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>
47 #include "current_stateid.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
53 #define all_ones {{~0,~0},~0}
54 static const stateid_t one_stateid = {
56 .si_opaque = all_ones,
58 static const stateid_t zero_stateid = {
61 static const stateid_t currentstateid = {
65 static u64 current_sessionid = 1;
67 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
68 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
69 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
71 /* forward declarations */
72 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
76 /* Currently used for almost all code touching nfsv4 state: */
77 static DEFINE_MUTEX(client_mutex);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(recall_lock);
86 static struct kmem_cache *openowner_slab = NULL;
87 static struct kmem_cache *lockowner_slab = NULL;
88 static struct kmem_cache *file_slab = NULL;
89 static struct kmem_cache *stateid_slab = NULL;
90 static struct kmem_cache *deleg_slab = NULL;
95 mutex_lock(&client_mutex);
98 static void free_session(struct nfsd4_session *);
100 void nfsd4_put_session(struct nfsd4_session *ses)
102 atomic_dec(&ses->se_ref);
105 static bool is_session_dead(struct nfsd4_session *ses)
107 return ses->se_flags & NFS4_SESSION_DEAD;
110 static __be32 mark_session_dead_locked(struct nfsd4_session *ses)
112 if (atomic_read(&ses->se_ref))
113 return nfserr_jukebox;
114 ses->se_flags |= NFS4_SESSION_DEAD;
118 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
120 if (is_session_dead(ses))
121 return nfserr_badsession;
122 atomic_inc(&ses->se_ref);
127 nfs4_unlock_state(void)
129 mutex_unlock(&client_mutex);
132 static bool is_client_expired(struct nfs4_client *clp)
134 return clp->cl_time == 0;
137 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
139 if (atomic_read(&clp->cl_refcount))
140 return nfserr_jukebox;
145 static __be32 mark_client_expired(struct nfs4_client *clp)
147 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
150 spin_lock(&nn->client_lock);
151 ret = mark_client_expired_locked(clp);
152 spin_unlock(&nn->client_lock);
156 static __be32 get_client_locked(struct nfs4_client *clp)
158 if (is_client_expired(clp))
159 return nfserr_expired;
160 atomic_inc(&clp->cl_refcount);
164 /* must be called under the client_lock */
166 renew_client_locked(struct nfs4_client *clp)
168 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
170 if (is_client_expired(clp)) {
172 printk("%s: client (clientid %08x/%08x) already expired\n",
174 clp->cl_clientid.cl_boot,
175 clp->cl_clientid.cl_id);
179 dprintk("renewing client (clientid %08x/%08x)\n",
180 clp->cl_clientid.cl_boot,
181 clp->cl_clientid.cl_id);
182 list_move_tail(&clp->cl_lru, &nn->client_lru);
183 clp->cl_time = get_seconds();
187 renew_client(struct nfs4_client *clp)
189 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
191 spin_lock(&nn->client_lock);
192 renew_client_locked(clp);
193 spin_unlock(&nn->client_lock);
196 void put_client_renew_locked(struct nfs4_client *clp)
198 if (!atomic_dec_and_test(&clp->cl_refcount))
200 if (!is_client_expired(clp))
201 renew_client_locked(clp);
204 void put_client_renew(struct nfs4_client *clp)
206 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
208 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
210 if (!is_client_expired(clp))
211 renew_client_locked(clp);
212 spin_unlock(&nn->client_lock);
217 opaque_hashval(const void *ptr, int nbytes)
219 unsigned char *cptr = (unsigned char *) ptr;
229 static void nfsd4_free_file(struct nfs4_file *f)
231 kmem_cache_free(file_slab, f);
235 put_nfs4_file(struct nfs4_file *fi)
237 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
238 hlist_del(&fi->fi_hash);
239 spin_unlock(&recall_lock);
246 get_nfs4_file(struct nfs4_file *fi)
248 atomic_inc(&fi->fi_ref);
251 static int num_delegations;
252 unsigned long max_delegations;
255 * Open owner state (share locks)
258 /* hash tables for lock and open owners */
259 #define OWNER_HASH_BITS 8
260 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
261 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
263 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
267 ret = opaque_hashval(ownername->data, ownername->len);
269 return ret & OWNER_HASH_MASK;
272 /* hash table for nfs4_file */
273 #define FILE_HASH_BITS 8
274 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
276 static unsigned int file_hashval(struct inode *ino)
278 /* XXX: why are we hashing on inode pointer, anyway? */
279 return hash_ptr(ino, FILE_HASH_BITS);
282 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
284 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
286 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
287 atomic_inc(&fp->fi_access[oflag]);
290 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
292 if (oflag == O_RDWR) {
293 __nfs4_file_get_access(fp, O_RDONLY);
294 __nfs4_file_get_access(fp, O_WRONLY);
296 __nfs4_file_get_access(fp, oflag);
299 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
301 if (fp->fi_fds[oflag]) {
302 fput(fp->fi_fds[oflag]);
303 fp->fi_fds[oflag] = NULL;
307 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
309 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
310 nfs4_file_put_fd(fp, oflag);
311 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
312 nfs4_file_put_fd(fp, O_RDWR);
316 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
318 if (oflag == O_RDWR) {
319 __nfs4_file_put_access(fp, O_RDONLY);
320 __nfs4_file_put_access(fp, O_WRONLY);
322 __nfs4_file_put_access(fp, oflag);
325 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
328 struct idr *stateids = &cl->cl_stateids;
329 static int min_stateid = 0;
330 struct nfs4_stid *stid;
333 stid = kmem_cache_alloc(slab, GFP_KERNEL);
337 new_id = idr_alloc(stateids, stid, min_stateid, 0, GFP_KERNEL);
340 stid->sc_client = cl;
342 stid->sc_stateid.si_opaque.so_id = new_id;
343 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
344 /* Will be incremented before return to client: */
345 stid->sc_stateid.si_generation = 0;
348 * It shouldn't be a problem to reuse an opaque stateid value.
349 * I don't think it is for 4.1. But with 4.0 I worry that, for
350 * example, a stray write retransmission could be accepted by
351 * the server when it should have been rejected. Therefore,
352 * adopt a trick from the sctp code to attempt to maximize the
353 * amount of time until an id is reused, by ensuring they always
354 * "increase" (mod INT_MAX):
357 min_stateid = new_id+1;
358 if (min_stateid == INT_MAX)
362 kmem_cache_free(slab, stid);
366 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
368 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
371 static struct nfs4_delegation *
372 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
374 struct nfs4_delegation *dp;
375 struct nfs4_file *fp = stp->st_file;
377 dprintk("NFSD alloc_init_deleg\n");
379 * Major work on the lease subsystem (for example, to support
380 * calbacks on stat) will be required before we can support
381 * write delegations properly.
383 if (type != NFS4_OPEN_DELEGATE_READ)
385 if (fp->fi_had_conflict)
387 if (num_delegations > max_delegations)
389 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
392 dp->dl_stid.sc_type = NFS4_DELEG_STID;
394 * delegation seqid's are never incremented. The 4.1 special
395 * meaning of seqid 0 isn't meaningful, really, but let's avoid
396 * 0 anyway just for consistency and use 1:
398 dp->dl_stid.sc_stateid.si_generation = 1;
400 INIT_LIST_HEAD(&dp->dl_perfile);
401 INIT_LIST_HEAD(&dp->dl_perclnt);
402 INIT_LIST_HEAD(&dp->dl_recall_lru);
406 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
408 atomic_set(&dp->dl_count, 1);
409 nfsd4_init_callback(&dp->dl_recall);
413 static void remove_stid(struct nfs4_stid *s)
415 struct idr *stateids = &s->sc_client->cl_stateids;
417 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
421 nfs4_put_delegation(struct nfs4_delegation *dp)
423 if (atomic_dec_and_test(&dp->dl_count)) {
424 kmem_cache_free(deleg_slab, dp);
429 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
431 if (atomic_dec_and_test(&fp->fi_delegees)) {
432 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
434 fput(fp->fi_deleg_file);
435 fp->fi_deleg_file = NULL;
439 static void unhash_stid(struct nfs4_stid *s)
444 /* Called under the state lock. */
446 unhash_delegation(struct nfs4_delegation *dp)
448 unhash_stid(&dp->dl_stid);
449 list_del_init(&dp->dl_perclnt);
450 spin_lock(&recall_lock);
451 list_del_init(&dp->dl_perfile);
452 list_del_init(&dp->dl_recall_lru);
453 spin_unlock(&recall_lock);
454 nfs4_put_deleg_lease(dp->dl_file);
455 put_nfs4_file(dp->dl_file);
457 remove_stid(&dp->dl_stid);
458 nfs4_put_delegation(dp);
465 static unsigned int clientid_hashval(u32 id)
467 return id & CLIENT_HASH_MASK;
470 static unsigned int clientstr_hashval(const char *name)
472 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
476 * We store the NONE, READ, WRITE, and BOTH bits separately in the
477 * st_{access,deny}_bmap field of the stateid, in order to track not
478 * only what share bits are currently in force, but also what
479 * combinations of share bits previous opens have used. This allows us
480 * to enforce the recommendation of rfc 3530 14.2.19 that the server
481 * return an error if the client attempt to downgrade to a combination
482 * of share bits not explicable by closing some of its previous opens.
484 * XXX: This enforcement is actually incomplete, since we don't keep
485 * track of access/deny bit combinations; so, e.g., we allow:
487 * OPEN allow read, deny write
488 * OPEN allow both, deny none
489 * DOWNGRADE allow read, deny none
491 * which we should reject.
494 bmap_to_share_mode(unsigned long bmap) {
496 unsigned int access = 0;
498 for (i = 1; i < 4; i++) {
499 if (test_bit(i, &bmap))
506 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
507 unsigned int access, deny;
509 access = bmap_to_share_mode(stp->st_access_bmap);
510 deny = bmap_to_share_mode(stp->st_deny_bmap);
511 if ((access & open->op_share_deny) || (deny & open->op_share_access))
516 /* set share access for a given stateid */
518 set_access(u32 access, struct nfs4_ol_stateid *stp)
520 __set_bit(access, &stp->st_access_bmap);
523 /* clear share access for a given stateid */
525 clear_access(u32 access, struct nfs4_ol_stateid *stp)
527 __clear_bit(access, &stp->st_access_bmap);
530 /* test whether a given stateid has access */
532 test_access(u32 access, struct nfs4_ol_stateid *stp)
534 return test_bit(access, &stp->st_access_bmap);
537 /* set share deny for a given stateid */
539 set_deny(u32 access, struct nfs4_ol_stateid *stp)
541 __set_bit(access, &stp->st_deny_bmap);
544 /* clear share deny for a given stateid */
546 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
548 __clear_bit(access, &stp->st_deny_bmap);
551 /* test whether a given stateid is denying specific access */
553 test_deny(u32 access, struct nfs4_ol_stateid *stp)
555 return test_bit(access, &stp->st_deny_bmap);
558 static int nfs4_access_to_omode(u32 access)
560 switch (access & NFS4_SHARE_ACCESS_BOTH) {
561 case NFS4_SHARE_ACCESS_READ:
563 case NFS4_SHARE_ACCESS_WRITE:
565 case NFS4_SHARE_ACCESS_BOTH:
572 /* release all access and file references for a given stateid */
574 release_all_access(struct nfs4_ol_stateid *stp)
578 for (i = 1; i < 4; i++) {
579 if (test_access(i, stp))
580 nfs4_file_put_access(stp->st_file,
581 nfs4_access_to_omode(i));
582 clear_access(i, stp);
586 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
588 list_del(&stp->st_perfile);
589 list_del(&stp->st_perstateowner);
592 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
594 release_all_access(stp);
595 put_nfs4_file(stp->st_file);
599 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
601 remove_stid(&stp->st_stid);
602 kmem_cache_free(stateid_slab, stp);
605 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
609 unhash_generic_stateid(stp);
610 unhash_stid(&stp->st_stid);
611 file = find_any_file(stp->st_file);
613 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
614 close_generic_stateid(stp);
615 free_generic_stateid(stp);
618 static void unhash_lockowner(struct nfs4_lockowner *lo)
620 struct nfs4_ol_stateid *stp;
622 list_del(&lo->lo_owner.so_strhash);
623 list_del(&lo->lo_perstateid);
624 list_del(&lo->lo_owner_ino_hash);
625 while (!list_empty(&lo->lo_owner.so_stateids)) {
626 stp = list_first_entry(&lo->lo_owner.so_stateids,
627 struct nfs4_ol_stateid, st_perstateowner);
628 release_lock_stateid(stp);
632 static void release_lockowner(struct nfs4_lockowner *lo)
634 unhash_lockowner(lo);
635 nfs4_free_lockowner(lo);
639 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
641 struct nfs4_lockowner *lo;
643 while (!list_empty(&open_stp->st_lockowners)) {
644 lo = list_entry(open_stp->st_lockowners.next,
645 struct nfs4_lockowner, lo_perstateid);
646 release_lockowner(lo);
650 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
652 unhash_generic_stateid(stp);
653 release_stateid_lockowners(stp);
654 close_generic_stateid(stp);
657 static void release_open_stateid(struct nfs4_ol_stateid *stp)
659 unhash_open_stateid(stp);
660 unhash_stid(&stp->st_stid);
661 free_generic_stateid(stp);
664 static void unhash_openowner(struct nfs4_openowner *oo)
666 struct nfs4_ol_stateid *stp;
668 list_del(&oo->oo_owner.so_strhash);
669 list_del(&oo->oo_perclient);
670 while (!list_empty(&oo->oo_owner.so_stateids)) {
671 stp = list_first_entry(&oo->oo_owner.so_stateids,
672 struct nfs4_ol_stateid, st_perstateowner);
673 release_open_stateid(stp);
677 static void release_last_closed_stateid(struct nfs4_openowner *oo)
679 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
682 unhash_stid(&s->st_stid);
683 free_generic_stateid(s);
684 oo->oo_last_closed_stid = NULL;
688 static void release_openowner(struct nfs4_openowner *oo)
690 unhash_openowner(oo);
691 list_del(&oo->oo_close_lru);
692 release_last_closed_stateid(oo);
693 nfs4_free_openowner(oo);
697 hash_sessionid(struct nfs4_sessionid *sessionid)
699 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
701 return sid->sequence % SESSION_HASH_SIZE;
706 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
708 u32 *ptr = (u32 *)(&sessionid->data[0]);
709 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
713 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
719 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
720 * won't be used for replay.
722 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
724 struct nfs4_stateowner *so = cstate->replay_owner;
726 if (nfserr == nfserr_replay_me)
729 if (!seqid_mutating_err(ntohl(nfserr))) {
730 cstate->replay_owner = NULL;
735 if (so->so_is_open_owner)
736 release_last_closed_stateid(openowner(so));
742 gen_sessionid(struct nfsd4_session *ses)
744 struct nfs4_client *clp = ses->se_client;
745 struct nfsd4_sessionid *sid;
747 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
748 sid->clientid = clp->cl_clientid;
749 sid->sequence = current_sessionid++;
754 * The protocol defines ca_maxresponssize_cached to include the size of
755 * the rpc header, but all we need to cache is the data starting after
756 * the end of the initial SEQUENCE operation--the rest we regenerate
757 * each time. Therefore we can advertise a ca_maxresponssize_cached
758 * value that is the number of bytes in our cache plus a few additional
759 * bytes. In order to stay on the safe side, and not promise more than
760 * we can cache, those additional bytes must be the minimum possible: 24
761 * bytes of rpc header (xid through accept state, with AUTH_NULL
762 * verifier), 12 for the compound header (with zero-length tag), and 44
763 * for the SEQUENCE op response:
765 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
768 free_session_slots(struct nfsd4_session *ses)
772 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
773 kfree(ses->se_slots[i]);
777 * We don't actually need to cache the rpc and session headers, so we
778 * can allocate a little less for each slot:
780 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
784 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
787 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
788 return size + sizeof(struct nfsd4_slot);
792 * XXX: If we run out of reserved DRC memory we could (up to a point)
793 * re-negotiate active sessions and reduce their slot usage to make
794 * room for new connections. For now we just fail the create session.
796 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
798 u32 slotsize = slot_bytes(ca);
799 u32 num = ca->maxreqs;
802 spin_lock(&nfsd_drc_lock);
803 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
804 nfsd_drc_max_mem - nfsd_drc_mem_used);
805 num = min_t(int, num, avail / slotsize);
806 nfsd_drc_mem_used += num * slotsize;
807 spin_unlock(&nfsd_drc_lock);
812 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
814 int slotsize = slot_bytes(ca);
816 spin_lock(&nfsd_drc_lock);
817 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
818 spin_unlock(&nfsd_drc_lock);
821 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *attrs)
823 int numslots = attrs->maxreqs;
824 int slotsize = slot_bytes(attrs);
825 struct nfsd4_session *new;
828 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
829 + sizeof(struct nfsd4_session) > PAGE_SIZE);
830 mem = numslots * sizeof(struct nfsd4_slot *);
832 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
835 /* allocate each struct nfsd4_slot and data cache in one piece */
836 for (i = 0; i < numslots; i++) {
837 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
838 if (!new->se_slots[i])
844 kfree(new->se_slots[i]);
849 static void free_conn(struct nfsd4_conn *c)
851 svc_xprt_put(c->cn_xprt);
855 static void nfsd4_conn_lost(struct svc_xpt_user *u)
857 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
858 struct nfs4_client *clp = c->cn_session->se_client;
860 spin_lock(&clp->cl_lock);
861 if (!list_empty(&c->cn_persession)) {
862 list_del(&c->cn_persession);
865 nfsd4_probe_callback(clp);
866 spin_unlock(&clp->cl_lock);
869 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
871 struct nfsd4_conn *conn;
873 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
876 svc_xprt_get(rqstp->rq_xprt);
877 conn->cn_xprt = rqstp->rq_xprt;
878 conn->cn_flags = flags;
879 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
883 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
885 conn->cn_session = ses;
886 list_add(&conn->cn_persession, &ses->se_conns);
889 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
891 struct nfs4_client *clp = ses->se_client;
893 spin_lock(&clp->cl_lock);
894 __nfsd4_hash_conn(conn, ses);
895 spin_unlock(&clp->cl_lock);
898 static int nfsd4_register_conn(struct nfsd4_conn *conn)
900 conn->cn_xpt_user.callback = nfsd4_conn_lost;
901 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
904 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
908 nfsd4_hash_conn(conn, ses);
909 ret = nfsd4_register_conn(conn);
911 /* oops; xprt is already down: */
912 nfsd4_conn_lost(&conn->cn_xpt_user);
913 if (conn->cn_flags & NFS4_CDFC4_BACK) {
914 /* callback channel may be back up */
915 nfsd4_probe_callback(ses->se_client);
919 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
921 u32 dir = NFS4_CDFC4_FORE;
923 if (cses->flags & SESSION4_BACK_CHAN)
924 dir |= NFS4_CDFC4_BACK;
925 return alloc_conn(rqstp, dir);
928 /* must be called under client_lock */
929 static void nfsd4_del_conns(struct nfsd4_session *s)
931 struct nfs4_client *clp = s->se_client;
932 struct nfsd4_conn *c;
934 spin_lock(&clp->cl_lock);
935 while (!list_empty(&s->se_conns)) {
936 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
937 list_del_init(&c->cn_persession);
938 spin_unlock(&clp->cl_lock);
940 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
943 spin_lock(&clp->cl_lock);
945 spin_unlock(&clp->cl_lock);
948 static void __free_session(struct nfsd4_session *ses)
950 free_session_slots(ses);
954 static void free_session(struct nfsd4_session *ses)
956 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
958 lockdep_assert_held(&nn->client_lock);
959 nfsd4_del_conns(ses);
960 nfsd4_put_drc_mem(&ses->se_fchannel);
964 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
967 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
969 new->se_client = clp;
972 INIT_LIST_HEAD(&new->se_conns);
974 new->se_cb_seq_nr = 1;
975 new->se_flags = cses->flags;
976 new->se_cb_prog = cses->callback_prog;
977 new->se_cb_sec = cses->cb_sec;
978 atomic_set(&new->se_ref, 0);
979 idx = hash_sessionid(&new->se_sessionid);
980 spin_lock(&nn->client_lock);
981 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
982 spin_lock(&clp->cl_lock);
983 list_add(&new->se_perclnt, &clp->cl_sessions);
984 spin_unlock(&clp->cl_lock);
985 spin_unlock(&nn->client_lock);
986 memcpy(&new->se_fchannel, &cses->fore_channel,
987 sizeof(struct nfsd4_channel_attrs));
988 if (cses->flags & SESSION4_BACK_CHAN) {
989 struct sockaddr *sa = svc_addr(rqstp);
991 * This is a little silly; with sessions there's no real
992 * use for the callback address. Use the peer address
993 * as a reasonable default for now, but consider fixing
994 * the rpc client not to require an address in the
997 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
998 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1002 /* caller must hold client_lock */
1003 static struct nfsd4_session *
1004 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1006 struct nfsd4_session *elem;
1008 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1010 dump_sessionid(__func__, sessionid);
1011 idx = hash_sessionid(sessionid);
1012 /* Search in the appropriate list */
1013 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1014 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1015 NFS4_MAX_SESSIONID_LEN)) {
1020 dprintk("%s: session not found\n", __func__);
1024 /* caller must hold client_lock */
1026 unhash_session(struct nfsd4_session *ses)
1028 list_del(&ses->se_hash);
1029 spin_lock(&ses->se_client->cl_lock);
1030 list_del(&ses->se_perclnt);
1031 spin_unlock(&ses->se_client->cl_lock);
1034 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1036 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1038 if (clid->cl_boot == nn->boot_time)
1040 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1041 clid->cl_boot, clid->cl_id, nn->boot_time);
1046 * XXX Should we use a slab cache ?
1047 * This type of memory management is somewhat inefficient, but we use it
1048 * anyway since SETCLIENTID is not a common operation.
1050 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1052 struct nfs4_client *clp;
1054 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1057 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1058 if (clp->cl_name.data == NULL) {
1062 clp->cl_name.len = name.len;
1067 free_client(struct nfs4_client *clp)
1069 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1071 lockdep_assert_held(&nn->client_lock);
1072 while (!list_empty(&clp->cl_sessions)) {
1073 struct nfsd4_session *ses;
1074 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1076 list_del(&ses->se_perclnt);
1077 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1080 free_svc_cred(&clp->cl_cred);
1081 kfree(clp->cl_name.data);
1082 idr_destroy(&clp->cl_stateids);
1086 /* must be called under the client_lock */
1088 unhash_client_locked(struct nfs4_client *clp)
1090 struct nfsd4_session *ses;
1092 list_del(&clp->cl_lru);
1093 spin_lock(&clp->cl_lock);
1094 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1095 list_del_init(&ses->se_hash);
1096 spin_unlock(&clp->cl_lock);
1100 destroy_client(struct nfs4_client *clp)
1102 struct nfs4_openowner *oo;
1103 struct nfs4_delegation *dp;
1104 struct list_head reaplist;
1105 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1107 INIT_LIST_HEAD(&reaplist);
1108 spin_lock(&recall_lock);
1109 while (!list_empty(&clp->cl_delegations)) {
1110 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1111 list_del_init(&dp->dl_perclnt);
1112 list_move(&dp->dl_recall_lru, &reaplist);
1114 spin_unlock(&recall_lock);
1115 while (!list_empty(&reaplist)) {
1116 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1117 unhash_delegation(dp);
1119 while (!list_empty(&clp->cl_openowners)) {
1120 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1121 release_openowner(oo);
1123 nfsd4_shutdown_callback(clp);
1124 if (clp->cl_cb_conn.cb_xprt)
1125 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1126 list_del(&clp->cl_idhash);
1127 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1128 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1130 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1131 spin_lock(&nn->client_lock);
1132 unhash_client_locked(clp);
1133 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1135 spin_unlock(&nn->client_lock);
1138 static void expire_client(struct nfs4_client *clp)
1140 nfsd4_client_record_remove(clp);
1141 destroy_client(clp);
1144 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1146 memcpy(target->cl_verifier.data, source->data,
1147 sizeof(target->cl_verifier.data));
1150 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1152 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1153 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1156 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1158 if (source->cr_principal) {
1159 target->cr_principal =
1160 kstrdup(source->cr_principal, GFP_KERNEL);
1161 if (target->cr_principal == NULL)
1164 target->cr_principal = NULL;
1165 target->cr_flavor = source->cr_flavor;
1166 target->cr_uid = source->cr_uid;
1167 target->cr_gid = source->cr_gid;
1168 target->cr_group_info = source->cr_group_info;
1169 get_group_info(target->cr_group_info);
1174 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1178 res = o1->len - o2->len;
1181 return (long long)memcmp(o1->data, o2->data, o1->len);
1184 static int same_name(const char *n1, const char *n2)
1186 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1190 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1192 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1196 same_clid(clientid_t *cl1, clientid_t *cl2)
1198 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1201 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1205 if (g1->ngroups != g2->ngroups)
1207 for (i=0; i<g1->ngroups; i++)
1208 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1214 * RFC 3530 language requires clid_inuse be returned when the
1215 * "principal" associated with a requests differs from that previously
1216 * used. We use uid, gid's, and gss principal string as our best
1217 * approximation. We also don't want to allow non-gss use of a client
1218 * established using gss: in theory cr_principal should catch that
1219 * change, but in practice cr_principal can be null even in the gss case
1220 * since gssd doesn't always pass down a principal string.
1222 static bool is_gss_cred(struct svc_cred *cr)
1224 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1225 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1230 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1232 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1233 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1234 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1235 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1237 if (cr1->cr_principal == cr2->cr_principal)
1239 if (!cr1->cr_principal || !cr2->cr_principal)
1241 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1244 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1246 static u32 current_clientid = 1;
1248 clp->cl_clientid.cl_boot = nn->boot_time;
1249 clp->cl_clientid.cl_id = current_clientid++;
1252 static void gen_confirm(struct nfs4_client *clp)
1257 verf[0] = (__be32)get_seconds();
1258 verf[1] = (__be32)i++;
1259 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1262 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1264 struct nfs4_stid *ret;
1266 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1267 if (!ret || !ret->sc_type)
1272 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1274 struct nfs4_stid *s;
1276 s = find_stateid(cl, t);
1279 if (typemask & s->sc_type)
1284 static struct nfs4_client *create_client(struct xdr_netobj name,
1285 struct svc_rqst *rqstp, nfs4_verifier *verf)
1287 struct nfs4_client *clp;
1288 struct sockaddr *sa = svc_addr(rqstp);
1290 struct net *net = SVC_NET(rqstp);
1291 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1293 clp = alloc_client(name);
1297 INIT_LIST_HEAD(&clp->cl_sessions);
1298 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1300 spin_lock(&nn->client_lock);
1302 spin_unlock(&nn->client_lock);
1305 idr_init(&clp->cl_stateids);
1306 atomic_set(&clp->cl_refcount, 0);
1307 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1308 INIT_LIST_HEAD(&clp->cl_idhash);
1309 INIT_LIST_HEAD(&clp->cl_openowners);
1310 INIT_LIST_HEAD(&clp->cl_delegations);
1311 INIT_LIST_HEAD(&clp->cl_lru);
1312 INIT_LIST_HEAD(&clp->cl_callbacks);
1313 spin_lock_init(&clp->cl_lock);
1314 nfsd4_init_callback(&clp->cl_cb_null);
1315 clp->cl_time = get_seconds();
1316 clear_bit(0, &clp->cl_cb_slot_busy);
1317 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1318 copy_verf(clp, verf);
1319 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1321 clp->cl_cb_session = NULL;
1327 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1329 struct rb_node **new = &(root->rb_node), *parent = NULL;
1330 struct nfs4_client *clp;
1333 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1336 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1337 new = &((*new)->rb_left);
1339 new = &((*new)->rb_right);
1342 rb_link_node(&new_clp->cl_namenode, parent, new);
1343 rb_insert_color(&new_clp->cl_namenode, root);
1346 static struct nfs4_client *
1347 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1350 struct rb_node *node = root->rb_node;
1351 struct nfs4_client *clp;
1354 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1355 cmp = compare_blob(&clp->cl_name, name);
1357 node = node->rb_left;
1359 node = node->rb_right;
1367 add_to_unconfirmed(struct nfs4_client *clp)
1369 unsigned int idhashval;
1370 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1372 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1373 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1374 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1375 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1380 move_to_confirmed(struct nfs4_client *clp)
1382 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1383 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1385 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1386 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1387 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1388 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1389 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1393 static struct nfs4_client *
1394 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1396 struct nfs4_client *clp;
1397 unsigned int idhashval = clientid_hashval(clid->cl_id);
1399 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1400 if (same_clid(&clp->cl_clientid, clid)) {
1401 if ((bool)clp->cl_minorversion != sessions)
1410 static struct nfs4_client *
1411 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1413 struct list_head *tbl = nn->conf_id_hashtbl;
1415 return find_client_in_id_table(tbl, clid, sessions);
1418 static struct nfs4_client *
1419 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1421 struct list_head *tbl = nn->unconf_id_hashtbl;
1423 return find_client_in_id_table(tbl, clid, sessions);
1426 static bool clp_used_exchangeid(struct nfs4_client *clp)
1428 return clp->cl_exchange_flags != 0;
1431 static struct nfs4_client *
1432 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1434 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1437 static struct nfs4_client *
1438 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1440 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1444 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1446 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1447 struct sockaddr *sa = svc_addr(rqstp);
1448 u32 scopeid = rpc_get_scope_id(sa);
1449 unsigned short expected_family;
1451 /* Currently, we only support tcp and tcp6 for the callback channel */
1452 if (se->se_callback_netid_len == 3 &&
1453 !memcmp(se->se_callback_netid_val, "tcp", 3))
1454 expected_family = AF_INET;
1455 else if (se->se_callback_netid_len == 4 &&
1456 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1457 expected_family = AF_INET6;
1461 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1462 se->se_callback_addr_len,
1463 (struct sockaddr *)&conn->cb_addr,
1464 sizeof(conn->cb_addr));
1466 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1469 if (conn->cb_addr.ss_family == AF_INET6)
1470 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1472 conn->cb_prog = se->se_callback_prog;
1473 conn->cb_ident = se->se_callback_ident;
1474 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1477 conn->cb_addr.ss_family = AF_UNSPEC;
1478 conn->cb_addrlen = 0;
1479 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1480 "will not receive delegations\n",
1481 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1487 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1490 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1492 struct nfsd4_slot *slot = resp->cstate.slot;
1495 dprintk("--> %s slot %p\n", __func__, slot);
1497 slot->sl_opcnt = resp->opcnt;
1498 slot->sl_status = resp->cstate.status;
1500 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1501 if (nfsd4_not_cached(resp)) {
1502 slot->sl_datalen = 0;
1505 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1506 base = (char *)resp->cstate.datap -
1507 (char *)resp->xbuf->head[0].iov_base;
1508 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1510 WARN("%s: sessions DRC could not cache compound\n", __func__);
1515 * Encode the replay sequence operation from the slot values.
1516 * If cachethis is FALSE encode the uncached rep error on the next
1517 * operation which sets resp->p and increments resp->opcnt for
1518 * nfs4svc_encode_compoundres.
1522 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1523 struct nfsd4_compoundres *resp)
1525 struct nfsd4_op *op;
1526 struct nfsd4_slot *slot = resp->cstate.slot;
1528 /* Encode the replayed sequence operation */
1529 op = &args->ops[resp->opcnt - 1];
1530 nfsd4_encode_operation(resp, op);
1532 /* Return nfserr_retry_uncached_rep in next operation. */
1533 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1534 op = &args->ops[resp->opcnt++];
1535 op->status = nfserr_retry_uncached_rep;
1536 nfsd4_encode_operation(resp, op);
1542 * The sequence operation is not cached because we can use the slot and
1546 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1547 struct nfsd4_sequence *seq)
1549 struct nfsd4_slot *slot = resp->cstate.slot;
1552 dprintk("--> %s slot %p\n", __func__, slot);
1554 /* Either returns 0 or nfserr_retry_uncached */
1555 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1556 if (status == nfserr_retry_uncached_rep)
1559 /* The sequence operation has been encoded, cstate->datap set. */
1560 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1562 resp->opcnt = slot->sl_opcnt;
1563 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1564 status = slot->sl_status;
1570 * Set the exchange_id flags returned by the server.
1573 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1575 /* pNFS is not supported */
1576 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1578 /* Referrals are supported, Migration is not. */
1579 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1581 /* set the wire flags to return to client. */
1582 clid->flags = new->cl_exchange_flags;
1585 static bool client_has_state(struct nfs4_client *clp)
1588 * Note clp->cl_openowners check isn't quite right: there's no
1589 * need to count owners without stateid's.
1591 * Also note we should probably be using this in 4.0 case too.
1593 return !list_empty(&clp->cl_openowners)
1594 || !list_empty(&clp->cl_delegations)
1595 || !list_empty(&clp->cl_sessions);
1599 nfsd4_exchange_id(struct svc_rqst *rqstp,
1600 struct nfsd4_compound_state *cstate,
1601 struct nfsd4_exchange_id *exid)
1603 struct nfs4_client *unconf, *conf, *new;
1605 char addr_str[INET6_ADDRSTRLEN];
1606 nfs4_verifier verf = exid->verifier;
1607 struct sockaddr *sa = svc_addr(rqstp);
1608 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1609 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1611 rpc_ntop(sa, addr_str, sizeof(addr_str));
1612 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1613 "ip_addr=%s flags %x, spa_how %d\n",
1614 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1615 addr_str, exid->flags, exid->spa_how);
1617 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1618 return nfserr_inval;
1620 /* Currently only support SP4_NONE */
1621 switch (exid->spa_how) {
1624 default: /* checked by xdr code */
1628 return nfserr_serverfault; /* no excuse :-/ */
1631 /* Cases below refer to rfc 5661 section 18.35.4: */
1633 conf = find_confirmed_client_by_name(&exid->clname, nn);
1635 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1636 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1639 if (!clp_used_exchangeid(conf)) { /* buggy client */
1640 status = nfserr_inval;
1643 if (!creds_match) { /* case 9 */
1644 status = nfserr_perm;
1647 if (!verfs_match) { /* case 8 */
1648 status = nfserr_not_same;
1652 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1656 if (!creds_match) { /* case 3 */
1657 if (client_has_state(conf)) {
1658 status = nfserr_clid_inuse;
1661 expire_client(conf);
1664 if (verfs_match) { /* case 2 */
1665 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1669 /* case 5, client reboot */
1673 if (update) { /* case 7 */
1674 status = nfserr_noent;
1678 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1679 if (unconf) /* case 4, possible retry or client restart */
1680 expire_client(unconf);
1682 /* case 1 (normal case) */
1684 new = create_client(exid->clname, rqstp, &verf);
1686 status = nfserr_jukebox;
1689 new->cl_minorversion = 1;
1692 add_to_unconfirmed(new);
1694 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1695 exid->clientid.cl_id = new->cl_clientid.cl_id;
1697 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1698 nfsd4_set_ex_flags(new, exid);
1700 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1701 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1705 nfs4_unlock_state();
1710 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1712 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1715 /* The slot is in use, and no response has been sent. */
1717 if (seqid == slot_seqid)
1718 return nfserr_jukebox;
1720 return nfserr_seq_misordered;
1722 /* Note unsigned 32-bit arithmetic handles wraparound: */
1723 if (likely(seqid == slot_seqid + 1))
1725 if (seqid == slot_seqid)
1726 return nfserr_replay_cache;
1727 return nfserr_seq_misordered;
1731 * Cache the create session result into the create session single DRC
1732 * slot cache by saving the xdr structure. sl_seqid has been set.
1733 * Do this for solo or embedded create session operations.
1736 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1737 struct nfsd4_clid_slot *slot, __be32 nfserr)
1739 slot->sl_status = nfserr;
1740 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1744 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1745 struct nfsd4_clid_slot *slot)
1747 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1748 return slot->sl_status;
1751 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1752 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1753 1 + /* MIN tag is length with zero, only length */ \
1754 3 + /* version, opcount, opcode */ \
1755 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1756 /* seqid, slotID, slotID, cache */ \
1757 4 ) * sizeof(__be32))
1759 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1760 2 + /* verifier: AUTH_NULL, length 0 */\
1762 1 + /* MIN tag is length with zero, only length */ \
1763 3 + /* opcount, opcode, opstatus*/ \
1764 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1765 /* seqid, slotID, slotID, slotID, status */ \
1766 5 ) * sizeof(__be32))
1768 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1770 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
1772 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
1773 return nfserr_toosmall;
1774 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
1775 return nfserr_toosmall;
1776 ca->headerpadsz = 0;
1777 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
1778 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
1779 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
1780 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
1781 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
1782 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
1784 * Note decreasing slot size below client's request may make it
1785 * difficult for client to function correctly, whereas
1786 * decreasing the number of slots will (just?) affect
1787 * performance. When short on memory we therefore prefer to
1788 * decrease number of slots instead of their size. Clients that
1789 * request larger slots than they need will get poor results:
1791 ca->maxreqs = nfsd4_get_drc_mem(ca);
1793 return nfserr_jukebox;
1798 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
1800 ca->headerpadsz = 0;
1803 * These RPC_MAX_HEADER macros are overkill, especially since we
1804 * don't even do gss on the backchannel yet. But this is still
1805 * less than 1k. Tighten up this estimate in the unlikely event
1806 * it turns out to be a problem for some client:
1808 if (ca->maxreq_sz < NFS4_enc_cb_recall_sz + RPC_MAX_HEADER_WITH_AUTH)
1809 return nfserr_toosmall;
1810 if (ca->maxresp_sz < NFS4_dec_cb_recall_sz + RPC_MAX_REPHEADER_WITH_AUTH)
1811 return nfserr_toosmall;
1812 ca->maxresp_cached = 0;
1814 return nfserr_toosmall;
1820 nfsd4_create_session(struct svc_rqst *rqstp,
1821 struct nfsd4_compound_state *cstate,
1822 struct nfsd4_create_session *cr_ses)
1824 struct sockaddr *sa = svc_addr(rqstp);
1825 struct nfs4_client *conf, *unconf;
1826 struct nfsd4_session *new;
1827 struct nfsd4_conn *conn;
1828 struct nfsd4_clid_slot *cs_slot = NULL;
1830 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1832 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1833 return nfserr_inval;
1834 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
1837 status = check_backchannel_attrs(&cr_ses->back_channel);
1840 status = nfserr_jukebox;
1841 new = alloc_session(&cr_ses->fore_channel);
1843 goto out_release_drc_mem;
1844 conn = alloc_conn_from_crses(rqstp, cr_ses);
1846 goto out_free_session;
1849 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1850 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1851 WARN_ON_ONCE(conf && unconf);
1854 cs_slot = &conf->cl_cs_slot;
1855 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1856 if (status == nfserr_replay_cache) {
1857 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1859 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1860 status = nfserr_seq_misordered;
1863 } else if (unconf) {
1864 struct nfs4_client *old;
1865 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1866 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1867 status = nfserr_clid_inuse;
1870 cs_slot = &unconf->cl_cs_slot;
1871 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1873 /* an unconfirmed replay returns misordered */
1874 status = nfserr_seq_misordered;
1877 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1879 status = mark_client_expired(old);
1884 move_to_confirmed(unconf);
1887 status = nfserr_stale_clientid;
1892 * We do not support RDMA or persistent sessions
1894 cr_ses->flags &= ~SESSION4_PERSIST;
1895 cr_ses->flags &= ~SESSION4_RDMA;
1897 init_session(rqstp, new, conf, cr_ses);
1898 nfsd4_init_conn(rqstp, conn, new);
1900 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1901 NFS4_MAX_SESSIONID_LEN);
1902 cs_slot->sl_seqid++;
1903 cr_ses->seqid = cs_slot->sl_seqid;
1905 /* cache solo and embedded create sessions under the state lock */
1906 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1907 nfs4_unlock_state();
1910 nfs4_unlock_state();
1913 __free_session(new);
1914 out_release_drc_mem:
1915 nfsd4_put_drc_mem(&cr_ses->fore_channel);
1919 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1922 case NFS4_CDFC4_FORE:
1923 case NFS4_CDFC4_BACK:
1925 case NFS4_CDFC4_FORE_OR_BOTH:
1926 case NFS4_CDFC4_BACK_OR_BOTH:
1927 *dir = NFS4_CDFC4_BOTH;
1930 return nfserr_inval;
1933 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1935 struct nfsd4_session *session = cstate->session;
1936 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1938 spin_lock(&nn->client_lock);
1939 session->se_cb_prog = bc->bc_cb_program;
1940 session->se_cb_sec = bc->bc_cb_sec;
1941 spin_unlock(&nn->client_lock);
1943 nfsd4_probe_callback(session->se_client);
1948 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1949 struct nfsd4_compound_state *cstate,
1950 struct nfsd4_bind_conn_to_session *bcts)
1953 struct nfsd4_conn *conn;
1954 struct nfsd4_session *session;
1955 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1957 if (!nfsd4_last_compound_op(rqstp))
1958 return nfserr_not_only_op;
1960 spin_lock(&nn->client_lock);
1961 session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1962 spin_unlock(&nn->client_lock);
1963 status = nfserr_badsession;
1966 status = nfsd4_map_bcts_dir(&bcts->dir);
1969 conn = alloc_conn(rqstp, bcts->dir);
1970 status = nfserr_jukebox;
1973 nfsd4_init_conn(rqstp, conn, session);
1976 nfs4_unlock_state();
1980 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1984 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1988 nfsd4_destroy_session(struct svc_rqst *r,
1989 struct nfsd4_compound_state *cstate,
1990 struct nfsd4_destroy_session *sessionid)
1992 struct nfsd4_session *ses;
1994 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
1997 status = nfserr_not_only_op;
1998 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1999 if (!nfsd4_last_compound_op(r))
2002 dump_sessionid(__func__, &sessionid->sessionid);
2003 spin_lock(&nn->client_lock);
2004 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
2005 status = nfserr_badsession;
2007 goto out_client_lock;
2008 status = mark_session_dead_locked(ses);
2010 goto out_client_lock;
2011 unhash_session(ses);
2012 spin_unlock(&nn->client_lock);
2014 nfsd4_probe_callback_sync(ses->se_client);
2016 spin_lock(&nn->client_lock);
2020 spin_unlock(&nn->client_lock);
2022 nfs4_unlock_state();
2026 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2028 struct nfsd4_conn *c;
2030 list_for_each_entry(c, &s->se_conns, cn_persession) {
2031 if (c->cn_xprt == xpt) {
2038 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2040 struct nfs4_client *clp = ses->se_client;
2041 struct nfsd4_conn *c;
2044 spin_lock(&clp->cl_lock);
2045 c = __nfsd4_find_conn(new->cn_xprt, ses);
2047 spin_unlock(&clp->cl_lock);
2051 __nfsd4_hash_conn(new, ses);
2052 spin_unlock(&clp->cl_lock);
2053 ret = nfsd4_register_conn(new);
2055 /* oops; xprt is already down: */
2056 nfsd4_conn_lost(&new->cn_xpt_user);
2060 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2062 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2064 return args->opcnt > session->se_fchannel.maxops;
2067 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2068 struct nfsd4_session *session)
2070 struct xdr_buf *xb = &rqstp->rq_arg;
2072 return xb->len > session->se_fchannel.maxreq_sz;
2076 nfsd4_sequence(struct svc_rqst *rqstp,
2077 struct nfsd4_compound_state *cstate,
2078 struct nfsd4_sequence *seq)
2080 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2081 struct nfsd4_session *session;
2082 struct nfs4_client *clp;
2083 struct nfsd4_slot *slot;
2084 struct nfsd4_conn *conn;
2086 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2088 if (resp->opcnt != 1)
2089 return nfserr_sequence_pos;
2092 * Will be either used or freed by nfsd4_sequence_check_conn
2095 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2097 return nfserr_jukebox;
2099 spin_lock(&nn->client_lock);
2100 status = nfserr_badsession;
2101 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2103 goto out_no_session;
2104 clp = session->se_client;
2105 status = get_client_locked(clp);
2107 goto out_no_session;
2108 status = nfsd4_get_session_locked(session);
2110 goto out_put_client;
2112 status = nfserr_too_many_ops;
2113 if (nfsd4_session_too_many_ops(rqstp, session))
2114 goto out_put_session;
2116 status = nfserr_req_too_big;
2117 if (nfsd4_request_too_big(rqstp, session))
2118 goto out_put_session;
2120 status = nfserr_badslot;
2121 if (seq->slotid >= session->se_fchannel.maxreqs)
2122 goto out_put_session;
2124 slot = session->se_slots[seq->slotid];
2125 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2127 /* We do not negotiate the number of slots yet, so set the
2128 * maxslots to the session maxreqs which is used to encode
2129 * sr_highest_slotid and the sr_target_slot id to maxslots */
2130 seq->maxslots = session->se_fchannel.maxreqs;
2132 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2133 slot->sl_flags & NFSD4_SLOT_INUSE);
2134 if (status == nfserr_replay_cache) {
2135 status = nfserr_seq_misordered;
2136 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2137 goto out_put_session;
2138 cstate->slot = slot;
2139 cstate->session = session;
2140 /* Return the cached reply status and set cstate->status
2141 * for nfsd4_proc_compound processing */
2142 status = nfsd4_replay_cache_entry(resp, seq);
2143 cstate->status = nfserr_replay_cache;
2147 goto out_put_session;
2149 nfsd4_sequence_check_conn(conn, session);
2152 /* Success! bump slot seqid */
2153 slot->sl_seqid = seq->seqid;
2154 slot->sl_flags |= NFSD4_SLOT_INUSE;
2156 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2158 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2160 cstate->slot = slot;
2161 cstate->session = session;
2164 switch (clp->cl_cb_state) {
2166 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2168 case NFSD4_CB_FAULT:
2169 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2172 seq->status_flags = 0;
2176 spin_unlock(&nn->client_lock);
2179 nfsd4_put_session(session);
2181 put_client_renew_locked(clp);
2182 goto out_no_session;
2186 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2188 struct nfs4_client *conf, *unconf, *clp;
2190 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2193 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2194 conf = find_confirmed_client(&dc->clientid, true, nn);
2195 WARN_ON_ONCE(conf && unconf);
2200 if (client_has_state(conf)) {
2201 status = nfserr_clientid_busy;
2207 status = nfserr_stale_clientid;
2213 nfs4_unlock_state();
2218 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2222 if (rc->rca_one_fs) {
2223 if (!cstate->current_fh.fh_dentry)
2224 return nfserr_nofilehandle;
2226 * We don't take advantage of the rca_one_fs case.
2227 * That's OK, it's optional, we can safely ignore it.
2233 status = nfserr_complete_already;
2234 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2235 &cstate->session->se_client->cl_flags))
2238 status = nfserr_stale_clientid;
2239 if (is_client_expired(cstate->session->se_client))
2241 * The following error isn't really legal.
2242 * But we only get here if the client just explicitly
2243 * destroyed the client. Surely it no longer cares what
2244 * error it gets back on an operation for the dead
2250 nfsd4_client_record_create(cstate->session->se_client);
2252 nfs4_unlock_state();
2257 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2258 struct nfsd4_setclientid *setclid)
2260 struct xdr_netobj clname = setclid->se_name;
2261 nfs4_verifier clverifier = setclid->se_verf;
2262 struct nfs4_client *conf, *unconf, *new;
2264 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2266 /* Cases below refer to rfc 3530 section 14.2.33: */
2268 conf = find_confirmed_client_by_name(&clname, nn);
2271 status = nfserr_clid_inuse;
2272 if (clp_used_exchangeid(conf))
2274 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2275 char addr_str[INET6_ADDRSTRLEN];
2276 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2278 dprintk("NFSD: setclientid: string in use by client "
2279 "at %s\n", addr_str);
2283 unconf = find_unconfirmed_client_by_name(&clname, nn);
2285 expire_client(unconf);
2286 status = nfserr_jukebox;
2287 new = create_client(clname, rqstp, &clverifier);
2290 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2291 /* case 1: probable callback update */
2292 copy_clid(new, conf);
2293 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2295 new->cl_minorversion = 0;
2296 gen_callback(new, setclid, rqstp);
2297 add_to_unconfirmed(new);
2298 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2299 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2300 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2303 nfs4_unlock_state();
2309 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2310 struct nfsd4_compound_state *cstate,
2311 struct nfsd4_setclientid_confirm *setclientid_confirm)
2313 struct nfs4_client *conf, *unconf;
2314 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2315 clientid_t * clid = &setclientid_confirm->sc_clientid;
2317 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2319 if (STALE_CLIENTID(clid, nn))
2320 return nfserr_stale_clientid;
2323 conf = find_confirmed_client(clid, false, nn);
2324 unconf = find_unconfirmed_client(clid, false, nn);
2326 * We try hard to give out unique clientid's, so if we get an
2327 * attempt to confirm the same clientid with a different cred,
2328 * there's a bug somewhere. Let's charitably assume it's our
2331 status = nfserr_serverfault;
2332 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2334 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2336 /* cases below refer to rfc 3530 section 14.2.34: */
2337 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2338 if (conf && !unconf) /* case 2: probable retransmit */
2340 else /* case 4: client hasn't noticed we rebooted yet? */
2341 status = nfserr_stale_clientid;
2345 if (conf) { /* case 1: callback update */
2346 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2347 nfsd4_probe_callback(conf);
2348 expire_client(unconf);
2349 } else { /* case 3: normal case; new or rebooted client */
2350 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2352 status = mark_client_expired(conf);
2355 expire_client(conf);
2357 move_to_confirmed(unconf);
2358 nfsd4_probe_callback(unconf);
2361 nfs4_unlock_state();
2365 static struct nfs4_file *nfsd4_alloc_file(void)
2367 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2370 /* OPEN Share state helper functions */
2371 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2373 unsigned int hashval = file_hashval(ino);
2375 atomic_set(&fp->fi_ref, 1);
2376 INIT_LIST_HEAD(&fp->fi_stateids);
2377 INIT_LIST_HEAD(&fp->fi_delegations);
2378 fp->fi_inode = igrab(ino);
2379 fp->fi_had_conflict = false;
2380 fp->fi_lease = NULL;
2381 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2382 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2383 spin_lock(&recall_lock);
2384 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2385 spin_unlock(&recall_lock);
2389 nfsd4_free_slab(struct kmem_cache **slab)
2393 kmem_cache_destroy(*slab);
2398 nfsd4_free_slabs(void)
2400 nfsd4_free_slab(&openowner_slab);
2401 nfsd4_free_slab(&lockowner_slab);
2402 nfsd4_free_slab(&file_slab);
2403 nfsd4_free_slab(&stateid_slab);
2404 nfsd4_free_slab(&deleg_slab);
2408 nfsd4_init_slabs(void)
2410 openowner_slab = kmem_cache_create("nfsd4_openowners",
2411 sizeof(struct nfs4_openowner), 0, 0, NULL);
2412 if (openowner_slab == NULL)
2414 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2415 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2416 if (lockowner_slab == NULL)
2418 file_slab = kmem_cache_create("nfsd4_files",
2419 sizeof(struct nfs4_file), 0, 0, NULL);
2420 if (file_slab == NULL)
2422 stateid_slab = kmem_cache_create("nfsd4_stateids",
2423 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2424 if (stateid_slab == NULL)
2426 deleg_slab = kmem_cache_create("nfsd4_delegations",
2427 sizeof(struct nfs4_delegation), 0, 0, NULL);
2428 if (deleg_slab == NULL)
2433 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2437 void nfs4_free_openowner(struct nfs4_openowner *oo)
2439 kfree(oo->oo_owner.so_owner.data);
2440 kmem_cache_free(openowner_slab, oo);
2443 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2445 kfree(lo->lo_owner.so_owner.data);
2446 kmem_cache_free(lockowner_slab, lo);
2449 static void init_nfs4_replay(struct nfs4_replay *rp)
2451 rp->rp_status = nfserr_serverfault;
2453 rp->rp_buf = rp->rp_ibuf;
2456 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2458 struct nfs4_stateowner *sop;
2460 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2464 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2465 if (!sop->so_owner.data) {
2466 kmem_cache_free(slab, sop);
2469 sop->so_owner.len = owner->len;
2471 INIT_LIST_HEAD(&sop->so_stateids);
2472 sop->so_client = clp;
2473 init_nfs4_replay(&sop->so_replay);
2477 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2479 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2481 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2482 list_add(&oo->oo_perclient, &clp->cl_openowners);
2485 static struct nfs4_openowner *
2486 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2487 struct nfs4_openowner *oo;
2489 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2492 oo->oo_owner.so_is_open_owner = 1;
2493 oo->oo_owner.so_seqid = open->op_seqid;
2494 oo->oo_flags = NFS4_OO_NEW;
2496 oo->oo_last_closed_stid = NULL;
2497 INIT_LIST_HEAD(&oo->oo_close_lru);
2498 hash_openowner(oo, clp, strhashval);
2502 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2503 struct nfs4_openowner *oo = open->op_openowner;
2505 stp->st_stid.sc_type = NFS4_OPEN_STID;
2506 INIT_LIST_HEAD(&stp->st_lockowners);
2507 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2508 list_add(&stp->st_perfile, &fp->fi_stateids);
2509 stp->st_stateowner = &oo->oo_owner;
2512 stp->st_access_bmap = 0;
2513 stp->st_deny_bmap = 0;
2514 set_access(open->op_share_access, stp);
2515 set_deny(open->op_share_deny, stp);
2516 stp->st_openstp = NULL;
2520 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2522 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2524 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2526 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2527 oo->oo_time = get_seconds();
2531 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2534 return (sop->so_owner.len == owner->len) &&
2535 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2536 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2539 static struct nfs4_openowner *
2540 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2541 bool sessions, struct nfsd_net *nn)
2543 struct nfs4_stateowner *so;
2544 struct nfs4_openowner *oo;
2545 struct nfs4_client *clp;
2547 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2548 if (!so->so_is_open_owner)
2550 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2552 clp = oo->oo_owner.so_client;
2553 if ((bool)clp->cl_minorversion != sessions)
2555 renew_client(oo->oo_owner.so_client);
2562 /* search file_hashtbl[] for file */
2563 static struct nfs4_file *
2564 find_file(struct inode *ino)
2566 unsigned int hashval = file_hashval(ino);
2567 struct nfs4_file *fp;
2569 spin_lock(&recall_lock);
2570 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2571 if (fp->fi_inode == ino) {
2573 spin_unlock(&recall_lock);
2577 spin_unlock(&recall_lock);
2582 * Called to check deny when READ with all zero stateid or
2583 * WRITE with all zero or all one stateid
2586 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2588 struct inode *ino = current_fh->fh_dentry->d_inode;
2589 struct nfs4_file *fp;
2590 struct nfs4_ol_stateid *stp;
2593 fp = find_file(ino);
2596 ret = nfserr_locked;
2597 /* Search for conflicting share reservations */
2598 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2599 if (test_deny(deny_type, stp) ||
2600 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2609 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2611 struct nfs4_client *clp = dp->dl_stid.sc_client;
2612 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2614 /* We're assuming the state code never drops its reference
2615 * without first removing the lease. Since we're in this lease
2616 * callback (and since the lease code is serialized by the kernel
2617 * lock) we know the server hasn't removed the lease yet, we know
2618 * it's safe to take a reference: */
2619 atomic_inc(&dp->dl_count);
2621 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
2623 /* only place dl_time is set. protected by lock_flocks*/
2624 dp->dl_time = get_seconds();
2626 nfsd4_cb_recall(dp);
2629 /* Called from break_lease() with lock_flocks() held. */
2630 static void nfsd_break_deleg_cb(struct file_lock *fl)
2632 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2633 struct nfs4_delegation *dp;
2636 WARN(1, "(%p)->fl_owner NULL\n", fl);
2639 if (fp->fi_had_conflict) {
2640 WARN(1, "duplicate break on %p\n", fp);
2644 * We don't want the locks code to timeout the lease for us;
2645 * we'll remove it ourself if a delegation isn't returned
2648 fl->fl_break_time = 0;
2650 spin_lock(&recall_lock);
2651 fp->fi_had_conflict = true;
2652 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2653 nfsd_break_one_deleg(dp);
2654 spin_unlock(&recall_lock);
2658 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2661 return lease_modify(onlist, arg);
2666 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2667 .lm_break = nfsd_break_deleg_cb,
2668 .lm_change = nfsd_change_deleg_cb,
2671 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2673 if (nfsd4_has_session(cstate))
2675 if (seqid == so->so_seqid - 1)
2676 return nfserr_replay_me;
2677 if (seqid == so->so_seqid)
2679 return nfserr_bad_seqid;
2683 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2684 struct nfsd4_open *open, struct nfsd_net *nn)
2686 clientid_t *clientid = &open->op_clientid;
2687 struct nfs4_client *clp = NULL;
2688 unsigned int strhashval;
2689 struct nfs4_openowner *oo = NULL;
2692 if (STALE_CLIENTID(&open->op_clientid, nn))
2693 return nfserr_stale_clientid;
2695 * In case we need it later, after we've already created the
2696 * file and don't want to risk a further failure:
2698 open->op_file = nfsd4_alloc_file();
2699 if (open->op_file == NULL)
2700 return nfserr_jukebox;
2702 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2703 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2704 open->op_openowner = oo;
2706 clp = find_confirmed_client(clientid, cstate->minorversion,
2709 return nfserr_expired;
2712 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2713 /* Replace unconfirmed owners without checking for replay. */
2714 clp = oo->oo_owner.so_client;
2715 release_openowner(oo);
2716 open->op_openowner = NULL;
2719 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2722 clp = oo->oo_owner.so_client;
2725 oo = alloc_init_open_stateowner(strhashval, clp, open);
2727 return nfserr_jukebox;
2728 open->op_openowner = oo;
2730 open->op_stp = nfs4_alloc_stateid(clp);
2732 return nfserr_jukebox;
2736 static inline __be32
2737 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2739 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2740 return nfserr_openmode;
2745 static int share_access_to_flags(u32 share_access)
2747 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2750 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2752 struct nfs4_stid *ret;
2754 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2757 return delegstateid(ret);
2760 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2762 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2763 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2767 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
2768 struct nfs4_delegation **dp)
2771 __be32 status = nfserr_bad_stateid;
2773 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2776 flags = share_access_to_flags(open->op_share_access);
2777 status = nfs4_check_delegmode(*dp, flags);
2781 if (!nfsd4_is_deleg_cur(open))
2785 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2790 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2792 struct nfs4_ol_stateid *local;
2793 struct nfs4_openowner *oo = open->op_openowner;
2795 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2796 /* ignore lock owners */
2797 if (local->st_stateowner->so_is_open_owner == 0)
2799 /* remember if we have seen this open owner */
2800 if (local->st_stateowner == &oo->oo_owner)
2802 /* check for conflicting share reservations */
2803 if (!test_share(local, open))
2804 return nfserr_share_denied;
2809 static inline int nfs4_access_to_access(u32 nfs4_access)
2813 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2814 flags |= NFSD_MAY_READ;
2815 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2816 flags |= NFSD_MAY_WRITE;
2820 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2821 struct svc_fh *cur_fh, struct nfsd4_open *open)
2824 int oflag = nfs4_access_to_omode(open->op_share_access);
2825 int access = nfs4_access_to_access(open->op_share_access);
2827 if (!fp->fi_fds[oflag]) {
2828 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2829 &fp->fi_fds[oflag]);
2833 nfs4_file_get_access(fp, oflag);
2838 static inline __be32
2839 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2840 struct nfsd4_open *open)
2842 struct iattr iattr = {
2843 .ia_valid = ATTR_SIZE,
2846 if (!open->op_truncate)
2848 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2849 return nfserr_inval;
2850 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2854 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)
2856 u32 op_share_access = open->op_share_access;
2860 new_access = !test_access(op_share_access, stp);
2862 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2866 status = nfsd4_truncate(rqstp, cur_fh, open);
2869 int oflag = nfs4_access_to_omode(op_share_access);
2870 nfs4_file_put_access(fp, oflag);
2874 /* remember the open */
2875 set_access(op_share_access, stp);
2876 set_deny(open->op_share_deny, stp);
2883 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2885 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2888 /* Should we give out recallable state?: */
2889 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2891 if (clp->cl_cb_state == NFSD4_CB_UP)
2894 * In the sessions case, since we don't have to establish a
2895 * separate connection for callbacks, we assume it's OK
2896 * until we hear otherwise:
2898 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2901 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2903 struct file_lock *fl;
2905 fl = locks_alloc_lock();
2908 locks_init_lock(fl);
2909 fl->fl_lmops = &nfsd_lease_mng_ops;
2910 fl->fl_flags = FL_LEASE;
2911 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2912 fl->fl_end = OFFSET_MAX;
2913 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2914 fl->fl_pid = current->tgid;
2918 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2920 struct nfs4_file *fp = dp->dl_file;
2921 struct file_lock *fl;
2924 fl = nfs4_alloc_init_lease(dp, flag);
2927 fl->fl_file = find_readable_file(fp);
2928 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2929 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2931 list_del_init(&dp->dl_perclnt);
2932 locks_free_lock(fl);
2936 fp->fi_deleg_file = get_file(fl->fl_file);
2937 atomic_set(&fp->fi_delegees, 1);
2938 list_add(&dp->dl_perfile, &fp->fi_delegations);
2942 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2944 struct nfs4_file *fp = dp->dl_file;
2947 return nfs4_setlease(dp, flag);
2948 spin_lock(&recall_lock);
2949 if (fp->fi_had_conflict) {
2950 spin_unlock(&recall_lock);
2953 atomic_inc(&fp->fi_delegees);
2954 list_add(&dp->dl_perfile, &fp->fi_delegations);
2955 spin_unlock(&recall_lock);
2956 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2960 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2962 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2963 if (status == -EAGAIN)
2964 open->op_why_no_deleg = WND4_CONTENTION;
2966 open->op_why_no_deleg = WND4_RESOURCE;
2967 switch (open->op_deleg_want) {
2968 case NFS4_SHARE_WANT_READ_DELEG:
2969 case NFS4_SHARE_WANT_WRITE_DELEG:
2970 case NFS4_SHARE_WANT_ANY_DELEG:
2972 case NFS4_SHARE_WANT_CANCEL:
2973 open->op_why_no_deleg = WND4_CANCELLED;
2975 case NFS4_SHARE_WANT_NO_DELEG:
2982 * Attempt to hand out a delegation.
2985 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2986 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2988 struct nfs4_delegation *dp;
2989 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2991 int status = 0, flag = 0;
2993 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2994 flag = NFS4_OPEN_DELEGATE_NONE;
2995 open->op_recall = 0;
2996 switch (open->op_claim_type) {
2997 case NFS4_OPEN_CLAIM_PREVIOUS:
2999 open->op_recall = 1;
3000 flag = open->op_delegate_type;
3001 if (flag == NFS4_OPEN_DELEGATE_NONE)
3004 case NFS4_OPEN_CLAIM_NULL:
3005 /* Let's not give out any delegations till everyone's
3006 * had the chance to reclaim theirs.... */
3007 if (locks_in_grace(net))
3009 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3011 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3012 flag = NFS4_OPEN_DELEGATE_WRITE;
3014 flag = NFS4_OPEN_DELEGATE_READ;
3020 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
3023 status = nfs4_set_delegation(dp, flag);
3027 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3029 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3030 STATEID_VAL(&dp->dl_stid.sc_stateid));
3032 open->op_delegate_type = flag;
3033 if (flag == NFS4_OPEN_DELEGATE_NONE) {
3034 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3035 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
3036 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3038 /* 4.1 client asking for a delegation? */
3039 if (open->op_deleg_want)
3040 nfsd4_open_deleg_none_ext(open, status);
3044 unhash_stid(&dp->dl_stid);
3045 nfs4_put_delegation(dp);
3047 flag = NFS4_OPEN_DELEGATE_NONE;
3051 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3052 struct nfs4_delegation *dp)
3054 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3055 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3056 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3057 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3058 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3059 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3060 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3061 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3063 /* Otherwise the client must be confused wanting a delegation
3064 * it already has, therefore we don't return
3065 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3070 * called with nfs4_lock_state() held.
3073 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3075 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3076 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3077 struct nfs4_file *fp = NULL;
3078 struct inode *ino = current_fh->fh_dentry->d_inode;
3079 struct nfs4_ol_stateid *stp = NULL;
3080 struct nfs4_delegation *dp = NULL;
3084 * Lookup file; if found, lookup stateid and check open request,
3085 * and check for delegations in the process of being recalled.
3086 * If not found, create the nfs4_file struct
3088 fp = find_file(ino);
3090 if ((status = nfs4_check_open(fp, open, &stp)))
3092 status = nfs4_check_deleg(cl, open, &dp);
3096 status = nfserr_bad_stateid;
3097 if (nfsd4_is_deleg_cur(open))
3099 status = nfserr_jukebox;
3101 open->op_file = NULL;
3102 nfsd4_init_file(fp, ino);
3106 * OPEN the file, or upgrade an existing OPEN.
3107 * If truncate fails, the OPEN fails.
3110 /* Stateid was found, this is an OPEN upgrade */
3111 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3115 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3118 status = nfsd4_truncate(rqstp, current_fh, open);
3122 open->op_stp = NULL;
3123 init_open_stateid(stp, fp, open);
3125 update_stateid(&stp->st_stid.sc_stateid);
3126 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3128 if (nfsd4_has_session(&resp->cstate)) {
3129 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3131 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3132 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3133 open->op_why_no_deleg = WND4_NOT_WANTED;
3139 * Attempt to hand out a delegation. No error return, because the
3140 * OPEN succeeds even if we fail.
3142 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3146 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3147 STATEID_VAL(&stp->st_stid.sc_stateid));
3149 /* 4.1 client trying to upgrade/downgrade delegation? */
3150 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3151 open->op_deleg_want)
3152 nfsd4_deleg_xgrade_none_ext(open, dp);
3156 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3157 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3159 * To finish the open response, we just need to set the rflags.
3161 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3162 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3163 !nfsd4_has_session(&resp->cstate))
3164 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3169 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3171 if (open->op_openowner) {
3172 struct nfs4_openowner *oo = open->op_openowner;
3174 if (!list_empty(&oo->oo_owner.so_stateids))
3175 list_del_init(&oo->oo_close_lru);
3176 if (oo->oo_flags & NFS4_OO_NEW) {
3178 release_openowner(oo);
3179 open->op_openowner = NULL;
3181 oo->oo_flags &= ~NFS4_OO_NEW;
3185 nfsd4_free_file(open->op_file);
3187 free_generic_stateid(open->op_stp);
3190 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3192 struct nfs4_client *found;
3194 if (STALE_CLIENTID(clid, nn))
3195 return nfserr_stale_clientid;
3196 found = find_confirmed_client(clid, session, nn);
3199 return found ? nfs_ok : nfserr_expired;
3203 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3206 struct nfs4_client *clp;
3208 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3211 dprintk("process_renew(%08x/%08x): starting\n",
3212 clid->cl_boot, clid->cl_id);
3213 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3216 status = nfserr_cb_path_down;
3217 if (!list_empty(&clp->cl_delegations)
3218 && clp->cl_cb_state != NFSD4_CB_UP)
3222 nfs4_unlock_state();
3227 nfsd4_end_grace(struct nfsd_net *nn)
3229 /* do nothing if grace period already ended */
3230 if (nn->grace_ended)
3233 dprintk("NFSD: end of grace period\n");
3234 nn->grace_ended = true;
3235 nfsd4_record_grace_done(nn, nn->boot_time);
3236 locks_end_grace(&nn->nfsd4_manager);
3238 * Now that every NFSv4 client has had the chance to recover and
3239 * to see the (possibly new, possibly shorter) lease time, we
3240 * can safely set the next grace time to the current lease time:
3242 nn->nfsd4_grace = nn->nfsd4_lease;
3246 nfs4_laundromat(struct nfsd_net *nn)
3248 struct nfs4_client *clp;
3249 struct nfs4_openowner *oo;
3250 struct nfs4_delegation *dp;
3251 struct list_head *pos, *next, reaplist;
3252 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3253 time_t t, clientid_val = nn->nfsd4_lease;
3254 time_t u, test_val = nn->nfsd4_lease;
3258 dprintk("NFSD: laundromat service - starting\n");
3259 nfsd4_end_grace(nn);
3260 INIT_LIST_HEAD(&reaplist);
3261 spin_lock(&nn->client_lock);
3262 list_for_each_safe(pos, next, &nn->client_lru) {
3263 clp = list_entry(pos, struct nfs4_client, cl_lru);
3264 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3265 t = clp->cl_time - cutoff;
3266 if (clientid_val > t)
3270 if (mark_client_expired_locked(clp)) {
3271 dprintk("NFSD: client in use (clientid %08x)\n",
3272 clp->cl_clientid.cl_id);
3275 list_move(&clp->cl_lru, &reaplist);
3277 spin_unlock(&nn->client_lock);
3278 list_for_each_safe(pos, next, &reaplist) {
3279 clp = list_entry(pos, struct nfs4_client, cl_lru);
3280 dprintk("NFSD: purging unused client (clientid %08x)\n",
3281 clp->cl_clientid.cl_id);
3284 spin_lock(&recall_lock);
3285 list_for_each_safe(pos, next, &nn->del_recall_lru) {
3286 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3287 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3289 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3290 u = dp->dl_time - cutoff;
3295 list_move(&dp->dl_recall_lru, &reaplist);
3297 spin_unlock(&recall_lock);
3298 list_for_each_safe(pos, next, &reaplist) {
3299 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3300 unhash_delegation(dp);
3302 test_val = nn->nfsd4_lease;
3303 list_for_each_safe(pos, next, &nn->close_lru) {
3304 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3305 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3306 u = oo->oo_time - cutoff;
3311 release_openowner(oo);
3313 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3314 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3315 nfs4_unlock_state();
3316 return clientid_val;
3319 static struct workqueue_struct *laundry_wq;
3320 static void laundromat_main(struct work_struct *);
3323 laundromat_main(struct work_struct *laundry)
3326 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3328 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3331 t = nfs4_laundromat(nn);
3332 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3333 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3336 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3338 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3339 return nfserr_bad_stateid;
3344 access_permit_read(struct nfs4_ol_stateid *stp)
3346 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3347 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3348 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3352 access_permit_write(struct nfs4_ol_stateid *stp)
3354 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3355 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3359 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3361 __be32 status = nfserr_openmode;
3363 /* For lock stateid's, we test the parent open, not the lock: */
3364 if (stp->st_openstp)
3365 stp = stp->st_openstp;
3366 if ((flags & WR_STATE) && !access_permit_write(stp))
3368 if ((flags & RD_STATE) && !access_permit_read(stp))
3375 static inline __be32
3376 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3378 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3380 else if (locks_in_grace(net)) {
3381 /* Answer in remaining cases depends on existence of
3382 * conflicting state; so we must wait out the grace period. */
3383 return nfserr_grace;
3384 } else if (flags & WR_STATE)
3385 return nfs4_share_conflict(current_fh,
3386 NFS4_SHARE_DENY_WRITE);
3387 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3388 return nfs4_share_conflict(current_fh,
3389 NFS4_SHARE_DENY_READ);
3393 * Allow READ/WRITE during grace period on recovered state only for files
3394 * that are not able to provide mandatory locking.
3397 grace_disallows_io(struct net *net, struct inode *inode)
3399 return locks_in_grace(net) && mandatory_lock(inode);
3402 /* Returns true iff a is later than b: */
3403 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3405 return (s32)a->si_generation - (s32)b->si_generation > 0;
3408 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3411 * When sessions are used the stateid generation number is ignored
3414 if (has_session && in->si_generation == 0)
3417 if (in->si_generation == ref->si_generation)
3420 /* If the client sends us a stateid from the future, it's buggy: */
3421 if (stateid_generation_after(in, ref))
3422 return nfserr_bad_stateid;
3424 * However, we could see a stateid from the past, even from a
3425 * non-buggy client. For example, if the client sends a lock
3426 * while some IO is outstanding, the lock may bump si_generation
3427 * while the IO is still in flight. The client could avoid that
3428 * situation by waiting for responses on all the IO requests,
3429 * but better performance may result in retrying IO that
3430 * receives an old_stateid error if requests are rarely
3431 * reordered in flight:
3433 return nfserr_old_stateid;
3436 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3438 struct nfs4_stid *s;
3439 struct nfs4_ol_stateid *ols;
3442 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3443 return nfserr_bad_stateid;
3444 /* Client debugging aid. */
3445 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3446 char addr_str[INET6_ADDRSTRLEN];
3447 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3449 pr_warn_ratelimited("NFSD: client %s testing state ID "
3450 "with incorrect client ID\n", addr_str);
3451 return nfserr_bad_stateid;
3453 s = find_stateid(cl, stateid);
3455 return nfserr_bad_stateid;
3456 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3459 switch (s->sc_type) {
3460 case NFS4_DELEG_STID:
3462 case NFS4_OPEN_STID:
3463 case NFS4_LOCK_STID:
3464 ols = openlockstateid(s);
3465 if (ols->st_stateowner->so_is_open_owner
3466 && !(openowner(ols->st_stateowner)->oo_flags
3467 & NFS4_OO_CONFIRMED))
3468 return nfserr_bad_stateid;
3471 printk("unknown stateid type %x\n", s->sc_type);
3472 case NFS4_CLOSED_STID:
3473 return nfserr_bad_stateid;
3477 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3478 struct nfs4_stid **s, bool sessions,
3479 struct nfsd_net *nn)
3481 struct nfs4_client *cl;
3484 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3485 return nfserr_bad_stateid;
3486 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3488 if (status == nfserr_stale_clientid)
3489 return nfserr_stale_stateid;
3492 *s = find_stateid_by_type(cl, stateid, typemask);
3494 return nfserr_bad_stateid;
3499 * Checks for stateid operations
3502 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3503 stateid_t *stateid, int flags, struct file **filpp)
3505 struct nfs4_stid *s;
3506 struct nfs4_ol_stateid *stp = NULL;
3507 struct nfs4_delegation *dp = NULL;
3508 struct svc_fh *current_fh = &cstate->current_fh;
3509 struct inode *ino = current_fh->fh_dentry->d_inode;
3510 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3516 if (grace_disallows_io(net, ino))
3517 return nfserr_grace;
3519 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3520 return check_special_stateids(net, current_fh, stateid, flags);
3522 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3523 &s, cstate->minorversion, nn);
3526 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3529 switch (s->sc_type) {
3530 case NFS4_DELEG_STID:
3531 dp = delegstateid(s);
3532 status = nfs4_check_delegmode(dp, flags);
3536 *filpp = dp->dl_file->fi_deleg_file;
3539 status = nfserr_serverfault;
3544 case NFS4_OPEN_STID:
3545 case NFS4_LOCK_STID:
3546 stp = openlockstateid(s);
3547 status = nfs4_check_fh(current_fh, stp);
3550 if (stp->st_stateowner->so_is_open_owner
3551 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3553 status = nfs4_check_openmode(stp, flags);
3557 if (flags & RD_STATE)
3558 *filpp = find_readable_file(stp->st_file);
3560 *filpp = find_writeable_file(stp->st_file);
3564 return nfserr_bad_stateid;
3572 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3574 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3575 return nfserr_locks_held;
3576 release_lock_stateid(stp);
3581 * Test if the stateid is valid
3584 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3585 struct nfsd4_test_stateid *test_stateid)
3587 struct nfsd4_test_stateid_id *stateid;
3588 struct nfs4_client *cl = cstate->session->se_client;
3591 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3592 stateid->ts_id_status =
3593 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3594 nfs4_unlock_state();
3600 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3601 struct nfsd4_free_stateid *free_stateid)
3603 stateid_t *stateid = &free_stateid->fr_stateid;
3604 struct nfs4_stid *s;
3605 struct nfs4_client *cl = cstate->session->se_client;
3606 __be32 ret = nfserr_bad_stateid;
3609 s = find_stateid(cl, stateid);
3612 switch (s->sc_type) {
3613 case NFS4_DELEG_STID:
3614 ret = nfserr_locks_held;
3616 case NFS4_OPEN_STID:
3617 case NFS4_LOCK_STID:
3618 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3621 if (s->sc_type == NFS4_LOCK_STID)
3622 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3624 ret = nfserr_locks_held;
3627 ret = nfserr_bad_stateid;
3630 nfs4_unlock_state();
3637 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3638 RD_STATE : WR_STATE;
3641 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3643 struct svc_fh *current_fh = &cstate->current_fh;
3644 struct nfs4_stateowner *sop = stp->st_stateowner;
3647 status = nfsd4_check_seqid(cstate, sop, seqid);
3650 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3652 * "Closed" stateid's exist *only* to return
3653 * nfserr_replay_me from the previous step.
3655 return nfserr_bad_stateid;
3656 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3659 return nfs4_check_fh(current_fh, stp);
3663 * Checks for sequence id mutating operations.
3666 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3667 stateid_t *stateid, char typemask,
3668 struct nfs4_ol_stateid **stpp,
3669 struct nfsd_net *nn)
3672 struct nfs4_stid *s;
3674 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3675 seqid, STATEID_VAL(stateid));
3678 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3679 cstate->minorversion, nn);
3682 *stpp = openlockstateid(s);
3683 if (!nfsd4_has_session(cstate))
3684 cstate->replay_owner = (*stpp)->st_stateowner;
3686 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3689 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3690 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3693 struct nfs4_openowner *oo;
3695 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3696 NFS4_OPEN_STID, stpp, nn);
3699 oo = openowner((*stpp)->st_stateowner);
3700 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3701 return nfserr_bad_stateid;
3706 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3707 struct nfsd4_open_confirm *oc)
3710 struct nfs4_openowner *oo;
3711 struct nfs4_ol_stateid *stp;
3712 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3714 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3715 (int)cstate->current_fh.fh_dentry->d_name.len,
3716 cstate->current_fh.fh_dentry->d_name.name);
3718 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3724 status = nfs4_preprocess_seqid_op(cstate,
3725 oc->oc_seqid, &oc->oc_req_stateid,
3726 NFS4_OPEN_STID, &stp, nn);
3729 oo = openowner(stp->st_stateowner);
3730 status = nfserr_bad_stateid;
3731 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3733 oo->oo_flags |= NFS4_OO_CONFIRMED;
3734 update_stateid(&stp->st_stid.sc_stateid);
3735 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3736 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3737 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3739 nfsd4_client_record_create(oo->oo_owner.so_client);
3742 nfsd4_bump_seqid(cstate, status);
3743 if (!cstate->replay_owner)
3744 nfs4_unlock_state();
3748 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3750 if (!test_access(access, stp))
3752 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3753 clear_access(access, stp);
3756 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3758 switch (to_access) {
3759 case NFS4_SHARE_ACCESS_READ:
3760 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3761 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3763 case NFS4_SHARE_ACCESS_WRITE:
3764 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3765 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3767 case NFS4_SHARE_ACCESS_BOTH:
3775 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3778 for (i = 0; i < 4; i++) {
3779 if ((i & deny) != i)
3785 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3786 struct nfsd4_compound_state *cstate,
3787 struct nfsd4_open_downgrade *od)
3790 struct nfs4_ol_stateid *stp;
3791 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3793 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3794 (int)cstate->current_fh.fh_dentry->d_name.len,
3795 cstate->current_fh.fh_dentry->d_name.name);
3797 /* We don't yet support WANT bits: */
3798 if (od->od_deleg_want)
3799 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3803 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3804 &od->od_stateid, &stp, nn);
3807 status = nfserr_inval;
3808 if (!test_access(od->od_share_access, stp)) {
3809 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3810 stp->st_access_bmap, od->od_share_access);
3813 if (!test_deny(od->od_share_deny, stp)) {
3814 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3815 stp->st_deny_bmap, od->od_share_deny);
3818 nfs4_stateid_downgrade(stp, od->od_share_access);
3820 reset_union_bmap_deny(od->od_share_deny, stp);
3822 update_stateid(&stp->st_stid.sc_stateid);
3823 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3826 nfsd4_bump_seqid(cstate, status);
3827 if (!cstate->replay_owner)
3828 nfs4_unlock_state();
3832 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3834 unhash_open_stateid(s);
3835 s->st_stid.sc_type = NFS4_CLOSED_STID;
3839 * nfs4_unlock_state() called after encode
3842 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3843 struct nfsd4_close *close)
3846 struct nfs4_openowner *oo;
3847 struct nfs4_ol_stateid *stp;
3848 struct net *net = SVC_NET(rqstp);
3849 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3851 dprintk("NFSD: nfsd4_close on file %.*s\n",
3852 (int)cstate->current_fh.fh_dentry->d_name.len,
3853 cstate->current_fh.fh_dentry->d_name.name);
3856 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3858 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3860 nfsd4_bump_seqid(cstate, status);
3863 oo = openowner(stp->st_stateowner);
3864 update_stateid(&stp->st_stid.sc_stateid);
3865 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3867 nfsd4_close_open_stateid(stp);
3869 if (cstate->minorversion) {
3870 unhash_stid(&stp->st_stid);
3871 free_generic_stateid(stp);
3873 oo->oo_last_closed_stid = stp;
3875 if (list_empty(&oo->oo_owner.so_stateids)) {
3876 if (cstate->minorversion)
3877 release_openowner(oo);
3880 * In the 4.0 case we need to keep the owners around a
3881 * little while to handle CLOSE replay.
3883 if (list_empty(&oo->oo_owner.so_stateids))
3884 move_to_close_lru(oo, SVC_NET(rqstp));
3888 if (!cstate->replay_owner)
3889 nfs4_unlock_state();
3894 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3895 struct nfsd4_delegreturn *dr)
3897 struct nfs4_delegation *dp;
3898 stateid_t *stateid = &dr->dr_stateid;
3899 struct nfs4_stid *s;
3901 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3903 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3907 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3908 cstate->minorversion, nn);
3911 dp = delegstateid(s);
3912 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3916 unhash_delegation(dp);
3918 nfs4_unlock_state();
3924 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3926 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3929 end_offset(u64 start, u64 len)
3934 return end >= start ? end: NFS4_MAX_UINT64;
3937 /* last octet in a range */
3939 last_byte_offset(u64 start, u64 len)
3945 return end > start ? end - 1: NFS4_MAX_UINT64;
3948 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3950 return (file_hashval(inode) + cl_id
3951 + opaque_hashval(ownername->data, ownername->len))
3952 & LOCKOWNER_INO_HASH_MASK;
3956 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3957 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3958 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3959 * locking, this prevents us from being completely protocol-compliant. The
3960 * real solution to this problem is to start using unsigned file offsets in
3961 * the VFS, but this is a very deep change!
3964 nfs4_transform_lock_offset(struct file_lock *lock)
3966 if (lock->fl_start < 0)
3967 lock->fl_start = OFFSET_MAX;
3968 if (lock->fl_end < 0)
3969 lock->fl_end = OFFSET_MAX;
3972 /* Hack!: For now, we're defining this just so we can use a pointer to it
3973 * as a unique cookie to identify our (NFSv4's) posix locks. */
3974 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3978 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3980 struct nfs4_lockowner *lo;
3982 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3983 lo = (struct nfs4_lockowner *) fl->fl_owner;
3984 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3985 lo->lo_owner.so_owner.len, GFP_KERNEL);
3986 if (!deny->ld_owner.data)
3987 /* We just don't care that much */
3989 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3990 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3993 deny->ld_owner.len = 0;
3994 deny->ld_owner.data = NULL;
3995 deny->ld_clientid.cl_boot = 0;
3996 deny->ld_clientid.cl_id = 0;
3998 deny->ld_start = fl->fl_start;
3999 deny->ld_length = NFS4_MAX_UINT64;
4000 if (fl->fl_end != NFS4_MAX_UINT64)
4001 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4002 deny->ld_type = NFS4_READ_LT;
4003 if (fl->fl_type != F_RDLCK)
4004 deny->ld_type = NFS4_WRITE_LT;
4007 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
4009 struct nfs4_ol_stateid *lst;
4011 if (!same_owner_str(&lo->lo_owner, owner, clid))
4013 lst = list_first_entry(&lo->lo_owner.so_stateids,
4014 struct nfs4_ol_stateid, st_perstateowner);
4015 return lst->st_file->fi_inode == inode;
4018 static struct nfs4_lockowner *
4019 find_lockowner_str(struct inode *inode, clientid_t *clid,
4020 struct xdr_netobj *owner, struct nfsd_net *nn)
4022 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
4023 struct nfs4_lockowner *lo;
4025 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
4026 if (same_lockowner_ino(lo, inode, clid, owner))
4032 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
4034 struct inode *inode = open_stp->st_file->fi_inode;
4035 unsigned int inohash = lockowner_ino_hashval(inode,
4036 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4037 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4039 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4040 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4041 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4045 * Alloc a lock owner structure.
4046 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4049 * strhashval = ownerstr_hashval
4052 static struct nfs4_lockowner *
4053 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4054 struct nfs4_lockowner *lo;
4056 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4059 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4060 lo->lo_owner.so_is_open_owner = 0;
4061 /* It is the openowner seqid that will be incremented in encode in the
4062 * case of new lockowners; so increment the lock seqid manually: */
4063 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4064 hash_lockowner(lo, strhashval, clp, open_stp);
4068 static struct nfs4_ol_stateid *
4069 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4071 struct nfs4_ol_stateid *stp;
4072 struct nfs4_client *clp = lo->lo_owner.so_client;
4074 stp = nfs4_alloc_stateid(clp);
4077 stp->st_stid.sc_type = NFS4_LOCK_STID;
4078 list_add(&stp->st_perfile, &fp->fi_stateids);
4079 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4080 stp->st_stateowner = &lo->lo_owner;
4083 stp->st_access_bmap = 0;
4084 stp->st_deny_bmap = open_stp->st_deny_bmap;
4085 stp->st_openstp = open_stp;
4090 check_lock_length(u64 offset, u64 length)
4092 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4093 LOFF_OVERFLOW(offset, length)));
4096 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4098 struct nfs4_file *fp = lock_stp->st_file;
4099 int oflag = nfs4_access_to_omode(access);
4101 if (test_access(access, lock_stp))
4103 nfs4_file_get_access(fp, oflag);
4104 set_access(access, lock_stp);
4107 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)
4109 struct nfs4_file *fi = ost->st_file;
4110 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4111 struct nfs4_client *cl = oo->oo_owner.so_client;
4112 struct nfs4_lockowner *lo;
4113 unsigned int strhashval;
4114 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4116 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4117 &lock->v.new.owner, nn);
4119 if (!cstate->minorversion)
4120 return nfserr_bad_seqid;
4121 /* XXX: a lockowner always has exactly one stateid: */
4122 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4123 struct nfs4_ol_stateid, st_perstateowner);
4126 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4127 &lock->v.new.owner);
4128 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4130 return nfserr_jukebox;
4131 *lst = alloc_init_lock_stateid(lo, fi, ost);
4133 release_lockowner(lo);
4134 return nfserr_jukebox;
4144 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4145 struct nfsd4_lock *lock)
4147 struct nfs4_openowner *open_sop = NULL;
4148 struct nfs4_lockowner *lock_sop = NULL;
4149 struct nfs4_ol_stateid *lock_stp;
4150 struct file *filp = NULL;
4151 struct file_lock *file_lock = NULL;
4152 struct file_lock *conflock = NULL;
4154 bool new_state = false;
4157 struct net *net = SVC_NET(rqstp);
4158 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4160 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4161 (long long) lock->lk_offset,
4162 (long long) lock->lk_length);
4164 if (check_lock_length(lock->lk_offset, lock->lk_length))
4165 return nfserr_inval;
4167 if ((status = fh_verify(rqstp, &cstate->current_fh,
4168 S_IFREG, NFSD_MAY_LOCK))) {
4169 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4175 if (lock->lk_is_new) {
4176 struct nfs4_ol_stateid *open_stp = NULL;
4178 if (nfsd4_has_session(cstate))
4179 /* See rfc 5661 18.10.3: given clientid is ignored: */
4180 memcpy(&lock->v.new.clientid,
4181 &cstate->session->se_client->cl_clientid,
4182 sizeof(clientid_t));
4184 status = nfserr_stale_clientid;
4185 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4188 /* validate and update open stateid and open seqid */
4189 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4190 lock->lk_new_open_seqid,
4191 &lock->lk_new_open_stateid,
4195 open_sop = openowner(open_stp->st_stateowner);
4196 status = nfserr_bad_stateid;
4197 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4198 &lock->v.new.clientid))
4200 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4201 &lock_stp, &new_state);
4203 status = nfs4_preprocess_seqid_op(cstate,
4204 lock->lk_old_lock_seqid,
4205 &lock->lk_old_lock_stateid,
4206 NFS4_LOCK_STID, &lock_stp, nn);
4209 lock_sop = lockowner(lock_stp->st_stateowner);
4211 lkflg = setlkflg(lock->lk_type);
4212 status = nfs4_check_openmode(lock_stp, lkflg);
4216 status = nfserr_grace;
4217 if (locks_in_grace(net) && !lock->lk_reclaim)
4219 status = nfserr_no_grace;
4220 if (!locks_in_grace(net) && lock->lk_reclaim)
4223 file_lock = locks_alloc_lock();
4225 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4226 status = nfserr_jukebox;
4230 locks_init_lock(file_lock);
4231 switch (lock->lk_type) {
4234 filp = find_readable_file(lock_stp->st_file);
4236 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4237 file_lock->fl_type = F_RDLCK;
4240 case NFS4_WRITEW_LT:
4241 filp = find_writeable_file(lock_stp->st_file);
4243 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4244 file_lock->fl_type = F_WRLCK;
4247 status = nfserr_inval;
4251 status = nfserr_openmode;
4254 file_lock->fl_owner = (fl_owner_t)lock_sop;
4255 file_lock->fl_pid = current->tgid;
4256 file_lock->fl_file = filp;
4257 file_lock->fl_flags = FL_POSIX;
4258 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4259 file_lock->fl_start = lock->lk_offset;
4260 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4261 nfs4_transform_lock_offset(file_lock);
4263 conflock = locks_alloc_lock();
4265 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4266 status = nfserr_jukebox;
4270 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4272 case 0: /* success! */
4273 update_stateid(&lock_stp->st_stid.sc_stateid);
4274 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4278 case (EAGAIN): /* conflock holds conflicting lock */
4279 status = nfserr_denied;
4280 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4281 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4284 status = nfserr_deadlock;
4287 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4288 status = nfserrno(err);
4292 if (status && new_state)
4293 release_lockowner(lock_sop);
4294 nfsd4_bump_seqid(cstate, status);
4295 if (!cstate->replay_owner)
4296 nfs4_unlock_state();
4298 locks_free_lock(file_lock);
4300 locks_free_lock(conflock);
4305 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4306 * so we do a temporary open here just to get an open file to pass to
4307 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4310 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4313 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4315 err = nfserrno(vfs_test_lock(file, lock));
4325 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4326 struct nfsd4_lockt *lockt)
4328 struct inode *inode;
4329 struct file_lock *file_lock = NULL;
4330 struct nfs4_lockowner *lo;
4332 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4334 if (locks_in_grace(SVC_NET(rqstp)))
4335 return nfserr_grace;
4337 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4338 return nfserr_inval;
4342 if (!nfsd4_has_session(cstate)) {
4343 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4348 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4351 inode = cstate->current_fh.fh_dentry->d_inode;
4352 file_lock = locks_alloc_lock();
4354 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4355 status = nfserr_jukebox;
4358 locks_init_lock(file_lock);
4359 switch (lockt->lt_type) {
4362 file_lock->fl_type = F_RDLCK;
4365 case NFS4_WRITEW_LT:
4366 file_lock->fl_type = F_WRLCK;
4369 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4370 status = nfserr_inval;
4374 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4376 file_lock->fl_owner = (fl_owner_t)lo;
4377 file_lock->fl_pid = current->tgid;
4378 file_lock->fl_flags = FL_POSIX;
4380 file_lock->fl_start = lockt->lt_offset;
4381 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4383 nfs4_transform_lock_offset(file_lock);
4385 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4389 if (file_lock->fl_type != F_UNLCK) {
4390 status = nfserr_denied;
4391 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4394 nfs4_unlock_state();
4396 locks_free_lock(file_lock);
4401 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4402 struct nfsd4_locku *locku)
4404 struct nfs4_lockowner *lo;
4405 struct nfs4_ol_stateid *stp;
4406 struct file *filp = NULL;
4407 struct file_lock *file_lock = NULL;
4410 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4412 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4413 (long long) locku->lu_offset,
4414 (long long) locku->lu_length);
4416 if (check_lock_length(locku->lu_offset, locku->lu_length))
4417 return nfserr_inval;
4421 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4422 &locku->lu_stateid, NFS4_LOCK_STID,
4426 filp = find_any_file(stp->st_file);
4428 status = nfserr_lock_range;
4431 file_lock = locks_alloc_lock();
4433 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4434 status = nfserr_jukebox;
4437 lo = lockowner(stp->st_stateowner);
4438 locks_init_lock(file_lock);
4439 file_lock->fl_type = F_UNLCK;
4440 file_lock->fl_owner = (fl_owner_t)lo;
4441 file_lock->fl_pid = current->tgid;
4442 file_lock->fl_file = filp;
4443 file_lock->fl_flags = FL_POSIX;
4444 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4445 file_lock->fl_start = locku->lu_offset;
4447 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4449 nfs4_transform_lock_offset(file_lock);
4451 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4453 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4456 update_stateid(&stp->st_stid.sc_stateid);
4457 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4459 if (nfsd4_has_session(cstate) && !check_for_locks(stp->st_file, lo)) {
4460 WARN_ON_ONCE(cstate->replay_owner);
4461 release_lockowner(lo);
4465 nfsd4_bump_seqid(cstate, status);
4466 if (!cstate->replay_owner)
4467 nfs4_unlock_state();
4469 locks_free_lock(file_lock);
4473 status = nfserrno(err);
4479 * 1: locks held by lockowner
4480 * 0: no locks held by lockowner
4483 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4485 struct file_lock **flpp;
4486 struct inode *inode = filp->fi_inode;
4490 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4491 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4502 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4503 struct nfsd4_compound_state *cstate,
4504 struct nfsd4_release_lockowner *rlockowner)
4506 clientid_t *clid = &rlockowner->rl_clientid;
4507 struct nfs4_stateowner *sop;
4508 struct nfs4_lockowner *lo;
4509 struct nfs4_ol_stateid *stp;
4510 struct xdr_netobj *owner = &rlockowner->rl_owner;
4511 struct list_head matches;
4512 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4514 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4516 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4517 clid->cl_boot, clid->cl_id);
4521 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4525 status = nfserr_locks_held;
4526 INIT_LIST_HEAD(&matches);
4528 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4529 if (sop->so_is_open_owner)
4531 if (!same_owner_str(sop, owner, clid))
4533 list_for_each_entry(stp, &sop->so_stateids,
4535 lo = lockowner(sop);
4536 if (check_for_locks(stp->st_file, lo))
4538 list_add(&lo->lo_list, &matches);
4541 /* Clients probably won't expect us to return with some (but not all)
4542 * of the lockowner state released; so don't release any until all
4543 * have been checked. */
4545 while (!list_empty(&matches)) {
4546 lo = list_entry(matches.next, struct nfs4_lockowner,
4548 /* unhash_stateowner deletes so_perclient only
4549 * for openowners. */
4550 list_del(&lo->lo_list);
4551 release_lockowner(lo);
4554 nfs4_unlock_state();
4558 static inline struct nfs4_client_reclaim *
4561 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4565 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4567 struct nfs4_client_reclaim *crp;
4569 crp = nfsd4_find_reclaim_client(name, nn);
4570 return (crp && crp->cr_clp);
4574 * failure => all reset bets are off, nfserr_no_grace...
4576 struct nfs4_client_reclaim *
4577 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4579 unsigned int strhashval;
4580 struct nfs4_client_reclaim *crp;
4582 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4583 crp = alloc_reclaim();
4585 strhashval = clientstr_hashval(name);
4586 INIT_LIST_HEAD(&crp->cr_strhash);
4587 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4588 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4590 nn->reclaim_str_hashtbl_size++;
4596 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4598 list_del(&crp->cr_strhash);
4600 nn->reclaim_str_hashtbl_size--;
4604 nfs4_release_reclaim(struct nfsd_net *nn)
4606 struct nfs4_client_reclaim *crp = NULL;
4609 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4610 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4611 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4612 struct nfs4_client_reclaim, cr_strhash);
4613 nfs4_remove_reclaim_record(crp, nn);
4616 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4620 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4621 struct nfs4_client_reclaim *
4622 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4624 unsigned int strhashval;
4625 struct nfs4_client_reclaim *crp = NULL;
4627 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4629 strhashval = clientstr_hashval(recdir);
4630 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4631 if (same_name(crp->cr_recdir, recdir)) {
4639 * Called from OPEN. Look for clientid in reclaim list.
4642 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4644 struct nfs4_client *clp;
4646 /* find clientid in conf_id_hashtbl */
4647 clp = find_confirmed_client(clid, sessions, nn);
4649 return nfserr_reclaim_bad;
4651 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4654 #ifdef CONFIG_NFSD_FAULT_INJECTION
4656 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4658 if (mark_client_expired(clp))
4664 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4666 char buf[INET6_ADDRSTRLEN];
4667 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4668 printk(KERN_INFO "NFS Client: %s\n", buf);
4672 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4675 char buf[INET6_ADDRSTRLEN];
4676 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4677 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4680 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4682 struct nfs4_openowner *oop;
4683 struct nfs4_lockowner *lop, *lo_next;
4684 struct nfs4_ol_stateid *stp, *st_next;
4687 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4688 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4689 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4701 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4703 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4706 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4708 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4709 nfsd_print_count(clp, count, "locked files");
4713 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4715 struct nfs4_openowner *oop, *next;
4718 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4728 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4730 return nfsd_foreach_client_open(clp, max, release_openowner);
4733 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4735 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4736 nfsd_print_count(clp, count, "open files");
4740 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4741 struct list_head *victims)
4743 struct nfs4_delegation *dp, *next;
4746 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4748 list_move(&dp->dl_recall_lru, victims);
4755 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4757 struct nfs4_delegation *dp, *next;
4761 spin_lock(&recall_lock);
4762 count = nfsd_find_all_delegations(clp, max, &victims);
4763 spin_unlock(&recall_lock);
4765 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4766 unhash_delegation(dp);
4771 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4773 struct nfs4_delegation *dp, *next;
4777 spin_lock(&recall_lock);
4778 count = nfsd_find_all_delegations(clp, max, &victims);
4779 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4780 nfsd_break_one_deleg(dp);
4781 spin_unlock(&recall_lock);
4786 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4790 spin_lock(&recall_lock);
4791 count = nfsd_find_all_delegations(clp, max, NULL);
4792 spin_unlock(&recall_lock);
4794 nfsd_print_count(clp, count, "delegations");
4798 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4800 struct nfs4_client *clp, *next;
4802 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4804 if (!nfsd_netns_ready(nn))
4807 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4808 count += func(clp, max - count);
4809 if ((max != 0) && (count >= max))
4816 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4818 struct nfs4_client *clp;
4819 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4821 if (!nfsd_netns_ready(nn))
4824 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4825 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4831 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4833 /* initialization to perform at module load time: */
4836 nfs4_state_init(void)
4841 * Since the lifetime of a delegation isn't limited to that of an open, a
4842 * client may quite reasonably hang on to a delegation as long as it has
4843 * the inode cached. This becomes an obvious problem the first time a
4844 * client's inode cache approaches the size of the server's total memory.
4846 * For now we avoid this problem by imposing a hard limit on the number
4847 * of delegations, which varies according to the server's memory size.
4850 set_max_delegations(void)
4853 * Allow at most 4 delegations per megabyte of RAM. Quick
4854 * estimates suggest that in the worst case (where every delegation
4855 * is for a different inode), a delegation could take about 1.5K,
4856 * giving a worst case usage of about 6% of memory.
4858 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4861 static int nfs4_state_create_net(struct net *net)
4863 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4866 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4867 CLIENT_HASH_SIZE, GFP_KERNEL);
4868 if (!nn->conf_id_hashtbl)
4870 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4871 CLIENT_HASH_SIZE, GFP_KERNEL);
4872 if (!nn->unconf_id_hashtbl)
4874 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4875 OWNER_HASH_SIZE, GFP_KERNEL);
4876 if (!nn->ownerstr_hashtbl)
4878 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4879 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4880 if (!nn->lockowner_ino_hashtbl)
4881 goto err_lockowner_ino;
4882 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4883 SESSION_HASH_SIZE, GFP_KERNEL);
4884 if (!nn->sessionid_hashtbl)
4887 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4888 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4889 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4891 for (i = 0; i < OWNER_HASH_SIZE; i++)
4892 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4893 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4894 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4895 for (i = 0; i < SESSION_HASH_SIZE; i++)
4896 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4897 nn->conf_name_tree = RB_ROOT;
4898 nn->unconf_name_tree = RB_ROOT;
4899 INIT_LIST_HEAD(&nn->client_lru);
4900 INIT_LIST_HEAD(&nn->close_lru);
4901 INIT_LIST_HEAD(&nn->del_recall_lru);
4902 spin_lock_init(&nn->client_lock);
4904 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
4910 kfree(nn->lockowner_ino_hashtbl);
4912 kfree(nn->ownerstr_hashtbl);
4914 kfree(nn->unconf_id_hashtbl);
4916 kfree(nn->conf_id_hashtbl);
4922 nfs4_state_destroy_net(struct net *net)
4925 struct nfs4_client *clp = NULL;
4926 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4927 struct rb_node *node, *tmp;
4929 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4930 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4931 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4932 destroy_client(clp);
4936 node = rb_first(&nn->unconf_name_tree);
4937 while (node != NULL) {
4939 node = rb_next(tmp);
4940 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4941 rb_erase(tmp, &nn->unconf_name_tree);
4942 destroy_client(clp);
4945 kfree(nn->sessionid_hashtbl);
4946 kfree(nn->lockowner_ino_hashtbl);
4947 kfree(nn->ownerstr_hashtbl);
4948 kfree(nn->unconf_id_hashtbl);
4949 kfree(nn->conf_id_hashtbl);
4954 nfs4_state_start_net(struct net *net)
4956 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4959 ret = nfs4_state_create_net(net);
4962 nfsd4_client_tracking_init(net);
4963 nn->boot_time = get_seconds();
4964 locks_start_grace(net, &nn->nfsd4_manager);
4965 nn->grace_ended = false;
4966 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
4967 nn->nfsd4_grace, net);
4968 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
4972 /* initialization to perform when the nfsd service is started: */
4975 nfs4_state_start(void)
4979 ret = set_callback_cred();
4982 laundry_wq = create_singlethread_workqueue("nfsd4");
4983 if (laundry_wq == NULL) {
4987 ret = nfsd4_create_callback_queue();
4989 goto out_free_laundry;
4991 set_max_delegations();
4996 destroy_workqueue(laundry_wq);
5001 /* should be called with the state lock held */
5003 nfs4_state_shutdown_net(struct net *net)
5005 struct nfs4_delegation *dp = NULL;
5006 struct list_head *pos, *next, reaplist;
5007 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5009 cancel_delayed_work_sync(&nn->laundromat_work);
5010 locks_end_grace(&nn->nfsd4_manager);
5012 INIT_LIST_HEAD(&reaplist);
5013 spin_lock(&recall_lock);
5014 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5015 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5016 list_move(&dp->dl_recall_lru, &reaplist);
5018 spin_unlock(&recall_lock);
5019 list_for_each_safe(pos, next, &reaplist) {
5020 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5021 unhash_delegation(dp);
5024 nfsd4_client_tracking_exit(net);
5025 nfs4_state_destroy_net(net);
5029 nfs4_state_shutdown(void)
5031 destroy_workqueue(laundry_wq);
5032 nfsd4_destroy_callback_queue();
5036 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5038 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5039 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5043 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5045 if (cstate->minorversion) {
5046 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5047 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5052 clear_current_stateid(struct nfsd4_compound_state *cstate)
5054 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5058 * functions to set current state id
5061 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5063 put_stateid(cstate, &odp->od_stateid);
5067 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5069 put_stateid(cstate, &open->op_stateid);
5073 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5075 put_stateid(cstate, &close->cl_stateid);
5079 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5081 put_stateid(cstate, &lock->lk_resp_stateid);
5085 * functions to consume current state id
5089 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5091 get_stateid(cstate, &odp->od_stateid);
5095 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5097 get_stateid(cstate, &drp->dr_stateid);
5101 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5103 get_stateid(cstate, &fsp->fr_stateid);
5107 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5109 get_stateid(cstate, &setattr->sa_stateid);
5113 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5115 get_stateid(cstate, &close->cl_stateid);
5119 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5121 get_stateid(cstate, &locku->lu_stateid);
5125 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5127 get_stateid(cstate, &read->rd_stateid);
5131 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5133 get_stateid(cstate, &write->wr_stateid);