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 static 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 struct nfs4_stid *stid;
332 stid = kmem_cache_alloc(slab, GFP_KERNEL);
336 new_id = idr_alloc_cyclic(stateids, stid, 0, 0, GFP_KERNEL);
339 stid->sc_client = cl;
341 stid->sc_stateid.si_opaque.so_id = new_id;
342 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
343 /* Will be incremented before return to client: */
344 stid->sc_stateid.si_generation = 0;
347 * It shouldn't be a problem to reuse an opaque stateid value.
348 * I don't think it is for 4.1. But with 4.0 I worry that, for
349 * example, a stray write retransmission could be accepted by
350 * the server when it should have been rejected. Therefore,
351 * adopt a trick from the sctp code to attempt to maximize the
352 * amount of time until an id is reused, by ensuring they always
353 * "increase" (mod INT_MAX):
357 kmem_cache_free(slab, stid);
361 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
363 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
366 static struct nfs4_delegation *
367 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
369 struct nfs4_delegation *dp;
370 struct nfs4_file *fp = stp->st_file;
372 dprintk("NFSD alloc_init_deleg\n");
374 * Major work on the lease subsystem (for example, to support
375 * calbacks on stat) will be required before we can support
376 * write delegations properly.
378 if (type != NFS4_OPEN_DELEGATE_READ)
380 if (fp->fi_had_conflict)
382 if (num_delegations > max_delegations)
384 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
387 dp->dl_stid.sc_type = NFS4_DELEG_STID;
389 * delegation seqid's are never incremented. The 4.1 special
390 * meaning of seqid 0 isn't meaningful, really, but let's avoid
391 * 0 anyway just for consistency and use 1:
393 dp->dl_stid.sc_stateid.si_generation = 1;
395 INIT_LIST_HEAD(&dp->dl_perfile);
396 INIT_LIST_HEAD(&dp->dl_perclnt);
397 INIT_LIST_HEAD(&dp->dl_recall_lru);
401 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
403 atomic_set(&dp->dl_count, 1);
404 nfsd4_init_callback(&dp->dl_recall);
408 static void remove_stid(struct nfs4_stid *s)
410 struct idr *stateids = &s->sc_client->cl_stateids;
412 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
416 nfs4_put_delegation(struct nfs4_delegation *dp)
418 if (atomic_dec_and_test(&dp->dl_count)) {
419 kmem_cache_free(deleg_slab, dp);
424 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
426 if (atomic_dec_and_test(&fp->fi_delegees)) {
427 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
429 fput(fp->fi_deleg_file);
430 fp->fi_deleg_file = NULL;
434 static void unhash_stid(struct nfs4_stid *s)
439 /* Called under the state lock. */
441 unhash_delegation(struct nfs4_delegation *dp)
443 list_del_init(&dp->dl_perclnt);
444 spin_lock(&recall_lock);
445 list_del_init(&dp->dl_perfile);
446 list_del_init(&dp->dl_recall_lru);
447 spin_unlock(&recall_lock);
448 nfs4_put_deleg_lease(dp->dl_file);
449 put_nfs4_file(dp->dl_file);
455 static void destroy_revoked_delegation(struct nfs4_delegation *dp)
457 list_del_init(&dp->dl_recall_lru);
458 remove_stid(&dp->dl_stid);
459 nfs4_put_delegation(dp);
462 static void destroy_delegation(struct nfs4_delegation *dp)
464 unhash_delegation(dp);
465 remove_stid(&dp->dl_stid);
466 nfs4_put_delegation(dp);
469 static void revoke_delegation(struct nfs4_delegation *dp)
471 struct nfs4_client *clp = dp->dl_stid.sc_client;
473 if (clp->cl_minorversion == 0)
474 destroy_delegation(dp);
476 unhash_delegation(dp);
477 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
478 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
486 static unsigned int clientid_hashval(u32 id)
488 return id & CLIENT_HASH_MASK;
491 static unsigned int clientstr_hashval(const char *name)
493 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
497 * We store the NONE, READ, WRITE, and BOTH bits separately in the
498 * st_{access,deny}_bmap field of the stateid, in order to track not
499 * only what share bits are currently in force, but also what
500 * combinations of share bits previous opens have used. This allows us
501 * to enforce the recommendation of rfc 3530 14.2.19 that the server
502 * return an error if the client attempt to downgrade to a combination
503 * of share bits not explicable by closing some of its previous opens.
505 * XXX: This enforcement is actually incomplete, since we don't keep
506 * track of access/deny bit combinations; so, e.g., we allow:
508 * OPEN allow read, deny write
509 * OPEN allow both, deny none
510 * DOWNGRADE allow read, deny none
512 * which we should reject.
515 bmap_to_share_mode(unsigned long bmap) {
517 unsigned int access = 0;
519 for (i = 1; i < 4; i++) {
520 if (test_bit(i, &bmap))
527 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
528 unsigned int access, deny;
530 access = bmap_to_share_mode(stp->st_access_bmap);
531 deny = bmap_to_share_mode(stp->st_deny_bmap);
532 if ((access & open->op_share_deny) || (deny & open->op_share_access))
537 /* set share access for a given stateid */
539 set_access(u32 access, struct nfs4_ol_stateid *stp)
541 __set_bit(access, &stp->st_access_bmap);
544 /* clear share access for a given stateid */
546 clear_access(u32 access, struct nfs4_ol_stateid *stp)
548 __clear_bit(access, &stp->st_access_bmap);
551 /* test whether a given stateid has access */
553 test_access(u32 access, struct nfs4_ol_stateid *stp)
555 return test_bit(access, &stp->st_access_bmap);
558 /* set share deny for a given stateid */
560 set_deny(u32 access, struct nfs4_ol_stateid *stp)
562 __set_bit(access, &stp->st_deny_bmap);
565 /* clear share deny for a given stateid */
567 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
569 __clear_bit(access, &stp->st_deny_bmap);
572 /* test whether a given stateid is denying specific access */
574 test_deny(u32 access, struct nfs4_ol_stateid *stp)
576 return test_bit(access, &stp->st_deny_bmap);
579 static int nfs4_access_to_omode(u32 access)
581 switch (access & NFS4_SHARE_ACCESS_BOTH) {
582 case NFS4_SHARE_ACCESS_READ:
584 case NFS4_SHARE_ACCESS_WRITE:
586 case NFS4_SHARE_ACCESS_BOTH:
593 /* release all access and file references for a given stateid */
595 release_all_access(struct nfs4_ol_stateid *stp)
599 for (i = 1; i < 4; i++) {
600 if (test_access(i, stp))
601 nfs4_file_put_access(stp->st_file,
602 nfs4_access_to_omode(i));
603 clear_access(i, stp);
607 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
609 list_del(&stp->st_perfile);
610 list_del(&stp->st_perstateowner);
613 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
615 release_all_access(stp);
616 put_nfs4_file(stp->st_file);
620 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
622 remove_stid(&stp->st_stid);
623 kmem_cache_free(stateid_slab, stp);
626 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
630 unhash_generic_stateid(stp);
631 unhash_stid(&stp->st_stid);
632 file = find_any_file(stp->st_file);
634 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
635 close_generic_stateid(stp);
636 free_generic_stateid(stp);
639 static void unhash_lockowner(struct nfs4_lockowner *lo)
641 struct nfs4_ol_stateid *stp;
643 list_del(&lo->lo_owner.so_strhash);
644 list_del(&lo->lo_perstateid);
645 list_del(&lo->lo_owner_ino_hash);
646 while (!list_empty(&lo->lo_owner.so_stateids)) {
647 stp = list_first_entry(&lo->lo_owner.so_stateids,
648 struct nfs4_ol_stateid, st_perstateowner);
649 release_lock_stateid(stp);
653 static void release_lockowner(struct nfs4_lockowner *lo)
655 unhash_lockowner(lo);
656 nfs4_free_lockowner(lo);
660 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
662 struct nfs4_lockowner *lo;
664 while (!list_empty(&open_stp->st_lockowners)) {
665 lo = list_entry(open_stp->st_lockowners.next,
666 struct nfs4_lockowner, lo_perstateid);
667 release_lockowner(lo);
671 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
673 unhash_generic_stateid(stp);
674 release_stateid_lockowners(stp);
675 close_generic_stateid(stp);
678 static void release_open_stateid(struct nfs4_ol_stateid *stp)
680 unhash_open_stateid(stp);
681 unhash_stid(&stp->st_stid);
682 free_generic_stateid(stp);
685 static void unhash_openowner(struct nfs4_openowner *oo)
687 struct nfs4_ol_stateid *stp;
689 list_del(&oo->oo_owner.so_strhash);
690 list_del(&oo->oo_perclient);
691 while (!list_empty(&oo->oo_owner.so_stateids)) {
692 stp = list_first_entry(&oo->oo_owner.so_stateids,
693 struct nfs4_ol_stateid, st_perstateowner);
694 release_open_stateid(stp);
698 static void release_last_closed_stateid(struct nfs4_openowner *oo)
700 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
703 unhash_stid(&s->st_stid);
704 free_generic_stateid(s);
705 oo->oo_last_closed_stid = NULL;
709 static void release_openowner(struct nfs4_openowner *oo)
711 unhash_openowner(oo);
712 list_del(&oo->oo_close_lru);
713 release_last_closed_stateid(oo);
714 nfs4_free_openowner(oo);
718 hash_sessionid(struct nfs4_sessionid *sessionid)
720 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
722 return sid->sequence % SESSION_HASH_SIZE;
727 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
729 u32 *ptr = (u32 *)(&sessionid->data[0]);
730 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
734 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
740 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
741 * won't be used for replay.
743 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
745 struct nfs4_stateowner *so = cstate->replay_owner;
747 if (nfserr == nfserr_replay_me)
750 if (!seqid_mutating_err(ntohl(nfserr))) {
751 cstate->replay_owner = NULL;
756 if (so->so_is_open_owner)
757 release_last_closed_stateid(openowner(so));
763 gen_sessionid(struct nfsd4_session *ses)
765 struct nfs4_client *clp = ses->se_client;
766 struct nfsd4_sessionid *sid;
768 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
769 sid->clientid = clp->cl_clientid;
770 sid->sequence = current_sessionid++;
775 * The protocol defines ca_maxresponssize_cached to include the size of
776 * the rpc header, but all we need to cache is the data starting after
777 * the end of the initial SEQUENCE operation--the rest we regenerate
778 * each time. Therefore we can advertise a ca_maxresponssize_cached
779 * value that is the number of bytes in our cache plus a few additional
780 * bytes. In order to stay on the safe side, and not promise more than
781 * we can cache, those additional bytes must be the minimum possible: 24
782 * bytes of rpc header (xid through accept state, with AUTH_NULL
783 * verifier), 12 for the compound header (with zero-length tag), and 44
784 * for the SEQUENCE op response:
786 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
789 free_session_slots(struct nfsd4_session *ses)
793 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
794 kfree(ses->se_slots[i]);
798 * We don't actually need to cache the rpc and session headers, so we
799 * can allocate a little less for each slot:
801 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
805 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
808 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
809 return size + sizeof(struct nfsd4_slot);
813 * XXX: If we run out of reserved DRC memory we could (up to a point)
814 * re-negotiate active sessions and reduce their slot usage to make
815 * room for new connections. For now we just fail the create session.
817 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
819 u32 slotsize = slot_bytes(ca);
820 u32 num = ca->maxreqs;
823 spin_lock(&nfsd_drc_lock);
824 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
825 nfsd_drc_max_mem - nfsd_drc_mem_used);
826 num = min_t(int, num, avail / slotsize);
827 nfsd_drc_mem_used += num * slotsize;
828 spin_unlock(&nfsd_drc_lock);
833 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
835 int slotsize = slot_bytes(ca);
837 spin_lock(&nfsd_drc_lock);
838 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
839 spin_unlock(&nfsd_drc_lock);
842 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *attrs)
844 int numslots = attrs->maxreqs;
845 int slotsize = slot_bytes(attrs);
846 struct nfsd4_session *new;
849 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
850 + sizeof(struct nfsd4_session) > PAGE_SIZE);
851 mem = numslots * sizeof(struct nfsd4_slot *);
853 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
856 /* allocate each struct nfsd4_slot and data cache in one piece */
857 for (i = 0; i < numslots; i++) {
858 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
859 if (!new->se_slots[i])
865 kfree(new->se_slots[i]);
870 static void free_conn(struct nfsd4_conn *c)
872 svc_xprt_put(c->cn_xprt);
876 static void nfsd4_conn_lost(struct svc_xpt_user *u)
878 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
879 struct nfs4_client *clp = c->cn_session->se_client;
881 spin_lock(&clp->cl_lock);
882 if (!list_empty(&c->cn_persession)) {
883 list_del(&c->cn_persession);
886 nfsd4_probe_callback(clp);
887 spin_unlock(&clp->cl_lock);
890 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
892 struct nfsd4_conn *conn;
894 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
897 svc_xprt_get(rqstp->rq_xprt);
898 conn->cn_xprt = rqstp->rq_xprt;
899 conn->cn_flags = flags;
900 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
904 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
906 conn->cn_session = ses;
907 list_add(&conn->cn_persession, &ses->se_conns);
910 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
912 struct nfs4_client *clp = ses->se_client;
914 spin_lock(&clp->cl_lock);
915 __nfsd4_hash_conn(conn, ses);
916 spin_unlock(&clp->cl_lock);
919 static int nfsd4_register_conn(struct nfsd4_conn *conn)
921 conn->cn_xpt_user.callback = nfsd4_conn_lost;
922 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
925 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
929 nfsd4_hash_conn(conn, ses);
930 ret = nfsd4_register_conn(conn);
932 /* oops; xprt is already down: */
933 nfsd4_conn_lost(&conn->cn_xpt_user);
934 if (conn->cn_flags & NFS4_CDFC4_BACK) {
935 /* callback channel may be back up */
936 nfsd4_probe_callback(ses->se_client);
940 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
942 u32 dir = NFS4_CDFC4_FORE;
944 if (cses->flags & SESSION4_BACK_CHAN)
945 dir |= NFS4_CDFC4_BACK;
946 return alloc_conn(rqstp, dir);
949 /* must be called under client_lock */
950 static void nfsd4_del_conns(struct nfsd4_session *s)
952 struct nfs4_client *clp = s->se_client;
953 struct nfsd4_conn *c;
955 spin_lock(&clp->cl_lock);
956 while (!list_empty(&s->se_conns)) {
957 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
958 list_del_init(&c->cn_persession);
959 spin_unlock(&clp->cl_lock);
961 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
964 spin_lock(&clp->cl_lock);
966 spin_unlock(&clp->cl_lock);
969 static void __free_session(struct nfsd4_session *ses)
971 free_session_slots(ses);
975 static void free_session(struct nfsd4_session *ses)
977 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
979 lockdep_assert_held(&nn->client_lock);
980 nfsd4_del_conns(ses);
981 nfsd4_put_drc_mem(&ses->se_fchannel);
985 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
988 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
990 new->se_client = clp;
993 INIT_LIST_HEAD(&new->se_conns);
995 new->se_cb_seq_nr = 1;
996 new->se_flags = cses->flags;
997 new->se_cb_prog = cses->callback_prog;
998 new->se_cb_sec = cses->cb_sec;
999 atomic_set(&new->se_ref, 0);
1000 idx = hash_sessionid(&new->se_sessionid);
1001 spin_lock(&nn->client_lock);
1002 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1003 spin_lock(&clp->cl_lock);
1004 list_add(&new->se_perclnt, &clp->cl_sessions);
1005 spin_unlock(&clp->cl_lock);
1006 spin_unlock(&nn->client_lock);
1007 memcpy(&new->se_fchannel, &cses->fore_channel,
1008 sizeof(struct nfsd4_channel_attrs));
1009 if (cses->flags & SESSION4_BACK_CHAN) {
1010 struct sockaddr *sa = svc_addr(rqstp);
1012 * This is a little silly; with sessions there's no real
1013 * use for the callback address. Use the peer address
1014 * as a reasonable default for now, but consider fixing
1015 * the rpc client not to require an address in the
1018 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1019 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1023 /* caller must hold client_lock */
1024 static struct nfsd4_session *
1025 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1027 struct nfsd4_session *elem;
1029 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1031 dump_sessionid(__func__, sessionid);
1032 idx = hash_sessionid(sessionid);
1033 /* Search in the appropriate list */
1034 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1035 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1036 NFS4_MAX_SESSIONID_LEN)) {
1041 dprintk("%s: session not found\n", __func__);
1045 /* caller must hold client_lock */
1047 unhash_session(struct nfsd4_session *ses)
1049 list_del(&ses->se_hash);
1050 spin_lock(&ses->se_client->cl_lock);
1051 list_del(&ses->se_perclnt);
1052 spin_unlock(&ses->se_client->cl_lock);
1055 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1057 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1059 if (clid->cl_boot == nn->boot_time)
1061 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1062 clid->cl_boot, clid->cl_id, nn->boot_time);
1067 * XXX Should we use a slab cache ?
1068 * This type of memory management is somewhat inefficient, but we use it
1069 * anyway since SETCLIENTID is not a common operation.
1071 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1073 struct nfs4_client *clp;
1075 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1078 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1079 if (clp->cl_name.data == NULL) {
1083 clp->cl_name.len = name.len;
1088 free_client(struct nfs4_client *clp)
1090 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1092 lockdep_assert_held(&nn->client_lock);
1093 while (!list_empty(&clp->cl_sessions)) {
1094 struct nfsd4_session *ses;
1095 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1097 list_del(&ses->se_perclnt);
1098 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1101 free_svc_cred(&clp->cl_cred);
1102 kfree(clp->cl_name.data);
1103 idr_destroy(&clp->cl_stateids);
1107 /* must be called under the client_lock */
1109 unhash_client_locked(struct nfs4_client *clp)
1111 struct nfsd4_session *ses;
1113 list_del(&clp->cl_lru);
1114 spin_lock(&clp->cl_lock);
1115 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1116 list_del_init(&ses->se_hash);
1117 spin_unlock(&clp->cl_lock);
1121 destroy_client(struct nfs4_client *clp)
1123 struct nfs4_openowner *oo;
1124 struct nfs4_delegation *dp;
1125 struct list_head reaplist;
1126 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1128 INIT_LIST_HEAD(&reaplist);
1129 spin_lock(&recall_lock);
1130 while (!list_empty(&clp->cl_delegations)) {
1131 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1132 list_del_init(&dp->dl_perclnt);
1133 list_move(&dp->dl_recall_lru, &reaplist);
1135 spin_unlock(&recall_lock);
1136 while (!list_empty(&reaplist)) {
1137 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1138 destroy_delegation(dp);
1140 while (!list_empty(&clp->cl_openowners)) {
1141 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1142 release_openowner(oo);
1144 nfsd4_shutdown_callback(clp);
1145 if (clp->cl_cb_conn.cb_xprt)
1146 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1147 list_del(&clp->cl_idhash);
1148 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1149 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1151 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1152 spin_lock(&nn->client_lock);
1153 unhash_client_locked(clp);
1154 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1156 spin_unlock(&nn->client_lock);
1159 static void expire_client(struct nfs4_client *clp)
1161 nfsd4_client_record_remove(clp);
1162 destroy_client(clp);
1165 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1167 memcpy(target->cl_verifier.data, source->data,
1168 sizeof(target->cl_verifier.data));
1171 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1173 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1174 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1177 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1179 if (source->cr_principal) {
1180 target->cr_principal =
1181 kstrdup(source->cr_principal, GFP_KERNEL);
1182 if (target->cr_principal == NULL)
1185 target->cr_principal = NULL;
1186 target->cr_flavor = source->cr_flavor;
1187 target->cr_uid = source->cr_uid;
1188 target->cr_gid = source->cr_gid;
1189 target->cr_group_info = source->cr_group_info;
1190 get_group_info(target->cr_group_info);
1191 target->cr_gss_mech = source->cr_gss_mech;
1192 if (source->cr_gss_mech)
1193 gss_mech_get(source->cr_gss_mech);
1198 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1202 res = o1->len - o2->len;
1205 return (long long)memcmp(o1->data, o2->data, o1->len);
1208 static int same_name(const char *n1, const char *n2)
1210 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1214 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1216 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1220 same_clid(clientid_t *cl1, clientid_t *cl2)
1222 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1225 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1229 if (g1->ngroups != g2->ngroups)
1231 for (i=0; i<g1->ngroups; i++)
1232 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1238 * RFC 3530 language requires clid_inuse be returned when the
1239 * "principal" associated with a requests differs from that previously
1240 * used. We use uid, gid's, and gss principal string as our best
1241 * approximation. We also don't want to allow non-gss use of a client
1242 * established using gss: in theory cr_principal should catch that
1243 * change, but in practice cr_principal can be null even in the gss case
1244 * since gssd doesn't always pass down a principal string.
1246 static bool is_gss_cred(struct svc_cred *cr)
1248 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1249 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1254 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1256 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1257 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1258 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1259 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1261 if (cr1->cr_principal == cr2->cr_principal)
1263 if (!cr1->cr_principal || !cr2->cr_principal)
1265 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1268 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1270 struct svc_cred *cr = &rqstp->rq_cred;
1273 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1274 return service == RPC_GSS_SVC_INTEGRITY ||
1275 service == RPC_GSS_SVC_PRIVACY;
1278 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1280 struct svc_cred *cr = &rqstp->rq_cred;
1282 if (!cl->cl_mach_cred)
1284 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1286 if (!svc_rqst_integrity_protected(rqstp))
1288 if (!cr->cr_principal)
1290 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1293 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1295 static u32 current_clientid = 1;
1297 clp->cl_clientid.cl_boot = nn->boot_time;
1298 clp->cl_clientid.cl_id = current_clientid++;
1301 static void gen_confirm(struct nfs4_client *clp)
1306 verf[0] = (__be32)get_seconds();
1307 verf[1] = (__be32)i++;
1308 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1311 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1313 struct nfs4_stid *ret;
1315 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1316 if (!ret || !ret->sc_type)
1321 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1323 struct nfs4_stid *s;
1325 s = find_stateid(cl, t);
1328 if (typemask & s->sc_type)
1333 static struct nfs4_client *create_client(struct xdr_netobj name,
1334 struct svc_rqst *rqstp, nfs4_verifier *verf)
1336 struct nfs4_client *clp;
1337 struct sockaddr *sa = svc_addr(rqstp);
1339 struct net *net = SVC_NET(rqstp);
1340 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1342 clp = alloc_client(name);
1346 INIT_LIST_HEAD(&clp->cl_sessions);
1347 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1349 spin_lock(&nn->client_lock);
1351 spin_unlock(&nn->client_lock);
1354 idr_init(&clp->cl_stateids);
1355 atomic_set(&clp->cl_refcount, 0);
1356 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1357 INIT_LIST_HEAD(&clp->cl_idhash);
1358 INIT_LIST_HEAD(&clp->cl_openowners);
1359 INIT_LIST_HEAD(&clp->cl_delegations);
1360 INIT_LIST_HEAD(&clp->cl_lru);
1361 INIT_LIST_HEAD(&clp->cl_callbacks);
1362 INIT_LIST_HEAD(&clp->cl_revoked);
1363 spin_lock_init(&clp->cl_lock);
1364 nfsd4_init_callback(&clp->cl_cb_null);
1365 clp->cl_time = get_seconds();
1366 clear_bit(0, &clp->cl_cb_slot_busy);
1367 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1368 copy_verf(clp, verf);
1369 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1371 clp->cl_cb_session = NULL;
1377 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1379 struct rb_node **new = &(root->rb_node), *parent = NULL;
1380 struct nfs4_client *clp;
1383 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1386 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1387 new = &((*new)->rb_left);
1389 new = &((*new)->rb_right);
1392 rb_link_node(&new_clp->cl_namenode, parent, new);
1393 rb_insert_color(&new_clp->cl_namenode, root);
1396 static struct nfs4_client *
1397 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1400 struct rb_node *node = root->rb_node;
1401 struct nfs4_client *clp;
1404 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1405 cmp = compare_blob(&clp->cl_name, name);
1407 node = node->rb_left;
1409 node = node->rb_right;
1417 add_to_unconfirmed(struct nfs4_client *clp)
1419 unsigned int idhashval;
1420 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1422 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1423 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1424 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1425 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1430 move_to_confirmed(struct nfs4_client *clp)
1432 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1433 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1435 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1436 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1437 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1438 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1439 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1443 static struct nfs4_client *
1444 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1446 struct nfs4_client *clp;
1447 unsigned int idhashval = clientid_hashval(clid->cl_id);
1449 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1450 if (same_clid(&clp->cl_clientid, clid)) {
1451 if ((bool)clp->cl_minorversion != sessions)
1460 static struct nfs4_client *
1461 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1463 struct list_head *tbl = nn->conf_id_hashtbl;
1465 return find_client_in_id_table(tbl, clid, sessions);
1468 static struct nfs4_client *
1469 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1471 struct list_head *tbl = nn->unconf_id_hashtbl;
1473 return find_client_in_id_table(tbl, clid, sessions);
1476 static bool clp_used_exchangeid(struct nfs4_client *clp)
1478 return clp->cl_exchange_flags != 0;
1481 static struct nfs4_client *
1482 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1484 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1487 static struct nfs4_client *
1488 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1490 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1494 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1496 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1497 struct sockaddr *sa = svc_addr(rqstp);
1498 u32 scopeid = rpc_get_scope_id(sa);
1499 unsigned short expected_family;
1501 /* Currently, we only support tcp and tcp6 for the callback channel */
1502 if (se->se_callback_netid_len == 3 &&
1503 !memcmp(se->se_callback_netid_val, "tcp", 3))
1504 expected_family = AF_INET;
1505 else if (se->se_callback_netid_len == 4 &&
1506 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1507 expected_family = AF_INET6;
1511 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1512 se->se_callback_addr_len,
1513 (struct sockaddr *)&conn->cb_addr,
1514 sizeof(conn->cb_addr));
1516 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1519 if (conn->cb_addr.ss_family == AF_INET6)
1520 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1522 conn->cb_prog = se->se_callback_prog;
1523 conn->cb_ident = se->se_callback_ident;
1524 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1527 conn->cb_addr.ss_family = AF_UNSPEC;
1528 conn->cb_addrlen = 0;
1529 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1530 "will not receive delegations\n",
1531 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1537 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1540 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1542 struct nfsd4_slot *slot = resp->cstate.slot;
1545 dprintk("--> %s slot %p\n", __func__, slot);
1547 slot->sl_opcnt = resp->opcnt;
1548 slot->sl_status = resp->cstate.status;
1550 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1551 if (nfsd4_not_cached(resp)) {
1552 slot->sl_datalen = 0;
1555 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1556 base = (char *)resp->cstate.datap -
1557 (char *)resp->xbuf->head[0].iov_base;
1558 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1560 WARN("%s: sessions DRC could not cache compound\n", __func__);
1565 * Encode the replay sequence operation from the slot values.
1566 * If cachethis is FALSE encode the uncached rep error on the next
1567 * operation which sets resp->p and increments resp->opcnt for
1568 * nfs4svc_encode_compoundres.
1572 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1573 struct nfsd4_compoundres *resp)
1575 struct nfsd4_op *op;
1576 struct nfsd4_slot *slot = resp->cstate.slot;
1578 /* Encode the replayed sequence operation */
1579 op = &args->ops[resp->opcnt - 1];
1580 nfsd4_encode_operation(resp, op);
1582 /* Return nfserr_retry_uncached_rep in next operation. */
1583 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1584 op = &args->ops[resp->opcnt++];
1585 op->status = nfserr_retry_uncached_rep;
1586 nfsd4_encode_operation(resp, op);
1592 * The sequence operation is not cached because we can use the slot and
1596 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1597 struct nfsd4_sequence *seq)
1599 struct nfsd4_slot *slot = resp->cstate.slot;
1602 dprintk("--> %s slot %p\n", __func__, slot);
1604 /* Either returns 0 or nfserr_retry_uncached */
1605 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1606 if (status == nfserr_retry_uncached_rep)
1609 /* The sequence operation has been encoded, cstate->datap set. */
1610 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1612 resp->opcnt = slot->sl_opcnt;
1613 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1614 status = slot->sl_status;
1620 * Set the exchange_id flags returned by the server.
1623 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1625 /* pNFS is not supported */
1626 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1628 /* Referrals are supported, Migration is not. */
1629 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1631 /* set the wire flags to return to client. */
1632 clid->flags = new->cl_exchange_flags;
1635 static bool client_has_state(struct nfs4_client *clp)
1638 * Note clp->cl_openowners check isn't quite right: there's no
1639 * need to count owners without stateid's.
1641 * Also note we should probably be using this in 4.0 case too.
1643 return !list_empty(&clp->cl_openowners)
1644 || !list_empty(&clp->cl_delegations)
1645 || !list_empty(&clp->cl_sessions);
1649 nfsd4_exchange_id(struct svc_rqst *rqstp,
1650 struct nfsd4_compound_state *cstate,
1651 struct nfsd4_exchange_id *exid)
1653 struct nfs4_client *unconf, *conf, *new;
1655 char addr_str[INET6_ADDRSTRLEN];
1656 nfs4_verifier verf = exid->verifier;
1657 struct sockaddr *sa = svc_addr(rqstp);
1658 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1659 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1661 rpc_ntop(sa, addr_str, sizeof(addr_str));
1662 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1663 "ip_addr=%s flags %x, spa_how %d\n",
1664 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1665 addr_str, exid->flags, exid->spa_how);
1667 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1668 return nfserr_inval;
1670 switch (exid->spa_how) {
1672 if (!svc_rqst_integrity_protected(rqstp))
1673 return nfserr_inval;
1676 default: /* checked by xdr code */
1679 return nfserr_encr_alg_unsupp;
1682 /* Cases below refer to rfc 5661 section 18.35.4: */
1684 conf = find_confirmed_client_by_name(&exid->clname, nn);
1686 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1687 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1690 if (!clp_used_exchangeid(conf)) { /* buggy client */
1691 status = nfserr_inval;
1694 if (!mach_creds_match(conf, rqstp)) {
1695 status = nfserr_wrong_cred;
1698 if (!creds_match) { /* case 9 */
1699 status = nfserr_perm;
1702 if (!verfs_match) { /* case 8 */
1703 status = nfserr_not_same;
1707 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1711 if (!creds_match) { /* case 3 */
1712 if (client_has_state(conf)) {
1713 status = nfserr_clid_inuse;
1716 expire_client(conf);
1719 if (verfs_match) { /* case 2 */
1720 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1724 /* case 5, client reboot */
1728 if (update) { /* case 7 */
1729 status = nfserr_noent;
1733 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1734 if (unconf) /* case 4, possible retry or client restart */
1735 expire_client(unconf);
1737 /* case 1 (normal case) */
1739 new = create_client(exid->clname, rqstp, &verf);
1741 status = nfserr_jukebox;
1744 new->cl_minorversion = cstate->minorversion;
1745 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
1748 add_to_unconfirmed(new);
1750 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1751 exid->clientid.cl_id = new->cl_clientid.cl_id;
1753 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1754 nfsd4_set_ex_flags(new, exid);
1756 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1757 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1761 nfs4_unlock_state();
1766 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1768 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1771 /* The slot is in use, and no response has been sent. */
1773 if (seqid == slot_seqid)
1774 return nfserr_jukebox;
1776 return nfserr_seq_misordered;
1778 /* Note unsigned 32-bit arithmetic handles wraparound: */
1779 if (likely(seqid == slot_seqid + 1))
1781 if (seqid == slot_seqid)
1782 return nfserr_replay_cache;
1783 return nfserr_seq_misordered;
1787 * Cache the create session result into the create session single DRC
1788 * slot cache by saving the xdr structure. sl_seqid has been set.
1789 * Do this for solo or embedded create session operations.
1792 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1793 struct nfsd4_clid_slot *slot, __be32 nfserr)
1795 slot->sl_status = nfserr;
1796 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1800 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1801 struct nfsd4_clid_slot *slot)
1803 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1804 return slot->sl_status;
1807 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1808 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1809 1 + /* MIN tag is length with zero, only length */ \
1810 3 + /* version, opcount, opcode */ \
1811 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1812 /* seqid, slotID, slotID, cache */ \
1813 4 ) * sizeof(__be32))
1815 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1816 2 + /* verifier: AUTH_NULL, length 0 */\
1818 1 + /* MIN tag is length with zero, only length */ \
1819 3 + /* opcount, opcode, opstatus*/ \
1820 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1821 /* seqid, slotID, slotID, slotID, status */ \
1822 5 ) * sizeof(__be32))
1824 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1826 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
1828 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
1829 return nfserr_toosmall;
1830 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
1831 return nfserr_toosmall;
1832 ca->headerpadsz = 0;
1833 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
1834 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
1835 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
1836 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
1837 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
1838 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
1840 * Note decreasing slot size below client's request may make it
1841 * difficult for client to function correctly, whereas
1842 * decreasing the number of slots will (just?) affect
1843 * performance. When short on memory we therefore prefer to
1844 * decrease number of slots instead of their size. Clients that
1845 * request larger slots than they need will get poor results:
1847 ca->maxreqs = nfsd4_get_drc_mem(ca);
1849 return nfserr_jukebox;
1854 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
1856 ca->headerpadsz = 0;
1859 * These RPC_MAX_HEADER macros are overkill, especially since we
1860 * don't even do gss on the backchannel yet. But this is still
1861 * less than 1k. Tighten up this estimate in the unlikely event
1862 * it turns out to be a problem for some client:
1864 if (ca->maxreq_sz < NFS4_enc_cb_recall_sz + RPC_MAX_HEADER_WITH_AUTH)
1865 return nfserr_toosmall;
1866 if (ca->maxresp_sz < NFS4_dec_cb_recall_sz + RPC_MAX_REPHEADER_WITH_AUTH)
1867 return nfserr_toosmall;
1868 ca->maxresp_cached = 0;
1870 return nfserr_toosmall;
1875 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
1877 switch (cbs->flavor) {
1883 * GSS case: the spec doesn't allow us to return this
1884 * error. But it also doesn't allow us not to support
1886 * I'd rather this fail hard than return some error the
1887 * client might think it can already handle:
1889 return nfserr_encr_alg_unsupp;
1894 nfsd4_create_session(struct svc_rqst *rqstp,
1895 struct nfsd4_compound_state *cstate,
1896 struct nfsd4_create_session *cr_ses)
1898 struct sockaddr *sa = svc_addr(rqstp);
1899 struct nfs4_client *conf, *unconf;
1900 struct nfsd4_session *new;
1901 struct nfsd4_conn *conn;
1902 struct nfsd4_clid_slot *cs_slot = NULL;
1904 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1906 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1907 return nfserr_inval;
1908 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
1911 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
1914 status = check_backchannel_attrs(&cr_ses->back_channel);
1917 status = nfserr_jukebox;
1918 new = alloc_session(&cr_ses->fore_channel);
1920 goto out_release_drc_mem;
1921 conn = alloc_conn_from_crses(rqstp, cr_ses);
1923 goto out_free_session;
1926 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1927 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1928 WARN_ON_ONCE(conf && unconf);
1931 status = nfserr_wrong_cred;
1932 if (!mach_creds_match(conf, rqstp))
1934 cs_slot = &conf->cl_cs_slot;
1935 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1936 if (status == nfserr_replay_cache) {
1937 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1939 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1940 status = nfserr_seq_misordered;
1943 } else if (unconf) {
1944 struct nfs4_client *old;
1945 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1946 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1947 status = nfserr_clid_inuse;
1950 status = nfserr_wrong_cred;
1951 if (!mach_creds_match(unconf, rqstp))
1953 cs_slot = &unconf->cl_cs_slot;
1954 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1956 /* an unconfirmed replay returns misordered */
1957 status = nfserr_seq_misordered;
1960 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1962 status = mark_client_expired(old);
1967 move_to_confirmed(unconf);
1970 status = nfserr_stale_clientid;
1975 * We do not support RDMA or persistent sessions
1977 cr_ses->flags &= ~SESSION4_PERSIST;
1978 cr_ses->flags &= ~SESSION4_RDMA;
1980 init_session(rqstp, new, conf, cr_ses);
1981 nfsd4_init_conn(rqstp, conn, new);
1983 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1984 NFS4_MAX_SESSIONID_LEN);
1985 cs_slot->sl_seqid++;
1986 cr_ses->seqid = cs_slot->sl_seqid;
1988 /* cache solo and embedded create sessions under the state lock */
1989 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1990 nfs4_unlock_state();
1993 nfs4_unlock_state();
1996 __free_session(new);
1997 out_release_drc_mem:
1998 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2002 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2005 case NFS4_CDFC4_FORE:
2006 case NFS4_CDFC4_BACK:
2008 case NFS4_CDFC4_FORE_OR_BOTH:
2009 case NFS4_CDFC4_BACK_OR_BOTH:
2010 *dir = NFS4_CDFC4_BOTH;
2013 return nfserr_inval;
2016 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2018 struct nfsd4_session *session = cstate->session;
2019 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2022 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2025 spin_lock(&nn->client_lock);
2026 session->se_cb_prog = bc->bc_cb_program;
2027 session->se_cb_sec = bc->bc_cb_sec;
2028 spin_unlock(&nn->client_lock);
2030 nfsd4_probe_callback(session->se_client);
2035 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2036 struct nfsd4_compound_state *cstate,
2037 struct nfsd4_bind_conn_to_session *bcts)
2040 struct nfsd4_conn *conn;
2041 struct nfsd4_session *session;
2042 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2044 if (!nfsd4_last_compound_op(rqstp))
2045 return nfserr_not_only_op;
2047 spin_lock(&nn->client_lock);
2048 session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
2049 spin_unlock(&nn->client_lock);
2050 status = nfserr_badsession;
2053 status = nfserr_wrong_cred;
2054 if (!mach_creds_match(session->se_client, rqstp))
2056 status = nfsd4_map_bcts_dir(&bcts->dir);
2059 conn = alloc_conn(rqstp, bcts->dir);
2060 status = nfserr_jukebox;
2063 nfsd4_init_conn(rqstp, conn, session);
2066 nfs4_unlock_state();
2070 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2074 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2078 nfsd4_destroy_session(struct svc_rqst *r,
2079 struct nfsd4_compound_state *cstate,
2080 struct nfsd4_destroy_session *sessionid)
2082 struct nfsd4_session *ses;
2084 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
2087 status = nfserr_not_only_op;
2088 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2089 if (!nfsd4_last_compound_op(r))
2092 dump_sessionid(__func__, &sessionid->sessionid);
2093 spin_lock(&nn->client_lock);
2094 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
2095 status = nfserr_badsession;
2097 goto out_client_lock;
2098 status = nfserr_wrong_cred;
2099 if (!mach_creds_match(ses->se_client, r))
2100 goto out_client_lock;
2101 status = mark_session_dead_locked(ses);
2103 goto out_client_lock;
2104 unhash_session(ses);
2105 spin_unlock(&nn->client_lock);
2107 nfsd4_probe_callback_sync(ses->se_client);
2109 spin_lock(&nn->client_lock);
2113 spin_unlock(&nn->client_lock);
2115 nfs4_unlock_state();
2119 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2121 struct nfsd4_conn *c;
2123 list_for_each_entry(c, &s->se_conns, cn_persession) {
2124 if (c->cn_xprt == xpt) {
2131 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2133 struct nfs4_client *clp = ses->se_client;
2134 struct nfsd4_conn *c;
2135 __be32 status = nfs_ok;
2138 spin_lock(&clp->cl_lock);
2139 c = __nfsd4_find_conn(new->cn_xprt, ses);
2142 status = nfserr_conn_not_bound_to_session;
2143 if (clp->cl_mach_cred)
2145 __nfsd4_hash_conn(new, ses);
2146 spin_unlock(&clp->cl_lock);
2147 ret = nfsd4_register_conn(new);
2149 /* oops; xprt is already down: */
2150 nfsd4_conn_lost(&new->cn_xpt_user);
2153 spin_unlock(&clp->cl_lock);
2158 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2160 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2162 return args->opcnt > session->se_fchannel.maxops;
2165 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2166 struct nfsd4_session *session)
2168 struct xdr_buf *xb = &rqstp->rq_arg;
2170 return xb->len > session->se_fchannel.maxreq_sz;
2174 nfsd4_sequence(struct svc_rqst *rqstp,
2175 struct nfsd4_compound_state *cstate,
2176 struct nfsd4_sequence *seq)
2178 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2179 struct nfsd4_session *session;
2180 struct nfs4_client *clp;
2181 struct nfsd4_slot *slot;
2182 struct nfsd4_conn *conn;
2184 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2186 if (resp->opcnt != 1)
2187 return nfserr_sequence_pos;
2190 * Will be either used or freed by nfsd4_sequence_check_conn
2193 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2195 return nfserr_jukebox;
2197 spin_lock(&nn->client_lock);
2198 status = nfserr_badsession;
2199 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2201 goto out_no_session;
2202 clp = session->se_client;
2203 status = get_client_locked(clp);
2205 goto out_no_session;
2206 status = nfsd4_get_session_locked(session);
2208 goto out_put_client;
2210 status = nfserr_too_many_ops;
2211 if (nfsd4_session_too_many_ops(rqstp, session))
2212 goto out_put_session;
2214 status = nfserr_req_too_big;
2215 if (nfsd4_request_too_big(rqstp, session))
2216 goto out_put_session;
2218 status = nfserr_badslot;
2219 if (seq->slotid >= session->se_fchannel.maxreqs)
2220 goto out_put_session;
2222 slot = session->se_slots[seq->slotid];
2223 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2225 /* We do not negotiate the number of slots yet, so set the
2226 * maxslots to the session maxreqs which is used to encode
2227 * sr_highest_slotid and the sr_target_slot id to maxslots */
2228 seq->maxslots = session->se_fchannel.maxreqs;
2230 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2231 slot->sl_flags & NFSD4_SLOT_INUSE);
2232 if (status == nfserr_replay_cache) {
2233 status = nfserr_seq_misordered;
2234 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2235 goto out_put_session;
2236 cstate->slot = slot;
2237 cstate->session = session;
2238 /* Return the cached reply status and set cstate->status
2239 * for nfsd4_proc_compound processing */
2240 status = nfsd4_replay_cache_entry(resp, seq);
2241 cstate->status = nfserr_replay_cache;
2245 goto out_put_session;
2247 status = nfsd4_sequence_check_conn(conn, session);
2250 goto out_put_session;
2252 /* Success! bump slot seqid */
2253 slot->sl_seqid = seq->seqid;
2254 slot->sl_flags |= NFSD4_SLOT_INUSE;
2256 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2258 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2260 cstate->slot = slot;
2261 cstate->session = session;
2264 switch (clp->cl_cb_state) {
2266 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2268 case NFSD4_CB_FAULT:
2269 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2272 seq->status_flags = 0;
2274 if (!list_empty(&clp->cl_revoked))
2275 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2278 spin_unlock(&nn->client_lock);
2281 nfsd4_put_session(session);
2283 put_client_renew_locked(clp);
2284 goto out_no_session;
2288 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2290 struct nfs4_client *conf, *unconf, *clp;
2292 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2295 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2296 conf = find_confirmed_client(&dc->clientid, true, nn);
2297 WARN_ON_ONCE(conf && unconf);
2302 if (client_has_state(conf)) {
2303 status = nfserr_clientid_busy;
2309 status = nfserr_stale_clientid;
2312 if (!mach_creds_match(clp, rqstp)) {
2313 status = nfserr_wrong_cred;
2318 nfs4_unlock_state();
2323 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2327 if (rc->rca_one_fs) {
2328 if (!cstate->current_fh.fh_dentry)
2329 return nfserr_nofilehandle;
2331 * We don't take advantage of the rca_one_fs case.
2332 * That's OK, it's optional, we can safely ignore it.
2338 status = nfserr_complete_already;
2339 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2340 &cstate->session->se_client->cl_flags))
2343 status = nfserr_stale_clientid;
2344 if (is_client_expired(cstate->session->se_client))
2346 * The following error isn't really legal.
2347 * But we only get here if the client just explicitly
2348 * destroyed the client. Surely it no longer cares what
2349 * error it gets back on an operation for the dead
2355 nfsd4_client_record_create(cstate->session->se_client);
2357 nfs4_unlock_state();
2362 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2363 struct nfsd4_setclientid *setclid)
2365 struct xdr_netobj clname = setclid->se_name;
2366 nfs4_verifier clverifier = setclid->se_verf;
2367 struct nfs4_client *conf, *unconf, *new;
2369 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2371 /* Cases below refer to rfc 3530 section 14.2.33: */
2373 conf = find_confirmed_client_by_name(&clname, nn);
2376 status = nfserr_clid_inuse;
2377 if (clp_used_exchangeid(conf))
2379 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2380 char addr_str[INET6_ADDRSTRLEN];
2381 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2383 dprintk("NFSD: setclientid: string in use by client "
2384 "at %s\n", addr_str);
2388 unconf = find_unconfirmed_client_by_name(&clname, nn);
2390 expire_client(unconf);
2391 status = nfserr_jukebox;
2392 new = create_client(clname, rqstp, &clverifier);
2395 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2396 /* case 1: probable callback update */
2397 copy_clid(new, conf);
2398 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2400 new->cl_minorversion = 0;
2401 gen_callback(new, setclid, rqstp);
2402 add_to_unconfirmed(new);
2403 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2404 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2405 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2408 nfs4_unlock_state();
2414 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2415 struct nfsd4_compound_state *cstate,
2416 struct nfsd4_setclientid_confirm *setclientid_confirm)
2418 struct nfs4_client *conf, *unconf;
2419 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2420 clientid_t * clid = &setclientid_confirm->sc_clientid;
2422 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2424 if (STALE_CLIENTID(clid, nn))
2425 return nfserr_stale_clientid;
2428 conf = find_confirmed_client(clid, false, nn);
2429 unconf = find_unconfirmed_client(clid, false, nn);
2431 * We try hard to give out unique clientid's, so if we get an
2432 * attempt to confirm the same clientid with a different cred,
2433 * there's a bug somewhere. Let's charitably assume it's our
2436 status = nfserr_serverfault;
2437 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2439 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2441 /* cases below refer to rfc 3530 section 14.2.34: */
2442 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2443 if (conf && !unconf) /* case 2: probable retransmit */
2445 else /* case 4: client hasn't noticed we rebooted yet? */
2446 status = nfserr_stale_clientid;
2450 if (conf) { /* case 1: callback update */
2451 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2452 nfsd4_probe_callback(conf);
2453 expire_client(unconf);
2454 } else { /* case 3: normal case; new or rebooted client */
2455 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2457 status = mark_client_expired(conf);
2460 expire_client(conf);
2462 move_to_confirmed(unconf);
2463 nfsd4_probe_callback(unconf);
2466 nfs4_unlock_state();
2470 static struct nfs4_file *nfsd4_alloc_file(void)
2472 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2475 /* OPEN Share state helper functions */
2476 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2478 unsigned int hashval = file_hashval(ino);
2480 atomic_set(&fp->fi_ref, 1);
2481 INIT_LIST_HEAD(&fp->fi_stateids);
2482 INIT_LIST_HEAD(&fp->fi_delegations);
2483 fp->fi_inode = igrab(ino);
2484 fp->fi_had_conflict = false;
2485 fp->fi_lease = NULL;
2486 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2487 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2488 spin_lock(&recall_lock);
2489 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2490 spin_unlock(&recall_lock);
2494 nfsd4_free_slab(struct kmem_cache **slab)
2498 kmem_cache_destroy(*slab);
2503 nfsd4_free_slabs(void)
2505 nfsd4_free_slab(&openowner_slab);
2506 nfsd4_free_slab(&lockowner_slab);
2507 nfsd4_free_slab(&file_slab);
2508 nfsd4_free_slab(&stateid_slab);
2509 nfsd4_free_slab(&deleg_slab);
2513 nfsd4_init_slabs(void)
2515 openowner_slab = kmem_cache_create("nfsd4_openowners",
2516 sizeof(struct nfs4_openowner), 0, 0, NULL);
2517 if (openowner_slab == NULL)
2519 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2520 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2521 if (lockowner_slab == NULL)
2523 file_slab = kmem_cache_create("nfsd4_files",
2524 sizeof(struct nfs4_file), 0, 0, NULL);
2525 if (file_slab == NULL)
2527 stateid_slab = kmem_cache_create("nfsd4_stateids",
2528 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2529 if (stateid_slab == NULL)
2531 deleg_slab = kmem_cache_create("nfsd4_delegations",
2532 sizeof(struct nfs4_delegation), 0, 0, NULL);
2533 if (deleg_slab == NULL)
2538 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2542 void nfs4_free_openowner(struct nfs4_openowner *oo)
2544 kfree(oo->oo_owner.so_owner.data);
2545 kmem_cache_free(openowner_slab, oo);
2548 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2550 kfree(lo->lo_owner.so_owner.data);
2551 kmem_cache_free(lockowner_slab, lo);
2554 static void init_nfs4_replay(struct nfs4_replay *rp)
2556 rp->rp_status = nfserr_serverfault;
2558 rp->rp_buf = rp->rp_ibuf;
2561 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2563 struct nfs4_stateowner *sop;
2565 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2569 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2570 if (!sop->so_owner.data) {
2571 kmem_cache_free(slab, sop);
2574 sop->so_owner.len = owner->len;
2576 INIT_LIST_HEAD(&sop->so_stateids);
2577 sop->so_client = clp;
2578 init_nfs4_replay(&sop->so_replay);
2582 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2584 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2586 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2587 list_add(&oo->oo_perclient, &clp->cl_openowners);
2590 static struct nfs4_openowner *
2591 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2592 struct nfs4_openowner *oo;
2594 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2597 oo->oo_owner.so_is_open_owner = 1;
2598 oo->oo_owner.so_seqid = open->op_seqid;
2599 oo->oo_flags = NFS4_OO_NEW;
2601 oo->oo_last_closed_stid = NULL;
2602 INIT_LIST_HEAD(&oo->oo_close_lru);
2603 hash_openowner(oo, clp, strhashval);
2607 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2608 struct nfs4_openowner *oo = open->op_openowner;
2610 stp->st_stid.sc_type = NFS4_OPEN_STID;
2611 INIT_LIST_HEAD(&stp->st_lockowners);
2612 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2613 list_add(&stp->st_perfile, &fp->fi_stateids);
2614 stp->st_stateowner = &oo->oo_owner;
2617 stp->st_access_bmap = 0;
2618 stp->st_deny_bmap = 0;
2619 set_access(open->op_share_access, stp);
2620 set_deny(open->op_share_deny, stp);
2621 stp->st_openstp = NULL;
2625 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2627 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2629 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2631 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2632 oo->oo_time = get_seconds();
2636 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2639 return (sop->so_owner.len == owner->len) &&
2640 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2641 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2644 static struct nfs4_openowner *
2645 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2646 bool sessions, struct nfsd_net *nn)
2648 struct nfs4_stateowner *so;
2649 struct nfs4_openowner *oo;
2650 struct nfs4_client *clp;
2652 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2653 if (!so->so_is_open_owner)
2655 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2657 clp = oo->oo_owner.so_client;
2658 if ((bool)clp->cl_minorversion != sessions)
2660 renew_client(oo->oo_owner.so_client);
2667 /* search file_hashtbl[] for file */
2668 static struct nfs4_file *
2669 find_file(struct inode *ino)
2671 unsigned int hashval = file_hashval(ino);
2672 struct nfs4_file *fp;
2674 spin_lock(&recall_lock);
2675 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2676 if (fp->fi_inode == ino) {
2678 spin_unlock(&recall_lock);
2682 spin_unlock(&recall_lock);
2687 * Called to check deny when READ with all zero stateid or
2688 * WRITE with all zero or all one stateid
2691 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2693 struct inode *ino = current_fh->fh_dentry->d_inode;
2694 struct nfs4_file *fp;
2695 struct nfs4_ol_stateid *stp;
2698 fp = find_file(ino);
2701 ret = nfserr_locked;
2702 /* Search for conflicting share reservations */
2703 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2704 if (test_deny(deny_type, stp) ||
2705 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2714 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2716 struct nfs4_client *clp = dp->dl_stid.sc_client;
2717 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2719 /* We're assuming the state code never drops its reference
2720 * without first removing the lease. Since we're in this lease
2721 * callback (and since the lease code is serialized by the kernel
2722 * lock) we know the server hasn't removed the lease yet, we know
2723 * it's safe to take a reference: */
2724 atomic_inc(&dp->dl_count);
2726 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
2728 /* only place dl_time is set. protected by lock_flocks*/
2729 dp->dl_time = get_seconds();
2731 nfsd4_cb_recall(dp);
2734 /* Called from break_lease() with lock_flocks() held. */
2735 static void nfsd_break_deleg_cb(struct file_lock *fl)
2737 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2738 struct nfs4_delegation *dp;
2741 WARN(1, "(%p)->fl_owner NULL\n", fl);
2744 if (fp->fi_had_conflict) {
2745 WARN(1, "duplicate break on %p\n", fp);
2749 * We don't want the locks code to timeout the lease for us;
2750 * we'll remove it ourself if a delegation isn't returned
2753 fl->fl_break_time = 0;
2755 spin_lock(&recall_lock);
2756 fp->fi_had_conflict = true;
2757 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2758 nfsd_break_one_deleg(dp);
2759 spin_unlock(&recall_lock);
2763 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2766 return lease_modify(onlist, arg);
2771 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2772 .lm_break = nfsd_break_deleg_cb,
2773 .lm_change = nfsd_change_deleg_cb,
2776 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2778 if (nfsd4_has_session(cstate))
2780 if (seqid == so->so_seqid - 1)
2781 return nfserr_replay_me;
2782 if (seqid == so->so_seqid)
2784 return nfserr_bad_seqid;
2788 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2789 struct nfsd4_open *open, struct nfsd_net *nn)
2791 clientid_t *clientid = &open->op_clientid;
2792 struct nfs4_client *clp = NULL;
2793 unsigned int strhashval;
2794 struct nfs4_openowner *oo = NULL;
2797 if (STALE_CLIENTID(&open->op_clientid, nn))
2798 return nfserr_stale_clientid;
2800 * In case we need it later, after we've already created the
2801 * file and don't want to risk a further failure:
2803 open->op_file = nfsd4_alloc_file();
2804 if (open->op_file == NULL)
2805 return nfserr_jukebox;
2807 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2808 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2809 open->op_openowner = oo;
2811 clp = find_confirmed_client(clientid, cstate->minorversion,
2814 return nfserr_expired;
2817 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2818 /* Replace unconfirmed owners without checking for replay. */
2819 clp = oo->oo_owner.so_client;
2820 release_openowner(oo);
2821 open->op_openowner = NULL;
2824 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2827 clp = oo->oo_owner.so_client;
2830 oo = alloc_init_open_stateowner(strhashval, clp, open);
2832 return nfserr_jukebox;
2833 open->op_openowner = oo;
2835 open->op_stp = nfs4_alloc_stateid(clp);
2837 return nfserr_jukebox;
2841 static inline __be32
2842 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2844 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2845 return nfserr_openmode;
2850 static int share_access_to_flags(u32 share_access)
2852 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2855 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2857 struct nfs4_stid *ret;
2859 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2862 return delegstateid(ret);
2865 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2867 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2868 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2872 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
2873 struct nfs4_delegation **dp)
2876 __be32 status = nfserr_bad_stateid;
2878 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2881 flags = share_access_to_flags(open->op_share_access);
2882 status = nfs4_check_delegmode(*dp, flags);
2886 if (!nfsd4_is_deleg_cur(open))
2890 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2895 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2897 struct nfs4_ol_stateid *local;
2898 struct nfs4_openowner *oo = open->op_openowner;
2900 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2901 /* ignore lock owners */
2902 if (local->st_stateowner->so_is_open_owner == 0)
2904 /* remember if we have seen this open owner */
2905 if (local->st_stateowner == &oo->oo_owner)
2907 /* check for conflicting share reservations */
2908 if (!test_share(local, open))
2909 return nfserr_share_denied;
2914 static inline int nfs4_access_to_access(u32 nfs4_access)
2918 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2919 flags |= NFSD_MAY_READ;
2920 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2921 flags |= NFSD_MAY_WRITE;
2925 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2926 struct svc_fh *cur_fh, struct nfsd4_open *open)
2929 int oflag = nfs4_access_to_omode(open->op_share_access);
2930 int access = nfs4_access_to_access(open->op_share_access);
2932 if (!fp->fi_fds[oflag]) {
2933 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2934 &fp->fi_fds[oflag]);
2938 nfs4_file_get_access(fp, oflag);
2943 static inline __be32
2944 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2945 struct nfsd4_open *open)
2947 struct iattr iattr = {
2948 .ia_valid = ATTR_SIZE,
2951 if (!open->op_truncate)
2953 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2954 return nfserr_inval;
2955 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2959 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)
2961 u32 op_share_access = open->op_share_access;
2965 new_access = !test_access(op_share_access, stp);
2967 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2971 status = nfsd4_truncate(rqstp, cur_fh, open);
2974 int oflag = nfs4_access_to_omode(op_share_access);
2975 nfs4_file_put_access(fp, oflag);
2979 /* remember the open */
2980 set_access(op_share_access, stp);
2981 set_deny(open->op_share_deny, stp);
2988 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2990 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2993 /* Should we give out recallable state?: */
2994 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2996 if (clp->cl_cb_state == NFSD4_CB_UP)
2999 * In the sessions case, since we don't have to establish a
3000 * separate connection for callbacks, we assume it's OK
3001 * until we hear otherwise:
3003 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3006 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
3008 struct file_lock *fl;
3010 fl = locks_alloc_lock();
3013 locks_init_lock(fl);
3014 fl->fl_lmops = &nfsd_lease_mng_ops;
3015 fl->fl_flags = FL_LEASE;
3016 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3017 fl->fl_end = OFFSET_MAX;
3018 fl->fl_owner = (fl_owner_t)(dp->dl_file);
3019 fl->fl_pid = current->tgid;
3023 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
3025 struct nfs4_file *fp = dp->dl_file;
3026 struct file_lock *fl;
3029 fl = nfs4_alloc_init_lease(dp, flag);
3032 fl->fl_file = find_readable_file(fp);
3033 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
3034 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
3036 list_del_init(&dp->dl_perclnt);
3037 locks_free_lock(fl);
3041 fp->fi_deleg_file = get_file(fl->fl_file);
3042 atomic_set(&fp->fi_delegees, 1);
3043 list_add(&dp->dl_perfile, &fp->fi_delegations);
3047 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
3049 struct nfs4_file *fp = dp->dl_file;
3052 return nfs4_setlease(dp, flag);
3053 spin_lock(&recall_lock);
3054 if (fp->fi_had_conflict) {
3055 spin_unlock(&recall_lock);
3058 atomic_inc(&fp->fi_delegees);
3059 list_add(&dp->dl_perfile, &fp->fi_delegations);
3060 spin_unlock(&recall_lock);
3061 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
3065 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3067 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3068 if (status == -EAGAIN)
3069 open->op_why_no_deleg = WND4_CONTENTION;
3071 open->op_why_no_deleg = WND4_RESOURCE;
3072 switch (open->op_deleg_want) {
3073 case NFS4_SHARE_WANT_READ_DELEG:
3074 case NFS4_SHARE_WANT_WRITE_DELEG:
3075 case NFS4_SHARE_WANT_ANY_DELEG:
3077 case NFS4_SHARE_WANT_CANCEL:
3078 open->op_why_no_deleg = WND4_CANCELLED;
3080 case NFS4_SHARE_WANT_NO_DELEG:
3087 * Attempt to hand out a delegation.
3090 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
3091 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
3093 struct nfs4_delegation *dp;
3094 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
3096 int status = 0, flag = 0;
3098 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3099 flag = NFS4_OPEN_DELEGATE_NONE;
3100 open->op_recall = 0;
3101 switch (open->op_claim_type) {
3102 case NFS4_OPEN_CLAIM_PREVIOUS:
3104 open->op_recall = 1;
3105 flag = open->op_delegate_type;
3106 if (flag == NFS4_OPEN_DELEGATE_NONE)
3109 case NFS4_OPEN_CLAIM_NULL:
3110 /* Let's not give out any delegations till everyone's
3111 * had the chance to reclaim theirs.... */
3112 if (locks_in_grace(net))
3114 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3116 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3117 flag = NFS4_OPEN_DELEGATE_WRITE;
3119 flag = NFS4_OPEN_DELEGATE_READ;
3125 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
3128 status = nfs4_set_delegation(dp, flag);
3132 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3134 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3135 STATEID_VAL(&dp->dl_stid.sc_stateid));
3137 open->op_delegate_type = flag;
3138 if (flag == NFS4_OPEN_DELEGATE_NONE) {
3139 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3140 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
3141 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3143 /* 4.1 client asking for a delegation? */
3144 if (open->op_deleg_want)
3145 nfsd4_open_deleg_none_ext(open, status);
3149 unhash_stid(&dp->dl_stid);
3150 nfs4_put_delegation(dp);
3152 flag = NFS4_OPEN_DELEGATE_NONE;
3156 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3157 struct nfs4_delegation *dp)
3159 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3160 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3161 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3162 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3163 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3164 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3165 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3166 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3168 /* Otherwise the client must be confused wanting a delegation
3169 * it already has, therefore we don't return
3170 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3175 * called with nfs4_lock_state() held.
3178 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3180 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3181 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3182 struct nfs4_file *fp = NULL;
3183 struct inode *ino = current_fh->fh_dentry->d_inode;
3184 struct nfs4_ol_stateid *stp = NULL;
3185 struct nfs4_delegation *dp = NULL;
3189 * Lookup file; if found, lookup stateid and check open request,
3190 * and check for delegations in the process of being recalled.
3191 * If not found, create the nfs4_file struct
3193 fp = find_file(ino);
3195 if ((status = nfs4_check_open(fp, open, &stp)))
3197 status = nfs4_check_deleg(cl, open, &dp);
3201 status = nfserr_bad_stateid;
3202 if (nfsd4_is_deleg_cur(open))
3204 status = nfserr_jukebox;
3206 open->op_file = NULL;
3207 nfsd4_init_file(fp, ino);
3211 * OPEN the file, or upgrade an existing OPEN.
3212 * If truncate fails, the OPEN fails.
3215 /* Stateid was found, this is an OPEN upgrade */
3216 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3220 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3223 status = nfsd4_truncate(rqstp, current_fh, open);
3227 open->op_stp = NULL;
3228 init_open_stateid(stp, fp, open);
3230 update_stateid(&stp->st_stid.sc_stateid);
3231 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3233 if (nfsd4_has_session(&resp->cstate)) {
3234 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3236 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3237 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3238 open->op_why_no_deleg = WND4_NOT_WANTED;
3244 * Attempt to hand out a delegation. No error return, because the
3245 * OPEN succeeds even if we fail.
3247 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3251 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3252 STATEID_VAL(&stp->st_stid.sc_stateid));
3254 /* 4.1 client trying to upgrade/downgrade delegation? */
3255 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3256 open->op_deleg_want)
3257 nfsd4_deleg_xgrade_none_ext(open, dp);
3261 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3262 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3264 * To finish the open response, we just need to set the rflags.
3266 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3267 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3268 !nfsd4_has_session(&resp->cstate))
3269 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3274 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3276 if (open->op_openowner) {
3277 struct nfs4_openowner *oo = open->op_openowner;
3279 if (!list_empty(&oo->oo_owner.so_stateids))
3280 list_del_init(&oo->oo_close_lru);
3281 if (oo->oo_flags & NFS4_OO_NEW) {
3283 release_openowner(oo);
3284 open->op_openowner = NULL;
3286 oo->oo_flags &= ~NFS4_OO_NEW;
3290 nfsd4_free_file(open->op_file);
3292 free_generic_stateid(open->op_stp);
3295 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3297 struct nfs4_client *found;
3299 if (STALE_CLIENTID(clid, nn))
3300 return nfserr_stale_clientid;
3301 found = find_confirmed_client(clid, session, nn);
3304 return found ? nfs_ok : nfserr_expired;
3308 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3311 struct nfs4_client *clp;
3313 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3316 dprintk("process_renew(%08x/%08x): starting\n",
3317 clid->cl_boot, clid->cl_id);
3318 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3321 status = nfserr_cb_path_down;
3322 if (!list_empty(&clp->cl_delegations)
3323 && clp->cl_cb_state != NFSD4_CB_UP)
3327 nfs4_unlock_state();
3332 nfsd4_end_grace(struct nfsd_net *nn)
3334 /* do nothing if grace period already ended */
3335 if (nn->grace_ended)
3338 dprintk("NFSD: end of grace period\n");
3339 nn->grace_ended = true;
3340 nfsd4_record_grace_done(nn, nn->boot_time);
3341 locks_end_grace(&nn->nfsd4_manager);
3343 * Now that every NFSv4 client has had the chance to recover and
3344 * to see the (possibly new, possibly shorter) lease time, we
3345 * can safely set the next grace time to the current lease time:
3347 nn->nfsd4_grace = nn->nfsd4_lease;
3351 nfs4_laundromat(struct nfsd_net *nn)
3353 struct nfs4_client *clp;
3354 struct nfs4_openowner *oo;
3355 struct nfs4_delegation *dp;
3356 struct list_head *pos, *next, reaplist;
3357 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3358 time_t t, clientid_val = nn->nfsd4_lease;
3359 time_t u, test_val = nn->nfsd4_lease;
3363 dprintk("NFSD: laundromat service - starting\n");
3364 nfsd4_end_grace(nn);
3365 INIT_LIST_HEAD(&reaplist);
3366 spin_lock(&nn->client_lock);
3367 list_for_each_safe(pos, next, &nn->client_lru) {
3368 clp = list_entry(pos, struct nfs4_client, cl_lru);
3369 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3370 t = clp->cl_time - cutoff;
3371 if (clientid_val > t)
3375 if (mark_client_expired_locked(clp)) {
3376 dprintk("NFSD: client in use (clientid %08x)\n",
3377 clp->cl_clientid.cl_id);
3380 list_move(&clp->cl_lru, &reaplist);
3382 spin_unlock(&nn->client_lock);
3383 list_for_each_safe(pos, next, &reaplist) {
3384 clp = list_entry(pos, struct nfs4_client, cl_lru);
3385 dprintk("NFSD: purging unused client (clientid %08x)\n",
3386 clp->cl_clientid.cl_id);
3389 spin_lock(&recall_lock);
3390 list_for_each_safe(pos, next, &nn->del_recall_lru) {
3391 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3392 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3394 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3395 u = dp->dl_time - cutoff;
3400 list_move(&dp->dl_recall_lru, &reaplist);
3402 spin_unlock(&recall_lock);
3403 list_for_each_safe(pos, next, &reaplist) {
3404 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3405 revoke_delegation(dp);
3407 test_val = nn->nfsd4_lease;
3408 list_for_each_safe(pos, next, &nn->close_lru) {
3409 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3410 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3411 u = oo->oo_time - cutoff;
3416 release_openowner(oo);
3418 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3419 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3420 nfs4_unlock_state();
3421 return clientid_val;
3424 static struct workqueue_struct *laundry_wq;
3425 static void laundromat_main(struct work_struct *);
3428 laundromat_main(struct work_struct *laundry)
3431 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3433 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3436 t = nfs4_laundromat(nn);
3437 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3438 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3441 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3443 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3444 return nfserr_bad_stateid;
3449 access_permit_read(struct nfs4_ol_stateid *stp)
3451 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3452 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3453 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3457 access_permit_write(struct nfs4_ol_stateid *stp)
3459 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3460 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3464 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3466 __be32 status = nfserr_openmode;
3468 /* For lock stateid's, we test the parent open, not the lock: */
3469 if (stp->st_openstp)
3470 stp = stp->st_openstp;
3471 if ((flags & WR_STATE) && !access_permit_write(stp))
3473 if ((flags & RD_STATE) && !access_permit_read(stp))
3480 static inline __be32
3481 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3483 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3485 else if (locks_in_grace(net)) {
3486 /* Answer in remaining cases depends on existence of
3487 * conflicting state; so we must wait out the grace period. */
3488 return nfserr_grace;
3489 } else if (flags & WR_STATE)
3490 return nfs4_share_conflict(current_fh,
3491 NFS4_SHARE_DENY_WRITE);
3492 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3493 return nfs4_share_conflict(current_fh,
3494 NFS4_SHARE_DENY_READ);
3498 * Allow READ/WRITE during grace period on recovered state only for files
3499 * that are not able to provide mandatory locking.
3502 grace_disallows_io(struct net *net, struct inode *inode)
3504 return locks_in_grace(net) && mandatory_lock(inode);
3507 /* Returns true iff a is later than b: */
3508 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3510 return (s32)(a->si_generation - b->si_generation) > 0;
3513 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3516 * When sessions are used the stateid generation number is ignored
3519 if (has_session && in->si_generation == 0)
3522 if (in->si_generation == ref->si_generation)
3525 /* If the client sends us a stateid from the future, it's buggy: */
3526 if (stateid_generation_after(in, ref))
3527 return nfserr_bad_stateid;
3529 * However, we could see a stateid from the past, even from a
3530 * non-buggy client. For example, if the client sends a lock
3531 * while some IO is outstanding, the lock may bump si_generation
3532 * while the IO is still in flight. The client could avoid that
3533 * situation by waiting for responses on all the IO requests,
3534 * but better performance may result in retrying IO that
3535 * receives an old_stateid error if requests are rarely
3536 * reordered in flight:
3538 return nfserr_old_stateid;
3541 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3543 struct nfs4_stid *s;
3544 struct nfs4_ol_stateid *ols;
3547 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3548 return nfserr_bad_stateid;
3549 /* Client debugging aid. */
3550 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3551 char addr_str[INET6_ADDRSTRLEN];
3552 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3554 pr_warn_ratelimited("NFSD: client %s testing state ID "
3555 "with incorrect client ID\n", addr_str);
3556 return nfserr_bad_stateid;
3558 s = find_stateid(cl, stateid);
3560 return nfserr_bad_stateid;
3561 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3564 switch (s->sc_type) {
3565 case NFS4_DELEG_STID:
3567 case NFS4_REVOKED_DELEG_STID:
3568 return nfserr_deleg_revoked;
3569 case NFS4_OPEN_STID:
3570 case NFS4_LOCK_STID:
3571 ols = openlockstateid(s);
3572 if (ols->st_stateowner->so_is_open_owner
3573 && !(openowner(ols->st_stateowner)->oo_flags
3574 & NFS4_OO_CONFIRMED))
3575 return nfserr_bad_stateid;
3578 printk("unknown stateid type %x\n", s->sc_type);
3579 case NFS4_CLOSED_STID:
3580 return nfserr_bad_stateid;
3584 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3585 struct nfs4_stid **s, bool sessions,
3586 struct nfsd_net *nn)
3588 struct nfs4_client *cl;
3591 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3592 return nfserr_bad_stateid;
3593 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3595 if (status == nfserr_stale_clientid)
3596 return nfserr_stale_stateid;
3599 *s = find_stateid_by_type(cl, stateid, typemask);
3601 return nfserr_bad_stateid;
3606 * Checks for stateid operations
3609 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3610 stateid_t *stateid, int flags, struct file **filpp)
3612 struct nfs4_stid *s;
3613 struct nfs4_ol_stateid *stp = NULL;
3614 struct nfs4_delegation *dp = NULL;
3615 struct svc_fh *current_fh = &cstate->current_fh;
3616 struct inode *ino = current_fh->fh_dentry->d_inode;
3617 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3623 if (grace_disallows_io(net, ino))
3624 return nfserr_grace;
3626 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3627 return check_special_stateids(net, current_fh, stateid, flags);
3629 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3630 &s, cstate->minorversion, nn);
3633 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3636 switch (s->sc_type) {
3637 case NFS4_DELEG_STID:
3638 dp = delegstateid(s);
3639 status = nfs4_check_delegmode(dp, flags);
3643 *filpp = dp->dl_file->fi_deleg_file;
3646 status = nfserr_serverfault;
3651 case NFS4_OPEN_STID:
3652 case NFS4_LOCK_STID:
3653 stp = openlockstateid(s);
3654 status = nfs4_check_fh(current_fh, stp);
3657 if (stp->st_stateowner->so_is_open_owner
3658 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3660 status = nfs4_check_openmode(stp, flags);
3664 if (flags & RD_STATE)
3665 *filpp = find_readable_file(stp->st_file);
3667 *filpp = find_writeable_file(stp->st_file);
3671 return nfserr_bad_stateid;
3679 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3681 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3682 return nfserr_locks_held;
3683 release_lock_stateid(stp);
3688 * Test if the stateid is valid
3691 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3692 struct nfsd4_test_stateid *test_stateid)
3694 struct nfsd4_test_stateid_id *stateid;
3695 struct nfs4_client *cl = cstate->session->se_client;
3698 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3699 stateid->ts_id_status =
3700 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3701 nfs4_unlock_state();
3707 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3708 struct nfsd4_free_stateid *free_stateid)
3710 stateid_t *stateid = &free_stateid->fr_stateid;
3711 struct nfs4_stid *s;
3712 struct nfs4_delegation *dp;
3713 struct nfs4_client *cl = cstate->session->se_client;
3714 __be32 ret = nfserr_bad_stateid;
3717 s = find_stateid(cl, stateid);
3720 switch (s->sc_type) {
3721 case NFS4_DELEG_STID:
3722 ret = nfserr_locks_held;
3724 case NFS4_OPEN_STID:
3725 case NFS4_LOCK_STID:
3726 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3729 if (s->sc_type == NFS4_LOCK_STID)
3730 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3732 ret = nfserr_locks_held;
3734 case NFS4_REVOKED_DELEG_STID:
3735 dp = delegstateid(s);
3736 destroy_revoked_delegation(dp);
3740 ret = nfserr_bad_stateid;
3743 nfs4_unlock_state();
3750 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3751 RD_STATE : WR_STATE;
3754 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3756 struct svc_fh *current_fh = &cstate->current_fh;
3757 struct nfs4_stateowner *sop = stp->st_stateowner;
3760 status = nfsd4_check_seqid(cstate, sop, seqid);
3763 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
3764 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
3766 * "Closed" stateid's exist *only* to return
3767 * nfserr_replay_me from the previous step, and
3768 * revoked delegations are kept only for free_stateid.
3770 return nfserr_bad_stateid;
3771 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3774 return nfs4_check_fh(current_fh, stp);
3778 * Checks for sequence id mutating operations.
3781 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3782 stateid_t *stateid, char typemask,
3783 struct nfs4_ol_stateid **stpp,
3784 struct nfsd_net *nn)
3787 struct nfs4_stid *s;
3789 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3790 seqid, STATEID_VAL(stateid));
3793 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3794 cstate->minorversion, nn);
3797 *stpp = openlockstateid(s);
3798 if (!nfsd4_has_session(cstate))
3799 cstate->replay_owner = (*stpp)->st_stateowner;
3801 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3804 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3805 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3808 struct nfs4_openowner *oo;
3810 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3811 NFS4_OPEN_STID, stpp, nn);
3814 oo = openowner((*stpp)->st_stateowner);
3815 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3816 return nfserr_bad_stateid;
3821 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3822 struct nfsd4_open_confirm *oc)
3825 struct nfs4_openowner *oo;
3826 struct nfs4_ol_stateid *stp;
3827 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3829 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3830 (int)cstate->current_fh.fh_dentry->d_name.len,
3831 cstate->current_fh.fh_dentry->d_name.name);
3833 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3839 status = nfs4_preprocess_seqid_op(cstate,
3840 oc->oc_seqid, &oc->oc_req_stateid,
3841 NFS4_OPEN_STID, &stp, nn);
3844 oo = openowner(stp->st_stateowner);
3845 status = nfserr_bad_stateid;
3846 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3848 oo->oo_flags |= NFS4_OO_CONFIRMED;
3849 update_stateid(&stp->st_stid.sc_stateid);
3850 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3851 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3852 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3854 nfsd4_client_record_create(oo->oo_owner.so_client);
3857 nfsd4_bump_seqid(cstate, status);
3858 if (!cstate->replay_owner)
3859 nfs4_unlock_state();
3863 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3865 if (!test_access(access, stp))
3867 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3868 clear_access(access, stp);
3871 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3873 switch (to_access) {
3874 case NFS4_SHARE_ACCESS_READ:
3875 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3876 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3878 case NFS4_SHARE_ACCESS_WRITE:
3879 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3880 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3882 case NFS4_SHARE_ACCESS_BOTH:
3890 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3893 for (i = 0; i < 4; i++) {
3894 if ((i & deny) != i)
3900 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3901 struct nfsd4_compound_state *cstate,
3902 struct nfsd4_open_downgrade *od)
3905 struct nfs4_ol_stateid *stp;
3906 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3908 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3909 (int)cstate->current_fh.fh_dentry->d_name.len,
3910 cstate->current_fh.fh_dentry->d_name.name);
3912 /* We don't yet support WANT bits: */
3913 if (od->od_deleg_want)
3914 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3918 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3919 &od->od_stateid, &stp, nn);
3922 status = nfserr_inval;
3923 if (!test_access(od->od_share_access, stp)) {
3924 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3925 stp->st_access_bmap, od->od_share_access);
3928 if (!test_deny(od->od_share_deny, stp)) {
3929 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3930 stp->st_deny_bmap, od->od_share_deny);
3933 nfs4_stateid_downgrade(stp, od->od_share_access);
3935 reset_union_bmap_deny(od->od_share_deny, stp);
3937 update_stateid(&stp->st_stid.sc_stateid);
3938 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3941 nfsd4_bump_seqid(cstate, status);
3942 if (!cstate->replay_owner)
3943 nfs4_unlock_state();
3947 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3949 unhash_open_stateid(s);
3950 s->st_stid.sc_type = NFS4_CLOSED_STID;
3954 * nfs4_unlock_state() called after encode
3957 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3958 struct nfsd4_close *close)
3961 struct nfs4_openowner *oo;
3962 struct nfs4_ol_stateid *stp;
3963 struct net *net = SVC_NET(rqstp);
3964 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3966 dprintk("NFSD: nfsd4_close on file %.*s\n",
3967 (int)cstate->current_fh.fh_dentry->d_name.len,
3968 cstate->current_fh.fh_dentry->d_name.name);
3971 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3973 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3975 nfsd4_bump_seqid(cstate, status);
3978 oo = openowner(stp->st_stateowner);
3979 update_stateid(&stp->st_stid.sc_stateid);
3980 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3982 nfsd4_close_open_stateid(stp);
3984 if (cstate->minorversion) {
3985 unhash_stid(&stp->st_stid);
3986 free_generic_stateid(stp);
3988 oo->oo_last_closed_stid = stp;
3990 if (list_empty(&oo->oo_owner.so_stateids)) {
3991 if (cstate->minorversion)
3992 release_openowner(oo);
3995 * In the 4.0 case we need to keep the owners around a
3996 * little while to handle CLOSE replay.
3998 move_to_close_lru(oo, SVC_NET(rqstp));
4002 if (!cstate->replay_owner)
4003 nfs4_unlock_state();
4008 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4009 struct nfsd4_delegreturn *dr)
4011 struct nfs4_delegation *dp;
4012 stateid_t *stateid = &dr->dr_stateid;
4013 struct nfs4_stid *s;
4015 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4017 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4021 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
4022 cstate->minorversion, nn);
4025 dp = delegstateid(s);
4026 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4030 destroy_delegation(dp);
4032 nfs4_unlock_state();
4038 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4040 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
4043 end_offset(u64 start, u64 len)
4048 return end >= start ? end: NFS4_MAX_UINT64;
4051 /* last octet in a range */
4053 last_byte_offset(u64 start, u64 len)
4059 return end > start ? end - 1: NFS4_MAX_UINT64;
4062 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
4064 return (file_hashval(inode) + cl_id
4065 + opaque_hashval(ownername->data, ownername->len))
4066 & LOCKOWNER_INO_HASH_MASK;
4070 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4071 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4072 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4073 * locking, this prevents us from being completely protocol-compliant. The
4074 * real solution to this problem is to start using unsigned file offsets in
4075 * the VFS, but this is a very deep change!
4078 nfs4_transform_lock_offset(struct file_lock *lock)
4080 if (lock->fl_start < 0)
4081 lock->fl_start = OFFSET_MAX;
4082 if (lock->fl_end < 0)
4083 lock->fl_end = OFFSET_MAX;
4086 /* Hack!: For now, we're defining this just so we can use a pointer to it
4087 * as a unique cookie to identify our (NFSv4's) posix locks. */
4088 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4092 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4094 struct nfs4_lockowner *lo;
4096 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4097 lo = (struct nfs4_lockowner *) fl->fl_owner;
4098 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4099 lo->lo_owner.so_owner.len, GFP_KERNEL);
4100 if (!deny->ld_owner.data)
4101 /* We just don't care that much */
4103 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4104 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4107 deny->ld_owner.len = 0;
4108 deny->ld_owner.data = NULL;
4109 deny->ld_clientid.cl_boot = 0;
4110 deny->ld_clientid.cl_id = 0;
4112 deny->ld_start = fl->fl_start;
4113 deny->ld_length = NFS4_MAX_UINT64;
4114 if (fl->fl_end != NFS4_MAX_UINT64)
4115 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4116 deny->ld_type = NFS4_READ_LT;
4117 if (fl->fl_type != F_RDLCK)
4118 deny->ld_type = NFS4_WRITE_LT;
4121 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
4123 struct nfs4_ol_stateid *lst;
4125 if (!same_owner_str(&lo->lo_owner, owner, clid))
4127 lst = list_first_entry(&lo->lo_owner.so_stateids,
4128 struct nfs4_ol_stateid, st_perstateowner);
4129 return lst->st_file->fi_inode == inode;
4132 static struct nfs4_lockowner *
4133 find_lockowner_str(struct inode *inode, clientid_t *clid,
4134 struct xdr_netobj *owner, struct nfsd_net *nn)
4136 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
4137 struct nfs4_lockowner *lo;
4139 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
4140 if (same_lockowner_ino(lo, inode, clid, owner))
4146 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
4148 struct inode *inode = open_stp->st_file->fi_inode;
4149 unsigned int inohash = lockowner_ino_hashval(inode,
4150 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4151 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4153 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4154 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4155 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4159 * Alloc a lock owner structure.
4160 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4163 * strhashval = ownerstr_hashval
4166 static struct nfs4_lockowner *
4167 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4168 struct nfs4_lockowner *lo;
4170 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4173 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4174 lo->lo_owner.so_is_open_owner = 0;
4175 /* It is the openowner seqid that will be incremented in encode in the
4176 * case of new lockowners; so increment the lock seqid manually: */
4177 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4178 hash_lockowner(lo, strhashval, clp, open_stp);
4182 static struct nfs4_ol_stateid *
4183 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4185 struct nfs4_ol_stateid *stp;
4186 struct nfs4_client *clp = lo->lo_owner.so_client;
4188 stp = nfs4_alloc_stateid(clp);
4191 stp->st_stid.sc_type = NFS4_LOCK_STID;
4192 list_add(&stp->st_perfile, &fp->fi_stateids);
4193 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4194 stp->st_stateowner = &lo->lo_owner;
4197 stp->st_access_bmap = 0;
4198 stp->st_deny_bmap = open_stp->st_deny_bmap;
4199 stp->st_openstp = open_stp;
4204 check_lock_length(u64 offset, u64 length)
4206 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4207 LOFF_OVERFLOW(offset, length)));
4210 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4212 struct nfs4_file *fp = lock_stp->st_file;
4213 int oflag = nfs4_access_to_omode(access);
4215 if (test_access(access, lock_stp))
4217 nfs4_file_get_access(fp, oflag);
4218 set_access(access, lock_stp);
4221 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)
4223 struct nfs4_file *fi = ost->st_file;
4224 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4225 struct nfs4_client *cl = oo->oo_owner.so_client;
4226 struct nfs4_lockowner *lo;
4227 unsigned int strhashval;
4228 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4230 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4231 &lock->v.new.owner, nn);
4233 if (!cstate->minorversion)
4234 return nfserr_bad_seqid;
4235 /* XXX: a lockowner always has exactly one stateid: */
4236 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4237 struct nfs4_ol_stateid, st_perstateowner);
4240 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4241 &lock->v.new.owner);
4242 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4244 return nfserr_jukebox;
4245 *lst = alloc_init_lock_stateid(lo, fi, ost);
4247 release_lockowner(lo);
4248 return nfserr_jukebox;
4258 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4259 struct nfsd4_lock *lock)
4261 struct nfs4_openowner *open_sop = NULL;
4262 struct nfs4_lockowner *lock_sop = NULL;
4263 struct nfs4_ol_stateid *lock_stp;
4264 struct file *filp = NULL;
4265 struct file_lock *file_lock = NULL;
4266 struct file_lock *conflock = NULL;
4268 bool new_state = false;
4271 struct net *net = SVC_NET(rqstp);
4272 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4274 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4275 (long long) lock->lk_offset,
4276 (long long) lock->lk_length);
4278 if (check_lock_length(lock->lk_offset, lock->lk_length))
4279 return nfserr_inval;
4281 if ((status = fh_verify(rqstp, &cstate->current_fh,
4282 S_IFREG, NFSD_MAY_LOCK))) {
4283 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4289 if (lock->lk_is_new) {
4290 struct nfs4_ol_stateid *open_stp = NULL;
4292 if (nfsd4_has_session(cstate))
4293 /* See rfc 5661 18.10.3: given clientid is ignored: */
4294 memcpy(&lock->v.new.clientid,
4295 &cstate->session->se_client->cl_clientid,
4296 sizeof(clientid_t));
4298 status = nfserr_stale_clientid;
4299 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4302 /* validate and update open stateid and open seqid */
4303 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4304 lock->lk_new_open_seqid,
4305 &lock->lk_new_open_stateid,
4309 open_sop = openowner(open_stp->st_stateowner);
4310 status = nfserr_bad_stateid;
4311 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4312 &lock->v.new.clientid))
4314 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4315 &lock_stp, &new_state);
4317 status = nfs4_preprocess_seqid_op(cstate,
4318 lock->lk_old_lock_seqid,
4319 &lock->lk_old_lock_stateid,
4320 NFS4_LOCK_STID, &lock_stp, nn);
4323 lock_sop = lockowner(lock_stp->st_stateowner);
4325 lkflg = setlkflg(lock->lk_type);
4326 status = nfs4_check_openmode(lock_stp, lkflg);
4330 status = nfserr_grace;
4331 if (locks_in_grace(net) && !lock->lk_reclaim)
4333 status = nfserr_no_grace;
4334 if (!locks_in_grace(net) && lock->lk_reclaim)
4337 file_lock = locks_alloc_lock();
4339 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4340 status = nfserr_jukebox;
4344 locks_init_lock(file_lock);
4345 switch (lock->lk_type) {
4348 filp = find_readable_file(lock_stp->st_file);
4350 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4351 file_lock->fl_type = F_RDLCK;
4354 case NFS4_WRITEW_LT:
4355 filp = find_writeable_file(lock_stp->st_file);
4357 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4358 file_lock->fl_type = F_WRLCK;
4361 status = nfserr_inval;
4365 status = nfserr_openmode;
4368 file_lock->fl_owner = (fl_owner_t)lock_sop;
4369 file_lock->fl_pid = current->tgid;
4370 file_lock->fl_file = filp;
4371 file_lock->fl_flags = FL_POSIX;
4372 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4373 file_lock->fl_start = lock->lk_offset;
4374 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4375 nfs4_transform_lock_offset(file_lock);
4377 conflock = locks_alloc_lock();
4379 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4380 status = nfserr_jukebox;
4384 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4386 case 0: /* success! */
4387 update_stateid(&lock_stp->st_stid.sc_stateid);
4388 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4392 case (EAGAIN): /* conflock holds conflicting lock */
4393 status = nfserr_denied;
4394 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4395 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4398 status = nfserr_deadlock;
4401 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4402 status = nfserrno(err);
4406 if (status && new_state)
4407 release_lockowner(lock_sop);
4408 nfsd4_bump_seqid(cstate, status);
4409 if (!cstate->replay_owner)
4410 nfs4_unlock_state();
4412 locks_free_lock(file_lock);
4414 locks_free_lock(conflock);
4419 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4420 * so we do a temporary open here just to get an open file to pass to
4421 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4424 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4427 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4429 err = nfserrno(vfs_test_lock(file, lock));
4439 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4440 struct nfsd4_lockt *lockt)
4442 struct inode *inode;
4443 struct file_lock *file_lock = NULL;
4444 struct nfs4_lockowner *lo;
4446 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4448 if (locks_in_grace(SVC_NET(rqstp)))
4449 return nfserr_grace;
4451 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4452 return nfserr_inval;
4456 if (!nfsd4_has_session(cstate)) {
4457 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4462 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4465 inode = cstate->current_fh.fh_dentry->d_inode;
4466 file_lock = locks_alloc_lock();
4468 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4469 status = nfserr_jukebox;
4472 locks_init_lock(file_lock);
4473 switch (lockt->lt_type) {
4476 file_lock->fl_type = F_RDLCK;
4479 case NFS4_WRITEW_LT:
4480 file_lock->fl_type = F_WRLCK;
4483 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4484 status = nfserr_inval;
4488 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4490 file_lock->fl_owner = (fl_owner_t)lo;
4491 file_lock->fl_pid = current->tgid;
4492 file_lock->fl_flags = FL_POSIX;
4494 file_lock->fl_start = lockt->lt_offset;
4495 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4497 nfs4_transform_lock_offset(file_lock);
4499 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4503 if (file_lock->fl_type != F_UNLCK) {
4504 status = nfserr_denied;
4505 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4508 nfs4_unlock_state();
4510 locks_free_lock(file_lock);
4515 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4516 struct nfsd4_locku *locku)
4518 struct nfs4_lockowner *lo;
4519 struct nfs4_ol_stateid *stp;
4520 struct file *filp = NULL;
4521 struct file_lock *file_lock = NULL;
4524 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4526 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4527 (long long) locku->lu_offset,
4528 (long long) locku->lu_length);
4530 if (check_lock_length(locku->lu_offset, locku->lu_length))
4531 return nfserr_inval;
4535 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4536 &locku->lu_stateid, NFS4_LOCK_STID,
4540 filp = find_any_file(stp->st_file);
4542 status = nfserr_lock_range;
4545 file_lock = locks_alloc_lock();
4547 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4548 status = nfserr_jukebox;
4551 lo = lockowner(stp->st_stateowner);
4552 locks_init_lock(file_lock);
4553 file_lock->fl_type = F_UNLCK;
4554 file_lock->fl_owner = (fl_owner_t)lo;
4555 file_lock->fl_pid = current->tgid;
4556 file_lock->fl_file = filp;
4557 file_lock->fl_flags = FL_POSIX;
4558 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4559 file_lock->fl_start = locku->lu_offset;
4561 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4563 nfs4_transform_lock_offset(file_lock);
4565 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4567 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4570 update_stateid(&stp->st_stid.sc_stateid);
4571 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4573 if (nfsd4_has_session(cstate) && !check_for_locks(stp->st_file, lo)) {
4574 WARN_ON_ONCE(cstate->replay_owner);
4575 release_lockowner(lo);
4579 nfsd4_bump_seqid(cstate, status);
4580 if (!cstate->replay_owner)
4581 nfs4_unlock_state();
4583 locks_free_lock(file_lock);
4587 status = nfserrno(err);
4593 * 1: locks held by lockowner
4594 * 0: no locks held by lockowner
4597 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4599 struct file_lock **flpp;
4600 struct inode *inode = filp->fi_inode;
4604 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4605 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4616 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4617 struct nfsd4_compound_state *cstate,
4618 struct nfsd4_release_lockowner *rlockowner)
4620 clientid_t *clid = &rlockowner->rl_clientid;
4621 struct nfs4_stateowner *sop;
4622 struct nfs4_lockowner *lo;
4623 struct nfs4_ol_stateid *stp;
4624 struct xdr_netobj *owner = &rlockowner->rl_owner;
4625 struct list_head matches;
4626 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4628 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4630 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4631 clid->cl_boot, clid->cl_id);
4635 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4639 status = nfserr_locks_held;
4640 INIT_LIST_HEAD(&matches);
4642 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4643 if (sop->so_is_open_owner)
4645 if (!same_owner_str(sop, owner, clid))
4647 list_for_each_entry(stp, &sop->so_stateids,
4649 lo = lockowner(sop);
4650 if (check_for_locks(stp->st_file, lo))
4652 list_add(&lo->lo_list, &matches);
4655 /* Clients probably won't expect us to return with some (but not all)
4656 * of the lockowner state released; so don't release any until all
4657 * have been checked. */
4659 while (!list_empty(&matches)) {
4660 lo = list_entry(matches.next, struct nfs4_lockowner,
4662 /* unhash_stateowner deletes so_perclient only
4663 * for openowners. */
4664 list_del(&lo->lo_list);
4665 release_lockowner(lo);
4668 nfs4_unlock_state();
4672 static inline struct nfs4_client_reclaim *
4675 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4679 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4681 struct nfs4_client_reclaim *crp;
4683 crp = nfsd4_find_reclaim_client(name, nn);
4684 return (crp && crp->cr_clp);
4688 * failure => all reset bets are off, nfserr_no_grace...
4690 struct nfs4_client_reclaim *
4691 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4693 unsigned int strhashval;
4694 struct nfs4_client_reclaim *crp;
4696 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4697 crp = alloc_reclaim();
4699 strhashval = clientstr_hashval(name);
4700 INIT_LIST_HEAD(&crp->cr_strhash);
4701 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4702 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4704 nn->reclaim_str_hashtbl_size++;
4710 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4712 list_del(&crp->cr_strhash);
4714 nn->reclaim_str_hashtbl_size--;
4718 nfs4_release_reclaim(struct nfsd_net *nn)
4720 struct nfs4_client_reclaim *crp = NULL;
4723 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4724 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4725 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4726 struct nfs4_client_reclaim, cr_strhash);
4727 nfs4_remove_reclaim_record(crp, nn);
4730 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4734 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4735 struct nfs4_client_reclaim *
4736 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4738 unsigned int strhashval;
4739 struct nfs4_client_reclaim *crp = NULL;
4741 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4743 strhashval = clientstr_hashval(recdir);
4744 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4745 if (same_name(crp->cr_recdir, recdir)) {
4753 * Called from OPEN. Look for clientid in reclaim list.
4756 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4758 struct nfs4_client *clp;
4760 /* find clientid in conf_id_hashtbl */
4761 clp = find_confirmed_client(clid, sessions, nn);
4763 return nfserr_reclaim_bad;
4765 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4768 #ifdef CONFIG_NFSD_FAULT_INJECTION
4770 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4772 if (mark_client_expired(clp))
4778 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4780 char buf[INET6_ADDRSTRLEN];
4781 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4782 printk(KERN_INFO "NFS Client: %s\n", buf);
4786 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4789 char buf[INET6_ADDRSTRLEN];
4790 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4791 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4794 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4796 struct nfs4_openowner *oop;
4797 struct nfs4_lockowner *lop, *lo_next;
4798 struct nfs4_ol_stateid *stp, *st_next;
4801 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4802 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4803 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4815 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4817 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4820 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4822 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4823 nfsd_print_count(clp, count, "locked files");
4827 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4829 struct nfs4_openowner *oop, *next;
4832 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4842 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4844 return nfsd_foreach_client_open(clp, max, release_openowner);
4847 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4849 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4850 nfsd_print_count(clp, count, "open files");
4854 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4855 struct list_head *victims)
4857 struct nfs4_delegation *dp, *next;
4860 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4862 list_move(&dp->dl_recall_lru, victims);
4869 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4871 struct nfs4_delegation *dp, *next;
4875 spin_lock(&recall_lock);
4876 count = nfsd_find_all_delegations(clp, max, &victims);
4877 spin_unlock(&recall_lock);
4879 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4880 revoke_delegation(dp);
4885 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4887 struct nfs4_delegation *dp, *next;
4891 spin_lock(&recall_lock);
4892 count = nfsd_find_all_delegations(clp, max, &victims);
4893 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4894 nfsd_break_one_deleg(dp);
4895 spin_unlock(&recall_lock);
4900 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4904 spin_lock(&recall_lock);
4905 count = nfsd_find_all_delegations(clp, max, NULL);
4906 spin_unlock(&recall_lock);
4908 nfsd_print_count(clp, count, "delegations");
4912 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4914 struct nfs4_client *clp, *next;
4916 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4918 if (!nfsd_netns_ready(nn))
4921 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4922 count += func(clp, max - count);
4923 if ((max != 0) && (count >= max))
4930 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4932 struct nfs4_client *clp;
4933 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4935 if (!nfsd_netns_ready(nn))
4938 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4939 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4945 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4947 /* initialization to perform at module load time: */
4950 nfs4_state_init(void)
4955 * Since the lifetime of a delegation isn't limited to that of an open, a
4956 * client may quite reasonably hang on to a delegation as long as it has
4957 * the inode cached. This becomes an obvious problem the first time a
4958 * client's inode cache approaches the size of the server's total memory.
4960 * For now we avoid this problem by imposing a hard limit on the number
4961 * of delegations, which varies according to the server's memory size.
4964 set_max_delegations(void)
4967 * Allow at most 4 delegations per megabyte of RAM. Quick
4968 * estimates suggest that in the worst case (where every delegation
4969 * is for a different inode), a delegation could take about 1.5K,
4970 * giving a worst case usage of about 6% of memory.
4972 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4975 static int nfs4_state_create_net(struct net *net)
4977 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4980 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4981 CLIENT_HASH_SIZE, GFP_KERNEL);
4982 if (!nn->conf_id_hashtbl)
4984 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4985 CLIENT_HASH_SIZE, GFP_KERNEL);
4986 if (!nn->unconf_id_hashtbl)
4988 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4989 OWNER_HASH_SIZE, GFP_KERNEL);
4990 if (!nn->ownerstr_hashtbl)
4992 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4993 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4994 if (!nn->lockowner_ino_hashtbl)
4995 goto err_lockowner_ino;
4996 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4997 SESSION_HASH_SIZE, GFP_KERNEL);
4998 if (!nn->sessionid_hashtbl)
5001 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5002 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5003 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5005 for (i = 0; i < OWNER_HASH_SIZE; i++)
5006 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
5007 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
5008 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
5009 for (i = 0; i < SESSION_HASH_SIZE; i++)
5010 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5011 nn->conf_name_tree = RB_ROOT;
5012 nn->unconf_name_tree = RB_ROOT;
5013 INIT_LIST_HEAD(&nn->client_lru);
5014 INIT_LIST_HEAD(&nn->close_lru);
5015 INIT_LIST_HEAD(&nn->del_recall_lru);
5016 spin_lock_init(&nn->client_lock);
5018 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5024 kfree(nn->lockowner_ino_hashtbl);
5026 kfree(nn->ownerstr_hashtbl);
5028 kfree(nn->unconf_id_hashtbl);
5030 kfree(nn->conf_id_hashtbl);
5036 nfs4_state_destroy_net(struct net *net)
5039 struct nfs4_client *clp = NULL;
5040 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5041 struct rb_node *node, *tmp;
5043 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5044 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5045 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5046 destroy_client(clp);
5050 node = rb_first(&nn->unconf_name_tree);
5051 while (node != NULL) {
5053 node = rb_next(tmp);
5054 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
5055 rb_erase(tmp, &nn->unconf_name_tree);
5056 destroy_client(clp);
5059 kfree(nn->sessionid_hashtbl);
5060 kfree(nn->lockowner_ino_hashtbl);
5061 kfree(nn->ownerstr_hashtbl);
5062 kfree(nn->unconf_id_hashtbl);
5063 kfree(nn->conf_id_hashtbl);
5068 nfs4_state_start_net(struct net *net)
5070 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5073 ret = nfs4_state_create_net(net);
5076 nfsd4_client_tracking_init(net);
5077 nn->boot_time = get_seconds();
5078 locks_start_grace(net, &nn->nfsd4_manager);
5079 nn->grace_ended = false;
5080 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5081 nn->nfsd4_grace, net);
5082 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5086 /* initialization to perform when the nfsd service is started: */
5089 nfs4_state_start(void)
5093 ret = set_callback_cred();
5096 laundry_wq = create_singlethread_workqueue("nfsd4");
5097 if (laundry_wq == NULL) {
5101 ret = nfsd4_create_callback_queue();
5103 goto out_free_laundry;
5105 set_max_delegations();
5110 destroy_workqueue(laundry_wq);
5115 /* should be called with the state lock held */
5117 nfs4_state_shutdown_net(struct net *net)
5119 struct nfs4_delegation *dp = NULL;
5120 struct list_head *pos, *next, reaplist;
5121 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5123 cancel_delayed_work_sync(&nn->laundromat_work);
5124 locks_end_grace(&nn->nfsd4_manager);
5126 INIT_LIST_HEAD(&reaplist);
5127 spin_lock(&recall_lock);
5128 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5129 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5130 list_move(&dp->dl_recall_lru, &reaplist);
5132 spin_unlock(&recall_lock);
5133 list_for_each_safe(pos, next, &reaplist) {
5134 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5135 destroy_delegation(dp);
5138 nfsd4_client_tracking_exit(net);
5139 nfs4_state_destroy_net(net);
5143 nfs4_state_shutdown(void)
5145 destroy_workqueue(laundry_wq);
5146 nfsd4_destroy_callback_queue();
5150 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5152 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5153 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5157 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5159 if (cstate->minorversion) {
5160 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5161 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5166 clear_current_stateid(struct nfsd4_compound_state *cstate)
5168 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5172 * functions to set current state id
5175 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5177 put_stateid(cstate, &odp->od_stateid);
5181 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5183 put_stateid(cstate, &open->op_stateid);
5187 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5189 put_stateid(cstate, &close->cl_stateid);
5193 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5195 put_stateid(cstate, &lock->lk_resp_stateid);
5199 * functions to consume current state id
5203 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5205 get_stateid(cstate, &odp->od_stateid);
5209 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5211 get_stateid(cstate, &drp->dr_stateid);
5215 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5217 get_stateid(cstate, &fsp->fr_stateid);
5221 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5223 get_stateid(cstate, &setattr->sa_stateid);
5227 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5229 get_stateid(cstate, &close->cl_stateid);
5233 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5235 get_stateid(cstate, &locku->lu_stateid);
5239 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5241 get_stateid(cstate, &read->rd_stateid);
5245 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5247 get_stateid(cstate, &write->wr_stateid);