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 void nfsd4_put_session(struct nfsd4_session *ses)
108 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
112 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
114 if (atomic_read(&ses->se_ref) > ref_held_by_me)
115 return nfserr_jukebox;
116 ses->se_flags |= NFS4_SESSION_DEAD;
120 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
122 if (is_session_dead(ses))
123 return nfserr_badsession;
124 atomic_inc(&ses->se_ref);
129 nfs4_unlock_state(void)
131 mutex_unlock(&client_mutex);
134 static bool is_client_expired(struct nfs4_client *clp)
136 return clp->cl_time == 0;
139 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
141 if (atomic_read(&clp->cl_refcount))
142 return nfserr_jukebox;
147 static __be32 mark_client_expired(struct nfs4_client *clp)
149 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
152 spin_lock(&nn->client_lock);
153 ret = mark_client_expired_locked(clp);
154 spin_unlock(&nn->client_lock);
158 static __be32 get_client_locked(struct nfs4_client *clp)
160 if (is_client_expired(clp))
161 return nfserr_expired;
162 atomic_inc(&clp->cl_refcount);
166 /* must be called under the client_lock */
168 renew_client_locked(struct nfs4_client *clp)
170 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
172 if (is_client_expired(clp)) {
174 printk("%s: client (clientid %08x/%08x) already expired\n",
176 clp->cl_clientid.cl_boot,
177 clp->cl_clientid.cl_id);
181 dprintk("renewing client (clientid %08x/%08x)\n",
182 clp->cl_clientid.cl_boot,
183 clp->cl_clientid.cl_id);
184 list_move_tail(&clp->cl_lru, &nn->client_lru);
185 clp->cl_time = get_seconds();
189 renew_client(struct nfs4_client *clp)
191 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
193 spin_lock(&nn->client_lock);
194 renew_client_locked(clp);
195 spin_unlock(&nn->client_lock);
198 static void put_client_renew_locked(struct nfs4_client *clp)
200 if (!atomic_dec_and_test(&clp->cl_refcount))
202 if (!is_client_expired(clp))
203 renew_client_locked(clp);
206 void put_client_renew(struct nfs4_client *clp)
208 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
210 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
212 if (!is_client_expired(clp))
213 renew_client_locked(clp);
214 spin_unlock(&nn->client_lock);
219 opaque_hashval(const void *ptr, int nbytes)
221 unsigned char *cptr = (unsigned char *) ptr;
231 static void nfsd4_free_file(struct nfs4_file *f)
233 kmem_cache_free(file_slab, f);
237 put_nfs4_file(struct nfs4_file *fi)
239 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
240 hlist_del(&fi->fi_hash);
241 spin_unlock(&state_lock);
248 get_nfs4_file(struct nfs4_file *fi)
250 atomic_inc(&fi->fi_ref);
253 static int num_delegations;
254 unsigned long max_delegations;
257 * Open owner state (share locks)
260 /* hash tables for lock and open owners */
261 #define OWNER_HASH_BITS 8
262 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
263 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
265 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
269 ret = opaque_hashval(ownername->data, ownername->len);
271 return ret & OWNER_HASH_MASK;
274 /* hash table for nfs4_file */
275 #define FILE_HASH_BITS 8
276 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
278 static unsigned int file_hashval(struct inode *ino)
280 /* XXX: why are we hashing on inode pointer, anyway? */
281 return hash_ptr(ino, FILE_HASH_BITS);
284 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
286 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
288 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
289 atomic_inc(&fp->fi_access[oflag]);
292 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
294 if (oflag == O_RDWR) {
295 __nfs4_file_get_access(fp, O_RDONLY);
296 __nfs4_file_get_access(fp, O_WRONLY);
298 __nfs4_file_get_access(fp, oflag);
301 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
303 if (fp->fi_fds[oflag]) {
304 fput(fp->fi_fds[oflag]);
305 fp->fi_fds[oflag] = NULL;
309 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
311 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
312 nfs4_file_put_fd(fp, oflag);
313 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
314 nfs4_file_put_fd(fp, O_RDWR);
318 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
320 if (oflag == O_RDWR) {
321 __nfs4_file_put_access(fp, O_RDONLY);
322 __nfs4_file_put_access(fp, O_WRONLY);
324 __nfs4_file_put_access(fp, oflag);
327 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
330 struct idr *stateids = &cl->cl_stateids;
331 struct nfs4_stid *stid;
334 stid = kmem_cache_alloc(slab, GFP_KERNEL);
338 new_id = idr_alloc_cyclic(stateids, stid, 0, 0, GFP_KERNEL);
341 stid->sc_client = cl;
343 stid->sc_stateid.si_opaque.so_id = new_id;
344 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
345 /* Will be incremented before return to client: */
346 stid->sc_stateid.si_generation = 0;
349 * It shouldn't be a problem to reuse an opaque stateid value.
350 * I don't think it is for 4.1. But with 4.0 I worry that, for
351 * example, a stray write retransmission could be accepted by
352 * the server when it should have been rejected. Therefore,
353 * adopt a trick from the sctp code to attempt to maximize the
354 * amount of time until an id is reused, by ensuring they always
355 * "increase" (mod INT_MAX):
359 kmem_cache_free(slab, stid);
363 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
365 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
369 * When we recall a delegation, we should be careful not to hand it
370 * out again straight away.
371 * To ensure this we keep a pair of bloom filters ('new' and 'old')
372 * in which the filehandles of recalled delegations are "stored".
373 * If a filehandle appear in either filter, a delegation is blocked.
374 * When a delegation is recalled, the filehandle is stored in the "new"
376 * Every 30 seconds we swap the filters and clear the "new" one,
377 * unless both are empty of course.
379 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
380 * low 3 bytes as hash-table indices.
382 * 'state_lock', which is always held when block_delegations() is called,
383 * is used to manage concurrent access. Testing does not need the lock
384 * except when swapping the two filters.
386 static struct bloom_pair {
387 int entries, old_entries;
389 int new; /* index into 'set' */
390 DECLARE_BITMAP(set[2], 256);
391 } blocked_delegations;
393 static int delegation_blocked(struct knfsd_fh *fh)
396 struct bloom_pair *bd = &blocked_delegations;
398 if (bd->entries == 0)
400 if (seconds_since_boot() - bd->swap_time > 30) {
401 spin_lock(&state_lock);
402 if (seconds_since_boot() - bd->swap_time > 30) {
403 bd->entries -= bd->old_entries;
404 bd->old_entries = bd->entries;
405 memset(bd->set[bd->new], 0,
408 bd->swap_time = seconds_since_boot();
410 spin_unlock(&state_lock);
412 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
413 if (test_bit(hash&255, bd->set[0]) &&
414 test_bit((hash>>8)&255, bd->set[0]) &&
415 test_bit((hash>>16)&255, bd->set[0]))
418 if (test_bit(hash&255, bd->set[1]) &&
419 test_bit((hash>>8)&255, bd->set[1]) &&
420 test_bit((hash>>16)&255, bd->set[1]))
426 static void block_delegations(struct knfsd_fh *fh)
429 struct bloom_pair *bd = &blocked_delegations;
431 hash = arch_fast_hash(&fh->fh_base, fh->fh_size, 0);
433 __set_bit(hash&255, bd->set[bd->new]);
434 __set_bit((hash>>8)&255, bd->set[bd->new]);
435 __set_bit((hash>>16)&255, bd->set[bd->new]);
436 if (bd->entries == 0)
437 bd->swap_time = seconds_since_boot();
441 static struct nfs4_delegation *
442 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh)
444 struct nfs4_delegation *dp;
446 dprintk("NFSD alloc_init_deleg\n");
447 if (num_delegations > max_delegations)
449 if (delegation_blocked(¤t_fh->fh_handle))
451 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
455 * delegation seqid's are never incremented. The 4.1 special
456 * meaning of seqid 0 isn't meaningful, really, but let's avoid
457 * 0 anyway just for consistency and use 1:
459 dp->dl_stid.sc_stateid.si_generation = 1;
461 INIT_LIST_HEAD(&dp->dl_perfile);
462 INIT_LIST_HEAD(&dp->dl_perclnt);
463 INIT_LIST_HEAD(&dp->dl_recall_lru);
465 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
466 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
468 atomic_set(&dp->dl_count, 1);
469 nfsd4_init_callback(&dp->dl_recall);
473 static void remove_stid(struct nfs4_stid *s)
475 struct idr *stateids = &s->sc_client->cl_stateids;
477 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
480 static void nfs4_free_stid(struct kmem_cache *slab, struct nfs4_stid *s)
482 kmem_cache_free(slab, s);
486 nfs4_put_delegation(struct nfs4_delegation *dp)
488 if (atomic_dec_and_test(&dp->dl_count)) {
489 nfs4_free_stid(deleg_slab, &dp->dl_stid);
494 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
498 if (atomic_dec_and_test(&fp->fi_delegees)) {
499 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
501 fput(fp->fi_deleg_file);
502 fp->fi_deleg_file = NULL;
506 static void unhash_stid(struct nfs4_stid *s)
512 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
514 lockdep_assert_held(&state_lock);
516 dp->dl_stid.sc_type = NFS4_DELEG_STID;
517 list_add(&dp->dl_perfile, &fp->fi_delegations);
518 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
521 /* Called under the state lock. */
523 unhash_delegation(struct nfs4_delegation *dp)
525 spin_lock(&state_lock);
526 list_del_init(&dp->dl_perclnt);
527 list_del_init(&dp->dl_perfile);
528 list_del_init(&dp->dl_recall_lru);
529 spin_unlock(&state_lock);
531 nfs4_put_deleg_lease(dp->dl_file);
532 put_nfs4_file(dp->dl_file);
539 static void destroy_revoked_delegation(struct nfs4_delegation *dp)
541 list_del_init(&dp->dl_recall_lru);
542 remove_stid(&dp->dl_stid);
543 nfs4_put_delegation(dp);
546 static void destroy_delegation(struct nfs4_delegation *dp)
548 unhash_delegation(dp);
549 remove_stid(&dp->dl_stid);
550 nfs4_put_delegation(dp);
553 static void revoke_delegation(struct nfs4_delegation *dp)
555 struct nfs4_client *clp = dp->dl_stid.sc_client;
557 if (clp->cl_minorversion == 0)
558 destroy_delegation(dp);
560 unhash_delegation(dp);
561 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
562 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
570 static unsigned int clientid_hashval(u32 id)
572 return id & CLIENT_HASH_MASK;
575 static unsigned int clientstr_hashval(const char *name)
577 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
581 * We store the NONE, READ, WRITE, and BOTH bits separately in the
582 * st_{access,deny}_bmap field of the stateid, in order to track not
583 * only what share bits are currently in force, but also what
584 * combinations of share bits previous opens have used. This allows us
585 * to enforce the recommendation of rfc 3530 14.2.19 that the server
586 * return an error if the client attempt to downgrade to a combination
587 * of share bits not explicable by closing some of its previous opens.
589 * XXX: This enforcement is actually incomplete, since we don't keep
590 * track of access/deny bit combinations; so, e.g., we allow:
592 * OPEN allow read, deny write
593 * OPEN allow both, deny none
594 * DOWNGRADE allow read, deny none
596 * which we should reject.
599 bmap_to_share_mode(unsigned long bmap) {
601 unsigned int access = 0;
603 for (i = 1; i < 4; i++) {
604 if (test_bit(i, &bmap))
611 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
612 unsigned int access, deny;
614 access = bmap_to_share_mode(stp->st_access_bmap);
615 deny = bmap_to_share_mode(stp->st_deny_bmap);
616 if ((access & open->op_share_deny) || (deny & open->op_share_access))
621 /* set share access for a given stateid */
623 set_access(u32 access, struct nfs4_ol_stateid *stp)
625 __set_bit(access, &stp->st_access_bmap);
628 /* clear share access for a given stateid */
630 clear_access(u32 access, struct nfs4_ol_stateid *stp)
632 __clear_bit(access, &stp->st_access_bmap);
635 /* test whether a given stateid has access */
637 test_access(u32 access, struct nfs4_ol_stateid *stp)
639 return test_bit(access, &stp->st_access_bmap);
642 /* set share deny for a given stateid */
644 set_deny(u32 access, struct nfs4_ol_stateid *stp)
646 __set_bit(access, &stp->st_deny_bmap);
649 /* clear share deny for a given stateid */
651 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
653 __clear_bit(access, &stp->st_deny_bmap);
656 /* test whether a given stateid is denying specific access */
658 test_deny(u32 access, struct nfs4_ol_stateid *stp)
660 return test_bit(access, &stp->st_deny_bmap);
663 static int nfs4_access_to_omode(u32 access)
665 switch (access & NFS4_SHARE_ACCESS_BOTH) {
666 case NFS4_SHARE_ACCESS_READ:
668 case NFS4_SHARE_ACCESS_WRITE:
670 case NFS4_SHARE_ACCESS_BOTH:
677 /* release all access and file references for a given stateid */
679 release_all_access(struct nfs4_ol_stateid *stp)
683 for (i = 1; i < 4; i++) {
684 if (test_access(i, stp))
685 nfs4_file_put_access(stp->st_file,
686 nfs4_access_to_omode(i));
687 clear_access(i, stp);
691 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
693 list_del(&stp->st_perfile);
694 list_del(&stp->st_perstateowner);
697 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
699 release_all_access(stp);
700 put_nfs4_file(stp->st_file);
704 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
706 remove_stid(&stp->st_stid);
707 nfs4_free_stid(stateid_slab, &stp->st_stid);
710 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
714 unhash_generic_stateid(stp);
715 unhash_stid(&stp->st_stid);
716 file = find_any_file(stp->st_file);
718 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
719 close_generic_stateid(stp);
720 free_generic_stateid(stp);
723 static void unhash_lockowner(struct nfs4_lockowner *lo)
725 struct nfs4_ol_stateid *stp;
727 list_del(&lo->lo_owner.so_strhash);
728 list_del(&lo->lo_perstateid);
729 list_del(&lo->lo_owner_ino_hash);
730 while (!list_empty(&lo->lo_owner.so_stateids)) {
731 stp = list_first_entry(&lo->lo_owner.so_stateids,
732 struct nfs4_ol_stateid, st_perstateowner);
733 release_lock_stateid(stp);
737 static void nfs4_free_lockowner(struct nfs4_lockowner *lo)
739 kfree(lo->lo_owner.so_owner.data);
740 kmem_cache_free(lockowner_slab, lo);
743 static void release_lockowner(struct nfs4_lockowner *lo)
745 unhash_lockowner(lo);
746 nfs4_free_lockowner(lo);
750 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
752 struct nfs4_lockowner *lo;
754 while (!list_empty(&open_stp->st_lockowners)) {
755 lo = list_entry(open_stp->st_lockowners.next,
756 struct nfs4_lockowner, lo_perstateid);
757 release_lockowner(lo);
761 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
763 unhash_generic_stateid(stp);
764 release_stateid_lockowners(stp);
765 close_generic_stateid(stp);
768 static void release_open_stateid(struct nfs4_ol_stateid *stp)
770 unhash_open_stateid(stp);
771 free_generic_stateid(stp);
774 static void unhash_openowner(struct nfs4_openowner *oo)
776 struct nfs4_ol_stateid *stp;
778 list_del(&oo->oo_owner.so_strhash);
779 list_del(&oo->oo_perclient);
780 while (!list_empty(&oo->oo_owner.so_stateids)) {
781 stp = list_first_entry(&oo->oo_owner.so_stateids,
782 struct nfs4_ol_stateid, st_perstateowner);
783 release_open_stateid(stp);
787 static void release_last_closed_stateid(struct nfs4_openowner *oo)
789 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
792 free_generic_stateid(s);
793 oo->oo_last_closed_stid = NULL;
797 static void nfs4_free_openowner(struct nfs4_openowner *oo)
799 kfree(oo->oo_owner.so_owner.data);
800 kmem_cache_free(openowner_slab, oo);
803 static void release_openowner(struct nfs4_openowner *oo)
805 unhash_openowner(oo);
806 list_del(&oo->oo_close_lru);
807 release_last_closed_stateid(oo);
808 nfs4_free_openowner(oo);
812 hash_sessionid(struct nfs4_sessionid *sessionid)
814 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
816 return sid->sequence % SESSION_HASH_SIZE;
821 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
823 u32 *ptr = (u32 *)(&sessionid->data[0]);
824 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
828 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
834 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
835 * won't be used for replay.
837 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
839 struct nfs4_stateowner *so = cstate->replay_owner;
841 if (nfserr == nfserr_replay_me)
844 if (!seqid_mutating_err(ntohl(nfserr))) {
845 cstate->replay_owner = NULL;
850 if (so->so_is_open_owner)
851 release_last_closed_stateid(openowner(so));
857 gen_sessionid(struct nfsd4_session *ses)
859 struct nfs4_client *clp = ses->se_client;
860 struct nfsd4_sessionid *sid;
862 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
863 sid->clientid = clp->cl_clientid;
864 sid->sequence = current_sessionid++;
869 * The protocol defines ca_maxresponssize_cached to include the size of
870 * the rpc header, but all we need to cache is the data starting after
871 * the end of the initial SEQUENCE operation--the rest we regenerate
872 * each time. Therefore we can advertise a ca_maxresponssize_cached
873 * value that is the number of bytes in our cache plus a few additional
874 * bytes. In order to stay on the safe side, and not promise more than
875 * we can cache, those additional bytes must be the minimum possible: 24
876 * bytes of rpc header (xid through accept state, with AUTH_NULL
877 * verifier), 12 for the compound header (with zero-length tag), and 44
878 * for the SEQUENCE op response:
880 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
883 free_session_slots(struct nfsd4_session *ses)
887 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
888 kfree(ses->se_slots[i]);
892 * We don't actually need to cache the rpc and session headers, so we
893 * can allocate a little less for each slot:
895 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
899 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
902 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
903 return size + sizeof(struct nfsd4_slot);
907 * XXX: If we run out of reserved DRC memory we could (up to a point)
908 * re-negotiate active sessions and reduce their slot usage to make
909 * room for new connections. For now we just fail the create session.
911 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
913 u32 slotsize = slot_bytes(ca);
914 u32 num = ca->maxreqs;
917 spin_lock(&nfsd_drc_lock);
918 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
919 nfsd_drc_max_mem - nfsd_drc_mem_used);
920 num = min_t(int, num, avail / slotsize);
921 nfsd_drc_mem_used += num * slotsize;
922 spin_unlock(&nfsd_drc_lock);
927 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
929 int slotsize = slot_bytes(ca);
931 spin_lock(&nfsd_drc_lock);
932 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
933 spin_unlock(&nfsd_drc_lock);
936 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
937 struct nfsd4_channel_attrs *battrs)
939 int numslots = fattrs->maxreqs;
940 int slotsize = slot_bytes(fattrs);
941 struct nfsd4_session *new;
944 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
945 + sizeof(struct nfsd4_session) > PAGE_SIZE);
946 mem = numslots * sizeof(struct nfsd4_slot *);
948 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
951 /* allocate each struct nfsd4_slot and data cache in one piece */
952 for (i = 0; i < numslots; i++) {
953 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
954 if (!new->se_slots[i])
958 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
959 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
964 kfree(new->se_slots[i]);
969 static void free_conn(struct nfsd4_conn *c)
971 svc_xprt_put(c->cn_xprt);
975 static void nfsd4_conn_lost(struct svc_xpt_user *u)
977 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
978 struct nfs4_client *clp = c->cn_session->se_client;
980 spin_lock(&clp->cl_lock);
981 if (!list_empty(&c->cn_persession)) {
982 list_del(&c->cn_persession);
985 nfsd4_probe_callback(clp);
986 spin_unlock(&clp->cl_lock);
989 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
991 struct nfsd4_conn *conn;
993 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
996 svc_xprt_get(rqstp->rq_xprt);
997 conn->cn_xprt = rqstp->rq_xprt;
998 conn->cn_flags = flags;
999 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1003 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1005 conn->cn_session = ses;
1006 list_add(&conn->cn_persession, &ses->se_conns);
1009 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1011 struct nfs4_client *clp = ses->se_client;
1013 spin_lock(&clp->cl_lock);
1014 __nfsd4_hash_conn(conn, ses);
1015 spin_unlock(&clp->cl_lock);
1018 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1020 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1021 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1024 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1028 nfsd4_hash_conn(conn, ses);
1029 ret = nfsd4_register_conn(conn);
1031 /* oops; xprt is already down: */
1032 nfsd4_conn_lost(&conn->cn_xpt_user);
1033 if (conn->cn_flags & NFS4_CDFC4_BACK) {
1034 /* callback channel may be back up */
1035 nfsd4_probe_callback(ses->se_client);
1039 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1041 u32 dir = NFS4_CDFC4_FORE;
1043 if (cses->flags & SESSION4_BACK_CHAN)
1044 dir |= NFS4_CDFC4_BACK;
1045 return alloc_conn(rqstp, dir);
1048 /* must be called under client_lock */
1049 static void nfsd4_del_conns(struct nfsd4_session *s)
1051 struct nfs4_client *clp = s->se_client;
1052 struct nfsd4_conn *c;
1054 spin_lock(&clp->cl_lock);
1055 while (!list_empty(&s->se_conns)) {
1056 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1057 list_del_init(&c->cn_persession);
1058 spin_unlock(&clp->cl_lock);
1060 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1063 spin_lock(&clp->cl_lock);
1065 spin_unlock(&clp->cl_lock);
1068 static void __free_session(struct nfsd4_session *ses)
1070 free_session_slots(ses);
1074 static void free_session(struct nfsd4_session *ses)
1076 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
1078 lockdep_assert_held(&nn->client_lock);
1079 nfsd4_del_conns(ses);
1080 nfsd4_put_drc_mem(&ses->se_fchannel);
1081 __free_session(ses);
1084 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1087 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1089 new->se_client = clp;
1092 INIT_LIST_HEAD(&new->se_conns);
1094 new->se_cb_seq_nr = 1;
1095 new->se_flags = cses->flags;
1096 new->se_cb_prog = cses->callback_prog;
1097 new->se_cb_sec = cses->cb_sec;
1098 atomic_set(&new->se_ref, 0);
1099 idx = hash_sessionid(&new->se_sessionid);
1100 spin_lock(&nn->client_lock);
1101 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1102 spin_lock(&clp->cl_lock);
1103 list_add(&new->se_perclnt, &clp->cl_sessions);
1104 spin_unlock(&clp->cl_lock);
1105 spin_unlock(&nn->client_lock);
1107 if (cses->flags & SESSION4_BACK_CHAN) {
1108 struct sockaddr *sa = svc_addr(rqstp);
1110 * This is a little silly; with sessions there's no real
1111 * use for the callback address. Use the peer address
1112 * as a reasonable default for now, but consider fixing
1113 * the rpc client not to require an address in the
1116 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1117 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1121 /* caller must hold client_lock */
1122 static struct nfsd4_session *
1123 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1125 struct nfsd4_session *elem;
1127 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1129 dump_sessionid(__func__, sessionid);
1130 idx = hash_sessionid(sessionid);
1131 /* Search in the appropriate list */
1132 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1133 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1134 NFS4_MAX_SESSIONID_LEN)) {
1139 dprintk("%s: session not found\n", __func__);
1143 /* caller must hold client_lock */
1145 unhash_session(struct nfsd4_session *ses)
1147 list_del(&ses->se_hash);
1148 spin_lock(&ses->se_client->cl_lock);
1149 list_del(&ses->se_perclnt);
1150 spin_unlock(&ses->se_client->cl_lock);
1153 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1155 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1157 if (clid->cl_boot == nn->boot_time)
1159 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1160 clid->cl_boot, clid->cl_id, nn->boot_time);
1165 * XXX Should we use a slab cache ?
1166 * This type of memory management is somewhat inefficient, but we use it
1167 * anyway since SETCLIENTID is not a common operation.
1169 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1171 struct nfs4_client *clp;
1173 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1176 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1177 if (clp->cl_name.data == NULL) {
1181 clp->cl_name.len = name.len;
1182 INIT_LIST_HEAD(&clp->cl_sessions);
1183 idr_init(&clp->cl_stateids);
1184 atomic_set(&clp->cl_refcount, 0);
1185 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1186 INIT_LIST_HEAD(&clp->cl_idhash);
1187 INIT_LIST_HEAD(&clp->cl_openowners);
1188 INIT_LIST_HEAD(&clp->cl_delegations);
1189 INIT_LIST_HEAD(&clp->cl_lru);
1190 INIT_LIST_HEAD(&clp->cl_callbacks);
1191 INIT_LIST_HEAD(&clp->cl_revoked);
1192 spin_lock_init(&clp->cl_lock);
1193 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1198 free_client(struct nfs4_client *clp)
1200 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1202 lockdep_assert_held(&nn->client_lock);
1203 while (!list_empty(&clp->cl_sessions)) {
1204 struct nfsd4_session *ses;
1205 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1207 list_del(&ses->se_perclnt);
1208 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1211 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1212 free_svc_cred(&clp->cl_cred);
1213 kfree(clp->cl_name.data);
1214 idr_destroy(&clp->cl_stateids);
1218 /* must be called under the client_lock */
1220 unhash_client_locked(struct nfs4_client *clp)
1222 struct nfsd4_session *ses;
1224 list_del(&clp->cl_lru);
1225 spin_lock(&clp->cl_lock);
1226 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1227 list_del_init(&ses->se_hash);
1228 spin_unlock(&clp->cl_lock);
1232 destroy_client(struct nfs4_client *clp)
1234 struct nfs4_openowner *oo;
1235 struct nfs4_delegation *dp;
1236 struct list_head reaplist;
1237 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1239 INIT_LIST_HEAD(&reaplist);
1240 spin_lock(&state_lock);
1241 while (!list_empty(&clp->cl_delegations)) {
1242 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1243 list_del_init(&dp->dl_perclnt);
1244 list_move(&dp->dl_recall_lru, &reaplist);
1246 spin_unlock(&state_lock);
1247 while (!list_empty(&reaplist)) {
1248 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1249 destroy_delegation(dp);
1251 list_splice_init(&clp->cl_revoked, &reaplist);
1252 while (!list_empty(&reaplist)) {
1253 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1254 destroy_revoked_delegation(dp);
1256 while (!list_empty(&clp->cl_openowners)) {
1257 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1258 release_openowner(oo);
1260 nfsd4_shutdown_callback(clp);
1261 if (clp->cl_cb_conn.cb_xprt)
1262 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1263 list_del(&clp->cl_idhash);
1264 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1265 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1267 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1268 spin_lock(&nn->client_lock);
1269 unhash_client_locked(clp);
1270 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1272 spin_unlock(&nn->client_lock);
1275 static void expire_client(struct nfs4_client *clp)
1277 nfsd4_client_record_remove(clp);
1278 destroy_client(clp);
1281 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1283 memcpy(target->cl_verifier.data, source->data,
1284 sizeof(target->cl_verifier.data));
1287 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1289 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1290 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1293 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1295 if (source->cr_principal) {
1296 target->cr_principal =
1297 kstrdup(source->cr_principal, GFP_KERNEL);
1298 if (target->cr_principal == NULL)
1301 target->cr_principal = NULL;
1302 target->cr_flavor = source->cr_flavor;
1303 target->cr_uid = source->cr_uid;
1304 target->cr_gid = source->cr_gid;
1305 target->cr_group_info = source->cr_group_info;
1306 get_group_info(target->cr_group_info);
1307 target->cr_gss_mech = source->cr_gss_mech;
1308 if (source->cr_gss_mech)
1309 gss_mech_get(source->cr_gss_mech);
1314 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1318 res = o1->len - o2->len;
1321 return (long long)memcmp(o1->data, o2->data, o1->len);
1324 static int same_name(const char *n1, const char *n2)
1326 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1330 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1332 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1336 same_clid(clientid_t *cl1, clientid_t *cl2)
1338 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1341 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1345 if (g1->ngroups != g2->ngroups)
1347 for (i=0; i<g1->ngroups; i++)
1348 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1354 * RFC 3530 language requires clid_inuse be returned when the
1355 * "principal" associated with a requests differs from that previously
1356 * used. We use uid, gid's, and gss principal string as our best
1357 * approximation. We also don't want to allow non-gss use of a client
1358 * established using gss: in theory cr_principal should catch that
1359 * change, but in practice cr_principal can be null even in the gss case
1360 * since gssd doesn't always pass down a principal string.
1362 static bool is_gss_cred(struct svc_cred *cr)
1364 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1365 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1370 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1372 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1373 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1374 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1375 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1377 if (cr1->cr_principal == cr2->cr_principal)
1379 if (!cr1->cr_principal || !cr2->cr_principal)
1381 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1384 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1386 struct svc_cred *cr = &rqstp->rq_cred;
1389 if (!cr->cr_gss_mech)
1391 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1392 return service == RPC_GSS_SVC_INTEGRITY ||
1393 service == RPC_GSS_SVC_PRIVACY;
1396 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1398 struct svc_cred *cr = &rqstp->rq_cred;
1400 if (!cl->cl_mach_cred)
1402 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1404 if (!svc_rqst_integrity_protected(rqstp))
1406 if (!cr->cr_principal)
1408 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1411 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1413 static u32 current_clientid = 1;
1415 clp->cl_clientid.cl_boot = nn->boot_time;
1416 clp->cl_clientid.cl_id = current_clientid++;
1419 static void gen_confirm(struct nfs4_client *clp)
1424 verf[0] = (__be32)get_seconds();
1425 verf[1] = (__be32)i++;
1426 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1429 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1431 struct nfs4_stid *ret;
1433 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1434 if (!ret || !ret->sc_type)
1439 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1441 struct nfs4_stid *s;
1443 s = find_stateid(cl, t);
1446 if (typemask & s->sc_type)
1451 static struct nfs4_client *create_client(struct xdr_netobj name,
1452 struct svc_rqst *rqstp, nfs4_verifier *verf)
1454 struct nfs4_client *clp;
1455 struct sockaddr *sa = svc_addr(rqstp);
1457 struct net *net = SVC_NET(rqstp);
1458 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1460 clp = alloc_client(name);
1464 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1466 spin_lock(&nn->client_lock);
1468 spin_unlock(&nn->client_lock);
1471 nfsd4_init_callback(&clp->cl_cb_null);
1472 clp->cl_time = get_seconds();
1473 clear_bit(0, &clp->cl_cb_slot_busy);
1474 copy_verf(clp, verf);
1475 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1477 clp->cl_cb_session = NULL;
1483 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1485 struct rb_node **new = &(root->rb_node), *parent = NULL;
1486 struct nfs4_client *clp;
1489 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1492 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1493 new = &((*new)->rb_left);
1495 new = &((*new)->rb_right);
1498 rb_link_node(&new_clp->cl_namenode, parent, new);
1499 rb_insert_color(&new_clp->cl_namenode, root);
1502 static struct nfs4_client *
1503 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1506 struct rb_node *node = root->rb_node;
1507 struct nfs4_client *clp;
1510 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1511 cmp = compare_blob(&clp->cl_name, name);
1513 node = node->rb_left;
1515 node = node->rb_right;
1523 add_to_unconfirmed(struct nfs4_client *clp)
1525 unsigned int idhashval;
1526 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1528 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1529 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1530 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1531 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1536 move_to_confirmed(struct nfs4_client *clp)
1538 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1539 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1541 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1542 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1543 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1544 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1545 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1549 static struct nfs4_client *
1550 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1552 struct nfs4_client *clp;
1553 unsigned int idhashval = clientid_hashval(clid->cl_id);
1555 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1556 if (same_clid(&clp->cl_clientid, clid)) {
1557 if ((bool)clp->cl_minorversion != sessions)
1566 static struct nfs4_client *
1567 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1569 struct list_head *tbl = nn->conf_id_hashtbl;
1571 return find_client_in_id_table(tbl, clid, sessions);
1574 static struct nfs4_client *
1575 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1577 struct list_head *tbl = nn->unconf_id_hashtbl;
1579 return find_client_in_id_table(tbl, clid, sessions);
1582 static bool clp_used_exchangeid(struct nfs4_client *clp)
1584 return clp->cl_exchange_flags != 0;
1587 static struct nfs4_client *
1588 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1590 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1593 static struct nfs4_client *
1594 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1596 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1600 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1602 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1603 struct sockaddr *sa = svc_addr(rqstp);
1604 u32 scopeid = rpc_get_scope_id(sa);
1605 unsigned short expected_family;
1607 /* Currently, we only support tcp and tcp6 for the callback channel */
1608 if (se->se_callback_netid_len == 3 &&
1609 !memcmp(se->se_callback_netid_val, "tcp", 3))
1610 expected_family = AF_INET;
1611 else if (se->se_callback_netid_len == 4 &&
1612 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1613 expected_family = AF_INET6;
1617 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1618 se->se_callback_addr_len,
1619 (struct sockaddr *)&conn->cb_addr,
1620 sizeof(conn->cb_addr));
1622 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1625 if (conn->cb_addr.ss_family == AF_INET6)
1626 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1628 conn->cb_prog = se->se_callback_prog;
1629 conn->cb_ident = se->se_callback_ident;
1630 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1633 conn->cb_addr.ss_family = AF_UNSPEC;
1634 conn->cb_addrlen = 0;
1635 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1636 "will not receive delegations\n",
1637 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1643 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
1646 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1648 struct xdr_buf *buf = resp->xdr.buf;
1649 struct nfsd4_slot *slot = resp->cstate.slot;
1652 dprintk("--> %s slot %p\n", __func__, slot);
1654 slot->sl_opcnt = resp->opcnt;
1655 slot->sl_status = resp->cstate.status;
1657 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1658 if (nfsd4_not_cached(resp)) {
1659 slot->sl_datalen = 0;
1662 base = resp->cstate.data_offset;
1663 slot->sl_datalen = buf->len - base;
1664 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
1665 WARN("%s: sessions DRC could not cache compound\n", __func__);
1670 * Encode the replay sequence operation from the slot values.
1671 * If cachethis is FALSE encode the uncached rep error on the next
1672 * operation which sets resp->p and increments resp->opcnt for
1673 * nfs4svc_encode_compoundres.
1677 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1678 struct nfsd4_compoundres *resp)
1680 struct nfsd4_op *op;
1681 struct nfsd4_slot *slot = resp->cstate.slot;
1683 /* Encode the replayed sequence operation */
1684 op = &args->ops[resp->opcnt - 1];
1685 nfsd4_encode_operation(resp, op);
1687 /* Return nfserr_retry_uncached_rep in next operation. */
1688 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1689 op = &args->ops[resp->opcnt++];
1690 op->status = nfserr_retry_uncached_rep;
1691 nfsd4_encode_operation(resp, op);
1697 * The sequence operation is not cached because we can use the slot and
1701 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1702 struct nfsd4_sequence *seq)
1704 struct nfsd4_slot *slot = resp->cstate.slot;
1705 struct xdr_stream *xdr = &resp->xdr;
1709 dprintk("--> %s slot %p\n", __func__, slot);
1711 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1715 p = xdr_reserve_space(xdr, slot->sl_datalen);
1718 return nfserr_serverfault;
1720 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
1721 xdr_commit_encode(xdr);
1723 resp->opcnt = slot->sl_opcnt;
1724 return slot->sl_status;
1728 * Set the exchange_id flags returned by the server.
1731 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1733 /* pNFS is not supported */
1734 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1736 /* Referrals are supported, Migration is not. */
1737 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1739 /* set the wire flags to return to client. */
1740 clid->flags = new->cl_exchange_flags;
1743 static bool client_has_state(struct nfs4_client *clp)
1746 * Note clp->cl_openowners check isn't quite right: there's no
1747 * need to count owners without stateid's.
1749 * Also note we should probably be using this in 4.0 case too.
1751 return !list_empty(&clp->cl_openowners)
1752 || !list_empty(&clp->cl_delegations)
1753 || !list_empty(&clp->cl_sessions);
1757 nfsd4_exchange_id(struct svc_rqst *rqstp,
1758 struct nfsd4_compound_state *cstate,
1759 struct nfsd4_exchange_id *exid)
1761 struct nfs4_client *unconf, *conf, *new;
1763 char addr_str[INET6_ADDRSTRLEN];
1764 nfs4_verifier verf = exid->verifier;
1765 struct sockaddr *sa = svc_addr(rqstp);
1766 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1767 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1769 rpc_ntop(sa, addr_str, sizeof(addr_str));
1770 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1771 "ip_addr=%s flags %x, spa_how %d\n",
1772 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1773 addr_str, exid->flags, exid->spa_how);
1775 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1776 return nfserr_inval;
1778 switch (exid->spa_how) {
1780 if (!svc_rqst_integrity_protected(rqstp))
1781 return nfserr_inval;
1784 default: /* checked by xdr code */
1787 return nfserr_encr_alg_unsupp;
1790 /* Cases below refer to rfc 5661 section 18.35.4: */
1792 conf = find_confirmed_client_by_name(&exid->clname, nn);
1794 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1795 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1798 if (!clp_used_exchangeid(conf)) { /* buggy client */
1799 status = nfserr_inval;
1802 if (!mach_creds_match(conf, rqstp)) {
1803 status = nfserr_wrong_cred;
1806 if (!creds_match) { /* case 9 */
1807 status = nfserr_perm;
1810 if (!verfs_match) { /* case 8 */
1811 status = nfserr_not_same;
1815 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1819 if (!creds_match) { /* case 3 */
1820 if (client_has_state(conf)) {
1821 status = nfserr_clid_inuse;
1824 expire_client(conf);
1827 if (verfs_match) { /* case 2 */
1828 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1832 /* case 5, client reboot */
1836 if (update) { /* case 7 */
1837 status = nfserr_noent;
1841 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1842 if (unconf) /* case 4, possible retry or client restart */
1843 expire_client(unconf);
1845 /* case 1 (normal case) */
1847 new = create_client(exid->clname, rqstp, &verf);
1849 status = nfserr_jukebox;
1852 new->cl_minorversion = cstate->minorversion;
1853 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
1856 add_to_unconfirmed(new);
1858 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1859 exid->clientid.cl_id = new->cl_clientid.cl_id;
1861 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1862 nfsd4_set_ex_flags(new, exid);
1864 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1865 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1869 nfs4_unlock_state();
1874 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1876 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1879 /* The slot is in use, and no response has been sent. */
1881 if (seqid == slot_seqid)
1882 return nfserr_jukebox;
1884 return nfserr_seq_misordered;
1886 /* Note unsigned 32-bit arithmetic handles wraparound: */
1887 if (likely(seqid == slot_seqid + 1))
1889 if (seqid == slot_seqid)
1890 return nfserr_replay_cache;
1891 return nfserr_seq_misordered;
1895 * Cache the create session result into the create session single DRC
1896 * slot cache by saving the xdr structure. sl_seqid has been set.
1897 * Do this for solo or embedded create session operations.
1900 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1901 struct nfsd4_clid_slot *slot, __be32 nfserr)
1903 slot->sl_status = nfserr;
1904 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1908 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1909 struct nfsd4_clid_slot *slot)
1911 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1912 return slot->sl_status;
1915 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1916 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1917 1 + /* MIN tag is length with zero, only length */ \
1918 3 + /* version, opcount, opcode */ \
1919 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1920 /* seqid, slotID, slotID, cache */ \
1921 4 ) * sizeof(__be32))
1923 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1924 2 + /* verifier: AUTH_NULL, length 0 */\
1926 1 + /* MIN tag is length with zero, only length */ \
1927 3 + /* opcount, opcode, opstatus*/ \
1928 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1929 /* seqid, slotID, slotID, slotID, status */ \
1930 5 ) * sizeof(__be32))
1932 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1934 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
1936 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
1937 return nfserr_toosmall;
1938 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
1939 return nfserr_toosmall;
1940 ca->headerpadsz = 0;
1941 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
1942 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
1943 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
1944 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
1945 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
1946 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
1948 * Note decreasing slot size below client's request may make it
1949 * difficult for client to function correctly, whereas
1950 * decreasing the number of slots will (just?) affect
1951 * performance. When short on memory we therefore prefer to
1952 * decrease number of slots instead of their size. Clients that
1953 * request larger slots than they need will get poor results:
1955 ca->maxreqs = nfsd4_get_drc_mem(ca);
1957 return nfserr_jukebox;
1962 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
1963 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
1964 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
1965 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
1967 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
1969 ca->headerpadsz = 0;
1972 * These RPC_MAX_HEADER macros are overkill, especially since we
1973 * don't even do gss on the backchannel yet. But this is still
1974 * less than 1k. Tighten up this estimate in the unlikely event
1975 * it turns out to be a problem for some client:
1977 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
1978 return nfserr_toosmall;
1979 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
1980 return nfserr_toosmall;
1981 ca->maxresp_cached = 0;
1983 return nfserr_toosmall;
1988 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
1990 switch (cbs->flavor) {
1996 * GSS case: the spec doesn't allow us to return this
1997 * error. But it also doesn't allow us not to support
1999 * I'd rather this fail hard than return some error the
2000 * client might think it can already handle:
2002 return nfserr_encr_alg_unsupp;
2007 nfsd4_create_session(struct svc_rqst *rqstp,
2008 struct nfsd4_compound_state *cstate,
2009 struct nfsd4_create_session *cr_ses)
2011 struct sockaddr *sa = svc_addr(rqstp);
2012 struct nfs4_client *conf, *unconf;
2013 struct nfsd4_session *new;
2014 struct nfsd4_conn *conn;
2015 struct nfsd4_clid_slot *cs_slot = NULL;
2017 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2019 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2020 return nfserr_inval;
2021 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2024 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2027 status = check_backchannel_attrs(&cr_ses->back_channel);
2029 goto out_release_drc_mem;
2030 status = nfserr_jukebox;
2031 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2033 goto out_release_drc_mem;
2034 conn = alloc_conn_from_crses(rqstp, cr_ses);
2036 goto out_free_session;
2039 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2040 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2041 WARN_ON_ONCE(conf && unconf);
2044 status = nfserr_wrong_cred;
2045 if (!mach_creds_match(conf, rqstp))
2047 cs_slot = &conf->cl_cs_slot;
2048 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2049 if (status == nfserr_replay_cache) {
2050 status = nfsd4_replay_create_session(cr_ses, cs_slot);
2052 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2053 status = nfserr_seq_misordered;
2056 } else if (unconf) {
2057 struct nfs4_client *old;
2058 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2059 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2060 status = nfserr_clid_inuse;
2063 status = nfserr_wrong_cred;
2064 if (!mach_creds_match(unconf, rqstp))
2066 cs_slot = &unconf->cl_cs_slot;
2067 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2069 /* an unconfirmed replay returns misordered */
2070 status = nfserr_seq_misordered;
2073 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2075 status = mark_client_expired(old);
2080 move_to_confirmed(unconf);
2083 status = nfserr_stale_clientid;
2088 * We do not support RDMA or persistent sessions
2090 cr_ses->flags &= ~SESSION4_PERSIST;
2091 cr_ses->flags &= ~SESSION4_RDMA;
2093 init_session(rqstp, new, conf, cr_ses);
2094 nfsd4_init_conn(rqstp, conn, new);
2096 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2097 NFS4_MAX_SESSIONID_LEN);
2098 cs_slot->sl_seqid++;
2099 cr_ses->seqid = cs_slot->sl_seqid;
2101 /* cache solo and embedded create sessions under the state lock */
2102 nfsd4_cache_create_session(cr_ses, cs_slot, status);
2103 nfs4_unlock_state();
2106 nfs4_unlock_state();
2109 __free_session(new);
2110 out_release_drc_mem:
2111 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2115 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2118 case NFS4_CDFC4_FORE:
2119 case NFS4_CDFC4_BACK:
2121 case NFS4_CDFC4_FORE_OR_BOTH:
2122 case NFS4_CDFC4_BACK_OR_BOTH:
2123 *dir = NFS4_CDFC4_BOTH;
2126 return nfserr_inval;
2129 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2131 struct nfsd4_session *session = cstate->session;
2132 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2135 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2138 spin_lock(&nn->client_lock);
2139 session->se_cb_prog = bc->bc_cb_program;
2140 session->se_cb_sec = bc->bc_cb_sec;
2141 spin_unlock(&nn->client_lock);
2143 nfsd4_probe_callback(session->se_client);
2148 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2149 struct nfsd4_compound_state *cstate,
2150 struct nfsd4_bind_conn_to_session *bcts)
2153 struct nfsd4_conn *conn;
2154 struct nfsd4_session *session;
2155 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2157 if (!nfsd4_last_compound_op(rqstp))
2158 return nfserr_not_only_op;
2160 spin_lock(&nn->client_lock);
2161 session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
2162 spin_unlock(&nn->client_lock);
2163 status = nfserr_badsession;
2166 status = nfserr_wrong_cred;
2167 if (!mach_creds_match(session->se_client, rqstp))
2169 status = nfsd4_map_bcts_dir(&bcts->dir);
2172 conn = alloc_conn(rqstp, bcts->dir);
2173 status = nfserr_jukebox;
2176 nfsd4_init_conn(rqstp, conn, session);
2179 nfs4_unlock_state();
2183 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2187 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2191 nfsd4_destroy_session(struct svc_rqst *r,
2192 struct nfsd4_compound_state *cstate,
2193 struct nfsd4_destroy_session *sessionid)
2195 struct nfsd4_session *ses;
2197 int ref_held_by_me = 0;
2198 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
2201 status = nfserr_not_only_op;
2202 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2203 if (!nfsd4_last_compound_op(r))
2207 dump_sessionid(__func__, &sessionid->sessionid);
2208 spin_lock(&nn->client_lock);
2209 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
2210 status = nfserr_badsession;
2212 goto out_client_lock;
2213 status = nfserr_wrong_cred;
2214 if (!mach_creds_match(ses->se_client, r))
2215 goto out_client_lock;
2216 nfsd4_get_session_locked(ses);
2217 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2219 goto out_put_session;
2220 unhash_session(ses);
2221 spin_unlock(&nn->client_lock);
2223 nfsd4_probe_callback_sync(ses->se_client);
2225 spin_lock(&nn->client_lock);
2228 nfsd4_put_session(ses);
2230 spin_unlock(&nn->client_lock);
2232 nfs4_unlock_state();
2236 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2238 struct nfsd4_conn *c;
2240 list_for_each_entry(c, &s->se_conns, cn_persession) {
2241 if (c->cn_xprt == xpt) {
2248 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2250 struct nfs4_client *clp = ses->se_client;
2251 struct nfsd4_conn *c;
2252 __be32 status = nfs_ok;
2255 spin_lock(&clp->cl_lock);
2256 c = __nfsd4_find_conn(new->cn_xprt, ses);
2259 status = nfserr_conn_not_bound_to_session;
2260 if (clp->cl_mach_cred)
2262 __nfsd4_hash_conn(new, ses);
2263 spin_unlock(&clp->cl_lock);
2264 ret = nfsd4_register_conn(new);
2266 /* oops; xprt is already down: */
2267 nfsd4_conn_lost(&new->cn_xpt_user);
2270 spin_unlock(&clp->cl_lock);
2275 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2277 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2279 return args->opcnt > session->se_fchannel.maxops;
2282 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2283 struct nfsd4_session *session)
2285 struct xdr_buf *xb = &rqstp->rq_arg;
2287 return xb->len > session->se_fchannel.maxreq_sz;
2291 nfsd4_sequence(struct svc_rqst *rqstp,
2292 struct nfsd4_compound_state *cstate,
2293 struct nfsd4_sequence *seq)
2295 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2296 struct xdr_stream *xdr = &resp->xdr;
2297 struct nfsd4_session *session;
2298 struct nfs4_client *clp;
2299 struct nfsd4_slot *slot;
2300 struct nfsd4_conn *conn;
2303 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2305 if (resp->opcnt != 1)
2306 return nfserr_sequence_pos;
2309 * Will be either used or freed by nfsd4_sequence_check_conn
2312 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2314 return nfserr_jukebox;
2316 spin_lock(&nn->client_lock);
2317 status = nfserr_badsession;
2318 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2320 goto out_no_session;
2321 clp = session->se_client;
2322 status = get_client_locked(clp);
2324 goto out_no_session;
2325 status = nfsd4_get_session_locked(session);
2327 goto out_put_client;
2329 status = nfserr_too_many_ops;
2330 if (nfsd4_session_too_many_ops(rqstp, session))
2331 goto out_put_session;
2333 status = nfserr_req_too_big;
2334 if (nfsd4_request_too_big(rqstp, session))
2335 goto out_put_session;
2337 status = nfserr_badslot;
2338 if (seq->slotid >= session->se_fchannel.maxreqs)
2339 goto out_put_session;
2341 slot = session->se_slots[seq->slotid];
2342 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2344 /* We do not negotiate the number of slots yet, so set the
2345 * maxslots to the session maxreqs which is used to encode
2346 * sr_highest_slotid and the sr_target_slot id to maxslots */
2347 seq->maxslots = session->se_fchannel.maxreqs;
2349 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2350 slot->sl_flags & NFSD4_SLOT_INUSE);
2351 if (status == nfserr_replay_cache) {
2352 status = nfserr_seq_misordered;
2353 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2354 goto out_put_session;
2355 cstate->slot = slot;
2356 cstate->session = session;
2357 /* Return the cached reply status and set cstate->status
2358 * for nfsd4_proc_compound processing */
2359 status = nfsd4_replay_cache_entry(resp, seq);
2360 cstate->status = nfserr_replay_cache;
2364 goto out_put_session;
2366 status = nfsd4_sequence_check_conn(conn, session);
2369 goto out_put_session;
2371 buflen = (seq->cachethis) ?
2372 session->se_fchannel.maxresp_cached :
2373 session->se_fchannel.maxresp_sz;
2374 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2376 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2377 goto out_put_session;
2378 svc_reserve(rqstp, buflen);
2381 /* Success! bump slot seqid */
2382 slot->sl_seqid = seq->seqid;
2383 slot->sl_flags |= NFSD4_SLOT_INUSE;
2385 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2387 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2389 cstate->slot = slot;
2390 cstate->session = session;
2393 switch (clp->cl_cb_state) {
2395 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2397 case NFSD4_CB_FAULT:
2398 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2401 seq->status_flags = 0;
2403 if (!list_empty(&clp->cl_revoked))
2404 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2408 spin_unlock(&nn->client_lock);
2411 nfsd4_put_session(session);
2413 put_client_renew_locked(clp);
2414 goto out_no_session;
2418 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2420 struct nfs4_client *conf, *unconf, *clp;
2422 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2425 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2426 conf = find_confirmed_client(&dc->clientid, true, nn);
2427 WARN_ON_ONCE(conf && unconf);
2432 if (client_has_state(conf)) {
2433 status = nfserr_clientid_busy;
2439 status = nfserr_stale_clientid;
2442 if (!mach_creds_match(clp, rqstp)) {
2443 status = nfserr_wrong_cred;
2448 nfs4_unlock_state();
2453 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2457 if (rc->rca_one_fs) {
2458 if (!cstate->current_fh.fh_dentry)
2459 return nfserr_nofilehandle;
2461 * We don't take advantage of the rca_one_fs case.
2462 * That's OK, it's optional, we can safely ignore it.
2468 status = nfserr_complete_already;
2469 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2470 &cstate->session->se_client->cl_flags))
2473 status = nfserr_stale_clientid;
2474 if (is_client_expired(cstate->session->se_client))
2476 * The following error isn't really legal.
2477 * But we only get here if the client just explicitly
2478 * destroyed the client. Surely it no longer cares what
2479 * error it gets back on an operation for the dead
2485 nfsd4_client_record_create(cstate->session->se_client);
2487 nfs4_unlock_state();
2492 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2493 struct nfsd4_setclientid *setclid)
2495 struct xdr_netobj clname = setclid->se_name;
2496 nfs4_verifier clverifier = setclid->se_verf;
2497 struct nfs4_client *conf, *unconf, *new;
2499 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2501 /* Cases below refer to rfc 3530 section 14.2.33: */
2503 conf = find_confirmed_client_by_name(&clname, nn);
2506 status = nfserr_clid_inuse;
2507 if (clp_used_exchangeid(conf))
2509 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2510 char addr_str[INET6_ADDRSTRLEN];
2511 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2513 dprintk("NFSD: setclientid: string in use by client "
2514 "at %s\n", addr_str);
2518 unconf = find_unconfirmed_client_by_name(&clname, nn);
2520 expire_client(unconf);
2521 status = nfserr_jukebox;
2522 new = create_client(clname, rqstp, &clverifier);
2525 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2526 /* case 1: probable callback update */
2527 copy_clid(new, conf);
2528 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2530 new->cl_minorversion = 0;
2531 gen_callback(new, setclid, rqstp);
2532 add_to_unconfirmed(new);
2533 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2534 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2535 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2538 nfs4_unlock_state();
2544 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2545 struct nfsd4_compound_state *cstate,
2546 struct nfsd4_setclientid_confirm *setclientid_confirm)
2548 struct nfs4_client *conf, *unconf;
2549 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2550 clientid_t * clid = &setclientid_confirm->sc_clientid;
2552 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2554 if (STALE_CLIENTID(clid, nn))
2555 return nfserr_stale_clientid;
2558 conf = find_confirmed_client(clid, false, nn);
2559 unconf = find_unconfirmed_client(clid, false, nn);
2561 * We try hard to give out unique clientid's, so if we get an
2562 * attempt to confirm the same clientid with a different cred,
2563 * there's a bug somewhere. Let's charitably assume it's our
2566 status = nfserr_serverfault;
2567 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2569 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2571 /* cases below refer to rfc 3530 section 14.2.34: */
2572 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2573 if (conf && !unconf) /* case 2: probable retransmit */
2575 else /* case 4: client hasn't noticed we rebooted yet? */
2576 status = nfserr_stale_clientid;
2580 if (conf) { /* case 1: callback update */
2581 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2582 nfsd4_probe_callback(conf);
2583 expire_client(unconf);
2584 } else { /* case 3: normal case; new or rebooted client */
2585 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2587 status = mark_client_expired(conf);
2590 expire_client(conf);
2592 move_to_confirmed(unconf);
2593 nfsd4_probe_callback(unconf);
2596 nfs4_unlock_state();
2600 static struct nfs4_file *nfsd4_alloc_file(void)
2602 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2605 /* OPEN Share state helper functions */
2606 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2608 unsigned int hashval = file_hashval(ino);
2610 atomic_set(&fp->fi_ref, 1);
2611 INIT_LIST_HEAD(&fp->fi_stateids);
2612 INIT_LIST_HEAD(&fp->fi_delegations);
2613 fp->fi_inode = igrab(ino);
2614 fp->fi_had_conflict = false;
2615 fp->fi_lease = NULL;
2616 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2617 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2618 spin_lock(&state_lock);
2619 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2620 spin_unlock(&state_lock);
2624 nfsd4_free_slabs(void)
2626 kmem_cache_destroy(openowner_slab);
2627 kmem_cache_destroy(lockowner_slab);
2628 kmem_cache_destroy(file_slab);
2629 kmem_cache_destroy(stateid_slab);
2630 kmem_cache_destroy(deleg_slab);
2634 nfsd4_init_slabs(void)
2636 openowner_slab = kmem_cache_create("nfsd4_openowners",
2637 sizeof(struct nfs4_openowner), 0, 0, NULL);
2638 if (openowner_slab == NULL)
2640 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2641 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2642 if (lockowner_slab == NULL)
2643 goto out_free_openowner_slab;
2644 file_slab = kmem_cache_create("nfsd4_files",
2645 sizeof(struct nfs4_file), 0, 0, NULL);
2646 if (file_slab == NULL)
2647 goto out_free_lockowner_slab;
2648 stateid_slab = kmem_cache_create("nfsd4_stateids",
2649 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2650 if (stateid_slab == NULL)
2651 goto out_free_file_slab;
2652 deleg_slab = kmem_cache_create("nfsd4_delegations",
2653 sizeof(struct nfs4_delegation), 0, 0, NULL);
2654 if (deleg_slab == NULL)
2655 goto out_free_stateid_slab;
2658 out_free_stateid_slab:
2659 kmem_cache_destroy(stateid_slab);
2661 kmem_cache_destroy(file_slab);
2662 out_free_lockowner_slab:
2663 kmem_cache_destroy(lockowner_slab);
2664 out_free_openowner_slab:
2665 kmem_cache_destroy(openowner_slab);
2667 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2671 static void init_nfs4_replay(struct nfs4_replay *rp)
2673 rp->rp_status = nfserr_serverfault;
2675 rp->rp_buf = rp->rp_ibuf;
2678 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2680 struct nfs4_stateowner *sop;
2682 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2686 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2687 if (!sop->so_owner.data) {
2688 kmem_cache_free(slab, sop);
2691 sop->so_owner.len = owner->len;
2693 INIT_LIST_HEAD(&sop->so_stateids);
2694 sop->so_client = clp;
2695 init_nfs4_replay(&sop->so_replay);
2699 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2701 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2703 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2704 list_add(&oo->oo_perclient, &clp->cl_openowners);
2707 static struct nfs4_openowner *
2708 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2709 struct nfs4_openowner *oo;
2711 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2714 oo->oo_owner.so_is_open_owner = 1;
2715 oo->oo_owner.so_seqid = open->op_seqid;
2716 oo->oo_flags = NFS4_OO_NEW;
2718 oo->oo_last_closed_stid = NULL;
2719 INIT_LIST_HEAD(&oo->oo_close_lru);
2720 hash_openowner(oo, clp, strhashval);
2724 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2725 struct nfs4_openowner *oo = open->op_openowner;
2727 stp->st_stid.sc_type = NFS4_OPEN_STID;
2728 INIT_LIST_HEAD(&stp->st_lockowners);
2729 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2730 list_add(&stp->st_perfile, &fp->fi_stateids);
2731 stp->st_stateowner = &oo->oo_owner;
2734 stp->st_access_bmap = 0;
2735 stp->st_deny_bmap = 0;
2736 set_access(open->op_share_access, stp);
2737 set_deny(open->op_share_deny, stp);
2738 stp->st_openstp = NULL;
2742 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2744 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2746 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2748 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2749 oo->oo_time = get_seconds();
2753 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2756 return (sop->so_owner.len == owner->len) &&
2757 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2758 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2761 static struct nfs4_openowner *
2762 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2763 bool sessions, struct nfsd_net *nn)
2765 struct nfs4_stateowner *so;
2766 struct nfs4_openowner *oo;
2767 struct nfs4_client *clp;
2769 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2770 if (!so->so_is_open_owner)
2772 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2774 clp = oo->oo_owner.so_client;
2775 if ((bool)clp->cl_minorversion != sessions)
2777 renew_client(oo->oo_owner.so_client);
2784 /* search file_hashtbl[] for file */
2785 static struct nfs4_file *
2786 find_file(struct inode *ino)
2788 unsigned int hashval = file_hashval(ino);
2789 struct nfs4_file *fp;
2791 spin_lock(&state_lock);
2792 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2793 if (fp->fi_inode == ino) {
2795 spin_unlock(&state_lock);
2799 spin_unlock(&state_lock);
2804 * Called to check deny when READ with all zero stateid or
2805 * WRITE with all zero or all one stateid
2808 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2810 struct inode *ino = current_fh->fh_dentry->d_inode;
2811 struct nfs4_file *fp;
2812 struct nfs4_ol_stateid *stp;
2815 fp = find_file(ino);
2818 ret = nfserr_locked;
2819 /* Search for conflicting share reservations */
2820 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2821 if (test_deny(deny_type, stp) ||
2822 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2831 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2833 struct nfs4_client *clp = dp->dl_stid.sc_client;
2834 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2836 lockdep_assert_held(&state_lock);
2837 /* We're assuming the state code never drops its reference
2838 * without first removing the lease. Since we're in this lease
2839 * callback (and since the lease code is serialized by the kernel
2840 * lock) we know the server hasn't removed the lease yet, we know
2841 * it's safe to take a reference: */
2842 atomic_inc(&dp->dl_count);
2844 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
2846 /* Only place dl_time is set; protected by i_lock: */
2847 dp->dl_time = get_seconds();
2849 block_delegations(&dp->dl_fh);
2851 nfsd4_cb_recall(dp);
2854 /* Called from break_lease() with i_lock held. */
2855 static void nfsd_break_deleg_cb(struct file_lock *fl)
2857 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2858 struct nfs4_delegation *dp;
2861 WARN(1, "(%p)->fl_owner NULL\n", fl);
2864 if (fp->fi_had_conflict) {
2865 WARN(1, "duplicate break on %p\n", fp);
2869 * We don't want the locks code to timeout the lease for us;
2870 * we'll remove it ourself if a delegation isn't returned
2873 fl->fl_break_time = 0;
2875 spin_lock(&state_lock);
2876 fp->fi_had_conflict = true;
2877 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2878 nfsd_break_one_deleg(dp);
2879 spin_unlock(&state_lock);
2883 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2886 return lease_modify(onlist, arg);
2891 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2892 .lm_break = nfsd_break_deleg_cb,
2893 .lm_change = nfsd_change_deleg_cb,
2896 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2898 if (nfsd4_has_session(cstate))
2900 if (seqid == so->so_seqid - 1)
2901 return nfserr_replay_me;
2902 if (seqid == so->so_seqid)
2904 return nfserr_bad_seqid;
2908 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2909 struct nfsd4_open *open, struct nfsd_net *nn)
2911 clientid_t *clientid = &open->op_clientid;
2912 struct nfs4_client *clp = NULL;
2913 unsigned int strhashval;
2914 struct nfs4_openowner *oo = NULL;
2917 if (STALE_CLIENTID(&open->op_clientid, nn))
2918 return nfserr_stale_clientid;
2920 * In case we need it later, after we've already created the
2921 * file and don't want to risk a further failure:
2923 open->op_file = nfsd4_alloc_file();
2924 if (open->op_file == NULL)
2925 return nfserr_jukebox;
2927 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2928 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2929 open->op_openowner = oo;
2931 clp = find_confirmed_client(clientid, cstate->minorversion,
2934 return nfserr_expired;
2937 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2938 /* Replace unconfirmed owners without checking for replay. */
2939 clp = oo->oo_owner.so_client;
2940 release_openowner(oo);
2941 open->op_openowner = NULL;
2944 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2947 clp = oo->oo_owner.so_client;
2950 oo = alloc_init_open_stateowner(strhashval, clp, open);
2952 return nfserr_jukebox;
2953 open->op_openowner = oo;
2955 open->op_stp = nfs4_alloc_stateid(clp);
2957 return nfserr_jukebox;
2961 static inline __be32
2962 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2964 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2965 return nfserr_openmode;
2970 static int share_access_to_flags(u32 share_access)
2972 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2975 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2977 struct nfs4_stid *ret;
2979 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2982 return delegstateid(ret);
2985 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2987 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2988 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2992 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
2993 struct nfs4_delegation **dp)
2996 __be32 status = nfserr_bad_stateid;
2998 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
3001 flags = share_access_to_flags(open->op_share_access);
3002 status = nfs4_check_delegmode(*dp, flags);
3006 if (!nfsd4_is_deleg_cur(open))
3010 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3015 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
3017 struct nfs4_ol_stateid *local;
3018 struct nfs4_openowner *oo = open->op_openowner;
3020 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3021 /* ignore lock owners */
3022 if (local->st_stateowner->so_is_open_owner == 0)
3024 /* remember if we have seen this open owner */
3025 if (local->st_stateowner == &oo->oo_owner)
3027 /* check for conflicting share reservations */
3028 if (!test_share(local, open))
3029 return nfserr_share_denied;
3034 static inline int nfs4_access_to_access(u32 nfs4_access)
3038 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3039 flags |= NFSD_MAY_READ;
3040 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3041 flags |= NFSD_MAY_WRITE;
3045 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3046 struct svc_fh *cur_fh, struct nfsd4_open *open)
3049 int oflag = nfs4_access_to_omode(open->op_share_access);
3050 int access = nfs4_access_to_access(open->op_share_access);
3052 if (!fp->fi_fds[oflag]) {
3053 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
3054 &fp->fi_fds[oflag]);
3058 nfs4_file_get_access(fp, oflag);
3063 static inline __be32
3064 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3065 struct nfsd4_open *open)
3067 struct iattr iattr = {
3068 .ia_valid = ATTR_SIZE,
3071 if (!open->op_truncate)
3073 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3074 return nfserr_inval;
3075 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3079 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)
3081 u32 op_share_access = open->op_share_access;
3085 new_access = !test_access(op_share_access, stp);
3087 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
3091 status = nfsd4_truncate(rqstp, cur_fh, open);
3094 int oflag = nfs4_access_to_omode(op_share_access);
3095 nfs4_file_put_access(fp, oflag);
3099 /* remember the open */
3100 set_access(op_share_access, stp);
3101 set_deny(open->op_share_deny, stp);
3108 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3110 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3113 /* Should we give out recallable state?: */
3114 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3116 if (clp->cl_cb_state == NFSD4_CB_UP)
3119 * In the sessions case, since we don't have to establish a
3120 * separate connection for callbacks, we assume it's OK
3121 * until we hear otherwise:
3123 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3126 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
3128 struct file_lock *fl;
3130 fl = locks_alloc_lock();
3133 locks_init_lock(fl);
3134 fl->fl_lmops = &nfsd_lease_mng_ops;
3135 fl->fl_flags = FL_DELEG;
3136 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3137 fl->fl_end = OFFSET_MAX;
3138 fl->fl_owner = (fl_owner_t)(dp->dl_file);
3139 fl->fl_pid = current->tgid;
3143 static int nfs4_setlease(struct nfs4_delegation *dp)
3145 struct nfs4_file *fp = dp->dl_file;
3146 struct file_lock *fl;
3149 fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
3152 fl->fl_file = find_readable_file(fp);
3153 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
3157 fp->fi_deleg_file = get_file(fl->fl_file);
3158 atomic_set(&fp->fi_delegees, 1);
3159 spin_lock(&state_lock);
3160 hash_delegation_locked(dp, fp);
3161 spin_unlock(&state_lock);
3164 locks_free_lock(fl);
3168 static int nfs4_set_delegation(struct nfs4_delegation *dp, struct nfs4_file *fp)
3170 if (fp->fi_had_conflict)
3175 return nfs4_setlease(dp);
3176 spin_lock(&state_lock);
3177 atomic_inc(&fp->fi_delegees);
3178 if (fp->fi_had_conflict) {
3179 spin_unlock(&state_lock);
3182 hash_delegation_locked(dp, fp);
3183 spin_unlock(&state_lock);
3187 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3189 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3190 if (status == -EAGAIN)
3191 open->op_why_no_deleg = WND4_CONTENTION;
3193 open->op_why_no_deleg = WND4_RESOURCE;
3194 switch (open->op_deleg_want) {
3195 case NFS4_SHARE_WANT_READ_DELEG:
3196 case NFS4_SHARE_WANT_WRITE_DELEG:
3197 case NFS4_SHARE_WANT_ANY_DELEG:
3199 case NFS4_SHARE_WANT_CANCEL:
3200 open->op_why_no_deleg = WND4_CANCELLED;
3202 case NFS4_SHARE_WANT_NO_DELEG:
3209 * Attempt to hand out a delegation.
3211 * Note we don't support write delegations, and won't until the vfs has
3212 * proper support for them.
3215 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
3216 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
3218 struct nfs4_delegation *dp;
3219 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
3223 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3224 open->op_recall = 0;
3225 switch (open->op_claim_type) {
3226 case NFS4_OPEN_CLAIM_PREVIOUS:
3228 open->op_recall = 1;
3229 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3232 case NFS4_OPEN_CLAIM_NULL:
3233 case NFS4_OPEN_CLAIM_FH:
3235 * Let's not give out any delegations till everyone's
3236 * had the chance to reclaim theirs....
3238 if (locks_in_grace(net))
3240 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3243 * Also, if the file was opened for write or
3244 * create, there's a good chance the client's
3245 * about to write to it, resulting in an
3246 * immediate recall (since we don't support
3247 * write delegations):
3249 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3251 if (open->op_create == NFS4_OPEN_CREATE)
3257 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh);
3260 status = nfs4_set_delegation(dp, stp->st_file);
3264 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3266 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3267 STATEID_VAL(&dp->dl_stid.sc_stateid));
3268 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3271 destroy_delegation(dp);
3273 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3274 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3275 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3276 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3277 open->op_recall = 1;
3280 /* 4.1 client asking for a delegation? */
3281 if (open->op_deleg_want)
3282 nfsd4_open_deleg_none_ext(open, status);
3286 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3287 struct nfs4_delegation *dp)
3289 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3290 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3291 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3292 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3293 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3294 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3295 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3296 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3298 /* Otherwise the client must be confused wanting a delegation
3299 * it already has, therefore we don't return
3300 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3305 * called with nfs4_lock_state() held.
3308 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3310 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3311 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3312 struct nfs4_file *fp = NULL;
3313 struct inode *ino = current_fh->fh_dentry->d_inode;
3314 struct nfs4_ol_stateid *stp = NULL;
3315 struct nfs4_delegation *dp = NULL;
3319 * Lookup file; if found, lookup stateid and check open request,
3320 * and check for delegations in the process of being recalled.
3321 * If not found, create the nfs4_file struct
3323 fp = find_file(ino);
3325 if ((status = nfs4_check_open(fp, open, &stp)))
3327 status = nfs4_check_deleg(cl, open, &dp);
3331 status = nfserr_bad_stateid;
3332 if (nfsd4_is_deleg_cur(open))
3334 status = nfserr_jukebox;
3336 open->op_file = NULL;
3337 nfsd4_init_file(fp, ino);
3341 * OPEN the file, or upgrade an existing OPEN.
3342 * If truncate fails, the OPEN fails.
3345 /* Stateid was found, this is an OPEN upgrade */
3346 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3350 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3353 status = nfsd4_truncate(rqstp, current_fh, open);
3357 open->op_stp = NULL;
3358 init_open_stateid(stp, fp, open);
3360 update_stateid(&stp->st_stid.sc_stateid);
3361 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3363 if (nfsd4_has_session(&resp->cstate)) {
3364 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3366 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3367 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3368 open->op_why_no_deleg = WND4_NOT_WANTED;
3374 * Attempt to hand out a delegation. No error return, because the
3375 * OPEN succeeds even if we fail.
3377 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3381 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3382 STATEID_VAL(&stp->st_stid.sc_stateid));
3384 /* 4.1 client trying to upgrade/downgrade delegation? */
3385 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3386 open->op_deleg_want)
3387 nfsd4_deleg_xgrade_none_ext(open, dp);
3391 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3392 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3394 * To finish the open response, we just need to set the rflags.
3396 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3397 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3398 !nfsd4_has_session(&resp->cstate))
3399 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3404 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3406 if (open->op_openowner) {
3407 struct nfs4_openowner *oo = open->op_openowner;
3409 if (!list_empty(&oo->oo_owner.so_stateids))
3410 list_del_init(&oo->oo_close_lru);
3411 if (oo->oo_flags & NFS4_OO_NEW) {
3413 release_openowner(oo);
3414 open->op_openowner = NULL;
3416 oo->oo_flags &= ~NFS4_OO_NEW;
3420 nfsd4_free_file(open->op_file);
3422 free_generic_stateid(open->op_stp);
3425 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3427 struct nfs4_client *found;
3429 if (STALE_CLIENTID(clid, nn))
3430 return nfserr_stale_clientid;
3431 found = find_confirmed_client(clid, session, nn);
3434 return found ? nfs_ok : nfserr_expired;
3438 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3441 struct nfs4_client *clp;
3443 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3446 dprintk("process_renew(%08x/%08x): starting\n",
3447 clid->cl_boot, clid->cl_id);
3448 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3451 status = nfserr_cb_path_down;
3452 if (!list_empty(&clp->cl_delegations)
3453 && clp->cl_cb_state != NFSD4_CB_UP)
3457 nfs4_unlock_state();
3462 nfsd4_end_grace(struct nfsd_net *nn)
3464 /* do nothing if grace period already ended */
3465 if (nn->grace_ended)
3468 dprintk("NFSD: end of grace period\n");
3469 nn->grace_ended = true;
3470 nfsd4_record_grace_done(nn, nn->boot_time);
3471 locks_end_grace(&nn->nfsd4_manager);
3473 * Now that every NFSv4 client has had the chance to recover and
3474 * to see the (possibly new, possibly shorter) lease time, we
3475 * can safely set the next grace time to the current lease time:
3477 nn->nfsd4_grace = nn->nfsd4_lease;
3481 nfs4_laundromat(struct nfsd_net *nn)
3483 struct nfs4_client *clp;
3484 struct nfs4_openowner *oo;
3485 struct nfs4_delegation *dp;
3486 struct list_head *pos, *next, reaplist;
3487 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3488 time_t t, new_timeo = nn->nfsd4_lease;
3492 dprintk("NFSD: laundromat service - starting\n");
3493 nfsd4_end_grace(nn);
3494 INIT_LIST_HEAD(&reaplist);
3495 spin_lock(&nn->client_lock);
3496 list_for_each_safe(pos, next, &nn->client_lru) {
3497 clp = list_entry(pos, struct nfs4_client, cl_lru);
3498 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3499 t = clp->cl_time - cutoff;
3500 new_timeo = min(new_timeo, t);
3503 if (mark_client_expired_locked(clp)) {
3504 dprintk("NFSD: client in use (clientid %08x)\n",
3505 clp->cl_clientid.cl_id);
3508 list_move(&clp->cl_lru, &reaplist);
3510 spin_unlock(&nn->client_lock);
3511 list_for_each_safe(pos, next, &reaplist) {
3512 clp = list_entry(pos, struct nfs4_client, cl_lru);
3513 dprintk("NFSD: purging unused client (clientid %08x)\n",
3514 clp->cl_clientid.cl_id);
3517 spin_lock(&state_lock);
3518 list_for_each_safe(pos, next, &nn->del_recall_lru) {
3519 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3520 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3522 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3523 t = dp->dl_time - cutoff;
3524 new_timeo = min(new_timeo, t);
3527 list_move(&dp->dl_recall_lru, &reaplist);
3529 spin_unlock(&state_lock);
3530 list_for_each_safe(pos, next, &reaplist) {
3531 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3532 revoke_delegation(dp);
3534 list_for_each_safe(pos, next, &nn->close_lru) {
3535 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3536 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3537 t = oo->oo_time - cutoff;
3538 new_timeo = min(new_timeo, t);
3541 release_openowner(oo);
3543 new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
3544 nfs4_unlock_state();
3548 static struct workqueue_struct *laundry_wq;
3549 static void laundromat_main(struct work_struct *);
3552 laundromat_main(struct work_struct *laundry)
3555 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3557 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3560 t = nfs4_laundromat(nn);
3561 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3562 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3565 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3567 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3568 return nfserr_bad_stateid;
3573 access_permit_read(struct nfs4_ol_stateid *stp)
3575 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3576 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3577 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3581 access_permit_write(struct nfs4_ol_stateid *stp)
3583 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3584 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3588 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3590 __be32 status = nfserr_openmode;
3592 /* For lock stateid's, we test the parent open, not the lock: */
3593 if (stp->st_openstp)
3594 stp = stp->st_openstp;
3595 if ((flags & WR_STATE) && !access_permit_write(stp))
3597 if ((flags & RD_STATE) && !access_permit_read(stp))
3604 static inline __be32
3605 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3607 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3609 else if (locks_in_grace(net)) {
3610 /* Answer in remaining cases depends on existence of
3611 * conflicting state; so we must wait out the grace period. */
3612 return nfserr_grace;
3613 } else if (flags & WR_STATE)
3614 return nfs4_share_conflict(current_fh,
3615 NFS4_SHARE_DENY_WRITE);
3616 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3617 return nfs4_share_conflict(current_fh,
3618 NFS4_SHARE_DENY_READ);
3622 * Allow READ/WRITE during grace period on recovered state only for files
3623 * that are not able to provide mandatory locking.
3626 grace_disallows_io(struct net *net, struct inode *inode)
3628 return locks_in_grace(net) && mandatory_lock(inode);
3631 /* Returns true iff a is later than b: */
3632 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3634 return (s32)(a->si_generation - b->si_generation) > 0;
3637 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3640 * When sessions are used the stateid generation number is ignored
3643 if (has_session && in->si_generation == 0)
3646 if (in->si_generation == ref->si_generation)
3649 /* If the client sends us a stateid from the future, it's buggy: */
3650 if (stateid_generation_after(in, ref))
3651 return nfserr_bad_stateid;
3653 * However, we could see a stateid from the past, even from a
3654 * non-buggy client. For example, if the client sends a lock
3655 * while some IO is outstanding, the lock may bump si_generation
3656 * while the IO is still in flight. The client could avoid that
3657 * situation by waiting for responses on all the IO requests,
3658 * but better performance may result in retrying IO that
3659 * receives an old_stateid error if requests are rarely
3660 * reordered in flight:
3662 return nfserr_old_stateid;
3665 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3667 struct nfs4_stid *s;
3668 struct nfs4_ol_stateid *ols;
3671 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3672 return nfserr_bad_stateid;
3673 /* Client debugging aid. */
3674 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3675 char addr_str[INET6_ADDRSTRLEN];
3676 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3678 pr_warn_ratelimited("NFSD: client %s testing state ID "
3679 "with incorrect client ID\n", addr_str);
3680 return nfserr_bad_stateid;
3682 s = find_stateid(cl, stateid);
3684 return nfserr_bad_stateid;
3685 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3688 switch (s->sc_type) {
3689 case NFS4_DELEG_STID:
3691 case NFS4_REVOKED_DELEG_STID:
3692 return nfserr_deleg_revoked;
3693 case NFS4_OPEN_STID:
3694 case NFS4_LOCK_STID:
3695 ols = openlockstateid(s);
3696 if (ols->st_stateowner->so_is_open_owner
3697 && !(openowner(ols->st_stateowner)->oo_flags
3698 & NFS4_OO_CONFIRMED))
3699 return nfserr_bad_stateid;
3702 printk("unknown stateid type %x\n", s->sc_type);
3703 case NFS4_CLOSED_STID:
3704 return nfserr_bad_stateid;
3708 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3709 struct nfs4_stid **s, bool sessions,
3710 struct nfsd_net *nn)
3712 struct nfs4_client *cl;
3715 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3716 return nfserr_bad_stateid;
3717 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3719 if (status == nfserr_stale_clientid) {
3721 return nfserr_bad_stateid;
3722 return nfserr_stale_stateid;
3726 *s = find_stateid_by_type(cl, stateid, typemask);
3728 return nfserr_bad_stateid;
3733 * Checks for stateid operations
3736 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3737 stateid_t *stateid, int flags, struct file **filpp)
3739 struct nfs4_stid *s;
3740 struct nfs4_ol_stateid *stp = NULL;
3741 struct nfs4_delegation *dp = NULL;
3742 struct svc_fh *current_fh = &cstate->current_fh;
3743 struct inode *ino = current_fh->fh_dentry->d_inode;
3744 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3745 struct file *file = NULL;
3751 if (grace_disallows_io(net, ino))
3752 return nfserr_grace;
3754 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3755 return check_special_stateids(net, current_fh, stateid, flags);
3759 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3760 &s, cstate->minorversion, nn);
3763 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3766 switch (s->sc_type) {
3767 case NFS4_DELEG_STID:
3768 dp = delegstateid(s);
3769 status = nfs4_check_delegmode(dp, flags);
3773 file = dp->dl_file->fi_deleg_file;
3776 status = nfserr_serverfault;
3781 case NFS4_OPEN_STID:
3782 case NFS4_LOCK_STID:
3783 stp = openlockstateid(s);
3784 status = nfs4_check_fh(current_fh, stp);
3787 if (stp->st_stateowner->so_is_open_owner
3788 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3790 status = nfs4_check_openmode(stp, flags);
3794 if (flags & RD_STATE)
3795 file = find_readable_file(stp->st_file);
3797 file = find_writeable_file(stp->st_file);
3801 status = nfserr_bad_stateid;
3806 *filpp = get_file(file);
3808 nfs4_unlock_state();
3813 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3815 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
3817 if (check_for_locks(stp->st_file, lo))
3818 return nfserr_locks_held;
3820 * Currently there's a 1-1 lock stateid<->lockowner
3821 * correspondance, and we have to delete the lockowner when we
3822 * delete the lock stateid:
3824 release_lockowner(lo);
3829 * Test if the stateid is valid
3832 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3833 struct nfsd4_test_stateid *test_stateid)
3835 struct nfsd4_test_stateid_id *stateid;
3836 struct nfs4_client *cl = cstate->session->se_client;
3839 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3840 stateid->ts_id_status =
3841 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3842 nfs4_unlock_state();
3848 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3849 struct nfsd4_free_stateid *free_stateid)
3851 stateid_t *stateid = &free_stateid->fr_stateid;
3852 struct nfs4_stid *s;
3853 struct nfs4_delegation *dp;
3854 struct nfs4_client *cl = cstate->session->se_client;
3855 __be32 ret = nfserr_bad_stateid;
3858 s = find_stateid(cl, stateid);
3861 switch (s->sc_type) {
3862 case NFS4_DELEG_STID:
3863 ret = nfserr_locks_held;
3865 case NFS4_OPEN_STID:
3866 case NFS4_LOCK_STID:
3867 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3870 if (s->sc_type == NFS4_LOCK_STID)
3871 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3873 ret = nfserr_locks_held;
3875 case NFS4_REVOKED_DELEG_STID:
3876 dp = delegstateid(s);
3877 destroy_revoked_delegation(dp);
3881 ret = nfserr_bad_stateid;
3884 nfs4_unlock_state();
3891 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3892 RD_STATE : WR_STATE;
3895 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3897 struct svc_fh *current_fh = &cstate->current_fh;
3898 struct nfs4_stateowner *sop = stp->st_stateowner;
3901 status = nfsd4_check_seqid(cstate, sop, seqid);
3904 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
3905 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
3907 * "Closed" stateid's exist *only* to return
3908 * nfserr_replay_me from the previous step, and
3909 * revoked delegations are kept only for free_stateid.
3911 return nfserr_bad_stateid;
3912 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3915 return nfs4_check_fh(current_fh, stp);
3919 * Checks for sequence id mutating operations.
3922 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3923 stateid_t *stateid, char typemask,
3924 struct nfs4_ol_stateid **stpp,
3925 struct nfsd_net *nn)
3928 struct nfs4_stid *s;
3930 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3931 seqid, STATEID_VAL(stateid));
3934 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3935 cstate->minorversion, nn);
3938 *stpp = openlockstateid(s);
3939 if (!nfsd4_has_session(cstate))
3940 cstate->replay_owner = (*stpp)->st_stateowner;
3942 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3945 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3946 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3949 struct nfs4_openowner *oo;
3951 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3952 NFS4_OPEN_STID, stpp, nn);
3955 oo = openowner((*stpp)->st_stateowner);
3956 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3957 return nfserr_bad_stateid;
3962 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3963 struct nfsd4_open_confirm *oc)
3966 struct nfs4_openowner *oo;
3967 struct nfs4_ol_stateid *stp;
3968 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3970 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
3971 cstate->current_fh.fh_dentry);
3973 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3979 status = nfs4_preprocess_seqid_op(cstate,
3980 oc->oc_seqid, &oc->oc_req_stateid,
3981 NFS4_OPEN_STID, &stp, nn);
3984 oo = openowner(stp->st_stateowner);
3985 status = nfserr_bad_stateid;
3986 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3988 oo->oo_flags |= NFS4_OO_CONFIRMED;
3989 update_stateid(&stp->st_stid.sc_stateid);
3990 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3991 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3992 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3994 nfsd4_client_record_create(oo->oo_owner.so_client);
3997 nfsd4_bump_seqid(cstate, status);
3998 if (!cstate->replay_owner)
3999 nfs4_unlock_state();
4003 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4005 if (!test_access(access, stp))
4007 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
4008 clear_access(access, stp);
4011 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4013 switch (to_access) {
4014 case NFS4_SHARE_ACCESS_READ:
4015 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4016 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4018 case NFS4_SHARE_ACCESS_WRITE:
4019 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4020 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4022 case NFS4_SHARE_ACCESS_BOTH:
4030 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
4033 for (i = 0; i < 4; i++) {
4034 if ((i & deny) != i)
4040 nfsd4_open_downgrade(struct svc_rqst *rqstp,
4041 struct nfsd4_compound_state *cstate,
4042 struct nfsd4_open_downgrade *od)
4045 struct nfs4_ol_stateid *stp;
4046 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4048 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
4049 cstate->current_fh.fh_dentry);
4051 /* We don't yet support WANT bits: */
4052 if (od->od_deleg_want)
4053 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4057 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4058 &od->od_stateid, &stp, nn);
4061 status = nfserr_inval;
4062 if (!test_access(od->od_share_access, stp)) {
4063 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
4064 stp->st_access_bmap, od->od_share_access);
4067 if (!test_deny(od->od_share_deny, stp)) {
4068 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
4069 stp->st_deny_bmap, od->od_share_deny);
4072 nfs4_stateid_downgrade(stp, od->od_share_access);
4074 reset_union_bmap_deny(od->od_share_deny, stp);
4076 update_stateid(&stp->st_stid.sc_stateid);
4077 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4080 nfsd4_bump_seqid(cstate, status);
4081 if (!cstate->replay_owner)
4082 nfs4_unlock_state();
4086 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4088 unhash_open_stateid(s);
4089 s->st_stid.sc_type = NFS4_CLOSED_STID;
4093 * nfs4_unlock_state() called after encode
4096 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4097 struct nfsd4_close *close)
4100 struct nfs4_openowner *oo;
4101 struct nfs4_ol_stateid *stp;
4102 struct net *net = SVC_NET(rqstp);
4103 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4105 dprintk("NFSD: nfsd4_close on file %pd\n",
4106 cstate->current_fh.fh_dentry);
4109 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4111 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4113 nfsd4_bump_seqid(cstate, status);
4116 oo = openowner(stp->st_stateowner);
4117 update_stateid(&stp->st_stid.sc_stateid);
4118 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4120 nfsd4_close_open_stateid(stp);
4122 if (cstate->minorversion)
4123 free_generic_stateid(stp);
4125 oo->oo_last_closed_stid = stp;
4127 if (list_empty(&oo->oo_owner.so_stateids)) {
4128 if (cstate->minorversion)
4129 release_openowner(oo);
4132 * In the 4.0 case we need to keep the owners around a
4133 * little while to handle CLOSE replay.
4135 move_to_close_lru(oo, SVC_NET(rqstp));
4139 if (!cstate->replay_owner)
4140 nfs4_unlock_state();
4145 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4146 struct nfsd4_delegreturn *dr)
4148 struct nfs4_delegation *dp;
4149 stateid_t *stateid = &dr->dr_stateid;
4150 struct nfs4_stid *s;
4152 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4154 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4158 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
4159 cstate->minorversion, nn);
4162 dp = delegstateid(s);
4163 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4167 destroy_delegation(dp);
4169 nfs4_unlock_state();
4175 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4177 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
4180 end_offset(u64 start, u64 len)
4185 return end >= start ? end: NFS4_MAX_UINT64;
4188 /* last octet in a range */
4190 last_byte_offset(u64 start, u64 len)
4196 return end > start ? end - 1: NFS4_MAX_UINT64;
4199 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
4201 return (file_hashval(inode) + cl_id
4202 + opaque_hashval(ownername->data, ownername->len))
4203 & LOCKOWNER_INO_HASH_MASK;
4207 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4208 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4209 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4210 * locking, this prevents us from being completely protocol-compliant. The
4211 * real solution to this problem is to start using unsigned file offsets in
4212 * the VFS, but this is a very deep change!
4215 nfs4_transform_lock_offset(struct file_lock *lock)
4217 if (lock->fl_start < 0)
4218 lock->fl_start = OFFSET_MAX;
4219 if (lock->fl_end < 0)
4220 lock->fl_end = OFFSET_MAX;
4223 /* Hack!: For now, we're defining this just so we can use a pointer to it
4224 * as a unique cookie to identify our (NFSv4's) posix locks. */
4225 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4229 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4231 struct nfs4_lockowner *lo;
4233 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4234 lo = (struct nfs4_lockowner *) fl->fl_owner;
4235 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4236 lo->lo_owner.so_owner.len, GFP_KERNEL);
4237 if (!deny->ld_owner.data)
4238 /* We just don't care that much */
4240 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4241 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4244 deny->ld_owner.len = 0;
4245 deny->ld_owner.data = NULL;
4246 deny->ld_clientid.cl_boot = 0;
4247 deny->ld_clientid.cl_id = 0;
4249 deny->ld_start = fl->fl_start;
4250 deny->ld_length = NFS4_MAX_UINT64;
4251 if (fl->fl_end != NFS4_MAX_UINT64)
4252 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4253 deny->ld_type = NFS4_READ_LT;
4254 if (fl->fl_type != F_RDLCK)
4255 deny->ld_type = NFS4_WRITE_LT;
4258 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
4260 struct nfs4_ol_stateid *lst;
4262 if (!same_owner_str(&lo->lo_owner, owner, clid))
4264 if (list_empty(&lo->lo_owner.so_stateids)) {
4268 lst = list_first_entry(&lo->lo_owner.so_stateids,
4269 struct nfs4_ol_stateid, st_perstateowner);
4270 return lst->st_file->fi_inode == inode;
4273 static struct nfs4_lockowner *
4274 find_lockowner_str(struct inode *inode, clientid_t *clid,
4275 struct xdr_netobj *owner, struct nfsd_net *nn)
4277 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
4278 struct nfs4_lockowner *lo;
4280 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
4281 if (same_lockowner_ino(lo, inode, clid, owner))
4287 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
4289 struct inode *inode = open_stp->st_file->fi_inode;
4290 unsigned int inohash = lockowner_ino_hashval(inode,
4291 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4292 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4294 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4295 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4296 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4300 * Alloc a lock owner structure.
4301 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4304 * strhashval = ownerstr_hashval
4307 static struct nfs4_lockowner *
4308 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4309 struct nfs4_lockowner *lo;
4311 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4314 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4315 lo->lo_owner.so_is_open_owner = 0;
4316 /* It is the openowner seqid that will be incremented in encode in the
4317 * case of new lockowners; so increment the lock seqid manually: */
4318 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4319 hash_lockowner(lo, strhashval, clp, open_stp);
4323 static struct nfs4_ol_stateid *
4324 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4326 struct nfs4_ol_stateid *stp;
4327 struct nfs4_client *clp = lo->lo_owner.so_client;
4329 stp = nfs4_alloc_stateid(clp);
4332 stp->st_stid.sc_type = NFS4_LOCK_STID;
4333 list_add(&stp->st_perfile, &fp->fi_stateids);
4334 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4335 stp->st_stateowner = &lo->lo_owner;
4338 stp->st_access_bmap = 0;
4339 stp->st_deny_bmap = open_stp->st_deny_bmap;
4340 stp->st_openstp = open_stp;
4345 check_lock_length(u64 offset, u64 length)
4347 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4348 LOFF_OVERFLOW(offset, length)));
4351 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4353 struct nfs4_file *fp = lock_stp->st_file;
4354 int oflag = nfs4_access_to_omode(access);
4356 if (test_access(access, lock_stp))
4358 nfs4_file_get_access(fp, oflag);
4359 set_access(access, lock_stp);
4362 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)
4364 struct nfs4_file *fi = ost->st_file;
4365 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4366 struct nfs4_client *cl = oo->oo_owner.so_client;
4367 struct nfs4_lockowner *lo;
4368 unsigned int strhashval;
4369 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4371 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4372 &lock->v.new.owner, nn);
4374 if (!cstate->minorversion)
4375 return nfserr_bad_seqid;
4376 /* XXX: a lockowner always has exactly one stateid: */
4377 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4378 struct nfs4_ol_stateid, st_perstateowner);
4381 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4382 &lock->v.new.owner);
4383 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4385 return nfserr_jukebox;
4386 *lst = alloc_init_lock_stateid(lo, fi, ost);
4388 release_lockowner(lo);
4389 return nfserr_jukebox;
4399 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4400 struct nfsd4_lock *lock)
4402 struct nfs4_openowner *open_sop = NULL;
4403 struct nfs4_lockowner *lock_sop = NULL;
4404 struct nfs4_ol_stateid *lock_stp;
4405 struct file *filp = NULL;
4406 struct file_lock *file_lock = NULL;
4407 struct file_lock *conflock = NULL;
4409 bool new_state = false;
4412 struct net *net = SVC_NET(rqstp);
4413 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4415 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4416 (long long) lock->lk_offset,
4417 (long long) lock->lk_length);
4419 if (check_lock_length(lock->lk_offset, lock->lk_length))
4420 return nfserr_inval;
4422 if ((status = fh_verify(rqstp, &cstate->current_fh,
4423 S_IFREG, NFSD_MAY_LOCK))) {
4424 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4430 if (lock->lk_is_new) {
4431 struct nfs4_ol_stateid *open_stp = NULL;
4433 if (nfsd4_has_session(cstate))
4434 /* See rfc 5661 18.10.3: given clientid is ignored: */
4435 memcpy(&lock->v.new.clientid,
4436 &cstate->session->se_client->cl_clientid,
4437 sizeof(clientid_t));
4439 status = nfserr_stale_clientid;
4440 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4443 /* validate and update open stateid and open seqid */
4444 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4445 lock->lk_new_open_seqid,
4446 &lock->lk_new_open_stateid,
4450 open_sop = openowner(open_stp->st_stateowner);
4451 status = nfserr_bad_stateid;
4452 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4453 &lock->v.new.clientid))
4455 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4456 &lock_stp, &new_state);
4458 status = nfs4_preprocess_seqid_op(cstate,
4459 lock->lk_old_lock_seqid,
4460 &lock->lk_old_lock_stateid,
4461 NFS4_LOCK_STID, &lock_stp, nn);
4464 lock_sop = lockowner(lock_stp->st_stateowner);
4466 lkflg = setlkflg(lock->lk_type);
4467 status = nfs4_check_openmode(lock_stp, lkflg);
4471 status = nfserr_grace;
4472 if (locks_in_grace(net) && !lock->lk_reclaim)
4474 status = nfserr_no_grace;
4475 if (!locks_in_grace(net) && lock->lk_reclaim)
4478 file_lock = locks_alloc_lock();
4480 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4481 status = nfserr_jukebox;
4485 locks_init_lock(file_lock);
4486 switch (lock->lk_type) {
4489 filp = find_readable_file(lock_stp->st_file);
4491 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4492 file_lock->fl_type = F_RDLCK;
4495 case NFS4_WRITEW_LT:
4496 filp = find_writeable_file(lock_stp->st_file);
4498 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4499 file_lock->fl_type = F_WRLCK;
4502 status = nfserr_inval;
4506 status = nfserr_openmode;
4509 file_lock->fl_owner = (fl_owner_t)lock_sop;
4510 file_lock->fl_pid = current->tgid;
4511 file_lock->fl_file = filp;
4512 file_lock->fl_flags = FL_POSIX;
4513 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4514 file_lock->fl_start = lock->lk_offset;
4515 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4516 nfs4_transform_lock_offset(file_lock);
4518 conflock = locks_alloc_lock();
4520 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4521 status = nfserr_jukebox;
4525 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4527 case 0: /* success! */
4528 update_stateid(&lock_stp->st_stid.sc_stateid);
4529 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4533 case (EAGAIN): /* conflock holds conflicting lock */
4534 status = nfserr_denied;
4535 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4536 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4539 status = nfserr_deadlock;
4542 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4543 status = nfserrno(err);
4547 if (status && new_state)
4548 release_lockowner(lock_sop);
4549 nfsd4_bump_seqid(cstate, status);
4550 if (!cstate->replay_owner)
4551 nfs4_unlock_state();
4553 locks_free_lock(file_lock);
4555 locks_free_lock(conflock);
4560 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4561 * so we do a temporary open here just to get an open file to pass to
4562 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4565 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4568 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4570 err = nfserrno(vfs_test_lock(file, lock));
4580 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4581 struct nfsd4_lockt *lockt)
4583 struct inode *inode;
4584 struct file_lock *file_lock = NULL;
4585 struct nfs4_lockowner *lo;
4587 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4589 if (locks_in_grace(SVC_NET(rqstp)))
4590 return nfserr_grace;
4592 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4593 return nfserr_inval;
4597 if (!nfsd4_has_session(cstate)) {
4598 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4603 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4606 inode = cstate->current_fh.fh_dentry->d_inode;
4607 file_lock = locks_alloc_lock();
4609 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4610 status = nfserr_jukebox;
4613 locks_init_lock(file_lock);
4614 switch (lockt->lt_type) {
4617 file_lock->fl_type = F_RDLCK;
4620 case NFS4_WRITEW_LT:
4621 file_lock->fl_type = F_WRLCK;
4624 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4625 status = nfserr_inval;
4629 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4631 file_lock->fl_owner = (fl_owner_t)lo;
4632 file_lock->fl_pid = current->tgid;
4633 file_lock->fl_flags = FL_POSIX;
4635 file_lock->fl_start = lockt->lt_offset;
4636 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4638 nfs4_transform_lock_offset(file_lock);
4640 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4644 if (file_lock->fl_type != F_UNLCK) {
4645 status = nfserr_denied;
4646 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4649 nfs4_unlock_state();
4651 locks_free_lock(file_lock);
4656 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4657 struct nfsd4_locku *locku)
4659 struct nfs4_ol_stateid *stp;
4660 struct file *filp = NULL;
4661 struct file_lock *file_lock = NULL;
4664 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4666 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4667 (long long) locku->lu_offset,
4668 (long long) locku->lu_length);
4670 if (check_lock_length(locku->lu_offset, locku->lu_length))
4671 return nfserr_inval;
4675 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4676 &locku->lu_stateid, NFS4_LOCK_STID,
4680 filp = find_any_file(stp->st_file);
4682 status = nfserr_lock_range;
4685 file_lock = locks_alloc_lock();
4687 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4688 status = nfserr_jukebox;
4691 locks_init_lock(file_lock);
4692 file_lock->fl_type = F_UNLCK;
4693 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4694 file_lock->fl_pid = current->tgid;
4695 file_lock->fl_file = filp;
4696 file_lock->fl_flags = FL_POSIX;
4697 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4698 file_lock->fl_start = locku->lu_offset;
4700 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4702 nfs4_transform_lock_offset(file_lock);
4704 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4706 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4709 update_stateid(&stp->st_stid.sc_stateid);
4710 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4713 nfsd4_bump_seqid(cstate, status);
4714 if (!cstate->replay_owner)
4715 nfs4_unlock_state();
4717 locks_free_lock(file_lock);
4721 status = nfserrno(err);
4727 * 1: locks held by lockowner
4728 * 0: no locks held by lockowner
4731 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4733 struct file_lock **flpp;
4734 struct inode *inode = filp->fi_inode;
4737 spin_lock(&inode->i_lock);
4738 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4739 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4745 spin_unlock(&inode->i_lock);
4750 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4751 struct nfsd4_compound_state *cstate,
4752 struct nfsd4_release_lockowner *rlockowner)
4754 clientid_t *clid = &rlockowner->rl_clientid;
4755 struct nfs4_stateowner *sop;
4756 struct nfs4_lockowner *lo;
4757 struct nfs4_ol_stateid *stp;
4758 struct xdr_netobj *owner = &rlockowner->rl_owner;
4759 struct list_head matches;
4760 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4762 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4764 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4765 clid->cl_boot, clid->cl_id);
4769 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4773 status = nfserr_locks_held;
4774 INIT_LIST_HEAD(&matches);
4776 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4777 if (sop->so_is_open_owner)
4779 if (!same_owner_str(sop, owner, clid))
4781 list_for_each_entry(stp, &sop->so_stateids,
4783 lo = lockowner(sop);
4784 if (check_for_locks(stp->st_file, lo))
4786 list_add(&lo->lo_list, &matches);
4789 /* Clients probably won't expect us to return with some (but not all)
4790 * of the lockowner state released; so don't release any until all
4791 * have been checked. */
4793 while (!list_empty(&matches)) {
4794 lo = list_entry(matches.next, struct nfs4_lockowner,
4796 /* unhash_stateowner deletes so_perclient only
4797 * for openowners. */
4798 list_del(&lo->lo_list);
4799 release_lockowner(lo);
4802 nfs4_unlock_state();
4806 static inline struct nfs4_client_reclaim *
4809 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4813 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4815 struct nfs4_client_reclaim *crp;
4817 crp = nfsd4_find_reclaim_client(name, nn);
4818 return (crp && crp->cr_clp);
4822 * failure => all reset bets are off, nfserr_no_grace...
4824 struct nfs4_client_reclaim *
4825 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4827 unsigned int strhashval;
4828 struct nfs4_client_reclaim *crp;
4830 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4831 crp = alloc_reclaim();
4833 strhashval = clientstr_hashval(name);
4834 INIT_LIST_HEAD(&crp->cr_strhash);
4835 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4836 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4838 nn->reclaim_str_hashtbl_size++;
4844 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4846 list_del(&crp->cr_strhash);
4848 nn->reclaim_str_hashtbl_size--;
4852 nfs4_release_reclaim(struct nfsd_net *nn)
4854 struct nfs4_client_reclaim *crp = NULL;
4857 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4858 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4859 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4860 struct nfs4_client_reclaim, cr_strhash);
4861 nfs4_remove_reclaim_record(crp, nn);
4864 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4868 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4869 struct nfs4_client_reclaim *
4870 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4872 unsigned int strhashval;
4873 struct nfs4_client_reclaim *crp = NULL;
4875 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4877 strhashval = clientstr_hashval(recdir);
4878 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4879 if (same_name(crp->cr_recdir, recdir)) {
4887 * Called from OPEN. Look for clientid in reclaim list.
4890 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4892 struct nfs4_client *clp;
4894 /* find clientid in conf_id_hashtbl */
4895 clp = find_confirmed_client(clid, sessions, nn);
4897 return nfserr_reclaim_bad;
4899 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4902 #ifdef CONFIG_NFSD_FAULT_INJECTION
4904 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4906 if (mark_client_expired(clp))
4912 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4914 char buf[INET6_ADDRSTRLEN];
4915 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4916 printk(KERN_INFO "NFS Client: %s\n", buf);
4920 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4923 char buf[INET6_ADDRSTRLEN];
4924 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4925 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4928 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4930 struct nfs4_openowner *oop;
4931 struct nfs4_lockowner *lop, *lo_next;
4932 struct nfs4_ol_stateid *stp, *st_next;
4935 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4936 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4937 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4949 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4951 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4954 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4956 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4957 nfsd_print_count(clp, count, "locked files");
4961 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4963 struct nfs4_openowner *oop, *next;
4966 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4976 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4978 return nfsd_foreach_client_open(clp, max, release_openowner);
4981 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4983 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4984 nfsd_print_count(clp, count, "open files");
4988 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4989 struct list_head *victims)
4991 struct nfs4_delegation *dp, *next;
4994 lockdep_assert_held(&state_lock);
4995 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4997 list_move(&dp->dl_recall_lru, victims);
5004 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
5006 struct nfs4_delegation *dp, *next;
5010 spin_lock(&state_lock);
5011 count = nfsd_find_all_delegations(clp, max, &victims);
5012 spin_unlock(&state_lock);
5014 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
5015 revoke_delegation(dp);
5020 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
5022 struct nfs4_delegation *dp, *next;
5026 spin_lock(&state_lock);
5027 count = nfsd_find_all_delegations(clp, max, &victims);
5028 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
5029 nfsd_break_one_deleg(dp);
5030 spin_unlock(&state_lock);
5035 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
5039 spin_lock(&state_lock);
5040 count = nfsd_find_all_delegations(clp, max, NULL);
5041 spin_unlock(&state_lock);
5043 nfsd_print_count(clp, count, "delegations");
5047 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
5049 struct nfs4_client *clp, *next;
5051 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5053 if (!nfsd_netns_ready(nn))
5056 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5057 count += func(clp, max - count);
5058 if ((max != 0) && (count >= max))
5065 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5067 struct nfs4_client *clp;
5068 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
5070 if (!nfsd_netns_ready(nn))
5073 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5074 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5080 #endif /* CONFIG_NFSD_FAULT_INJECTION */
5083 * Since the lifetime of a delegation isn't limited to that of an open, a
5084 * client may quite reasonably hang on to a delegation as long as it has
5085 * the inode cached. This becomes an obvious problem the first time a
5086 * client's inode cache approaches the size of the server's total memory.
5088 * For now we avoid this problem by imposing a hard limit on the number
5089 * of delegations, which varies according to the server's memory size.
5092 set_max_delegations(void)
5095 * Allow at most 4 delegations per megabyte of RAM. Quick
5096 * estimates suggest that in the worst case (where every delegation
5097 * is for a different inode), a delegation could take about 1.5K,
5098 * giving a worst case usage of about 6% of memory.
5100 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
5103 static int nfs4_state_create_net(struct net *net)
5105 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5108 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5109 CLIENT_HASH_SIZE, GFP_KERNEL);
5110 if (!nn->conf_id_hashtbl)
5112 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5113 CLIENT_HASH_SIZE, GFP_KERNEL);
5114 if (!nn->unconf_id_hashtbl)
5116 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
5117 OWNER_HASH_SIZE, GFP_KERNEL);
5118 if (!nn->ownerstr_hashtbl)
5120 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
5121 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
5122 if (!nn->lockowner_ino_hashtbl)
5123 goto err_lockowner_ino;
5124 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
5125 SESSION_HASH_SIZE, GFP_KERNEL);
5126 if (!nn->sessionid_hashtbl)
5129 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5130 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5131 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5133 for (i = 0; i < OWNER_HASH_SIZE; i++)
5134 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
5135 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
5136 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
5137 for (i = 0; i < SESSION_HASH_SIZE; i++)
5138 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5139 nn->conf_name_tree = RB_ROOT;
5140 nn->unconf_name_tree = RB_ROOT;
5141 INIT_LIST_HEAD(&nn->client_lru);
5142 INIT_LIST_HEAD(&nn->close_lru);
5143 INIT_LIST_HEAD(&nn->del_recall_lru);
5144 spin_lock_init(&nn->client_lock);
5146 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5152 kfree(nn->lockowner_ino_hashtbl);
5154 kfree(nn->ownerstr_hashtbl);
5156 kfree(nn->unconf_id_hashtbl);
5158 kfree(nn->conf_id_hashtbl);
5164 nfs4_state_destroy_net(struct net *net)
5167 struct nfs4_client *clp = NULL;
5168 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5170 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5171 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5172 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5173 destroy_client(clp);
5177 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5178 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
5179 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5180 destroy_client(clp);
5184 kfree(nn->sessionid_hashtbl);
5185 kfree(nn->lockowner_ino_hashtbl);
5186 kfree(nn->ownerstr_hashtbl);
5187 kfree(nn->unconf_id_hashtbl);
5188 kfree(nn->conf_id_hashtbl);
5193 nfs4_state_start_net(struct net *net)
5195 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5198 ret = nfs4_state_create_net(net);
5201 nfsd4_client_tracking_init(net);
5202 nn->boot_time = get_seconds();
5203 locks_start_grace(net, &nn->nfsd4_manager);
5204 nn->grace_ended = false;
5205 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5206 nn->nfsd4_grace, net);
5207 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5211 /* initialization to perform when the nfsd service is started: */
5214 nfs4_state_start(void)
5218 ret = set_callback_cred();
5221 laundry_wq = create_singlethread_workqueue("nfsd4");
5222 if (laundry_wq == NULL) {
5226 ret = nfsd4_create_callback_queue();
5228 goto out_free_laundry;
5230 set_max_delegations();
5235 destroy_workqueue(laundry_wq);
5241 nfs4_state_shutdown_net(struct net *net)
5243 struct nfs4_delegation *dp = NULL;
5244 struct list_head *pos, *next, reaplist;
5245 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5247 cancel_delayed_work_sync(&nn->laundromat_work);
5248 locks_end_grace(&nn->nfsd4_manager);
5251 INIT_LIST_HEAD(&reaplist);
5252 spin_lock(&state_lock);
5253 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5254 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5255 list_move(&dp->dl_recall_lru, &reaplist);
5257 spin_unlock(&state_lock);
5258 list_for_each_safe(pos, next, &reaplist) {
5259 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5260 destroy_delegation(dp);
5263 nfsd4_client_tracking_exit(net);
5264 nfs4_state_destroy_net(net);
5265 nfs4_unlock_state();
5269 nfs4_state_shutdown(void)
5271 destroy_workqueue(laundry_wq);
5272 nfsd4_destroy_callback_queue();
5276 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5278 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5279 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5283 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5285 if (cstate->minorversion) {
5286 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5287 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5292 clear_current_stateid(struct nfsd4_compound_state *cstate)
5294 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5298 * functions to set current state id
5301 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5303 put_stateid(cstate, &odp->od_stateid);
5307 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5309 put_stateid(cstate, &open->op_stateid);
5313 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5315 put_stateid(cstate, &close->cl_stateid);
5319 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5321 put_stateid(cstate, &lock->lk_resp_stateid);
5325 * functions to consume current state id
5329 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5331 get_stateid(cstate, &odp->od_stateid);
5335 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5337 get_stateid(cstate, &drp->dr_stateid);
5341 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5343 get_stateid(cstate, &fsp->fr_stateid);
5347 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5349 get_stateid(cstate, &setattr->sa_stateid);
5353 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5355 get_stateid(cstate, &close->cl_stateid);
5359 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5361 get_stateid(cstate, &locku->lu_stateid);
5365 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5367 get_stateid(cstate, &read->rd_stateid);
5371 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5373 get_stateid(cstate, &write->wr_stateid);