc_addr(rqstp);
2054 struct nfs4_client *conf, *unconf;
2055 struct nfsd4_session *new;
2056 struct nfsd4_conn *conn;
2057 struct nfsd4_clid_slot *cs_slot = NULL;
2059 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2061 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2062 return nfserr_inval;
2063 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2066 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2069 status = check_backchannel_attrs(&cr_ses->back_channel);
2071 goto out_release_drc_mem;
2072 status = nfserr_jukebox;
2073 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2075 goto out_release_drc_mem;
2076 conn = alloc_conn_from_crses(rqstp, cr_ses);
2078 goto out_free_session;
2081 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2082 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2083 WARN_ON_ONCE(conf && unconf);
2086 status = nfserr_wrong_cred;
2087 if (!mach_creds_match(conf, rqstp))
2089 cs_slot = &conf->cl_cs_slot;
2090 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2091 if (status == nfserr_replay_cache) {
2092 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2094 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2095 status = nfserr_seq_misordered;
2098 } else if (unconf) {
2099 struct nfs4_client *old;
2100 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2101 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2102 status = nfserr_clid_inuse;
2105 status = nfserr_wrong_cred;
2106 if (!mach_creds_match(unconf, rqstp))
2108 cs_slot = &unconf->cl_cs_slot;
2109 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2111 /* an unconfirmed replay returns misordered */
2112 status = nfserr_seq_misordered;
2115 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2117 status = mark_client_expired(old);
2122 move_to_confirmed(unconf);
2125 status = nfserr_stale_clientid;
2130 * We do not support RDMA or persistent sessions
2132 cr_ses->flags &= ~SESSION4_PERSIST;
2133 cr_ses->flags &= ~SESSION4_RDMA;
2135 init_session(rqstp, new, conf, cr_ses);
2136 nfsd4_init_conn(rqstp, conn, new);
2138 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2139 NFS4_MAX_SESSIONID_LEN);
2140 cs_slot->sl_seqid++;
2141 cr_ses->seqid = cs_slot->sl_seqid;
2143 /* cache solo and embedded create sessions under the state lock */
2144 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2145 nfs4_unlock_state();
2148 nfs4_unlock_state();
2151 __free_session(new);
2152 out_release_drc_mem:
2153 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2157 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2160 case NFS4_CDFC4_FORE:
2161 case NFS4_CDFC4_BACK:
2163 case NFS4_CDFC4_FORE_OR_BOTH:
2164 case NFS4_CDFC4_BACK_OR_BOTH:
2165 *dir = NFS4_CDFC4_BOTH;
2168 return nfserr_inval;
2171 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2173 struct nfsd4_session *session = cstate->session;
2174 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2177 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2180 spin_lock(&nn->client_lock);
2181 session->se_cb_prog = bc->bc_cb_program;
2182 session->se_cb_sec = bc->bc_cb_sec;
2183 spin_unlock(&nn->client_lock);
2185 nfsd4_probe_callback(session->se_client);
2190 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2191 struct nfsd4_compound_state *cstate,
2192 struct nfsd4_bind_conn_to_session *bcts)
2195 struct nfsd4_conn *conn;
2196 struct nfsd4_session *session;
2197 struct net *net = SVC_NET(rqstp);
2198 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2200 if (!nfsd4_last_compound_op(rqstp))
2201 return nfserr_not_only_op;
2203 spin_lock(&nn->client_lock);
2204 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2205 spin_unlock(&nn->client_lock);
2207 goto out_no_session;
2208 status = nfserr_wrong_cred;
2209 if (!mach_creds_match(session->se_client, rqstp))
2211 status = nfsd4_map_bcts_dir(&bcts->dir);
2214 conn = alloc_conn(rqstp, bcts->dir);
2215 status = nfserr_jukebox;
2218 nfsd4_init_conn(rqstp, conn, session);
2221 nfsd4_put_session(session);
2223 nfs4_unlock_state();
2227 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2231 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2235 nfsd4_destroy_session(struct svc_rqst *r,
2236 struct nfsd4_compound_state *cstate,
2237 struct nfsd4_destroy_session *sessionid)
2239 struct nfsd4_session *ses;
2241 int ref_held_by_me = 0;
2242 struct net *net = SVC_NET(r);
2243 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2246 status = nfserr_not_only_op;
2247 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2248 if (!nfsd4_last_compound_op(r))
2252 dump_sessionid(__func__, &sessionid->sessionid);
2253 spin_lock(&nn->client_lock);
2254 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2256 goto out_client_lock;
2257 status = nfserr_wrong_cred;
2258 if (!mach_creds_match(ses->se_client, r))
2259 goto out_put_session;
2260 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2262 goto out_put_session;
2263 unhash_session(ses);
2264 spin_unlock(&nn->client_lock);
2266 nfsd4_probe_callback_sync(ses->se_client);
2268 spin_lock(&nn->client_lock);
2271 nfsd4_put_session_locked(ses);
2273 spin_unlock(&nn->client_lock);
2275 nfs4_unlock_state();
2279 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2281 struct nfsd4_conn *c;
2283 list_for_each_entry(c, &s->se_conns, cn_persession) {
2284 if (c->cn_xprt == xpt) {
2291 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2293 struct nfs4_client *clp = ses->se_client;
2294 struct nfsd4_conn *c;
2295 __be32 status = nfs_ok;
2298 spin_lock(&clp->cl_lock);
2299 c = __nfsd4_find_conn(new->cn_xprt, ses);
2302 status = nfserr_conn_not_bound_to_session;
2303 if (clp->cl_mach_cred)
2305 __nfsd4_hash_conn(new, ses);
2306 spin_unlock(&clp->cl_lock);
2307 ret = nfsd4_register_conn(new);
2309 /* oops; xprt is already down: */
2310 nfsd4_conn_lost(&new->cn_xpt_user);
2313 spin_unlock(&clp->cl_lock);
2318 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2320 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2322 return args->opcnt > session->se_fchannel.maxops;
2325 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2326 struct nfsd4_session *session)
2328 struct xdr_buf *xb = &rqstp->rq_arg;
2330 return xb->len > session->se_fchannel.maxreq_sz;
2334 nfsd4_sequence(struct svc_rqst *rqstp,
2335 struct nfsd4_compound_state *cstate,
2336 struct nfsd4_sequence *seq)
2338 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2339 struct xdr_stream *xdr = &resp->xdr;
2340 struct nfsd4_session *session;
2341 struct nfs4_client *clp;
2342 struct nfsd4_slot *slot;
2343 struct nfsd4_conn *conn;
2346 struct net *net = SVC_NET(rqstp);
2347 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2349 if (resp->opcnt != 1)
2350 return nfserr_sequence_pos;
2353 * Will be either used or freed by nfsd4_sequence_check_conn
2356 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2358 return nfserr_jukebox;
2360 spin_lock(&nn->client_lock);
2361 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2363 goto out_no_session;
2364 clp = session->se_client;
2366 status = nfserr_too_many_ops;
2367 if (nfsd4_session_too_many_ops(rqstp, session))
2368 goto out_put_session;
2370 status = nfserr_req_too_big;
2371 if (nfsd4_request_too_big(rqstp, session))
2372 goto out_put_session;
2374 status = nfserr_badslot;
2375 if (seq->slotid >= session->se_fchannel.maxreqs)
2376 goto out_put_session;
2378 slot = session->se_slots[seq->slotid];
2379 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2381 /* We do not negotiate the number of slots yet, so set the
2382 * maxslots to the session maxreqs which is used to encode
2383 * sr_highest_slotid and the sr_target_slot id to maxslots */
2384 seq->maxslots = session->se_fchannel.maxreqs;
2386 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2387 slot->sl_flags & NFSD4_SLOT_INUSE);
2388 if (status == nfserr_replay_cache) {
2389 status = nfserr_seq_misordered;
2390 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2391 goto out_put_session;
2392 cstate->slot = slot;
2393 cstate->session = session;
2394 /* Return the cached reply status and set cstate->status
2395 * for nfsd4_proc_compound processing */
2396 status = nfsd4_replay_cache_entry(resp, seq);
2397 cstate->status = nfserr_replay_cache;
2401 goto out_put_session;
2403 status = nfsd4_sequence_check_conn(conn, session);
2406 goto out_put_session;
2408 buflen = (seq->cachethis) ?
2409 session->se_fchannel.maxresp_cached :
2410 session->se_fchannel.maxresp_sz;
2411 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2413 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2414 goto out_put_session;
2415 svc_reserve(rqstp, buflen);
2418 /* Success! bump slot seqid */
2419 slot->sl_seqid = seq->seqid;
2420 slot->sl_flags |= NFSD4_SLOT_INUSE;
2422 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2424 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2426 cstate->slot = slot;
2427 cstate->session = session;
2430 switch (clp->cl_cb_state) {
2432 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2434 case NFSD4_CB_FAULT:
2435 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2438 seq->status_flags = 0;
2440 if (!list_empty(&clp->cl_revoked))
2441 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2445 spin_unlock(&nn->client_lock);
2448 nfsd4_put_session_locked(session);
2449 goto out_no_session;
2453 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2455 struct nfsd4_compound_state *cs = &resp->cstate;
2457 if (nfsd4_has_session(cs)) {
2458 if (cs->status != nfserr_replay_cache) {
2459 nfsd4_store_cache_entry(resp);
2460 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2462 /* Drop session reference that was taken in nfsd4_sequence() */
2463 nfsd4_put_session(cs->session);
2468 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2470 struct nfs4_client *conf, *unconf, *clp;
2472 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2475 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2476 conf = find_confirmed_client(&dc->clientid, true, nn);
2477 WARN_ON_ONCE(conf && unconf);
2482 if (client_has_state(conf)) {
2483 status = nfserr_clientid_busy;
2489 status = nfserr_stale_clientid;
2492 if (!mach_creds_match(clp, rqstp)) {
2493 status = nfserr_wrong_cred;
2498 nfs4_unlock_state();
2503 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2507 if (rc->rca_one_fs) {
2508 if (!cstate->current_fh.fh_dentry)
2509 return nfserr_nofilehandle;
2511 * We don't take advantage of the rca_one_fs case.
2512 * That's OK, it's optional, we can safely ignore it.
2518 status = nfserr_complete_already;
2519 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2520 &cstate->session->se_client->cl_flags))
2523 status = nfserr_stale_clientid;
2524 if (is_client_expired(cstate->session->se_client))
2526 * The following error isn't really legal.
2527 * But we only get here if the client just explicitly
2528 * destroyed the client. Surely it no longer cares what
2529 * error it gets back on an operation for the dead
2535 nfsd4_client_record_create(cstate->session->se_client);
2537 nfs4_unlock_state();
2542 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2543 struct nfsd4_setclientid *setclid)
2545 struct xdr_netobj clname = setclid->se_name;
2546 nfs4_verifier clverifier = setclid->se_verf;
2547 struct nfs4_client *conf, *unconf, *new;
2549 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2551 /* Cases below refer to rfc 3530 section 14.2.33: */
2553 conf = find_confirmed_client_by_name(&clname, nn);
2556 status = nfserr_clid_inuse;
2557 if (clp_used_exchangeid(conf))
2559 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2560 char addr_str[INET6_ADDRSTRLEN];
2561 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2563 dprintk("NFSD: setclientid: string in use by client "
2564 "at %s\n", addr_str);
2568 unconf = find_unconfirmed_client_by_name(&clname, nn);
2570 expire_client(unconf);
2571 status = nfserr_jukebox;
2572 new = create_client(clname, rqstp, &clverifier);
2575 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2576 /* case 1: probable callback update */
2577 copy_clid(new, conf);
2578 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2580 new->cl_minorversion = 0;
2581 gen_callback(new, setclid, rqstp);
2582 add_to_unconfirmed(new);
2583 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2584 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2585 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2588 nfs4_unlock_state();
2594 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2595 struct nfsd4_compound_state *cstate,
2596 struct nfsd4_setclientid_confirm *setclientid_confirm)
2598 struct nfs4_client *conf, *unconf;
2599 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2600 clientid_t * clid = &setclientid_confirm->sc_clientid;
2602 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2604 if (STALE_CLIENTID(clid, nn))
2605 return nfserr_stale_clientid;
2608 conf = find_confirmed_client(clid, false, nn);
2609 unconf = find_unconfirmed_client(clid, false, nn);
2611 * We try hard to give out unique clientid's, so if we get an
2612 * attempt to confirm the same clientid with a different cred,
2613 * there's a bug somewhere. Let's charitably assume it's our
2616 status = nfserr_serverfault;
2617 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2619 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2621 /* cases below refer to rfc 3530 section 14.2.34: */
2622 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2623 if (conf && !unconf) /* case 2: probable retransmit */
2625 else /* case 4: client hasn't noticed we rebooted yet? */
2626 status = nfserr_stale_clientid;
2630 if (conf) { /* case 1: callback update */
2631 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2632 nfsd4_probe_callback(conf);
2633 expire_client(unconf);
2634 } else { /* case 3: normal case; new or rebooted client */
2635 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2637 status = mark_client_expired(conf);
2640 expire_client(conf);
2642 move_to_confirmed(unconf);
2643 nfsd4_probe_callback(unconf);
2646 nfs4_unlock_state();
2650 static struct nfs4_file *nfsd4_alloc_file(void)
2652 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2655 /* OPEN Share state helper functions */
2656 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2658 unsigned int hashval = file_hashval(ino);
2660 lockdep_assert_held(&state_lock);
2662 atomic_set(&fp->fi_ref, 1);
2663 INIT_LIST_HEAD(&fp->fi_stateids);
2664 INIT_LIST_HEAD(&fp->fi_delegations);
2667 fp->fi_had_conflict = false;
2668 fp->fi_lease = NULL;
2669 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2670 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2671 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2675 nfsd4_free_slabs(void)
2677 kmem_cache_destroy(openowner_slab);
2678 kmem_cache_destroy(lockowner_slab);
2679 kmem_cache_destroy(file_slab);
2680 kmem_cache_destroy(stateid_slab);
2681 kmem_cache_destroy(deleg_slab);
2685 nfsd4_init_slabs(void)
2687 openowner_slab = kmem_cache_create("nfsd4_openowners",
2688 sizeof(struct nfs4_openowner), 0, 0, NULL);
2689 if (openowner_slab == NULL)
2691 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2692 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2693 if (lockowner_slab == NULL)
2694 goto out_free_openowner_slab;
2695 file_slab = kmem_cache_create("nfsd4_files",
2696 sizeof(struct nfs4_file), 0, 0, NULL);
2697 if (file_slab == NULL)
2698 goto out_free_lockowner_slab;
2699 stateid_slab = kmem_cache_create("nfsd4_stateids",
2700 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2701 if (stateid_slab == NULL)
2702 goto out_free_file_slab;
2703 deleg_slab = kmem_cache_create("nfsd4_delegations",
2704 sizeof(struct nfs4_delegation), 0, 0, NULL);
2705 if (deleg_slab == NULL)
2706 goto out_free_stateid_slab;
2709 out_free_stateid_slab:
2710 kmem_cache_destroy(stateid_slab);
2712 kmem_cache_destroy(file_slab);
2713 out_free_lockowner_slab:
2714 kmem_cache_destroy(lockowner_slab);
2715 out_free_openowner_slab:
2716 kmem_cache_destroy(openowner_slab);
2718 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2722 static void init_nfs4_replay(struct nfs4_replay *rp)
2724 rp->rp_status = nfserr_serverfault;
2726 rp->rp_buf = rp->rp_ibuf;
2729 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2731 struct nfs4_stateowner *sop;
2733 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2737 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2738 if (!sop->so_owner.data) {
2739 kmem_cache_free(slab, sop);
2742 sop->so_owner.len = owner->len;
2744 INIT_LIST_HEAD(&sop->so_stateids);
2745 sop->so_client = clp;
2746 init_nfs4_replay(&sop->so_replay);
2750 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2752 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2754 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2755 list_add(&oo->oo_perclient, &clp->cl_openowners);
2758 static struct nfs4_openowner *
2759 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp,
2760 struct nfsd4_open *open,
2761 struct nfsd4_compound_state *cstate)
2763 struct nfs4_openowner *oo;
2765 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2768 oo->oo_owner.so_is_open_owner = 1;
2769 oo->oo_owner.so_seqid = open->op_seqid;
2770 oo->oo_flags = NFS4_OO_NEW;
2771 if (nfsd4_has_session(cstate))
2772 oo->oo_flags |= NFS4_OO_CONFIRMED;
2774 oo->oo_last_closed_stid = NULL;
2775 INIT_LIST_HEAD(&oo->oo_close_lru);
2776 hash_openowner(oo, clp, strhashval);
2780 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2781 struct nfs4_openowner *oo = open->op_openowner;
2783 stp->st_stid.sc_type = NFS4_OPEN_STID;
2784 INIT_LIST_HEAD(&stp->st_locks);
2785 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2786 list_add(&stp->st_perfile, &fp->fi_stateids);
2787 stp->st_stateowner = &oo->oo_owner;
2790 stp->st_access_bmap = 0;
2791 stp->st_deny_bmap = 0;
2792 set_access(open->op_share_access, stp);
2793 set_deny(open->op_share_deny, stp);
2794 stp->st_openstp = NULL;
2798 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2800 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2802 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2804 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2805 oo->oo_time = get_seconds();
2809 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2812 return (sop->so_owner.len == owner->len) &&
2813 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2814 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2817 static struct nfs4_openowner *
2818 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2819 bool sessions, struct nfsd_net *nn)
2821 struct nfs4_stateowner *so;
2822 struct nfs4_openowner *oo;
2823 struct nfs4_client *clp;
2825 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2826 if (!so->so_is_open_owner)
2828 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2830 clp = oo->oo_owner.so_client;
2831 if ((bool)clp->cl_minorversion != sessions)
2833 renew_client(oo->oo_owner.so_client);
2840 /* search file_hashtbl[] for file */
2841 static struct nfs4_file *
2842 find_file_locked(struct inode *ino)
2844 unsigned int hashval = file_hashval(ino);
2845 struct nfs4_file *fp;
2847 lockdep_assert_held(&state_lock);
2849 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2850 if (fp->fi_inode == ino) {
2858 static struct nfs4_file *
2859 find_file(struct inode *ino)
2861 struct nfs4_file *fp;
2863 spin_lock(&state_lock);
2864 fp = find_file_locked(ino);
2865 spin_unlock(&state_lock);
2869 static struct nfs4_file *
2870 find_or_add_file(struct inode *ino, struct nfs4_file *new)
2872 struct nfs4_file *fp;
2874 spin_lock(&state_lock);
2875 fp = find_file_locked(ino);
2877 nfsd4_init_file(new, ino);
2880 spin_unlock(&state_lock);
2886 * Called to check deny when READ with all zero stateid or
2887 * WRITE with all zero or all one stateid
2890 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2892 struct inode *ino = current_fh->fh_dentry->d_inode;
2893 struct nfs4_file *fp;
2894 struct nfs4_ol_stateid *stp;
2897 fp = find_file(ino);
2900 ret = nfserr_locked;
2901 /* Search for conflicting share reservations */
2902 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2903 if (test_deny(deny_type, stp) ||
2904 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2913 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2915 struct nfs4_client *clp = dp->dl_stid.sc_client;
2916 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2918 lockdep_assert_held(&state_lock);
2919 /* We're assuming the state code never drops its reference
2920 * without first removing the lease. Since we're in this lease
2921 * callback (and since the lease code is serialized by the kernel
2922 * lock) we know the server hasn't removed the lease yet, we know
2923 * it's safe to take a reference: */
2924 atomic_inc(&dp->dl_count);
2926 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
2928 /* Only place dl_time is set; protected by i_lock: */
2929 dp->dl_time = get_seconds();
2931 block_delegations(&dp->dl_fh);
2933 nfsd4_cb_recall(dp);
2936 /* Called from break_lease() with i_lock held. */
2937 static void nfsd_break_deleg_cb(struct file_lock *fl)
2939 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2940 struct nfs4_delegation *dp;
2943 WARN(1, "(%p)->fl_owner NULL\n", fl);
2946 if (fp->fi_had_conflict) {
2947 WARN(1, "duplicate break on %p\n", fp);
2951 * We don't want the locks code to timeout the lease for us;
2952 * we'll remove it ourself if a delegation isn't returned
2955 fl->fl_break_time = 0;
2957 spin_lock(&state_lock);
2958 fp->fi_had_conflict = true;
2959 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2960 nfsd_break_one_deleg(dp);
2961 spin_unlock(&state_lock);
2965 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2968 return lease_modify(onlist, arg);
2973 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2974 .lm_break = nfsd_break_deleg_cb,
2975 .lm_change = nfsd_change_deleg_cb,
2978 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2980 if (nfsd4_has_session(cstate))
2982 if (seqid == so->so_seqid - 1)
2983 return nfserr_replay_me;
2984 if (seqid == so->so_seqid)
2986 return nfserr_bad_seqid;
2990 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2991 struct nfsd4_open *open, struct nfsd_net *nn)
2993 clientid_t *clientid = &open->op_clientid;
2994 struct nfs4_client *clp = NULL;
2995 unsigned int strhashval;
2996 struct nfs4_openowner *oo = NULL;
2999 if (STALE_CLIENTID(&open->op_clientid, nn))
3000 return nfserr_stale_clientid;
3002 * In case we need it later, after we've already created the
3003 * file and don't want to risk a further failure:
3005 open->op_file = nfsd4_alloc_file();
3006 if (open->op_file == NULL)
3007 return nfserr_jukebox;
3009 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
3010 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
3011 open->op_openowner = oo;
3013 clp = find_confirmed_client(clientid, cstate->minorversion,
3016 return nfserr_expired;
3019 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3020 /* Replace unconfirmed owners without checking for replay. */
3021 clp = oo->oo_owner.so_client;
3022 release_openowner(oo);
3023 open->op_openowner = NULL;
3026 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3029 clp = oo->oo_owner.so_client;
3032 oo = alloc_init_open_stateowner(strhashval, clp, open, cstate);
3034 return nfserr_jukebox;
3035 open->op_openowner = oo;
3037 open->op_stp = nfs4_alloc_stateid(clp);
3039 return nfserr_jukebox;
3043 static inline __be32
3044 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3046 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3047 return nfserr_openmode;
3052 static int share_access_to_flags(u32 share_access)
3054 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3057 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3059 struct nfs4_stid *ret;
3061 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3064 return delegstateid(ret);
3067 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3069 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3070 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3074 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3075 struct nfs4_delegation **dp)
3078 __be32 status = nfserr_bad_stateid;
3080 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
3083 flags = share_access_to_flags(open->op_share_access);
3084 status = nfs4_check_delegmode(*dp, flags);
3088 if (!nfsd4_is_deleg_cur(open))
3092 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3097 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
3099 struct nfs4_ol_stateid *local;
3100 struct nfs4_openowner *oo = open->op_openowner;
3102 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3103 /* ignore lock owners */
3104 if (local->st_stateowner->so_is_open_owner == 0)
3106 /* remember if we have seen this open owner */
3107 if (local->st_stateowner == &oo->oo_owner)
3109 /* check for conflicting share reservations */
3110 if (!test_share(local, open))
3111 return nfserr_share_denied;
3116 static inline int nfs4_access_to_access(u32 nfs4_access)
3120 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3121 flags |= NFSD_MAY_READ;
3122 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3123 flags |= NFSD_MAY_WRITE;
3127 static inline __be32
3128 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3129 struct nfsd4_open *open)
3131 struct iattr iattr = {
3132 .ia_valid = ATTR_SIZE,
3135 if (!open->op_truncate)
3137 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3138 return nfserr_inval;
3139 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3142 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3143 struct svc_fh *cur_fh, struct nfsd4_open *open)
3146 int oflag = nfs4_access_to_omode(open->op_share_access);
3147 int access = nfs4_access_to_access(open->op_share_access);
3149 if (!fp->fi_fds[oflag]) {
3150 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
3151 &fp->fi_fds[oflag]);
3155 nfs4_file_get_access(fp, oflag);
3157 status = nfsd4_truncate(rqstp, cur_fh, open);
3159 goto out_put_access;
3164 nfs4_file_put_access(fp, oflag);
3170 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)
3172 u32 op_share_access = open->op_share_access;
3175 if (!test_access(op_share_access, stp))
3176 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
3178 status = nfsd4_truncate(rqstp, cur_fh, open);
3183 /* remember the open */
3184 set_access(op_share_access, stp);
3185 set_deny(open->op_share_deny, stp);
3191 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3193 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3196 /* Should we give out recallable state?: */
3197 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3199 if (clp->cl_cb_state == NFSD4_CB_UP)
3202 * In the sessions case, since we don't have to establish a
3203 * separate connection for callbacks, we assume it's OK
3204 * until we hear otherwise:
3206 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3209 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
3211 struct file_lock *fl;
3213 fl = locks_alloc_lock();
3216 locks_init_lock(fl);
3217 fl->fl_lmops = &nfsd_lease_mng_ops;
3218 fl->fl_flags = FL_DELEG;
3219 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3220 fl->fl_end = OFFSET_MAX;
3221 fl->fl_owner = (fl_owner_t)(dp->dl_file);
3222 fl->fl_pid = current->tgid;
3226 static int nfs4_setlease(struct nfs4_delegation *dp)
3228 struct nfs4_file *fp = dp->dl_file;
3229 struct file_lock *fl;
3232 fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
3235 fl->fl_file = find_readable_file(fp);
3236 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
3240 fp->fi_deleg_file = get_file(fl->fl_file);
3241 atomic_set(&fp->fi_delegees, 1);
3242 spin_lock(&state_lock);
3243 hash_delegation_locked(dp, fp);
3244 spin_unlock(&state_lock);
3247 locks_free_lock(fl);
3251 static int nfs4_set_delegation(struct nfs4_delegation *dp, struct nfs4_file *fp)
3253 if (fp->fi_had_conflict)
3258 return nfs4_setlease(dp);
3259 spin_lock(&state_lock);
3260 atomic_inc(&fp->fi_delegees);
3261 if (fp->fi_had_conflict) {
3262 spin_unlock(&state_lock);
3265 hash_delegation_locked(dp, fp);
3266 spin_unlock(&state_lock);
3270 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3272 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3273 if (status == -EAGAIN)
3274 open->op_why_no_deleg = WND4_CONTENTION;
3276 open->op_why_no_deleg = WND4_RESOURCE;
3277 switch (open->op_deleg_want) {
3278 case NFS4_SHARE_WANT_READ_DELEG:
3279 case NFS4_SHARE_WANT_WRITE_DELEG:
3280 case NFS4_SHARE_WANT_ANY_DELEG:
3282 case NFS4_SHARE_WANT_CANCEL:
3283 open->op_why_no_deleg = WND4_CANCELLED;
3285 case NFS4_SHARE_WANT_NO_DELEG:
3292 * Attempt to hand out a delegation.
3294 * Note we don't support write delegations, and won't until the vfs has
3295 * proper support for them.
3298 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
3299 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
3301 struct nfs4_delegation *dp;
3302 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
3306 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3307 open->op_recall = 0;
3308 switch (open->op_claim_type) {
3309 case NFS4_OPEN_CLAIM_PREVIOUS:
3311 open->op_recall = 1;
3312 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3315 case NFS4_OPEN_CLAIM_NULL:
3316 case NFS4_OPEN_CLAIM_FH:
3318 * Let's not give out any delegations till everyone's
3319 * had the chance to reclaim theirs....
3321 if (locks_in_grace(net))
3323 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3326 * Also, if the file was opened for write or
3327 * create, there's a good chance the client's
3328 * about to write to it, resulting in an
3329 * immediate recall (since we don't support
3330 * write delegations):
3332 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3334 if (open->op_create == NFS4_OPEN_CREATE)
3340 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh);
3343 status = nfs4_set_delegation(dp, stp->st_file);
3347 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3349 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3350 STATEID_VAL(&dp->dl_stid.sc_stateid));
3351 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3354 destroy_delegation(dp);
3356 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3357 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3358 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3359 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3360 open->op_recall = 1;
3363 /* 4.1 client asking for a delegation? */
3364 if (open->op_deleg_want)
3365 nfsd4_open_deleg_none_ext(open, status);
3369 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3370 struct nfs4_delegation *dp)
3372 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3373 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3374 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3375 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3376 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3377 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3378 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3379 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3381 /* Otherwise the client must be confused wanting a delegation
3382 * it already has, therefore we don't return
3383 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3388 * called with nfs4_lock_state() held.
3391 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3393 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3394 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3395 struct nfs4_file *fp = NULL;
3396 struct inode *ino = current_fh->fh_dentry->d_inode;
3397 struct nfs4_ol_stateid *stp = NULL;
3398 struct nfs4_delegation *dp = NULL;
3402 * Lookup file; if found, lookup stateid and check open request,
3403 * and check for delegations in the process of being recalled.
3404 * If not found, create the nfs4_file struct
3406 fp = find_or_add_file(ino, open->op_file);
3407 if (fp != open->op_file) {
3408 if ((status = nfs4_check_open(fp, open, &stp)))
3410 status = nfs4_check_deleg(cl, open, &dp);
3414 open->op_file = NULL;
3415 status = nfserr_bad_stateid;
3416 if (nfsd4_is_deleg_cur(open))
3418 status = nfserr_jukebox;
3422 * OPEN the file, or upgrade an existing OPEN.
3423 * If truncate fails, the OPEN fails.
3426 /* Stateid was found, this is an OPEN upgrade */
3427 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3431 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3435 open->op_stp = NULL;
3436 init_open_stateid(stp, fp, open);
3438 update_stateid(&stp->st_stid.sc_stateid);
3439 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3441 if (nfsd4_has_session(&resp->cstate)) {
3442 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3443 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3444 open->op_why_no_deleg = WND4_NOT_WANTED;
3450 * Attempt to hand out a delegation. No error return, because the
3451 * OPEN succeeds even if we fail.
3453 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3457 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3458 STATEID_VAL(&stp->st_stid.sc_stateid));
3460 /* 4.1 client trying to upgrade/downgrade delegation? */
3461 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3462 open->op_deleg_want)
3463 nfsd4_deleg_xgrade_none_ext(open, dp);
3467 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3468 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3470 * To finish the open response, we just need to set the rflags.
3472 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3473 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3474 !nfsd4_has_session(&resp->cstate))
3475 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3480 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3482 if (open->op_openowner) {
3483 struct nfs4_openowner *oo = open->op_openowner;
3485 if (!list_empty(&oo->oo_owner.so_stateids))
3486 list_del_init(&oo->oo_close_lru);
3487 if (oo->oo_flags & NFS4_OO_NEW) {
3489 release_openowner(oo);
3490 open->op_openowner = NULL;
3492 oo->oo_flags &= ~NFS4_OO_NEW;
3496 nfsd4_free_file(open->op_file);
3498 free_generic_stateid(open->op_stp);
3501 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3503 struct nfs4_client *found;
3505 if (STALE_CLIENTID(clid, nn))
3506 return nfserr_stale_clientid;
3507 found = find_confirmed_client(clid, session, nn);
3510 return found ? nfs_ok : nfserr_expired;
3514 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3517 struct nfs4_client *clp;
3519 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3522 dprintk("process_renew(%08x/%08x): starting\n",
3523 clid->cl_boot, clid->cl_id);
3524 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3527 status = nfserr_cb_path_down;
3528 if (!list_empty(&clp->cl_delegations)
3529 && clp->cl_cb_state != NFSD4_CB_UP)
3533 nfs4_unlock_state();
3538 nfsd4_end_grace(struct nfsd_net *nn)
3540 /* do nothing if grace period already ended */
3541 if (nn->grace_ended)
3544 dprintk("NFSD: end of grace period\n");
3545 nn->grace_ended = true;
3546 nfsd4_record_grace_done(nn, nn->boot_time);
3547 locks_end_grace(&nn->nfsd4_manager);
3549 * Now that every NFSv4 client has had the chance to recover and
3550 * to see the (possibly new, possibly shorter) lease time, we
3551 * can safely set the next grace time to the current lease time:
3553 nn->nfsd4_grace = nn->nfsd4_lease;
3557 nfs4_laundromat(struct nfsd_net *nn)
3559 struct nfs4_client *clp;
3560 struct nfs4_openowner *oo;
3561 struct nfs4_delegation *dp;
3562 struct list_head *pos, *next, reaplist;
3563 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3564 time_t t, new_timeo = nn->nfsd4_lease;
3568 dprintk("NFSD: laundromat service - starting\n");
3569 nfsd4_end_grace(nn);
3570 INIT_LIST_HEAD(&reaplist);
3571 spin_lock(&nn->client_lock);
3572 list_for_each_safe(pos, next, &nn->client_lru) {
3573 clp = list_entry(pos, struct nfs4_client, cl_lru);
3574 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3575 t = clp->cl_time - cutoff;
3576 new_timeo = min(new_timeo, t);
3579 if (mark_client_expired_locked(clp)) {
3580 dprintk("NFSD: client in use (clientid %08x)\n",
3581 clp->cl_clientid.cl_id);
3584 list_move(&clp->cl_lru, &reaplist);
3586 spin_unlock(&nn->client_lock);
3587 list_for_each_safe(pos, next, &reaplist) {
3588 clp = list_entry(pos, struct nfs4_client, cl_lru);
3589 dprintk("NFSD: purging unused client (clientid %08x)\n",
3590 clp->cl_clientid.cl_id);
3593 spin_lock(&state_lock);
3594 list_for_each_safe(pos, next, &nn->del_recall_lru) {
3595 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3596 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3598 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3599 t = dp->dl_time - cutoff;
3600 new_timeo = min(new_timeo, t);
3603 list_move(&dp->dl_recall_lru, &reaplist);
3605 spin_unlock(&state_lock);
3606 list_for_each_safe(pos, next, &reaplist) {
3607 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3608 revoke_delegation(dp);
3610 list_for_each_safe(pos, next, &nn->close_lru) {
3611 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3612 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3613 t = oo->oo_time - cutoff;
3614 new_timeo = min(new_timeo, t);
3617 release_openowner(oo);
3619 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
3620 nfs4_unlock_state();
3624 static struct workqueue_struct *laundry_wq;
3625 static void laundromat_main(struct work_struct *);
3628 laundromat_main(struct work_struct *laundry)
3631 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3633 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3636 t = nfs4_laundromat(nn);
3637 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3638 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3641 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3643 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3644 return nfserr_bad_stateid;
3649 access_permit_read(struct nfs4_ol_stateid *stp)
3651 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3652 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3653 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3657 access_permit_write(struct nfs4_ol_stateid *stp)
3659 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3660 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3664 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3666 __be32 status = nfserr_openmode;
3668 /* For lock stateid's, we test the parent open, not the lock: */
3669 if (stp->st_openstp)
3670 stp = stp->st_openstp;
3671 if ((flags & WR_STATE) && !access_permit_write(stp))
3673 if ((flags & RD_STATE) && !access_permit_read(stp))
3680 static inline __be32
3681 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3683 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3685 else if (locks_in_grace(net)) {
3686 /* Answer in remaining cases depends on existence of
3687 * conflicting state; so we must wait out the grace period. */
3688 return nfserr_grace;
3689 } else if (flags & WR_STATE)
3690 return nfs4_share_conflict(current_fh,
3691 NFS4_SHARE_DENY_WRITE);
3692 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3693 return nfs4_share_conflict(current_fh,
3694 NFS4_SHARE_DENY_READ);
3698 * Allow READ/WRITE during grace period on recovered state only for files
3699 * that are not able to provide mandatory locking.
3702 grace_disallows_io(struct net *net, struct inode *inode)
3704 return locks_in_grace(net) && mandatory_lock(inode);
3707 /* Returns true iff a is later than b: */
3708 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3710 return (s32)(a->si_generation - b->si_generation) > 0;
3713 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3716 * When sessions are used the stateid generation number is ignored
3719 if (has_session && in->si_generation == 0)
3722 if (in->si_generation == ref->si_generation)
3725 /* If the client sends us a stateid from the future, it's buggy: */
3726 if (stateid_generation_after(in, ref))
3727 return nfserr_bad_stateid;
3729 * However, we could see a stateid from the past, even from a
3730 * non-buggy client. For example, if the client sends a lock
3731 * while some IO is outstanding, the lock may bump si_generation
3732 * while the IO is still in flight. The client could avoid that
3733 * situation by waiting for responses on all the IO requests,
3734 * but better performance may result in retrying IO that
3735 * receives an old_stateid error if requests are rarely
3736 * reordered in flight:
3738 return nfserr_old_stateid;
3741 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3743 struct nfs4_stid *s;
3744 struct nfs4_ol_stateid *ols;
3747 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3748 return nfserr_bad_stateid;
3749 /* Client debugging aid. */
3750 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3751 char addr_str[INET6_ADDRSTRLEN];
3752 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3754 pr_warn_ratelimited("NFSD: client %s testing state ID "
3755 "with incorrect client ID\n", addr_str);
3756 return nfserr_bad_stateid;
3758 s = find_stateid(cl, stateid);
3760 return nfserr_bad_stateid;
3761 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3764 switch (s->sc_type) {
3765 case NFS4_DELEG_STID:
3767 case NFS4_REVOKED_DELEG_STID:
3768 return nfserr_deleg_revoked;
3769 case NFS4_OPEN_STID:
3770 case NFS4_LOCK_STID:
3771 ols = openlockstateid(s);
3772 if (ols->st_stateowner->so_is_open_owner
3773 && !(openowner(ols->st_stateowner)->oo_flags
3774 & NFS4_OO_CONFIRMED))
3775 return nfserr_bad_stateid;
3778 printk("unknown stateid type %x\n", s->sc_type);
3779 case NFS4_CLOSED_STID:
3780 return nfserr_bad_stateid;
3785 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
3786 stateid_t *stateid, unsigned char typemask,
3787 struct nfs4_stid **s, struct nfsd_net *nn)
3789 struct nfs4_client *cl;
3791 bool sessions = cstate->minorversion != 0;
3793 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3794 return nfserr_bad_stateid;
3795 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3797 if (status == nfserr_stale_clientid) {
3799 return nfserr_bad_stateid;
3800 return nfserr_stale_stateid;
3804 *s = find_stateid_by_type(cl, stateid, typemask);
3806 return nfserr_bad_stateid;
3811 * Checks for stateid operations
3814 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3815 stateid_t *stateid, int flags, struct file **filpp)
3817 struct nfs4_stid *s;
3818 struct nfs4_ol_stateid *stp = NULL;
3819 struct nfs4_delegation *dp = NULL;
3820 struct svc_fh *current_fh = &cstate->current_fh;
3821 struct inode *ino = current_fh->fh_dentry->d_inode;
3822 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3823 struct file *file = NULL;
3829 if (grace_disallows_io(net, ino))
3830 return nfserr_grace;
3832 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3833 return check_special_stateids(net, current_fh, stateid, flags);
3837 status = nfsd4_lookup_stateid(cstate, stateid,
3838 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3842 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3845 switch (s->sc_type) {
3846 case NFS4_DELEG_STID:
3847 dp = delegstateid(s);
3848 status = nfs4_check_delegmode(dp, flags);
3852 file = dp->dl_file->fi_deleg_file;
3855 status = nfserr_serverfault;
3860 case NFS4_OPEN_STID:
3861 case NFS4_LOCK_STID:
3862 stp = openlockstateid(s);
3863 status = nfs4_check_fh(current_fh, stp);
3866 if (stp->st_stateowner->so_is_open_owner
3867 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3869 status = nfs4_check_openmode(stp, flags);
3873 if (flags & RD_STATE)
3874 file = find_readable_file(stp->st_file);
3876 file = find_writeable_file(stp->st_file);
3880 status = nfserr_bad_stateid;
3885 *filpp = get_file(file);
3887 nfs4_unlock_state();
3892 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3894 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
3896 if (check_for_locks(stp->st_file, lo))
3897 return nfserr_locks_held;
3898 release_lockowner_if_empty(lo);
3903 * Test if the stateid is valid
3906 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3907 struct nfsd4_test_stateid *test_stateid)
3909 struct nfsd4_test_stateid_id *stateid;
3910 struct nfs4_client *cl = cstate->session->se_client;
3913 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3914 stateid->ts_id_status =
3915 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3916 nfs4_unlock_state();
3922 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3923 struct nfsd4_free_stateid *free_stateid)
3925 stateid_t *stateid = &free_stateid->fr_stateid;
3926 struct nfs4_stid *s;
3927 struct nfs4_delegation *dp;
3928 struct nfs4_client *cl = cstate->session->se_client;
3929 __be32 ret = nfserr_bad_stateid;
3932 s = find_stateid(cl, stateid);
3935 switch (s->sc_type) {
3936 case NFS4_DELEG_STID:
3937 ret = nfserr_locks_held;
3939 case NFS4_OPEN_STID:
3940 case NFS4_LOCK_STID:
3941 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3944 if (s->sc_type == NFS4_LOCK_STID)
3945 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3947 ret = nfserr_locks_held;
3949 case NFS4_REVOKED_DELEG_STID:
3950 dp = delegstateid(s);
3951 destroy_revoked_delegation(dp);
3955 ret = nfserr_bad_stateid;
3958 nfs4_unlock_state();
3965 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3966 RD_STATE : WR_STATE;
3969 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3971 struct svc_fh *current_fh = &cstate->current_fh;
3972 struct nfs4_stateowner *sop = stp->st_stateowner;
3975 status = nfsd4_check_seqid(cstate, sop, seqid);
3978 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
3979 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
3981 * "Closed" stateid's exist *only* to return
3982 * nfserr_replay_me from the previous step, and
3983 * revoked delegations are kept only for free_stateid.
3985 return nfserr_bad_stateid;
3986 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3989 return nfs4_check_fh(current_fh, stp);
3993 * Checks for sequence id mutating operations.
3996 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3997 stateid_t *stateid, char typemask,
3998 struct nfs4_ol_stateid **stpp,
3999 struct nfsd_net *nn)
4002 struct nfs4_stid *s;
4003 struct nfs4_ol_stateid *stp = NULL;
4005 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4006 seqid, STATEID_VAL(stateid));
4009 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4012 stp = openlockstateid(s);
4013 if (!nfsd4_has_session(cstate))
4014 cstate->replay_owner = stp->st_stateowner;
4016 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4022 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4023 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4026 struct nfs4_openowner *oo;
4028 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4029 NFS4_OPEN_STID, stpp, nn);
4032 oo = openowner((*stpp)->st_stateowner);
4033 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
4034 return nfserr_bad_stateid;
4039 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4040 struct nfsd4_open_confirm *oc)
4043 struct nfs4_openowner *oo;
4044 struct nfs4_ol_stateid *stp;
4045 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4047 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4048 cstate->current_fh.fh_dentry);
4050 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4056 status = nfs4_preprocess_seqid_op(cstate,
4057 oc->oc_seqid, &oc->oc_req_stateid,
4058 NFS4_OPEN_STID, &stp, nn);
4061 oo = openowner(stp->st_stateowner);
4062 status = nfserr_bad_stateid;
4063 if (oo->oo_flags & NFS4_OO_CONFIRMED)
4065 oo->oo_flags |= NFS4_OO_CONFIRMED;
4066 update_stateid(&stp->st_stid.sc_stateid);
4067 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4068 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4069 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4071 nfsd4_client_record_create(oo->oo_owner.so_client);
4074 nfsd4_bump_seqid(cstate, status);
4075 if (!cstate->replay_owner)
4076 nfs4_unlock_state();
4080 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4082 if (!test_access(access, stp))
4084 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
4085 clear_access(access, stp);
4088 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4090 switch (to_access) {
4091 case NFS4_SHARE_ACCESS_READ:
4092 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4093 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4095 case NFS4_SHARE_ACCESS_WRITE:
4096 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4097 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4099 case NFS4_SHARE_ACCESS_BOTH:
4107 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
4110 for (i = 0; i < 4; i++) {
4111 if ((i & deny) != i)
4117 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4118 struct nfsd4_compound_state *cstate,
4119 struct nfsd4_open_downgrade *od)
4122 struct nfs4_ol_stateid *stp;
4123 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4125 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4126 cstate->current_fh.fh_dentry);
4128 /* We don't yet support WANT bits: */
4129 if (od->od_deleg_want)
4130 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4134 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4135 &od->od_stateid, &stp, nn);
4138 status = nfserr_inval;
4139 if (!test_access(od->od_share_access, stp)) {
4140 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
4141 stp->st_access_bmap, od->od_share_access);
4144 if (!test_deny(od->od_share_deny, stp)) {
4145 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
4146 stp->st_deny_bmap, od->od_share_deny);
4149 nfs4_stateid_downgrade(stp, od->od_share_access);
4151 reset_union_bmap_deny(od->od_share_deny, stp);
4153 update_stateid(&stp->st_stid.sc_stateid);
4154 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4157 nfsd4_bump_seqid(cstate, status);
4158 if (!cstate->replay_owner)
4159 nfs4_unlock_state();
4163 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4165 struct nfs4_client *clp = s->st_stid.sc_client;
4166 struct nfs4_openowner *oo = openowner(s->st_stateowner);
4168 s->st_stid.sc_type = NFS4_CLOSED_STID;
4169 unhash_open_stateid(s);
4171 if (clp->cl_minorversion) {
4172 free_generic_stateid(s);
4173 if (list_empty(&oo->oo_owner.so_stateids))
4174 release_openowner(oo);
4176 oo->oo_last_closed_stid = s;
4178 * In the 4.0 case we need to keep the owners around a
4179 * little while to handle CLOSE replay.
4181 if (list_empty(&oo->oo_owner.so_stateids))
4182 move_to_close_lru(oo, clp->net);
4187 * nfs4_unlock_state() called after encode
4190 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4191 struct nfsd4_close *close)
4194 struct nfs4_ol_stateid *stp;
4195 struct net *net = SVC_NET(rqstp);
4196 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4198 dprintk("NFSD: nfsd4_close on file %pd\n",
4199 cstate->current_fh.fh_dentry);
4202 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4204 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4206 nfsd4_bump_seqid(cstate, status);
4209 update_stateid(&stp->st_stid.sc_stateid);
4210 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4212 nfsd4_close_open_stateid(stp);
4214 if (!cstate->replay_owner)
4215 nfs4_unlock_state();
4220 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4221 struct nfsd4_delegreturn *dr)
4223 struct nfs4_delegation *dp;
4224 stateid_t *stateid = &dr->dr_stateid;
4225 struct nfs4_stid *s;
4227 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4229 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4233 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
4236 dp = delegstateid(s);
4237 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4241 destroy_delegation(dp);
4243 nfs4_unlock_state();
4249 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4252 end_offset(u64 start, u64 len)
4257 return end >= start ? end: NFS4_MAX_UINT64;
4260 /* last octet in a range */
4262 last_byte_offset(u64 start, u64 len)
4268 return end > start ? end - 1: NFS4_MAX_UINT64;
4272 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4273 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4274 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4275 * locking, this prevents us from being completely protocol-compliant. The
4276 * real solution to this problem is to start using unsigned file offsets in
4277 * the VFS, but this is a very deep change!
4280 nfs4_transform_lock_offset(struct file_lock *lock)
4282 if (lock->fl_start < 0)
4283 lock->fl_start = OFFSET_MAX;
4284 if (lock->fl_end < 0)
4285 lock->fl_end = OFFSET_MAX;
4288 /* Hack!: For now, we're defining this just so we can use a pointer to it
4289 * as a unique cookie to identify our (NFSv4's) posix locks. */
4290 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4294 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4296 struct nfs4_lockowner *lo;
4298 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4299 lo = (struct nfs4_lockowner *) fl->fl_owner;
4300 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4301 lo->lo_owner.so_owner.len, GFP_KERNEL);
4302 if (!deny->ld_owner.data)
4303 /* We just don't care that much */
4305 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4306 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4309 deny->ld_owner.len = 0;
4310 deny->ld_owner.data = NULL;
4311 deny->ld_clientid.cl_boot = 0;
4312 deny->ld_clientid.cl_id = 0;
4314 deny->ld_start = fl->fl_start;
4315 deny->ld_length = NFS4_MAX_UINT64;
4316 if (fl->fl_end != NFS4_MAX_UINT64)
4317 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4318 deny->ld_type = NFS4_READ_LT;
4319 if (fl->fl_type != F_RDLCK)
4320 deny->ld_type = NFS4_WRITE_LT;
4323 static struct nfs4_lockowner *
4324 find_lockowner_str(clientid_t *clid, struct xdr_netobj *owner,
4325 struct nfsd_net *nn)
4327 unsigned int strhashval = ownerstr_hashval(clid->cl_id, owner);
4328 struct nfs4_stateowner *so;
4330 list_for_each_entry(so, &nn->ownerstr_hashtbl[strhashval], so_strhash) {
4331 if (so->so_is_open_owner)
4333 if (!same_owner_str(so, owner, clid))
4335 return lockowner(so);
4341 * Alloc a lock owner structure.
4342 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4345 * strhashval = ownerstr_hashval
4347 static struct nfs4_lockowner *
4348 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4349 struct nfs4_lockowner *lo;
4350 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4352 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4355 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4356 lo->lo_owner.so_is_open_owner = 0;
4357 /* It is the openowner seqid that will be incremented in encode in the
4358 * case of new lockowners; so increment the lock seqid manually: */
4359 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4360 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4364 static struct nfs4_ol_stateid *
4365 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4367 struct nfs4_ol_stateid *stp;
4368 struct nfs4_client *clp = lo->lo_owner.so_client;
4370 stp = nfs4_alloc_stateid(clp);
4373 stp->st_stid.sc_type = NFS4_LOCK_STID;
4374 list_add(&stp->st_perfile, &fp->fi_stateids);
4375 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4376 stp->st_stateowner = &lo->lo_owner;
4379 stp->st_access_bmap = 0;
4380 stp->st_deny_bmap = open_stp->st_deny_bmap;
4381 stp->st_openstp = open_stp;
4382 list_add(&stp->st_locks, &open_stp->st_locks);
4386 static struct nfs4_ol_stateid *
4387 find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
4389 struct nfs4_ol_stateid *lst;
4391 list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
4392 if (lst->st_file == fp)
4400 check_lock_length(u64 offset, u64 length)
4402 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4403 LOFF_OVERFLOW(offset, length)));
4406 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4408 struct nfs4_file *fp = lock_stp->st_file;
4409 int oflag = nfs4_access_to_omode(access);
4411 if (test_access(access, lock_stp))
4413 nfs4_file_get_access(fp, oflag);
4414 set_access(access, lock_stp);
4417 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)
4419 struct nfs4_file *fi = ost->st_file;
4420 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4421 struct nfs4_client *cl = oo->oo_owner.so_client;
4422 struct nfs4_lockowner *lo;
4423 unsigned int strhashval;
4424 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4426 lo = find_lockowner_str(&cl->cl_clientid, &lock->v.new.owner, nn);
4428 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4429 &lock->v.new.owner);
4430 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4432 return nfserr_jukebox;
4434 /* with an existing lockowner, seqids must be the same */
4435 if (!cstate->minorversion &&
4436 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
4437 return nfserr_bad_seqid;
4440 *lst = find_lock_stateid(lo, fi);
4442 *lst = alloc_init_lock_stateid(lo, fi, ost);
4444 release_lockowner_if_empty(lo);
4445 return nfserr_jukebox;
4456 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4457 struct nfsd4_lock *lock)
4459 struct nfs4_openowner *open_sop = NULL;
4460 struct nfs4_lockowner *lock_sop = NULL;
4461 struct nfs4_ol_stateid *lock_stp;
4462 struct file *filp = NULL;
4463 struct file_lock *file_lock = NULL;
4464 struct file_lock *conflock = NULL;
4466 bool new_state = false;
4469 struct net *net = SVC_NET(rqstp);
4470 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4472 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4473 (long long) lock->lk_offset,
4474 (long long) lock->lk_length);
4476 if (check_lock_length(lock->lk_offset, lock->lk_length))
4477 return nfserr_inval;
4479 if ((status = fh_verify(rqstp, &cstate->current_fh,
4480 S_IFREG, NFSD_MAY_LOCK))) {
4481 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4487 if (lock->lk_is_new) {
4488 struct nfs4_ol_stateid *open_stp = NULL;
4490 if (nfsd4_has_session(cstate))
4491 /* See rfc 5661 18.10.3: given clientid is ignored: */
4492 memcpy(&lock->v.new.clientid,
4493 &cstate->session->se_client->cl_clientid,
4494 sizeof(clientid_t));
4496 status = nfserr_stale_clientid;
4497 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4500 /* validate and update open stateid and open seqid */
4501 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4502 lock->lk_new_open_seqid,
4503 &lock->lk_new_open_stateid,
4507 open_sop = openowner(open_stp->st_stateowner);
4508 status = nfserr_bad_stateid;
4509 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4510 &lock->v.new.clientid))
4512 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4513 &lock_stp, &new_state);
4515 status = nfs4_preprocess_seqid_op(cstate,
4516 lock->lk_old_lock_seqid,
4517 &lock->lk_old_lock_stateid,
4518 NFS4_LOCK_STID, &lock_stp, nn);
4521 lock_sop = lockowner(lock_stp->st_stateowner);
4523 lkflg = setlkflg(lock->lk_type);
4524 status = nfs4_check_openmode(lock_stp, lkflg);
4528 status = nfserr_grace;
4529 if (locks_in_grace(net) && !lock->lk_reclaim)
4531 status = nfserr_no_grace;
4532 if (!locks_in_grace(net) && lock->lk_reclaim)
4535 file_lock = locks_alloc_lock();
4537 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4538 status = nfserr_jukebox;
4542 locks_init_lock(file_lock);
4543 switch (lock->lk_type) {
4546 filp = find_readable_file(lock_stp->st_file);
4548 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4549 file_lock->fl_type = F_RDLCK;
4552 case NFS4_WRITEW_LT:
4553 filp = find_writeable_file(lock_stp->st_file);
4555 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4556 file_lock->fl_type = F_WRLCK;
4559 status = nfserr_inval;
4563 status = nfserr_openmode;
4566 file_lock->fl_owner = (fl_owner_t)lock_sop;
4567 file_lock->fl_pid = current->tgid;
4568 file_lock->fl_file = filp;
4569 file_lock->fl_flags = FL_POSIX;
4570 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4571 file_lock->fl_start = lock->lk_offset;
4572 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4573 nfs4_transform_lock_offset(file_lock);
4575 conflock = locks_alloc_lock();
4577 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4578 status = nfserr_jukebox;
4582 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4584 case 0: /* success! */
4585 update_stateid(&lock_stp->st_stid.sc_stateid);
4586 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4590 case (EAGAIN): /* conflock holds conflicting lock */
4591 status = nfserr_denied;
4592 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4593 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4596 status = nfserr_deadlock;
4599 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4600 status = nfserrno(err);
4604 if (status && new_state)
4605 release_lock_stateid(lock_stp);
4606 nfsd4_bump_seqid(cstate, status);
4607 if (!cstate->replay_owner)
4608 nfs4_unlock_state();
4610 locks_free_lock(file_lock);
4612 locks_free_lock(conflock);
4617 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4618 * so we do a temporary open here just to get an open file to pass to
4619 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4622 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4625 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4627 err = nfserrno(vfs_test_lock(file, lock));
4637 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4638 struct nfsd4_lockt *lockt)
4640 struct file_lock *file_lock = NULL;
4641 struct nfs4_lockowner *lo;
4643 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4645 if (locks_in_grace(SVC_NET(rqstp)))
4646 return nfserr_grace;
4648 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4649 return nfserr_inval;
4653 if (!nfsd4_has_session(cstate)) {
4654 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4659 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4662 file_lock = locks_alloc_lock();
4664 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4665 status = nfserr_jukebox;
4668 locks_init_lock(file_lock);
4669 switch (lockt->lt_type) {
4672 file_lock->fl_type = F_RDLCK;
4675 case NFS4_WRITEW_LT:
4676 file_lock->fl_type = F_WRLCK;
4679 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4680 status = nfserr_inval;
4684 lo = find_lockowner_str(&lockt->lt_clientid, &lockt->lt_owner, nn);
4686 file_lock->fl_owner = (fl_owner_t)lo;
4687 file_lock->fl_pid = current->tgid;
4688 file_lock->fl_flags = FL_POSIX;
4690 file_lock->fl_start = lockt->lt_offset;
4691 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4693 nfs4_transform_lock_offset(file_lock);
4695 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4699 if (file_lock->fl_type != F_UNLCK) {
4700 status = nfserr_denied;
4701 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4704 nfs4_unlock_state();
4706 locks_free_lock(file_lock);
4711 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4712 struct nfsd4_locku *locku)
4714 struct nfs4_ol_stateid *stp;
4715 struct file *filp = NULL;
4716 struct file_lock *file_lock = NULL;
4719 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4721 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4722 (long long) locku->lu_offset,
4723 (long long) locku->lu_length);
4725 if (check_lock_length(locku->lu_offset, locku->lu_length))
4726 return nfserr_inval;
4730 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4731 &locku->lu_stateid, NFS4_LOCK_STID,
4735 filp = find_any_file(stp->st_file);
4737 status = nfserr_lock_range;
4740 file_lock = locks_alloc_lock();
4742 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4743 status = nfserr_jukebox;
4746 locks_init_lock(file_lock);
4747 file_lock->fl_type = F_UNLCK;
4748 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4749 file_lock->fl_pid = current->tgid;
4750 file_lock->fl_file = filp;
4751 file_lock->fl_flags = FL_POSIX;
4752 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4753 file_lock->fl_start = locku->lu_offset;
4755 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4757 nfs4_transform_lock_offset(file_lock);
4759 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4761 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4764 update_stateid(&stp->st_stid.sc_stateid);
4765 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4768 nfsd4_bump_seqid(cstate, status);
4769 if (!cstate->replay_owner)
4770 nfs4_unlock_state();
4772 locks_free_lock(file_lock);
4776 status = nfserrno(err);
4782 * 1: locks held by lockowner
4783 * 0: no locks held by lockowner
4786 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4788 struct file_lock **flpp;
4789 struct inode *inode = filp->fi_inode;
4792 spin_lock(&inode->i_lock);
4793 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4794 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4800 spin_unlock(&inode->i_lock);
4805 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4806 struct nfsd4_compound_state *cstate,
4807 struct nfsd4_release_lockowner *rlockowner)
4809 clientid_t *clid = &rlockowner->rl_clientid;
4810 struct nfs4_stateowner *sop = NULL, *tmp;
4811 struct nfs4_lockowner *lo;
4812 struct nfs4_ol_stateid *stp;
4813 struct xdr_netobj *owner = &rlockowner->rl_owner;
4814 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4816 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4818 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4819 clid->cl_boot, clid->cl_id);
4823 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4827 status = nfserr_locks_held;
4829 /* Find the matching lock stateowner */
4830 list_for_each_entry(tmp, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4831 if (tmp->so_is_open_owner)
4833 if (same_owner_str(tmp, owner, clid)) {
4839 /* No matching owner found, maybe a replay? Just declare victory... */
4845 lo = lockowner(sop);
4846 /* see if there are still any locks associated with it */
4847 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
4848 if (check_for_locks(stp->st_file, lo))
4853 release_lockowner(lo);
4855 nfs4_unlock_state();
4859 static inline struct nfs4_client_reclaim *
4862 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4866 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4868 struct nfs4_client_reclaim *crp;
4870 crp = nfsd4_find_reclaim_client(name, nn);
4871 return (crp && crp->cr_clp);
4875 * failure => all reset bets are off, nfserr_no_grace...
4877 struct nfs4_client_reclaim *
4878 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4880 unsigned int strhashval;
4881 struct nfs4_client_reclaim *crp;
4883 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4884 crp = alloc_reclaim();
4886 strhashval = clientstr_hashval(name);
4887 INIT_LIST_HEAD(&crp->cr_strhash);
4888 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4889 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4891 nn->reclaim_str_hashtbl_size++;
4897 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4899 list_del(&crp->cr_strhash);
4901 nn->reclaim_str_hashtbl_size--;
4905 nfs4_release_reclaim(struct nfsd_net *nn)
4907 struct nfs4_client_reclaim *crp = NULL;
4910 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4911 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4912 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4913 struct nfs4_client_reclaim, cr_strhash);
4914 nfs4_remove_reclaim_record(crp, nn);
4917 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4921 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4922 struct nfs4_client_reclaim *
4923 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4925 unsigned int strhashval;
4926 struct nfs4_client_reclaim *crp = NULL;
4928 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4930 strhashval = clientstr_hashval(recdir);
4931 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4932 if (same_name(crp->cr_recdir, recdir)) {
4940 * Called from OPEN. Look for clientid in reclaim list.
4943 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4945 struct nfs4_client *clp;
4947 /* find clientid in conf_id_hashtbl */
4948 clp = find_confirmed_client(clid, sessions, nn);
4950 return nfserr_reclaim_bad;
4952 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4955 #ifdef CONFIG_NFSD_FAULT_INJECTION
4957 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4959 if (mark_client_expired(clp))
4965 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4967 char buf[INET6_ADDRSTRLEN];
4968 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4969 printk(KERN_INFO "NFS Client: %s\n", buf);
4973 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4976 char buf[INET6_ADDRSTRLEN];
4977 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4978 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4981 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
4982 void (*func)(struct nfs4_ol_stateid *))
4984 struct nfs4_openowner *oop;
4985 struct nfs4_ol_stateid *stp, *st_next;
4986 struct nfs4_ol_stateid *lst, *lst_next;
4989 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4990 list_for_each_entry_safe(stp, st_next,
4991 &oop->oo_owner.so_stateids, st_perstateowner) {
4992 list_for_each_entry_safe(lst, lst_next,
4993 &stp->st_locks, st_locks) {
5005 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
5007 return nfsd_foreach_client_lock(clp, max, release_lock_stateid);
5010 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
5012 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
5013 nfsd_print_count(clp, count, "locked files");
5017 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
5019 struct nfs4_openowner *oop, *next;
5022 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
5032 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
5034 return nfsd_foreach_client_open(clp, max, release_openowner);
5037 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
5039 u64 count = nfsd_foreach_client_open(clp, max, NULL);
5040 nfsd_print_count(clp, count, "open files");
5044 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
5045 struct list_head *victims)
5047 struct nfs4_delegation *dp, *next;
5050 lockdep_assert_held(&state_lock);
5051 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
5053 list_move(&dp->dl_recall_lru, victims);
5060 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
5062 struct nfs4_delegation *dp, *next;
5066 spin_lock(&state_lock);
5067 count = nfsd_find_all_delegations(clp, max, &victims);
5068 spin_unlock(&state_lock);
5070 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
5071 revoke_delegation(dp);
5076 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
5078 struct nfs4_delegation *dp, *next;
5082 spin_lock(&state_lock);
5083 count = nfsd_find_all_delegations(clp, max, &victims);
5084 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
5085 nfsd_break_one_deleg(dp);
5086 spin_unlock(&state_lock);
5091 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
5095 spin_lock(&state_lock);
5096 count = nfsd_find_all_delegations(clp, max, NULL);
5097 spin_unlock(&state_lock);
5099 nfsd_print_count(clp, count, "delegations");
5103 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
5105 struct nfs4_client *clp, *next;
5107 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5109 if (!nfsd_netns_ready(nn))
5112 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5113 count += func(clp, max - count);
5114 if ((max != 0) && (count >= max))
5121 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5123 struct nfs4_client *clp;
5124 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5126 if (!nfsd_netns_ready(nn))
5129 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5130 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5136 #endif /* CONFIG_NFSD_FAULT_INJECTION */
5139 * Since the lifetime of a delegation isn't limited to that of an open, a
5140 * client may quite reasonably hang on to a delegation as long as it has
5141 * the inode cached. This becomes an obvious problem the first time a
5142 * client's inode cache approaches the size of the server's total memory.
5144 * For now we avoid this problem by imposing a hard limit on the number
5145 * of delegations, which varies according to the server's memory size.
5148 set_max_delegations(void)
5151 * Allow at most 4 delegations per megabyte of RAM. Quick
5152 * estimates suggest that in the worst case (where every delegation
5153 * is for a different inode), a delegation could take about 1.5K,
5154 * giving a worst case usage of about 6% of memory.
5156 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
5159 static int nfs4_state_create_net(struct net *net)
5161 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5164 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5165 CLIENT_HASH_SIZE, GFP_KERNEL);
5166 if (!nn->conf_id_hashtbl)
5168 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5169 CLIENT_HASH_SIZE, GFP_KERNEL);
5170 if (!nn->unconf_id_hashtbl)
5172 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
5173 OWNER_HASH_SIZE, GFP_KERNEL);
5174 if (!nn->ownerstr_hashtbl)
5176 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
5177 SESSION_HASH_SIZE, GFP_KERNEL);
5178 if (!nn->sessionid_hashtbl)
5181 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5182 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5183 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5185 for (i = 0; i < OWNER_HASH_SIZE; i++)
5186 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
5187 for (i = 0; i < SESSION_HASH_SIZE; i++)
5188 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5189 nn->conf_name_tree = RB_ROOT;
5190 nn->unconf_name_tree = RB_ROOT;
5191 INIT_LIST_HEAD(&nn->client_lru);
5192 INIT_LIST_HEAD(&nn->close_lru);
5193 INIT_LIST_HEAD(&nn->del_recall_lru);
5194 spin_lock_init(&nn->client_lock);
5196 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5202 kfree(nn->ownerstr_hashtbl);
5204 kfree(nn->unconf_id_hashtbl);
5206 kfree(nn->conf_id_hashtbl);
5212 nfs4_state_destroy_net(struct net *net)
5215 struct nfs4_client *clp = NULL;
5216 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5218 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5219 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5220 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5221 destroy_client(clp);
5225 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5226 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
5227 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5228 destroy_client(clp);
5232 kfree(nn->sessionid_hashtbl);
5233 kfree(nn->ownerstr_hashtbl);
5234 kfree(nn->unconf_id_hashtbl);
5235 kfree(nn->conf_id_hashtbl);
5240 nfs4_state_start_net(struct net *net)
5242 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5245 ret = nfs4_state_create_net(net);
5248 nfsd4_client_tracking_init(net);
5249 nn->boot_time = get_seconds();
5250 locks_start_grace(net, &nn->nfsd4_manager);
5251 nn->grace_ended = false;
5252 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5253 nn->nfsd4_grace, net);
5254 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5258 /* initialization to perform when the nfsd service is started: */
5261 nfs4_state_start(void)
5265 ret = set_callback_cred();
5268 laundry_wq = create_singlethread_workqueue("nfsd4");
5269 if (laundry_wq == NULL) {
5273 ret = nfsd4_create_callback_queue();
5275 goto out_free_laundry;
5277 set_max_delegations();
5282 destroy_workqueue(laundry_wq);
5288 nfs4_state_shutdown_net(struct net *net)
5290 struct nfs4_delegation *dp = NULL;
5291 struct list_head *pos, *next, reaplist;
5292 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5294 cancel_delayed_work_sync(&nn->laundromat_work);
5295 locks_end_grace(&nn->nfsd4_manager);
5298 INIT_LIST_HEAD(&reaplist);
5299 spin_lock(&state_lock);
5300 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5301 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5302 list_move(&dp->dl_recall_lru, &reaplist);
5304 spin_unlock(&state_lock);
5305 list_for_each_safe(pos, next, &reaplist) {
5306 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5307 destroy_delegation(dp);
5310 nfsd4_client_tracking_exit(net);
5311 nfs4_state_destroy_net(net);
5312 nfs4_unlock_state();
5316 nfs4_state_shutdown(void)
5318 destroy_workqueue(laundry_wq);
5319 nfsd4_destroy_callback_queue();
5323 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5325 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5326 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5330 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5332 if (cstate->minorversion) {
5333 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5334 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5339 clear_current_stateid(struct nfsd4_compound_state *cstate)
5341 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5345 * functions to set current state id
5348 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5350 put_stateid(cstate, &odp->od_stateid);
5354 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5356 put_stateid(cstate, &open->op_stateid);
5360 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5362 put_stateid(cstate, &close->cl_stateid);
5366 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5368 put_stateid(cstate, &lock->lk_resp_stateid);
5372 * functions to consume current state id
5376 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5378 get_stateid(cstate, &odp->od_stateid);
5382 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5384 get_stateid(cstate, &drp->dr_stateid);
5388 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5390 get_stateid(cstate, &fsp->fr_stateid);
5394 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5396 get_stateid(cstate, &setattr->sa_stateid);
5400 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5402 get_stateid(cstate, &close->cl_stateid);
5406 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5408 get_stateid(cstate, &locku->lu_stateid);
5412 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5414 get_stateid(cstate, &read->rd_stateid);
5418 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5420 get_stateid(cstate, &write->wr_stateid);