2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
46 #define NFSDDBG_FACILITY NFSDDBG_PROC
49 time_t nfsd4_lease = 90; /* default lease time */
50 time_t nfsd4_grace = 90;
51 static time_t boot_time;
52 static u32 current_ownerid = 1;
53 static u32 current_fileid = 1;
54 static u32 current_delegid = 1;
55 static stateid_t zerostateid; /* bits all 0 */
56 static stateid_t onestateid; /* bits all 1 */
57 static u64 current_sessionid = 1;
59 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
62 /* forward declarations */
63 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
67 /* Currently used for almost all code touching nfsv4 state: */
68 static DEFINE_MUTEX(client_mutex);
71 * Currently used for the del_recall_lru and file hash table. In an
72 * effort to decrease the scope of the client_mutex, this spinlock may
73 * eventually cover more:
75 static DEFINE_SPINLOCK(recall_lock);
77 static struct kmem_cache *openowner_slab = NULL;
78 static struct kmem_cache *lockowner_slab = NULL;
79 static struct kmem_cache *file_slab = NULL;
80 static struct kmem_cache *stateid_slab = NULL;
81 static struct kmem_cache *deleg_slab = NULL;
86 mutex_lock(&client_mutex);
90 nfs4_unlock_state(void)
92 mutex_unlock(&client_mutex);
96 opaque_hashval(const void *ptr, int nbytes)
98 unsigned char *cptr = (unsigned char *) ptr;
108 static struct list_head del_recall_lru;
111 put_nfs4_file(struct nfs4_file *fi)
113 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
114 list_del(&fi->fi_hash);
115 spin_unlock(&recall_lock);
117 kmem_cache_free(file_slab, fi);
122 get_nfs4_file(struct nfs4_file *fi)
124 atomic_inc(&fi->fi_ref);
127 static int num_delegations;
128 unsigned int max_delegations;
131 * Open owner state (share locks)
134 /* hash tables for open owners */
135 #define OPEN_OWNER_HASH_BITS 8
136 #define OPEN_OWNER_HASH_SIZE (1 << OPEN_OWNER_HASH_BITS)
137 #define OPEN_OWNER_HASH_MASK (OPEN_OWNER_HASH_SIZE - 1)
139 static unsigned int open_ownerid_hashval(const u32 id)
141 return id & OPEN_OWNER_HASH_MASK;
144 static unsigned int open_ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
148 ret = opaque_hashval(ownername->data, ownername->len);
150 return ret & OPEN_OWNER_HASH_MASK;
153 static struct list_head open_ownerid_hashtbl[OPEN_OWNER_HASH_SIZE];
154 static struct list_head open_ownerstr_hashtbl[OPEN_OWNER_HASH_SIZE];
156 /* hash table for nfs4_file */
157 #define FILE_HASH_BITS 8
158 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
160 /* hash table for (open)nfs4_ol_stateid */
161 #define STATEID_HASH_BITS 10
162 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
163 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
165 static unsigned int file_hashval(struct inode *ino)
167 /* XXX: why are we hashing on inode pointer, anyway? */
168 return hash_ptr(ino, FILE_HASH_BITS);
171 static unsigned int stateid_hashval(stateid_t *s)
173 return opaque_hashval(&s->si_opaque, sizeof(stateid_opaque_t)) & STATEID_HASH_MASK;
176 static struct list_head file_hashtbl[FILE_HASH_SIZE];
177 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
179 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
181 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
182 atomic_inc(&fp->fi_access[oflag]);
185 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
187 if (oflag == O_RDWR) {
188 __nfs4_file_get_access(fp, O_RDONLY);
189 __nfs4_file_get_access(fp, O_WRONLY);
191 __nfs4_file_get_access(fp, oflag);
194 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
196 if (fp->fi_fds[oflag]) {
197 fput(fp->fi_fds[oflag]);
198 fp->fi_fds[oflag] = NULL;
202 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
204 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
205 nfs4_file_put_fd(fp, O_RDWR);
206 nfs4_file_put_fd(fp, oflag);
210 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
212 if (oflag == O_RDWR) {
213 __nfs4_file_put_access(fp, O_RDONLY);
214 __nfs4_file_put_access(fp, O_WRONLY);
216 __nfs4_file_put_access(fp, oflag);
219 static inline void hash_stid(struct nfs4_stid *stid)
221 stateid_t *s = &stid->sc_stateid;
222 unsigned int hashval;
224 hashval = stateid_hashval(s);
225 list_add(&stid->sc_hash, &stateid_hashtbl[hashval]);
228 static struct nfs4_delegation *
229 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
231 struct nfs4_delegation *dp;
232 struct nfs4_file *fp = stp->st_file;
234 dprintk("NFSD alloc_init_deleg\n");
236 * Major work on the lease subsystem (for example, to support
237 * calbacks on stat) will be required before we can support
238 * write delegations properly.
240 if (type != NFS4_OPEN_DELEGATE_READ)
242 if (fp->fi_had_conflict)
244 if (num_delegations > max_delegations)
246 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
250 INIT_LIST_HEAD(&dp->dl_perfile);
251 INIT_LIST_HEAD(&dp->dl_perclnt);
252 INIT_LIST_HEAD(&dp->dl_recall_lru);
257 dp->dl_stid.sc_type = NFS4_DELEG_STID;
258 dp->dl_stid.sc_stateid.si_boot = boot_time;
259 dp->dl_stid.sc_stateid.si_stateownerid = current_delegid++;
260 dp->dl_stid.sc_stateid.si_fileid = 0;
261 dp->dl_stid.sc_stateid.si_generation = 1;
262 hash_stid(&dp->dl_stid);
263 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
265 atomic_set(&dp->dl_count, 1);
266 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
271 nfs4_put_delegation(struct nfs4_delegation *dp)
273 if (atomic_dec_and_test(&dp->dl_count)) {
274 dprintk("NFSD: freeing dp %p\n",dp);
275 put_nfs4_file(dp->dl_file);
276 kmem_cache_free(deleg_slab, dp);
281 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
283 if (atomic_dec_and_test(&fp->fi_delegees)) {
284 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
286 fput(fp->fi_deleg_file);
287 fp->fi_deleg_file = NULL;
291 /* Called under the state lock. */
293 unhash_delegation(struct nfs4_delegation *dp)
295 list_del_init(&dp->dl_stid.sc_hash);
296 list_del_init(&dp->dl_perclnt);
297 spin_lock(&recall_lock);
298 list_del_init(&dp->dl_perfile);
299 list_del_init(&dp->dl_recall_lru);
300 spin_unlock(&recall_lock);
301 nfs4_put_deleg_lease(dp->dl_file);
302 nfs4_put_delegation(dp);
309 /* client_lock protects the client lru list and session hash table */
310 static DEFINE_SPINLOCK(client_lock);
312 /* Hash tables for nfs4_clientid state */
313 #define CLIENT_HASH_BITS 4
314 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
315 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
317 static unsigned int clientid_hashval(u32 id)
319 return id & CLIENT_HASH_MASK;
322 static unsigned int clientstr_hashval(const char *name)
324 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
328 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
329 * used in reboot/reset lease grace period processing
331 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
332 * setclientid_confirmed info.
334 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
337 * client_lru holds client queue ordered by nfs4_client.cl_time
340 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
341 * for last close replay.
343 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
344 static int reclaim_str_hashtbl_size = 0;
345 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
346 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
347 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
348 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
349 static struct list_head client_lru;
350 static struct list_head close_lru;
353 * We store the NONE, READ, WRITE, and BOTH bits separately in the
354 * st_{access,deny}_bmap field of the stateid, in order to track not
355 * only what share bits are currently in force, but also what
356 * combinations of share bits previous opens have used. This allows us
357 * to enforce the recommendation of rfc 3530 14.2.19 that the server
358 * return an error if the client attempt to downgrade to a combination
359 * of share bits not explicable by closing some of its previous opens.
361 * XXX: This enforcement is actually incomplete, since we don't keep
362 * track of access/deny bit combinations; so, e.g., we allow:
364 * OPEN allow read, deny write
365 * OPEN allow both, deny none
366 * DOWNGRADE allow read, deny none
368 * which we should reject.
371 set_access(unsigned int *access, unsigned long bmap) {
375 for (i = 1; i < 4; i++) {
376 if (test_bit(i, &bmap))
382 set_deny(unsigned int *deny, unsigned long bmap) {
386 for (i = 0; i < 4; i++) {
387 if (test_bit(i, &bmap))
393 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
394 unsigned int access, deny;
396 set_access(&access, stp->st_access_bmap);
397 set_deny(&deny, stp->st_deny_bmap);
398 if ((access & open->op_share_deny) || (deny & open->op_share_access))
403 static int nfs4_access_to_omode(u32 access)
405 switch (access & NFS4_SHARE_ACCESS_BOTH) {
406 case NFS4_SHARE_ACCESS_READ:
408 case NFS4_SHARE_ACCESS_WRITE:
410 case NFS4_SHARE_ACCESS_BOTH:
416 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
418 list_del(&stp->st_stid.sc_hash);
419 list_del(&stp->st_perfile);
420 list_del(&stp->st_perstateowner);
423 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
427 if (stp->st_access_bmap) {
428 for (i = 1; i < 4; i++) {
429 if (test_bit(i, &stp->st_access_bmap))
430 nfs4_file_put_access(stp->st_file,
431 nfs4_access_to_omode(i));
432 __clear_bit(i, &stp->st_access_bmap);
435 put_nfs4_file(stp->st_file);
439 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
441 close_generic_stateid(stp);
442 kmem_cache_free(stateid_slab, stp);
445 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
449 unhash_generic_stateid(stp);
450 file = find_any_file(stp->st_file);
452 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
453 free_generic_stateid(stp);
456 static void unhash_lockowner(struct nfs4_lockowner *lo)
458 struct nfs4_ol_stateid *stp;
460 list_del(&lo->lo_owner.so_idhash);
461 list_del(&lo->lo_owner.so_strhash);
462 list_del(&lo->lo_perstateid);
463 while (!list_empty(&lo->lo_owner.so_stateids)) {
464 stp = list_first_entry(&lo->lo_owner.so_stateids,
465 struct nfs4_ol_stateid, st_perstateowner);
466 release_lock_stateid(stp);
470 static void release_lockowner(struct nfs4_lockowner *lo)
472 unhash_lockowner(lo);
473 nfs4_free_lockowner(lo);
477 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
479 struct nfs4_lockowner *lo;
481 while (!list_empty(&open_stp->st_lockowners)) {
482 lo = list_entry(open_stp->st_lockowners.next,
483 struct nfs4_lockowner, lo_perstateid);
484 release_lockowner(lo);
488 static void release_open_stateid(struct nfs4_ol_stateid *stp)
490 unhash_generic_stateid(stp);
491 release_stateid_lockowners(stp);
492 free_generic_stateid(stp);
495 static void unhash_openowner(struct nfs4_openowner *oo)
497 struct nfs4_ol_stateid *stp;
499 list_del(&oo->oo_owner.so_idhash);
500 list_del(&oo->oo_owner.so_strhash);
501 list_del(&oo->oo_perclient);
502 while (!list_empty(&oo->oo_owner.so_stateids)) {
503 stp = list_first_entry(&oo->oo_owner.so_stateids,
504 struct nfs4_ol_stateid, st_perstateowner);
505 release_open_stateid(stp);
509 static void release_openowner(struct nfs4_openowner *oo)
511 unhash_openowner(oo);
512 list_del(&oo->oo_close_lru);
513 nfs4_free_openowner(oo);
516 #define SESSION_HASH_SIZE 512
517 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
520 hash_sessionid(struct nfs4_sessionid *sessionid)
522 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
524 return sid->sequence % SESSION_HASH_SIZE;
528 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
530 u32 *ptr = (u32 *)(&sessionid->data[0]);
531 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
535 gen_sessionid(struct nfsd4_session *ses)
537 struct nfs4_client *clp = ses->se_client;
538 struct nfsd4_sessionid *sid;
540 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
541 sid->clientid = clp->cl_clientid;
542 sid->sequence = current_sessionid++;
547 * The protocol defines ca_maxresponssize_cached to include the size of
548 * the rpc header, but all we need to cache is the data starting after
549 * the end of the initial SEQUENCE operation--the rest we regenerate
550 * each time. Therefore we can advertise a ca_maxresponssize_cached
551 * value that is the number of bytes in our cache plus a few additional
552 * bytes. In order to stay on the safe side, and not promise more than
553 * we can cache, those additional bytes must be the minimum possible: 24
554 * bytes of rpc header (xid through accept state, with AUTH_NULL
555 * verifier), 12 for the compound header (with zero-length tag), and 44
556 * for the SEQUENCE op response:
558 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
561 free_session_slots(struct nfsd4_session *ses)
565 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
566 kfree(ses->se_slots[i]);
570 * We don't actually need to cache the rpc and session headers, so we
571 * can allocate a little less for each slot:
573 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
575 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
578 static int nfsd4_sanitize_slot_size(u32 size)
580 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
581 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
587 * XXX: If we run out of reserved DRC memory we could (up to a point)
588 * re-negotiate active sessions and reduce their slot usage to make
589 * rooom for new connections. For now we just fail the create session.
591 static int nfsd4_get_drc_mem(int slotsize, u32 num)
595 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
597 spin_lock(&nfsd_drc_lock);
598 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
599 nfsd_drc_max_mem - nfsd_drc_mem_used);
600 num = min_t(int, num, avail / slotsize);
601 nfsd_drc_mem_used += num * slotsize;
602 spin_unlock(&nfsd_drc_lock);
607 static void nfsd4_put_drc_mem(int slotsize, int num)
609 spin_lock(&nfsd_drc_lock);
610 nfsd_drc_mem_used -= slotsize * num;
611 spin_unlock(&nfsd_drc_lock);
614 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
616 struct nfsd4_session *new;
619 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
620 + sizeof(struct nfsd4_session) > PAGE_SIZE);
621 mem = numslots * sizeof(struct nfsd4_slot *);
623 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
626 /* allocate each struct nfsd4_slot and data cache in one piece */
627 for (i = 0; i < numslots; i++) {
628 mem = sizeof(struct nfsd4_slot) + slotsize;
629 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
630 if (!new->se_slots[i])
636 kfree(new->se_slots[i]);
641 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
643 u32 maxrpc = nfsd_serv->sv_max_mesg;
645 new->maxreqs = numslots;
646 new->maxresp_cached = min_t(u32, req->maxresp_cached,
647 slotsize + NFSD_MIN_HDR_SEQ_SZ);
648 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
649 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
650 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
653 static void free_conn(struct nfsd4_conn *c)
655 svc_xprt_put(c->cn_xprt);
659 static void nfsd4_conn_lost(struct svc_xpt_user *u)
661 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
662 struct nfs4_client *clp = c->cn_session->se_client;
664 spin_lock(&clp->cl_lock);
665 if (!list_empty(&c->cn_persession)) {
666 list_del(&c->cn_persession);
669 spin_unlock(&clp->cl_lock);
670 nfsd4_probe_callback(clp);
673 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
675 struct nfsd4_conn *conn;
677 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
680 svc_xprt_get(rqstp->rq_xprt);
681 conn->cn_xprt = rqstp->rq_xprt;
682 conn->cn_flags = flags;
683 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
687 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
689 conn->cn_session = ses;
690 list_add(&conn->cn_persession, &ses->se_conns);
693 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
695 struct nfs4_client *clp = ses->se_client;
697 spin_lock(&clp->cl_lock);
698 __nfsd4_hash_conn(conn, ses);
699 spin_unlock(&clp->cl_lock);
702 static int nfsd4_register_conn(struct nfsd4_conn *conn)
704 conn->cn_xpt_user.callback = nfsd4_conn_lost;
705 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
708 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
710 struct nfsd4_conn *conn;
713 conn = alloc_conn(rqstp, dir);
715 return nfserr_jukebox;
716 nfsd4_hash_conn(conn, ses);
717 ret = nfsd4_register_conn(conn);
719 /* oops; xprt is already down: */
720 nfsd4_conn_lost(&conn->cn_xpt_user);
724 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
726 u32 dir = NFS4_CDFC4_FORE;
728 if (ses->se_flags & SESSION4_BACK_CHAN)
729 dir |= NFS4_CDFC4_BACK;
731 return nfsd4_new_conn(rqstp, ses, dir);
734 /* must be called under client_lock */
735 static void nfsd4_del_conns(struct nfsd4_session *s)
737 struct nfs4_client *clp = s->se_client;
738 struct nfsd4_conn *c;
740 spin_lock(&clp->cl_lock);
741 while (!list_empty(&s->se_conns)) {
742 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
743 list_del_init(&c->cn_persession);
744 spin_unlock(&clp->cl_lock);
746 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
749 spin_lock(&clp->cl_lock);
751 spin_unlock(&clp->cl_lock);
754 void free_session(struct kref *kref)
756 struct nfsd4_session *ses;
759 ses = container_of(kref, struct nfsd4_session, se_ref);
760 nfsd4_del_conns(ses);
761 spin_lock(&nfsd_drc_lock);
762 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
763 nfsd_drc_mem_used -= mem;
764 spin_unlock(&nfsd_drc_lock);
765 free_session_slots(ses);
769 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
771 struct nfsd4_session *new;
772 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
773 int numslots, slotsize;
778 * Note decreasing slot size below client's request may
779 * make it difficult for client to function correctly, whereas
780 * decreasing the number of slots will (just?) affect
781 * performance. When short on memory we therefore prefer to
782 * decrease number of slots instead of their size.
784 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
785 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
789 new = alloc_session(slotsize, numslots);
791 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
794 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
796 new->se_client = clp;
799 INIT_LIST_HEAD(&new->se_conns);
801 new->se_cb_seq_nr = 1;
802 new->se_flags = cses->flags;
803 new->se_cb_prog = cses->callback_prog;
804 kref_init(&new->se_ref);
805 idx = hash_sessionid(&new->se_sessionid);
806 spin_lock(&client_lock);
807 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
808 spin_lock(&clp->cl_lock);
809 list_add(&new->se_perclnt, &clp->cl_sessions);
810 spin_unlock(&clp->cl_lock);
811 spin_unlock(&client_lock);
813 status = nfsd4_new_conn_from_crses(rqstp, new);
814 /* whoops: benny points out, status is ignored! (err, or bogus) */
816 free_session(&new->se_ref);
819 if (cses->flags & SESSION4_BACK_CHAN) {
820 struct sockaddr *sa = svc_addr(rqstp);
822 * This is a little silly; with sessions there's no real
823 * use for the callback address. Use the peer address
824 * as a reasonable default for now, but consider fixing
825 * the rpc client not to require an address in the
828 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
829 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
831 nfsd4_probe_callback(clp);
835 /* caller must hold client_lock */
836 static struct nfsd4_session *
837 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
839 struct nfsd4_session *elem;
842 dump_sessionid(__func__, sessionid);
843 idx = hash_sessionid(sessionid);
844 /* Search in the appropriate list */
845 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
846 if (!memcmp(elem->se_sessionid.data, sessionid->data,
847 NFS4_MAX_SESSIONID_LEN)) {
852 dprintk("%s: session not found\n", __func__);
856 /* caller must hold client_lock */
858 unhash_session(struct nfsd4_session *ses)
860 list_del(&ses->se_hash);
861 spin_lock(&ses->se_client->cl_lock);
862 list_del(&ses->se_perclnt);
863 spin_unlock(&ses->se_client->cl_lock);
866 /* must be called under the client_lock */
868 renew_client_locked(struct nfs4_client *clp)
870 if (is_client_expired(clp)) {
871 dprintk("%s: client (clientid %08x/%08x) already expired\n",
873 clp->cl_clientid.cl_boot,
874 clp->cl_clientid.cl_id);
879 * Move client to the end to the LRU list.
881 dprintk("renewing client (clientid %08x/%08x)\n",
882 clp->cl_clientid.cl_boot,
883 clp->cl_clientid.cl_id);
884 list_move_tail(&clp->cl_lru, &client_lru);
885 clp->cl_time = get_seconds();
889 renew_client(struct nfs4_client *clp)
891 spin_lock(&client_lock);
892 renew_client_locked(clp);
893 spin_unlock(&client_lock);
896 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
898 STALE_CLIENTID(clientid_t *clid)
900 if (clid->cl_boot == boot_time)
902 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
903 clid->cl_boot, clid->cl_id, boot_time);
908 * XXX Should we use a slab cache ?
909 * This type of memory management is somewhat inefficient, but we use it
910 * anyway since SETCLIENTID is not a common operation.
912 static struct nfs4_client *alloc_client(struct xdr_netobj name)
914 struct nfs4_client *clp;
916 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
919 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
920 if (clp->cl_name.data == NULL) {
924 memcpy(clp->cl_name.data, name.data, name.len);
925 clp->cl_name.len = name.len;
930 free_client(struct nfs4_client *clp)
932 while (!list_empty(&clp->cl_sessions)) {
933 struct nfsd4_session *ses;
934 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
936 list_del(&ses->se_perclnt);
937 nfsd4_put_session(ses);
939 if (clp->cl_cred.cr_group_info)
940 put_group_info(clp->cl_cred.cr_group_info);
941 kfree(clp->cl_principal);
942 kfree(clp->cl_name.data);
947 release_session_client(struct nfsd4_session *session)
949 struct nfs4_client *clp = session->se_client;
951 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
953 if (is_client_expired(clp)) {
955 session->se_client = NULL;
957 renew_client_locked(clp);
958 spin_unlock(&client_lock);
961 /* must be called under the client_lock */
963 unhash_client_locked(struct nfs4_client *clp)
965 struct nfsd4_session *ses;
967 mark_client_expired(clp);
968 list_del(&clp->cl_lru);
969 spin_lock(&clp->cl_lock);
970 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
971 list_del_init(&ses->se_hash);
972 spin_unlock(&clp->cl_lock);
976 expire_client(struct nfs4_client *clp)
978 struct nfs4_openowner *oo;
979 struct nfs4_delegation *dp;
980 struct list_head reaplist;
982 INIT_LIST_HEAD(&reaplist);
983 spin_lock(&recall_lock);
984 while (!list_empty(&clp->cl_delegations)) {
985 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
986 list_del_init(&dp->dl_perclnt);
987 list_move(&dp->dl_recall_lru, &reaplist);
989 spin_unlock(&recall_lock);
990 while (!list_empty(&reaplist)) {
991 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
992 list_del_init(&dp->dl_recall_lru);
993 unhash_delegation(dp);
995 while (!list_empty(&clp->cl_openowners)) {
996 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
997 release_openowner(oo);
999 nfsd4_shutdown_callback(clp);
1000 if (clp->cl_cb_conn.cb_xprt)
1001 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1002 list_del(&clp->cl_idhash);
1003 list_del(&clp->cl_strhash);
1004 spin_lock(&client_lock);
1005 unhash_client_locked(clp);
1006 if (atomic_read(&clp->cl_refcount) == 0)
1008 spin_unlock(&client_lock);
1011 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1013 memcpy(target->cl_verifier.data, source->data,
1014 sizeof(target->cl_verifier.data));
1017 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1019 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1020 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1023 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1025 target->cr_uid = source->cr_uid;
1026 target->cr_gid = source->cr_gid;
1027 target->cr_group_info = source->cr_group_info;
1028 get_group_info(target->cr_group_info);
1031 static int same_name(const char *n1, const char *n2)
1033 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1037 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1039 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1043 same_clid(clientid_t *cl1, clientid_t *cl2)
1045 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1048 /* XXX what about NGROUP */
1050 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1052 return cr1->cr_uid == cr2->cr_uid;
1055 static void gen_clid(struct nfs4_client *clp)
1057 static u32 current_clientid = 1;
1059 clp->cl_clientid.cl_boot = boot_time;
1060 clp->cl_clientid.cl_id = current_clientid++;
1063 static void gen_confirm(struct nfs4_client *clp)
1068 p = (u32 *)clp->cl_confirm.data;
1069 *p++ = get_seconds();
1074 same_stateid(stateid_t *id_one, stateid_t *id_two)
1076 if (id_one->si_stateownerid != id_two->si_stateownerid)
1078 return id_one->si_fileid == id_two->si_fileid;
1081 static struct nfs4_stid *find_stateid(stateid_t *t)
1083 struct nfs4_stid *s;
1084 unsigned int hashval;
1086 hashval = stateid_hashval(t);
1087 list_for_each_entry(s, &stateid_hashtbl[hashval], sc_hash)
1088 if (same_stateid(&s->sc_stateid, t))
1093 static struct nfs4_ol_stateid *find_ol_stateid(stateid_t *t)
1095 struct nfs4_stid *s;
1097 s = find_stateid(t);
1100 return openlockstateid(s);
1103 static struct nfs4_stid *find_stateid_by_type(stateid_t *t, char typemask)
1105 struct nfs4_stid *s;
1107 s = find_stateid(t);
1110 if (typemask & s->sc_type)
1115 static struct nfs4_ol_stateid *find_ol_stateid_by_type(stateid_t *t, char typemask)
1117 struct nfs4_stid *s;
1119 s = find_stateid_by_type(t, typemask);
1122 return openlockstateid(s);
1125 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1126 struct svc_rqst *rqstp, nfs4_verifier *verf)
1128 struct nfs4_client *clp;
1129 struct sockaddr *sa = svc_addr(rqstp);
1132 clp = alloc_client(name);
1136 INIT_LIST_HEAD(&clp->cl_sessions);
1138 princ = svc_gss_principal(rqstp);
1140 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1141 if (clp->cl_principal == NULL) {
1147 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1148 atomic_set(&clp->cl_refcount, 0);
1149 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1150 INIT_LIST_HEAD(&clp->cl_idhash);
1151 INIT_LIST_HEAD(&clp->cl_strhash);
1152 INIT_LIST_HEAD(&clp->cl_openowners);
1153 INIT_LIST_HEAD(&clp->cl_delegations);
1154 INIT_LIST_HEAD(&clp->cl_lru);
1155 INIT_LIST_HEAD(&clp->cl_callbacks);
1156 spin_lock_init(&clp->cl_lock);
1157 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1158 clp->cl_time = get_seconds();
1159 clear_bit(0, &clp->cl_cb_slot_busy);
1160 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1161 copy_verf(clp, verf);
1162 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1163 clp->cl_flavor = rqstp->rq_flavor;
1164 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1166 clp->cl_cb_session = NULL;
1170 static int check_name(struct xdr_netobj name)
1174 if (name.len > NFS4_OPAQUE_LIMIT) {
1175 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1182 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1184 unsigned int idhashval;
1186 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1187 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1188 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1193 move_to_confirmed(struct nfs4_client *clp)
1195 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1196 unsigned int strhashval;
1198 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1199 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1200 strhashval = clientstr_hashval(clp->cl_recdir);
1201 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1205 static struct nfs4_client *
1206 find_confirmed_client(clientid_t *clid)
1208 struct nfs4_client *clp;
1209 unsigned int idhashval = clientid_hashval(clid->cl_id);
1211 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1212 if (same_clid(&clp->cl_clientid, clid))
1218 static struct nfs4_client *
1219 find_unconfirmed_client(clientid_t *clid)
1221 struct nfs4_client *clp;
1222 unsigned int idhashval = clientid_hashval(clid->cl_id);
1224 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1225 if (same_clid(&clp->cl_clientid, clid))
1231 static bool clp_used_exchangeid(struct nfs4_client *clp)
1233 return clp->cl_exchange_flags != 0;
1236 static struct nfs4_client *
1237 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1239 struct nfs4_client *clp;
1241 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1242 if (same_name(clp->cl_recdir, dname))
1248 static struct nfs4_client *
1249 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1251 struct nfs4_client *clp;
1253 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1254 if (same_name(clp->cl_recdir, dname))
1261 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1263 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1264 struct sockaddr *sa = svc_addr(rqstp);
1265 u32 scopeid = rpc_get_scope_id(sa);
1266 unsigned short expected_family;
1268 /* Currently, we only support tcp and tcp6 for the callback channel */
1269 if (se->se_callback_netid_len == 3 &&
1270 !memcmp(se->se_callback_netid_val, "tcp", 3))
1271 expected_family = AF_INET;
1272 else if (se->se_callback_netid_len == 4 &&
1273 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1274 expected_family = AF_INET6;
1278 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1279 se->se_callback_addr_len,
1280 (struct sockaddr *)&conn->cb_addr,
1281 sizeof(conn->cb_addr));
1283 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1286 if (conn->cb_addr.ss_family == AF_INET6)
1287 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1289 conn->cb_prog = se->se_callback_prog;
1290 conn->cb_ident = se->se_callback_ident;
1291 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1294 conn->cb_addr.ss_family = AF_UNSPEC;
1295 conn->cb_addrlen = 0;
1296 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1297 "will not receive delegations\n",
1298 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1304 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1307 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1309 struct nfsd4_slot *slot = resp->cstate.slot;
1312 dprintk("--> %s slot %p\n", __func__, slot);
1314 slot->sl_opcnt = resp->opcnt;
1315 slot->sl_status = resp->cstate.status;
1317 if (nfsd4_not_cached(resp)) {
1318 slot->sl_datalen = 0;
1321 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1322 base = (char *)resp->cstate.datap -
1323 (char *)resp->xbuf->head[0].iov_base;
1324 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1326 WARN("%s: sessions DRC could not cache compound\n", __func__);
1331 * Encode the replay sequence operation from the slot values.
1332 * If cachethis is FALSE encode the uncached rep error on the next
1333 * operation which sets resp->p and increments resp->opcnt for
1334 * nfs4svc_encode_compoundres.
1338 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1339 struct nfsd4_compoundres *resp)
1341 struct nfsd4_op *op;
1342 struct nfsd4_slot *slot = resp->cstate.slot;
1344 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1345 resp->opcnt, resp->cstate.slot->sl_cachethis);
1347 /* Encode the replayed sequence operation */
1348 op = &args->ops[resp->opcnt - 1];
1349 nfsd4_encode_operation(resp, op);
1351 /* Return nfserr_retry_uncached_rep in next operation. */
1352 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1353 op = &args->ops[resp->opcnt++];
1354 op->status = nfserr_retry_uncached_rep;
1355 nfsd4_encode_operation(resp, op);
1361 * The sequence operation is not cached because we can use the slot and
1365 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1366 struct nfsd4_sequence *seq)
1368 struct nfsd4_slot *slot = resp->cstate.slot;
1371 dprintk("--> %s slot %p\n", __func__, slot);
1373 /* Either returns 0 or nfserr_retry_uncached */
1374 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1375 if (status == nfserr_retry_uncached_rep)
1378 /* The sequence operation has been encoded, cstate->datap set. */
1379 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1381 resp->opcnt = slot->sl_opcnt;
1382 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1383 status = slot->sl_status;
1389 * Set the exchange_id flags returned by the server.
1392 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1394 /* pNFS is not supported */
1395 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1397 /* Referrals are supported, Migration is not. */
1398 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1400 /* set the wire flags to return to client. */
1401 clid->flags = new->cl_exchange_flags;
1405 nfsd4_exchange_id(struct svc_rqst *rqstp,
1406 struct nfsd4_compound_state *cstate,
1407 struct nfsd4_exchange_id *exid)
1409 struct nfs4_client *unconf, *conf, *new;
1411 unsigned int strhashval;
1412 char dname[HEXDIR_LEN];
1413 char addr_str[INET6_ADDRSTRLEN];
1414 nfs4_verifier verf = exid->verifier;
1415 struct sockaddr *sa = svc_addr(rqstp);
1417 rpc_ntop(sa, addr_str, sizeof(addr_str));
1418 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1419 "ip_addr=%s flags %x, spa_how %d\n",
1420 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1421 addr_str, exid->flags, exid->spa_how);
1423 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1424 return nfserr_inval;
1426 /* Currently only support SP4_NONE */
1427 switch (exid->spa_how) {
1431 return nfserr_serverfault;
1433 BUG(); /* checked by xdr code */
1435 return nfserr_serverfault; /* no excuse :-/ */
1438 status = nfs4_make_rec_clidname(dname, &exid->clname);
1443 strhashval = clientstr_hashval(dname);
1448 conf = find_confirmed_client_by_str(dname, strhashval);
1450 if (!clp_used_exchangeid(conf)) {
1451 status = nfserr_clid_inuse; /* XXX: ? */
1454 if (!same_verf(&verf, &conf->cl_verifier)) {
1455 /* 18.35.4 case 8 */
1456 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1457 status = nfserr_not_same;
1460 /* Client reboot: destroy old state */
1461 expire_client(conf);
1464 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1465 /* 18.35.4 case 9 */
1466 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1467 status = nfserr_perm;
1470 expire_client(conf);
1474 * Set bit when the owner id and verifier map to an already
1475 * confirmed client id (18.35.3).
1477 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1480 * Falling into 18.35.4 case 2, possible router replay.
1481 * Leave confirmed record intact and return same result.
1483 copy_verf(conf, &verf);
1488 /* 18.35.4 case 7 */
1489 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1490 status = nfserr_noent;
1494 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1497 * Possible retry or client restart. Per 18.35.4 case 4,
1498 * a new unconfirmed record should be generated regardless
1499 * of whether any properties have changed.
1501 expire_client(unconf);
1506 new = create_client(exid->clname, dname, rqstp, &verf);
1508 status = nfserr_jukebox;
1513 add_to_unconfirmed(new, strhashval);
1515 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1516 exid->clientid.cl_id = new->cl_clientid.cl_id;
1519 nfsd4_set_ex_flags(new, exid);
1521 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1522 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1526 nfs4_unlock_state();
1528 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1533 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1535 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1538 /* The slot is in use, and no response has been sent. */
1540 if (seqid == slot_seqid)
1541 return nfserr_jukebox;
1543 return nfserr_seq_misordered;
1546 if (likely(seqid == slot_seqid + 1))
1549 if (seqid == slot_seqid)
1550 return nfserr_replay_cache;
1552 if (seqid == 1 && (slot_seqid + 1) == 0)
1554 /* Misordered replay or misordered new request */
1555 return nfserr_seq_misordered;
1559 * Cache the create session result into the create session single DRC
1560 * slot cache by saving the xdr structure. sl_seqid has been set.
1561 * Do this for solo or embedded create session operations.
1564 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1565 struct nfsd4_clid_slot *slot, int nfserr)
1567 slot->sl_status = nfserr;
1568 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1572 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1573 struct nfsd4_clid_slot *slot)
1575 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1576 return slot->sl_status;
1579 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1580 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1581 1 + /* MIN tag is length with zero, only length */ \
1582 3 + /* version, opcount, opcode */ \
1583 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1584 /* seqid, slotID, slotID, cache */ \
1585 4 ) * sizeof(__be32))
1587 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1588 2 + /* verifier: AUTH_NULL, length 0 */\
1590 1 + /* MIN tag is length with zero, only length */ \
1591 3 + /* opcount, opcode, opstatus*/ \
1592 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1593 /* seqid, slotID, slotID, slotID, status */ \
1594 5 ) * sizeof(__be32))
1596 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1598 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1599 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1603 nfsd4_create_session(struct svc_rqst *rqstp,
1604 struct nfsd4_compound_state *cstate,
1605 struct nfsd4_create_session *cr_ses)
1607 struct sockaddr *sa = svc_addr(rqstp);
1608 struct nfs4_client *conf, *unconf;
1609 struct nfsd4_session *new;
1610 struct nfsd4_clid_slot *cs_slot = NULL;
1611 bool confirm_me = false;
1614 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1615 return nfserr_inval;
1618 unconf = find_unconfirmed_client(&cr_ses->clientid);
1619 conf = find_confirmed_client(&cr_ses->clientid);
1622 cs_slot = &conf->cl_cs_slot;
1623 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1624 if (status == nfserr_replay_cache) {
1625 dprintk("Got a create_session replay! seqid= %d\n",
1627 /* Return the cached reply status */
1628 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1630 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1631 status = nfserr_seq_misordered;
1632 dprintk("Sequence misordered!\n");
1633 dprintk("Expected seqid= %d but got seqid= %d\n",
1634 cs_slot->sl_seqid, cr_ses->seqid);
1637 } else if (unconf) {
1638 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1639 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1640 status = nfserr_clid_inuse;
1644 cs_slot = &unconf->cl_cs_slot;
1645 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1647 /* an unconfirmed replay returns misordered */
1648 status = nfserr_seq_misordered;
1655 status = nfserr_stale_clientid;
1660 * XXX: we should probably set this at creation time, and check
1661 * for consistent minorversion use throughout:
1663 conf->cl_minorversion = 1;
1665 * We do not support RDMA or persistent sessions
1667 cr_ses->flags &= ~SESSION4_PERSIST;
1668 cr_ses->flags &= ~SESSION4_RDMA;
1670 status = nfserr_toosmall;
1671 if (check_forechannel_attrs(cr_ses->fore_channel))
1674 status = nfserr_jukebox;
1675 new = alloc_init_session(rqstp, conf, cr_ses);
1679 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1680 NFS4_MAX_SESSIONID_LEN);
1681 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1682 sizeof(struct nfsd4_channel_attrs));
1683 cs_slot->sl_seqid++;
1684 cr_ses->seqid = cs_slot->sl_seqid;
1686 /* cache solo and embedded create sessions under the state lock */
1687 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1689 move_to_confirmed(conf);
1691 nfs4_unlock_state();
1692 dprintk("%s returns %d\n", __func__, ntohl(status));
1696 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1698 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1699 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1701 return argp->opcnt == resp->opcnt;
1704 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1707 case NFS4_CDFC4_FORE:
1708 case NFS4_CDFC4_BACK:
1710 case NFS4_CDFC4_FORE_OR_BOTH:
1711 case NFS4_CDFC4_BACK_OR_BOTH:
1712 *dir = NFS4_CDFC4_BOTH;
1715 return nfserr_inval;
1718 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1719 struct nfsd4_compound_state *cstate,
1720 struct nfsd4_bind_conn_to_session *bcts)
1724 if (!nfsd4_last_compound_op(rqstp))
1725 return nfserr_not_only_op;
1726 spin_lock(&client_lock);
1727 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1728 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1729 * client_lock iself: */
1730 if (cstate->session) {
1731 nfsd4_get_session(cstate->session);
1732 atomic_inc(&cstate->session->se_client->cl_refcount);
1734 spin_unlock(&client_lock);
1735 if (!cstate->session)
1736 return nfserr_badsession;
1738 status = nfsd4_map_bcts_dir(&bcts->dir);
1740 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1744 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1748 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1752 nfsd4_destroy_session(struct svc_rqst *r,
1753 struct nfsd4_compound_state *cstate,
1754 struct nfsd4_destroy_session *sessionid)
1756 struct nfsd4_session *ses;
1757 u32 status = nfserr_badsession;
1760 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1761 * - Should we return nfserr_back_chan_busy if waiting for
1762 * callbacks on to-be-destroyed session?
1763 * - Do we need to clear any callback info from previous session?
1766 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1767 if (!nfsd4_last_compound_op(r))
1768 return nfserr_not_only_op;
1770 dump_sessionid(__func__, &sessionid->sessionid);
1771 spin_lock(&client_lock);
1772 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1774 spin_unlock(&client_lock);
1778 unhash_session(ses);
1779 spin_unlock(&client_lock);
1782 nfsd4_probe_callback_sync(ses->se_client);
1783 nfs4_unlock_state();
1785 nfsd4_del_conns(ses);
1787 nfsd4_put_session(ses);
1790 dprintk("%s returns %d\n", __func__, ntohl(status));
1794 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1796 struct nfsd4_conn *c;
1798 list_for_each_entry(c, &s->se_conns, cn_persession) {
1799 if (c->cn_xprt == xpt) {
1806 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1808 struct nfs4_client *clp = ses->se_client;
1809 struct nfsd4_conn *c;
1812 spin_lock(&clp->cl_lock);
1813 c = __nfsd4_find_conn(new->cn_xprt, ses);
1815 spin_unlock(&clp->cl_lock);
1819 __nfsd4_hash_conn(new, ses);
1820 spin_unlock(&clp->cl_lock);
1821 ret = nfsd4_register_conn(new);
1823 /* oops; xprt is already down: */
1824 nfsd4_conn_lost(&new->cn_xpt_user);
1828 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1830 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1832 return args->opcnt > session->se_fchannel.maxops;
1835 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1836 struct nfsd4_session *session)
1838 struct xdr_buf *xb = &rqstp->rq_arg;
1840 return xb->len > session->se_fchannel.maxreq_sz;
1844 nfsd4_sequence(struct svc_rqst *rqstp,
1845 struct nfsd4_compound_state *cstate,
1846 struct nfsd4_sequence *seq)
1848 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1849 struct nfsd4_session *session;
1850 struct nfsd4_slot *slot;
1851 struct nfsd4_conn *conn;
1854 if (resp->opcnt != 1)
1855 return nfserr_sequence_pos;
1858 * Will be either used or freed by nfsd4_sequence_check_conn
1861 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1863 return nfserr_jukebox;
1865 spin_lock(&client_lock);
1866 status = nfserr_badsession;
1867 session = find_in_sessionid_hashtbl(&seq->sessionid);
1871 status = nfserr_too_many_ops;
1872 if (nfsd4_session_too_many_ops(rqstp, session))
1875 status = nfserr_req_too_big;
1876 if (nfsd4_request_too_big(rqstp, session))
1879 status = nfserr_badslot;
1880 if (seq->slotid >= session->se_fchannel.maxreqs)
1883 slot = session->se_slots[seq->slotid];
1884 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1886 /* We do not negotiate the number of slots yet, so set the
1887 * maxslots to the session maxreqs which is used to encode
1888 * sr_highest_slotid and the sr_target_slot id to maxslots */
1889 seq->maxslots = session->se_fchannel.maxreqs;
1891 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1892 if (status == nfserr_replay_cache) {
1893 cstate->slot = slot;
1894 cstate->session = session;
1895 /* Return the cached reply status and set cstate->status
1896 * for nfsd4_proc_compound processing */
1897 status = nfsd4_replay_cache_entry(resp, seq);
1898 cstate->status = nfserr_replay_cache;
1904 nfsd4_sequence_check_conn(conn, session);
1907 /* Success! bump slot seqid */
1908 slot->sl_inuse = true;
1909 slot->sl_seqid = seq->seqid;
1910 slot->sl_cachethis = seq->cachethis;
1912 cstate->slot = slot;
1913 cstate->session = session;
1916 /* Hold a session reference until done processing the compound. */
1917 if (cstate->session) {
1918 struct nfs4_client *clp = session->se_client;
1920 nfsd4_get_session(cstate->session);
1921 atomic_inc(&clp->cl_refcount);
1922 if (clp->cl_cb_state == NFSD4_CB_DOWN)
1923 seq->status_flags |= SEQ4_STATUS_CB_PATH_DOWN;
1926 spin_unlock(&client_lock);
1927 dprintk("%s: return %d\n", __func__, ntohl(status));
1932 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1936 if (rc->rca_one_fs) {
1937 if (!cstate->current_fh.fh_dentry)
1938 return nfserr_nofilehandle;
1940 * We don't take advantage of the rca_one_fs case.
1941 * That's OK, it's optional, we can safely ignore it.
1947 status = nfserr_complete_already;
1948 if (cstate->session->se_client->cl_firststate)
1951 status = nfserr_stale_clientid;
1952 if (is_client_expired(cstate->session->se_client))
1954 * The following error isn't really legal.
1955 * But we only get here if the client just explicitly
1956 * destroyed the client. Surely it no longer cares what
1957 * error it gets back on an operation for the dead
1963 nfsd4_create_clid_dir(cstate->session->se_client);
1965 nfs4_unlock_state();
1970 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1971 struct nfsd4_setclientid *setclid)
1973 struct xdr_netobj clname = {
1974 .len = setclid->se_namelen,
1975 .data = setclid->se_name,
1977 nfs4_verifier clverifier = setclid->se_verf;
1978 unsigned int strhashval;
1979 struct nfs4_client *conf, *unconf, *new;
1981 char dname[HEXDIR_LEN];
1983 if (!check_name(clname))
1984 return nfserr_inval;
1986 status = nfs4_make_rec_clidname(dname, &clname);
1991 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1992 * We get here on a DRC miss.
1995 strhashval = clientstr_hashval(dname);
1998 conf = find_confirmed_client_by_str(dname, strhashval);
2000 /* RFC 3530 14.2.33 CASE 0: */
2001 status = nfserr_clid_inuse;
2002 if (clp_used_exchangeid(conf))
2004 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2005 char addr_str[INET6_ADDRSTRLEN];
2006 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2008 dprintk("NFSD: setclientid: string in use by client "
2009 "at %s\n", addr_str);
2014 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2015 * has a description of SETCLIENTID request processing consisting
2016 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2018 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2019 status = nfserr_jukebox;
2022 * RFC 3530 14.2.33 CASE 4:
2023 * placed first, because it is the normal case
2026 expire_client(unconf);
2027 new = create_client(clname, dname, rqstp, &clverifier);
2031 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2033 * RFC 3530 14.2.33 CASE 1:
2034 * probable callback update
2037 /* Note this is removing unconfirmed {*x***},
2038 * which is stronger than RFC recommended {vxc**}.
2039 * This has the advantage that there is at most
2040 * one {*x***} in either list at any time.
2042 expire_client(unconf);
2044 new = create_client(clname, dname, rqstp, &clverifier);
2047 copy_clid(new, conf);
2048 } else if (!unconf) {
2050 * RFC 3530 14.2.33 CASE 2:
2051 * probable client reboot; state will be removed if
2054 new = create_client(clname, dname, rqstp, &clverifier);
2060 * RFC 3530 14.2.33 CASE 3:
2061 * probable client reboot; state will be removed if
2064 expire_client(unconf);
2065 new = create_client(clname, dname, rqstp, &clverifier);
2071 * XXX: we should probably set this at creation time, and check
2072 * for consistent minorversion use throughout:
2074 new->cl_minorversion = 0;
2075 gen_callback(new, setclid, rqstp);
2076 add_to_unconfirmed(new, strhashval);
2077 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2078 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2079 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2082 nfs4_unlock_state();
2088 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2089 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2090 * bullets, labeled as CASE1 - CASE4 below.
2093 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2094 struct nfsd4_compound_state *cstate,
2095 struct nfsd4_setclientid_confirm *setclientid_confirm)
2097 struct sockaddr *sa = svc_addr(rqstp);
2098 struct nfs4_client *conf, *unconf;
2099 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2100 clientid_t * clid = &setclientid_confirm->sc_clientid;
2103 if (STALE_CLIENTID(clid))
2104 return nfserr_stale_clientid;
2106 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2107 * We get here on a DRC miss.
2112 conf = find_confirmed_client(clid);
2113 unconf = find_unconfirmed_client(clid);
2115 status = nfserr_clid_inuse;
2116 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2118 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2122 * section 14.2.34 of RFC 3530 has a description of
2123 * SETCLIENTID_CONFIRM request processing consisting
2124 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2126 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2128 * RFC 3530 14.2.34 CASE 1:
2131 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2132 status = nfserr_clid_inuse;
2134 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2135 nfsd4_probe_callback(conf);
2136 expire_client(unconf);
2140 } else if (conf && !unconf) {
2142 * RFC 3530 14.2.34 CASE 2:
2143 * probable retransmitted request; play it safe and
2146 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2147 status = nfserr_clid_inuse;
2150 } else if (!conf && unconf
2151 && same_verf(&unconf->cl_confirm, &confirm)) {
2153 * RFC 3530 14.2.34 CASE 3:
2154 * Normal case; new or rebooted client:
2156 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2157 status = nfserr_clid_inuse;
2160 clientstr_hashval(unconf->cl_recdir);
2161 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2164 nfsd4_remove_clid_dir(conf);
2165 expire_client(conf);
2167 move_to_confirmed(unconf);
2169 nfsd4_probe_callback(conf);
2172 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2173 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2176 * RFC 3530 14.2.34 CASE 4:
2177 * Client probably hasn't noticed that we rebooted yet.
2179 status = nfserr_stale_clientid;
2181 /* check that we have hit one of the cases...*/
2182 status = nfserr_clid_inuse;
2185 nfs4_unlock_state();
2189 /* OPEN Share state helper functions */
2190 static inline struct nfs4_file *
2191 alloc_init_file(struct inode *ino)
2193 struct nfs4_file *fp;
2194 unsigned int hashval = file_hashval(ino);
2196 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
2198 atomic_set(&fp->fi_ref, 1);
2199 INIT_LIST_HEAD(&fp->fi_hash);
2200 INIT_LIST_HEAD(&fp->fi_stateids);
2201 INIT_LIST_HEAD(&fp->fi_delegations);
2202 fp->fi_inode = igrab(ino);
2203 fp->fi_id = current_fileid++;
2204 fp->fi_had_conflict = false;
2205 fp->fi_lease = NULL;
2206 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2207 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2208 spin_lock(&recall_lock);
2209 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2210 spin_unlock(&recall_lock);
2217 nfsd4_free_slab(struct kmem_cache **slab)
2221 kmem_cache_destroy(*slab);
2226 nfsd4_free_slabs(void)
2228 nfsd4_free_slab(&openowner_slab);
2229 nfsd4_free_slab(&lockowner_slab);
2230 nfsd4_free_slab(&file_slab);
2231 nfsd4_free_slab(&stateid_slab);
2232 nfsd4_free_slab(&deleg_slab);
2236 nfsd4_init_slabs(void)
2238 openowner_slab = kmem_cache_create("nfsd4_openowners",
2239 sizeof(struct nfs4_openowner), 0, 0, NULL);
2240 if (openowner_slab == NULL)
2242 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2243 sizeof(struct nfs4_openowner), 0, 0, NULL);
2244 if (lockowner_slab == NULL)
2246 file_slab = kmem_cache_create("nfsd4_files",
2247 sizeof(struct nfs4_file), 0, 0, NULL);
2248 if (file_slab == NULL)
2250 stateid_slab = kmem_cache_create("nfsd4_stateids",
2251 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2252 if (stateid_slab == NULL)
2254 deleg_slab = kmem_cache_create("nfsd4_delegations",
2255 sizeof(struct nfs4_delegation), 0, 0, NULL);
2256 if (deleg_slab == NULL)
2261 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2265 void nfs4_free_openowner(struct nfs4_openowner *oo)
2267 kfree(oo->oo_owner.so_owner.data);
2268 kmem_cache_free(openowner_slab, oo);
2271 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2273 kfree(lo->lo_owner.so_owner.data);
2274 kmem_cache_free(lockowner_slab, lo);
2277 static void init_nfs4_replay(struct nfs4_replay *rp)
2279 rp->rp_status = nfserr_serverfault;
2281 rp->rp_buf = rp->rp_ibuf;
2284 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2286 struct nfs4_stateowner *sop;
2288 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2292 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2293 if (!sop->so_owner.data) {
2294 kmem_cache_free(slab, sop);
2297 sop->so_owner.len = owner->len;
2299 INIT_LIST_HEAD(&sop->so_stateids);
2300 sop->so_id = current_ownerid++;
2301 sop->so_client = clp;
2302 init_nfs4_replay(&sop->so_replay);
2306 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2308 unsigned int idhashval;
2310 idhashval = open_ownerid_hashval(oo->oo_owner.so_id);
2311 list_add(&oo->oo_owner.so_idhash, &open_ownerid_hashtbl[idhashval]);
2312 list_add(&oo->oo_owner.so_strhash, &open_ownerstr_hashtbl[strhashval]);
2313 list_add(&oo->oo_perclient, &clp->cl_openowners);
2316 static struct nfs4_openowner *
2317 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2318 struct nfs4_openowner *oo;
2320 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2323 oo->oo_owner.so_is_open_owner = 1;
2324 oo->oo_owner.so_seqid = open->op_seqid;
2327 INIT_LIST_HEAD(&oo->oo_close_lru);
2328 hash_openowner(oo, clp, strhashval);
2333 init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2334 struct nfs4_openowner *oo = open->op_openowner;
2336 INIT_LIST_HEAD(&stp->st_lockowners);
2337 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2338 list_add(&stp->st_perfile, &fp->fi_stateids);
2339 stp->st_stid.sc_type = NFS4_OPEN_STID;
2340 stp->st_stateowner = &oo->oo_owner;
2343 stp->st_stid.sc_stateid.si_boot = boot_time;
2344 stp->st_stid.sc_stateid.si_stateownerid = oo->oo_owner.so_id;
2345 stp->st_stid.sc_stateid.si_fileid = fp->fi_id;
2346 /* note will be incremented before first return to client: */
2347 stp->st_stid.sc_stateid.si_generation = 0;
2348 hash_stid(&stp->st_stid);
2349 stp->st_access_bmap = 0;
2350 stp->st_deny_bmap = 0;
2351 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
2352 &stp->st_access_bmap);
2353 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2354 stp->st_openstp = NULL;
2358 move_to_close_lru(struct nfs4_openowner *oo)
2360 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2362 list_move_tail(&oo->oo_close_lru, &close_lru);
2363 oo->oo_time = get_seconds();
2367 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2370 return (sop->so_owner.len == owner->len) &&
2371 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2372 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2375 static struct nfs4_openowner *
2376 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2378 struct nfs4_stateowner *so = NULL;
2380 list_for_each_entry(so, &open_ownerstr_hashtbl[hashval], so_strhash) {
2381 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
2382 return container_of(so, struct nfs4_openowner, oo_owner);
2387 /* search file_hashtbl[] for file */
2388 static struct nfs4_file *
2389 find_file(struct inode *ino)
2391 unsigned int hashval = file_hashval(ino);
2392 struct nfs4_file *fp;
2394 spin_lock(&recall_lock);
2395 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2396 if (fp->fi_inode == ino) {
2398 spin_unlock(&recall_lock);
2402 spin_unlock(&recall_lock);
2406 static inline int access_valid(u32 x, u32 minorversion)
2408 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
2410 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
2412 x &= ~NFS4_SHARE_ACCESS_MASK;
2413 if (minorversion && x) {
2414 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
2416 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
2418 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
2425 static inline int deny_valid(u32 x)
2427 /* Note: unlike access bits, deny bits may be zero. */
2428 return x <= NFS4_SHARE_DENY_BOTH;
2432 * Called to check deny when READ with all zero stateid or
2433 * WRITE with all zero or all one stateid
2436 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2438 struct inode *ino = current_fh->fh_dentry->d_inode;
2439 struct nfs4_file *fp;
2440 struct nfs4_ol_stateid *stp;
2443 dprintk("NFSD: nfs4_share_conflict\n");
2445 fp = find_file(ino);
2448 ret = nfserr_locked;
2449 /* Search for conflicting share reservations */
2450 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2451 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2452 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2461 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2463 /* We're assuming the state code never drops its reference
2464 * without first removing the lease. Since we're in this lease
2465 * callback (and since the lease code is serialized by the kernel
2466 * lock) we know the server hasn't removed the lease yet, we know
2467 * it's safe to take a reference: */
2468 atomic_inc(&dp->dl_count);
2470 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2472 /* only place dl_time is set. protected by lock_flocks*/
2473 dp->dl_time = get_seconds();
2475 nfsd4_cb_recall(dp);
2478 /* Called from break_lease() with lock_flocks() held. */
2479 static void nfsd_break_deleg_cb(struct file_lock *fl)
2481 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2482 struct nfs4_delegation *dp;
2485 /* We assume break_lease is only called once per lease: */
2486 BUG_ON(fp->fi_had_conflict);
2488 * We don't want the locks code to timeout the lease for us;
2489 * we'll remove it ourself if a delegation isn't returned
2492 fl->fl_break_time = 0;
2494 spin_lock(&recall_lock);
2495 fp->fi_had_conflict = true;
2496 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2497 nfsd_break_one_deleg(dp);
2498 spin_unlock(&recall_lock);
2502 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2505 return lease_modify(onlist, arg);
2510 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2511 .lm_break = nfsd_break_deleg_cb,
2512 .lm_change = nfsd_change_deleg_cb,
2515 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2517 if (nfsd4_has_session(cstate))
2519 if (seqid == so->so_seqid - 1)
2520 return nfserr_replay_me;
2521 if (seqid == so->so_seqid)
2523 return nfserr_bad_seqid;
2527 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2528 struct nfsd4_open *open)
2530 clientid_t *clientid = &open->op_clientid;
2531 struct nfs4_client *clp = NULL;
2532 unsigned int strhashval;
2533 struct nfs4_openowner *oo = NULL;
2536 if (!check_name(open->op_owner))
2537 return nfserr_inval;
2539 if (STALE_CLIENTID(&open->op_clientid))
2540 return nfserr_stale_clientid;
2542 strhashval = open_ownerstr_hashval(clientid->cl_id, &open->op_owner);
2543 oo = find_openstateowner_str(strhashval, open);
2544 open->op_openowner = oo;
2546 /* Make sure the client's lease hasn't expired. */
2547 clp = find_confirmed_client(clientid);
2549 return nfserr_expired;
2552 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2553 /* Replace unconfirmed owners without checking for replay. */
2554 clp = oo->oo_owner.so_client;
2555 release_openowner(oo);
2556 open->op_openowner = NULL;
2559 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2563 if (open->op_openowner == NULL) {
2564 oo = alloc_init_open_stateowner(strhashval, clp, open);
2566 return nfserr_jukebox;
2567 open->op_openowner = oo;
2569 list_del_init(&oo->oo_close_lru);
2570 renew_client(oo->oo_owner.so_client);
2574 static inline __be32
2575 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2577 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2578 return nfserr_openmode;
2583 static int share_access_to_flags(u32 share_access)
2585 share_access &= ~NFS4_SHARE_WANT_MASK;
2587 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2590 static struct nfs4_delegation *find_deleg_stateid(stateid_t *s)
2592 struct nfs4_stid *ret;
2594 ret = find_stateid_by_type(s, NFS4_DELEG_STID);
2597 return delegstateid(ret);
2601 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2602 struct nfs4_delegation **dp)
2605 __be32 status = nfserr_bad_stateid;
2607 *dp = find_deleg_stateid(&open->op_delegate_stateid);
2610 flags = share_access_to_flags(open->op_share_access);
2611 status = nfs4_check_delegmode(*dp, flags);
2615 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2619 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2624 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2626 struct nfs4_ol_stateid *local;
2627 struct nfs4_openowner *oo = open->op_openowner;
2629 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2630 /* ignore lock owners */
2631 if (local->st_stateowner->so_is_open_owner == 0)
2633 /* remember if we have seen this open owner */
2634 if (local->st_stateowner == &oo->oo_owner)
2636 /* check for conflicting share reservations */
2637 if (!test_share(local, open))
2638 return nfserr_share_denied;
2643 static inline struct nfs4_ol_stateid *
2644 nfs4_alloc_stateid(void)
2646 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2649 static inline int nfs4_access_to_access(u32 nfs4_access)
2653 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2654 flags |= NFSD_MAY_READ;
2655 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2656 flags |= NFSD_MAY_WRITE;
2660 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2661 struct svc_fh *cur_fh, struct nfsd4_open *open)
2664 int oflag = nfs4_access_to_omode(open->op_share_access);
2665 int access = nfs4_access_to_access(open->op_share_access);
2667 /* CLAIM_DELEGATE_CUR is used in response to a broken lease;
2668 * allowing it to break the lease and return EAGAIN leaves the
2669 * client unable to make progress in returning the delegation */
2670 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2671 access |= NFSD_MAY_NOT_BREAK_LEASE;
2673 if (!fp->fi_fds[oflag]) {
2674 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2675 &fp->fi_fds[oflag]);
2679 nfs4_file_get_access(fp, oflag);
2685 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_ol_stateid **stpp,
2686 struct nfs4_file *fp, struct svc_fh *cur_fh,
2687 struct nfsd4_open *open)
2689 struct nfs4_ol_stateid *stp;
2692 stp = nfs4_alloc_stateid();
2694 return nfserr_jukebox;
2696 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2698 kmem_cache_free(stateid_slab, stp);
2705 static inline __be32
2706 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2707 struct nfsd4_open *open)
2709 struct iattr iattr = {
2710 .ia_valid = ATTR_SIZE,
2713 if (!open->op_truncate)
2715 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2716 return nfserr_inval;
2717 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2721 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)
2723 u32 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2727 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2729 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2733 status = nfsd4_truncate(rqstp, cur_fh, open);
2736 int oflag = nfs4_access_to_omode(op_share_access);
2737 nfs4_file_put_access(fp, oflag);
2741 /* remember the open */
2742 __set_bit(op_share_access, &stp->st_access_bmap);
2743 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2750 nfs4_set_claim_prev(struct nfsd4_open *open)
2752 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2753 open->op_openowner->oo_owner.so_client->cl_firststate = 1;
2756 /* Should we give out recallable state?: */
2757 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2759 if (clp->cl_cb_state == NFSD4_CB_UP)
2762 * In the sessions case, since we don't have to establish a
2763 * separate connection for callbacks, we assume it's OK
2764 * until we hear otherwise:
2766 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2769 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2771 struct file_lock *fl;
2773 fl = locks_alloc_lock();
2776 locks_init_lock(fl);
2777 fl->fl_lmops = &nfsd_lease_mng_ops;
2778 fl->fl_flags = FL_LEASE;
2779 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2780 fl->fl_end = OFFSET_MAX;
2781 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2782 fl->fl_pid = current->tgid;
2786 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2788 struct nfs4_file *fp = dp->dl_file;
2789 struct file_lock *fl;
2792 fl = nfs4_alloc_init_lease(dp, flag);
2795 fl->fl_file = find_readable_file(fp);
2796 list_add(&dp->dl_perclnt, &dp->dl_client->cl_delegations);
2797 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2799 list_del_init(&dp->dl_perclnt);
2800 locks_free_lock(fl);
2804 fp->fi_deleg_file = fl->fl_file;
2805 get_file(fp->fi_deleg_file);
2806 atomic_set(&fp->fi_delegees, 1);
2807 list_add(&dp->dl_perfile, &fp->fi_delegations);
2811 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2813 struct nfs4_file *fp = dp->dl_file;
2816 return nfs4_setlease(dp, flag);
2817 spin_lock(&recall_lock);
2818 if (fp->fi_had_conflict) {
2819 spin_unlock(&recall_lock);
2822 atomic_inc(&fp->fi_delegees);
2823 list_add(&dp->dl_perfile, &fp->fi_delegations);
2824 spin_unlock(&recall_lock);
2825 list_add(&dp->dl_perclnt, &dp->dl_client->cl_delegations);
2830 * Attempt to hand out a delegation.
2833 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2835 struct nfs4_delegation *dp;
2836 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2838 int status, flag = 0;
2840 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2841 flag = NFS4_OPEN_DELEGATE_NONE;
2842 open->op_recall = 0;
2843 switch (open->op_claim_type) {
2844 case NFS4_OPEN_CLAIM_PREVIOUS:
2846 open->op_recall = 1;
2847 flag = open->op_delegate_type;
2848 if (flag == NFS4_OPEN_DELEGATE_NONE)
2851 case NFS4_OPEN_CLAIM_NULL:
2852 /* Let's not give out any delegations till everyone's
2853 * had the chance to reclaim theirs.... */
2854 if (locks_in_grace())
2856 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2858 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2859 flag = NFS4_OPEN_DELEGATE_WRITE;
2861 flag = NFS4_OPEN_DELEGATE_READ;
2867 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2870 status = nfs4_set_delegation(dp, flag);
2874 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2876 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2877 STATEID_VAL(&dp->dl_stid.sc_stateid));
2879 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2880 && flag == NFS4_OPEN_DELEGATE_NONE
2881 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2882 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2883 open->op_delegate_type = flag;
2886 nfs4_put_delegation(dp);
2888 flag = NFS4_OPEN_DELEGATE_NONE;
2893 * called with nfs4_lock_state() held.
2896 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2898 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2899 struct nfs4_file *fp = NULL;
2900 struct inode *ino = current_fh->fh_dentry->d_inode;
2901 struct nfs4_ol_stateid *stp = NULL;
2902 struct nfs4_delegation *dp = NULL;
2905 status = nfserr_inval;
2906 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2907 || !deny_valid(open->op_share_deny))
2910 * Lookup file; if found, lookup stateid and check open request,
2911 * and check for delegations in the process of being recalled.
2912 * If not found, create the nfs4_file struct
2914 fp = find_file(ino);
2916 if ((status = nfs4_check_open(fp, open, &stp)))
2918 status = nfs4_check_deleg(fp, open, &dp);
2922 status = nfserr_bad_stateid;
2923 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2925 status = nfserr_jukebox;
2926 fp = alloc_init_file(ino);
2932 * OPEN the file, or upgrade an existing OPEN.
2933 * If truncate fails, the OPEN fails.
2936 /* Stateid was found, this is an OPEN upgrade */
2937 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
2941 status = nfs4_new_open(rqstp, &stp, fp, current_fh, open);
2944 init_open_stateid(stp, fp, open);
2945 status = nfsd4_truncate(rqstp, current_fh, open);
2947 release_open_stateid(stp);
2951 update_stateid(&stp->st_stid.sc_stateid);
2952 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
2954 if (nfsd4_has_session(&resp->cstate))
2955 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2958 * Attempt to hand out a delegation. No error return, because the
2959 * OPEN succeeds even if we fail.
2961 nfs4_open_delegation(current_fh, open, stp);
2965 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2966 STATEID_VAL(&stp->st_stid.sc_stateid));
2970 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2971 nfs4_set_claim_prev(open);
2973 * To finish the open response, we just need to set the rflags.
2975 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2976 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
2977 !nfsd4_has_session(&resp->cstate))
2978 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2984 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2987 struct nfs4_client *clp;
2991 dprintk("process_renew(%08x/%08x): starting\n",
2992 clid->cl_boot, clid->cl_id);
2993 status = nfserr_stale_clientid;
2994 if (STALE_CLIENTID(clid))
2996 clp = find_confirmed_client(clid);
2997 status = nfserr_expired;
2999 /* We assume the client took too long to RENEW. */
3000 dprintk("nfsd4_renew: clientid not found!\n");
3004 status = nfserr_cb_path_down;
3005 if (!list_empty(&clp->cl_delegations)
3006 && clp->cl_cb_state != NFSD4_CB_UP)
3010 nfs4_unlock_state();
3014 static struct lock_manager nfsd4_manager = {
3018 nfsd4_end_grace(void)
3020 dprintk("NFSD: end of grace period\n");
3021 nfsd4_recdir_purge_old();
3022 locks_end_grace(&nfsd4_manager);
3024 * Now that every NFSv4 client has had the chance to recover and
3025 * to see the (possibly new, possibly shorter) lease time, we
3026 * can safely set the next grace time to the current lease time:
3028 nfsd4_grace = nfsd4_lease;
3032 nfs4_laundromat(void)
3034 struct nfs4_client *clp;
3035 struct nfs4_openowner *oo;
3036 struct nfs4_delegation *dp;
3037 struct list_head *pos, *next, reaplist;
3038 time_t cutoff = get_seconds() - nfsd4_lease;
3039 time_t t, clientid_val = nfsd4_lease;
3040 time_t u, test_val = nfsd4_lease;
3044 dprintk("NFSD: laundromat service - starting\n");
3045 if (locks_in_grace())
3047 INIT_LIST_HEAD(&reaplist);
3048 spin_lock(&client_lock);
3049 list_for_each_safe(pos, next, &client_lru) {
3050 clp = list_entry(pos, struct nfs4_client, cl_lru);
3051 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3052 t = clp->cl_time - cutoff;
3053 if (clientid_val > t)
3057 if (atomic_read(&clp->cl_refcount)) {
3058 dprintk("NFSD: client in use (clientid %08x)\n",
3059 clp->cl_clientid.cl_id);
3062 unhash_client_locked(clp);
3063 list_add(&clp->cl_lru, &reaplist);
3065 spin_unlock(&client_lock);
3066 list_for_each_safe(pos, next, &reaplist) {
3067 clp = list_entry(pos, struct nfs4_client, cl_lru);
3068 dprintk("NFSD: purging unused client (clientid %08x)\n",
3069 clp->cl_clientid.cl_id);
3070 nfsd4_remove_clid_dir(clp);
3073 spin_lock(&recall_lock);
3074 list_for_each_safe(pos, next, &del_recall_lru) {
3075 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3076 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3077 u = dp->dl_time - cutoff;
3082 list_move(&dp->dl_recall_lru, &reaplist);
3084 spin_unlock(&recall_lock);
3085 list_for_each_safe(pos, next, &reaplist) {
3086 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3087 list_del_init(&dp->dl_recall_lru);
3088 unhash_delegation(dp);
3090 test_val = nfsd4_lease;
3091 list_for_each_safe(pos, next, &close_lru) {
3092 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3093 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3094 u = oo->oo_time - cutoff;
3099 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
3100 oo->oo_owner.so_id);
3101 release_openowner(oo);
3103 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3104 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3105 nfs4_unlock_state();
3106 return clientid_val;
3109 static struct workqueue_struct *laundry_wq;
3110 static void laundromat_main(struct work_struct *);
3111 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3114 laundromat_main(struct work_struct *not_used)
3118 t = nfs4_laundromat();
3119 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3120 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3123 static struct nfs4_openowner * search_close_lru(u32 st_id)
3125 struct nfs4_openowner *local;
3127 list_for_each_entry(local, &close_lru, oo_close_lru) {
3128 if (local->oo_owner.so_id == st_id)
3135 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3137 return fhp->fh_dentry->d_inode != stp->st_file->fi_inode;
3141 STALE_STATEID(stateid_t *stateid)
3143 if (stateid->si_boot == boot_time)
3145 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3146 STATEID_VAL(stateid));
3151 access_permit_read(unsigned long access_bmap)
3153 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3154 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3155 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3159 access_permit_write(unsigned long access_bmap)
3161 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3162 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3166 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3168 __be32 status = nfserr_openmode;
3170 /* For lock stateid's, we test the parent open, not the lock: */
3171 if (stp->st_openstp)
3172 stp = stp->st_openstp;
3173 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3175 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3182 static inline __be32
3183 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3185 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3187 else if (locks_in_grace()) {
3188 /* Answer in remaining cases depends on existence of
3189 * conflicting state; so we must wait out the grace period. */
3190 return nfserr_grace;
3191 } else if (flags & WR_STATE)
3192 return nfs4_share_conflict(current_fh,
3193 NFS4_SHARE_DENY_WRITE);
3194 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3195 return nfs4_share_conflict(current_fh,
3196 NFS4_SHARE_DENY_READ);
3200 * Allow READ/WRITE during grace period on recovered state only for files
3201 * that are not able to provide mandatory locking.
3204 grace_disallows_io(struct inode *inode)
3206 return locks_in_grace() && mandatory_lock(inode);
3209 /* Returns true iff a is later than b: */
3210 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3212 return (s32)a->si_generation - (s32)b->si_generation > 0;
3215 static int check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3218 * When sessions are used the stateid generation number is ignored
3221 if (has_session && in->si_generation == 0)
3224 if (in->si_generation == ref->si_generation)
3227 /* If the client sends us a stateid from the future, it's buggy: */
3228 if (stateid_generation_after(in, ref))
3229 return nfserr_bad_stateid;
3231 * However, we could see a stateid from the past, even from a
3232 * non-buggy client. For example, if the client sends a lock
3233 * while some IO is outstanding, the lock may bump si_generation
3234 * while the IO is still in flight. The client could avoid that
3235 * situation by waiting for responses on all the IO requests,
3236 * but better performance may result in retrying IO that
3237 * receives an old_stateid error if requests are rarely
3238 * reordered in flight:
3240 return nfserr_old_stateid;
3243 static int is_delegation_stateid(stateid_t *stateid)
3245 return stateid->si_fileid == 0;
3248 __be32 nfs4_validate_stateid(stateid_t *stateid, bool has_session)
3250 struct nfs4_stid *s;
3251 struct nfs4_ol_stateid *ols;
3254 if (STALE_STATEID(stateid))
3255 return nfserr_stale_stateid;
3257 s = find_stateid(stateid);
3259 return nfserr_stale_stateid;
3260 status = check_stateid_generation(stateid, &s->sc_stateid, has_session);
3263 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3265 ols = openlockstateid(s);
3266 if (ols->st_stateowner->so_is_open_owner
3267 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3268 return nfserr_bad_stateid;
3273 * Checks for stateid operations
3276 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3277 stateid_t *stateid, int flags, struct file **filpp)
3279 struct nfs4_stid *s;
3280 struct nfs4_ol_stateid *stp = NULL;
3281 struct nfs4_delegation *dp = NULL;
3282 struct svc_fh *current_fh = &cstate->current_fh;
3283 struct inode *ino = current_fh->fh_dentry->d_inode;
3289 if (grace_disallows_io(ino))
3290 return nfserr_grace;
3292 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3293 return check_special_stateids(current_fh, stateid, flags);
3295 status = nfserr_stale_stateid;
3296 if (STALE_STATEID(stateid))
3300 * We assume that any stateid that has the current boot time,
3301 * but that we can't find, is expired:
3303 status = nfserr_expired;
3304 s = find_stateid(stateid);
3307 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3310 if (s->sc_type == NFS4_DELEG_STID) {
3311 dp = delegstateid(s);
3312 status = nfs4_check_delegmode(dp, flags);
3315 renew_client(dp->dl_client);
3317 *filpp = dp->dl_file->fi_deleg_file;
3320 } else { /* open or lock stateid */
3321 stp = openlockstateid(s);
3322 status = nfserr_bad_stateid;
3323 if (nfs4_check_fh(current_fh, stp))
3325 if (stp->st_stateowner->so_is_open_owner
3326 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3328 status = nfs4_check_openmode(stp, flags);
3331 renew_client(stp->st_stateowner->so_client);
3333 if (flags & RD_STATE)
3334 *filpp = find_readable_file(stp->st_file);
3336 *filpp = find_writeable_file(stp->st_file);
3345 nfsd4_free_delegation_stateid(stateid_t *stateid)
3347 struct nfs4_delegation *dp = find_deleg_stateid(stateid);
3349 return nfserr_locks_held;
3351 return nfserr_bad_stateid;
3355 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3357 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3358 return nfserr_locks_held;
3359 release_lock_stateid(stp);
3364 * Test if the stateid is valid
3367 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3368 struct nfsd4_test_stateid *test_stateid)
3370 test_stateid->ts_has_session = nfsd4_has_session(cstate);
3378 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3379 struct nfsd4_free_stateid *free_stateid)
3381 stateid_t *stateid = &free_stateid->fr_stateid;
3382 struct nfs4_ol_stateid *stp;
3386 if (is_delegation_stateid(stateid)) {
3387 ret = nfsd4_free_delegation_stateid(stateid);
3391 stp = find_ol_stateid(stateid);
3393 ret = nfserr_bad_stateid;
3396 ret = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, 1);
3400 if (stp->st_stid.sc_type == NFS4_OPEN_STID) {
3401 ret = nfserr_locks_held;
3404 ret = nfsd4_free_lock_stateid(stp);
3409 nfs4_unlock_state();
3416 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3417 RD_STATE : WR_STATE;
3420 static __be32 nfs4_nospecial_stateid_checks(stateid_t *stateid)
3422 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3423 return nfserr_bad_stateid;
3424 if (STALE_STATEID(stateid))
3425 return nfserr_stale_stateid;
3429 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3431 struct svc_fh *current_fh = &cstate->current_fh;
3432 struct nfs4_stateowner *sop = stp->st_stateowner;
3435 if (nfs4_check_fh(current_fh, stp))
3436 return nfserr_bad_stateid;
3437 status = nfsd4_check_seqid(cstate, sop, seqid);
3440 return check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3444 * Checks for sequence id mutating operations.
3447 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3448 stateid_t *stateid, char typemask,
3449 struct nfs4_ol_stateid **stpp)
3453 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3454 seqid, STATEID_VAL(stateid));
3457 status = nfs4_nospecial_stateid_checks(stateid);
3460 *stpp = find_ol_stateid_by_type(stateid, typemask);
3462 return nfserr_expired;
3463 cstate->replay_owner = (*stpp)->st_stateowner;
3464 renew_client((*stpp)->st_stateowner->so_client);
3466 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3469 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3472 struct nfs4_openowner *oo;
3474 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3475 NFS4_OPEN_STID, stpp);
3478 oo = openowner((*stpp)->st_stateowner);
3479 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3480 return nfserr_bad_stateid;
3485 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3486 struct nfsd4_open_confirm *oc)
3489 struct nfs4_openowner *oo;
3490 struct nfs4_ol_stateid *stp;
3492 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3493 (int)cstate->current_fh.fh_dentry->d_name.len,
3494 cstate->current_fh.fh_dentry->d_name.name);
3496 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3502 status = nfs4_preprocess_seqid_op(cstate,
3503 oc->oc_seqid, &oc->oc_req_stateid,
3504 NFS4_OPEN_STID, &stp);
3507 oo = openowner(stp->st_stateowner);
3508 status = nfserr_bad_stateid;
3509 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3511 oo->oo_flags |= NFS4_OO_CONFIRMED;
3512 update_stateid(&stp->st_stid.sc_stateid);
3513 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3514 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3515 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3517 nfsd4_create_clid_dir(oo->oo_owner.so_client);
3520 if (!cstate->replay_owner)
3521 nfs4_unlock_state();
3525 static inline void nfs4_file_downgrade(struct nfs4_ol_stateid *stp, unsigned int to_access)
3529 for (i = 1; i < 4; i++) {
3530 if (test_bit(i, &stp->st_access_bmap) && !(i & to_access)) {
3531 nfs4_file_put_access(stp->st_file, i);
3532 __clear_bit(i, &stp->st_access_bmap);
3538 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3541 for (i = 0; i < 4; i++) {
3542 if ((i & deny) != i)
3543 __clear_bit(i, bmap);
3548 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3549 struct nfsd4_compound_state *cstate,
3550 struct nfsd4_open_downgrade *od)
3553 struct nfs4_ol_stateid *stp;
3555 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3556 (int)cstate->current_fh.fh_dentry->d_name.len,
3557 cstate->current_fh.fh_dentry->d_name.name);
3559 if (!access_valid(od->od_share_access, cstate->minorversion)
3560 || !deny_valid(od->od_share_deny))
3561 return nfserr_inval;
3564 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3565 &od->od_stateid, &stp);
3568 status = nfserr_inval;
3569 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3570 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3571 stp->st_access_bmap, od->od_share_access);
3574 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3575 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3576 stp->st_deny_bmap, od->od_share_deny);
3579 nfs4_file_downgrade(stp, od->od_share_access);
3581 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3583 update_stateid(&stp->st_stid.sc_stateid);
3584 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3587 if (!cstate->replay_owner)
3588 nfs4_unlock_state();
3593 * nfs4_unlock_state() called after encode
3596 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3597 struct nfsd4_close *close)
3600 struct nfs4_openowner *oo;
3601 struct nfs4_ol_stateid *stp;
3603 dprintk("NFSD: nfsd4_close on file %.*s\n",
3604 (int)cstate->current_fh.fh_dentry->d_name.len,
3605 cstate->current_fh.fh_dentry->d_name.name);
3608 /* check close_lru for replay */
3609 status = nfs4_preprocess_confirmed_seqid_op(cstate, close->cl_seqid,
3610 &close->cl_stateid, &stp);
3611 if (stp == NULL && status == nfserr_expired) {
3613 * Also, we should make sure this isn't just the result of
3616 oo = search_close_lru(close->cl_stateid.si_stateownerid);
3617 /* It's not stale; let's assume it's expired: */
3620 cstate->replay_owner = &oo->oo_owner;
3621 status = nfsd4_check_seqid(cstate, &oo->oo_owner, close->cl_seqid);
3624 status = nfserr_bad_seqid;
3628 oo = openowner(stp->st_stateowner);
3630 update_stateid(&stp->st_stid.sc_stateid);
3631 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3633 /* release_stateid() calls nfsd_close() if needed */
3634 release_open_stateid(stp);
3636 /* place unused nfs4_stateowners on so_close_lru list to be
3637 * released by the laundromat service after the lease period
3638 * to enable us to handle CLOSE replay
3640 if (list_empty(&oo->oo_owner.so_stateids))
3641 move_to_close_lru(oo);
3643 if (!cstate->replay_owner)
3644 nfs4_unlock_state();
3649 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3650 struct nfsd4_delegreturn *dr)
3652 struct nfs4_delegation *dp;
3653 stateid_t *stateid = &dr->dr_stateid;
3654 struct inode *inode;
3657 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3659 inode = cstate->current_fh.fh_dentry->d_inode;
3662 status = nfserr_bad_stateid;
3663 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3665 status = nfserr_stale_stateid;
3666 if (STALE_STATEID(stateid))
3668 status = nfserr_bad_stateid;
3669 if (!is_delegation_stateid(stateid))
3671 status = nfserr_expired;
3672 dp = find_deleg_stateid(stateid);
3675 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3678 renew_client(dp->dl_client);
3680 unhash_delegation(dp);
3682 nfs4_unlock_state();
3689 * Lock owner state (byte-range locks)
3691 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3692 #define LOCK_HASH_BITS 8
3693 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3694 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3697 end_offset(u64 start, u64 len)
3702 return end >= start ? end: NFS4_MAX_UINT64;
3705 /* last octet in a range */
3707 last_byte_offset(u64 start, u64 len)
3713 return end > start ? end - 1: NFS4_MAX_UINT64;
3716 static unsigned int lockownerid_hashval(u32 id)
3718 return id & LOCK_HASH_MASK;
3721 static inline unsigned int
3722 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3723 struct xdr_netobj *ownername)
3725 return (file_hashval(inode) + cl_id
3726 + opaque_hashval(ownername->data, ownername->len))
3730 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3731 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3734 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3735 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3736 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3737 * locking, this prevents us from being completely protocol-compliant. The
3738 * real solution to this problem is to start using unsigned file offsets in
3739 * the VFS, but this is a very deep change!
3742 nfs4_transform_lock_offset(struct file_lock *lock)
3744 if (lock->fl_start < 0)
3745 lock->fl_start = OFFSET_MAX;
3746 if (lock->fl_end < 0)
3747 lock->fl_end = OFFSET_MAX;
3750 /* Hack!: For now, we're defining this just so we can use a pointer to it
3751 * as a unique cookie to identify our (NFSv4's) posix locks. */
3752 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3756 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3758 struct nfs4_lockowner *lo;
3760 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3761 lo = (struct nfs4_lockowner *) fl->fl_owner;
3762 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3763 lo->lo_owner.so_owner.len, GFP_KERNEL);
3764 if (!deny->ld_owner.data)
3765 /* We just don't care that much */
3767 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3768 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3771 deny->ld_owner.len = 0;
3772 deny->ld_owner.data = NULL;
3773 deny->ld_clientid.cl_boot = 0;
3774 deny->ld_clientid.cl_id = 0;
3776 deny->ld_start = fl->fl_start;
3777 deny->ld_length = NFS4_MAX_UINT64;
3778 if (fl->fl_end != NFS4_MAX_UINT64)
3779 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3780 deny->ld_type = NFS4_READ_LT;
3781 if (fl->fl_type != F_RDLCK)
3782 deny->ld_type = NFS4_WRITE_LT;
3785 static struct nfs4_lockowner *
3786 find_lockowner_str(struct inode *inode, clientid_t *clid,
3787 struct xdr_netobj *owner)
3789 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3790 struct nfs4_stateowner *op;
3792 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3793 if (same_owner_str(op, owner, clid))
3794 return lockowner(op);
3799 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3801 unsigned int idhashval;
3803 idhashval = lockownerid_hashval(lo->lo_owner.so_id);
3804 list_add(&lo->lo_owner.so_idhash, &lock_ownerid_hashtbl[idhashval]);
3805 list_add(&lo->lo_owner.so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3806 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3810 * Alloc a lock owner structure.
3811 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3814 * strhashval = lock_ownerstr_hashval
3817 static struct nfs4_lockowner *
3818 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3819 struct nfs4_lockowner *lo;
3821 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3824 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3825 lo->lo_owner.so_is_open_owner = 0;
3826 /* It is the openowner seqid that will be incremented in encode in the
3827 * case of new lockowners; so increment the lock seqid manually: */
3828 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3829 hash_lockowner(lo, strhashval, clp, open_stp);
3833 static struct nfs4_ol_stateid *
3834 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3836 struct nfs4_ol_stateid *stp;
3838 stp = nfs4_alloc_stateid();
3841 list_add(&stp->st_perfile, &fp->fi_stateids);
3842 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3843 stp->st_stateowner = &lo->lo_owner;
3844 stp->st_stid.sc_type = NFS4_LOCK_STID;
3847 stp->st_stid.sc_stateid.si_boot = boot_time;
3848 stp->st_stid.sc_stateid.si_stateownerid = lo->lo_owner.so_id;
3849 stp->st_stid.sc_stateid.si_fileid = fp->fi_id;
3850 /* note will be incremented before first return to client: */
3851 stp->st_stid.sc_stateid.si_generation = 0;
3852 hash_stid(&stp->st_stid);
3853 stp->st_access_bmap = 0;
3854 stp->st_deny_bmap = open_stp->st_deny_bmap;
3855 stp->st_openstp = open_stp;
3862 check_lock_length(u64 offset, u64 length)
3864 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3865 LOFF_OVERFLOW(offset, length)));
3868 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
3870 struct nfs4_file *fp = lock_stp->st_file;
3871 int oflag = nfs4_access_to_omode(access);
3873 if (test_bit(access, &lock_stp->st_access_bmap))
3875 nfs4_file_get_access(fp, oflag);
3876 __set_bit(access, &lock_stp->st_access_bmap);
3883 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3884 struct nfsd4_lock *lock)
3886 struct nfs4_openowner *open_sop = NULL;
3887 struct nfs4_lockowner *lock_sop = NULL;
3888 struct nfs4_ol_stateid *lock_stp;
3889 struct nfs4_file *fp;
3890 struct file *filp = NULL;
3891 struct file_lock file_lock;
3892 struct file_lock conflock;
3894 unsigned int strhashval;
3898 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3899 (long long) lock->lk_offset,
3900 (long long) lock->lk_length);
3902 if (check_lock_length(lock->lk_offset, lock->lk_length))
3903 return nfserr_inval;
3905 if ((status = fh_verify(rqstp, &cstate->current_fh,
3906 S_IFREG, NFSD_MAY_LOCK))) {
3907 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3913 if (lock->lk_is_new) {
3915 * Client indicates that this is a new lockowner.
3916 * Use open owner and open stateid to create lock owner and
3919 struct nfs4_ol_stateid *open_stp = NULL;
3921 status = nfserr_stale_clientid;
3922 if (!nfsd4_has_session(cstate) &&
3923 STALE_CLIENTID(&lock->lk_new_clientid))
3926 /* validate and update open stateid and open seqid */
3927 status = nfs4_preprocess_confirmed_seqid_op(cstate,
3928 lock->lk_new_open_seqid,
3929 &lock->lk_new_open_stateid,
3933 open_sop = openowner(open_stp->st_stateowner);
3934 status = nfserr_bad_stateid;
3935 if (!nfsd4_has_session(cstate) &&
3936 !same_clid(&open_sop->oo_owner.so_client->cl_clientid,
3937 &lock->v.new.clientid))
3939 /* create lockowner and lock stateid */
3940 fp = open_stp->st_file;
3941 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3942 open_sop->oo_owner.so_client->cl_clientid.cl_id,
3943 &lock->v.new.owner);
3944 /* XXX: Do we need to check for duplicate stateowners on
3945 * the same file, or should they just be allowed (and
3946 * create new stateids)? */
3947 status = nfserr_jukebox;
3948 lock_sop = alloc_init_lock_stateowner(strhashval,
3949 open_sop->oo_owner.so_client, open_stp, lock);
3950 if (lock_sop == NULL)
3952 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3953 if (lock_stp == NULL)
3956 /* lock (lock owner + lock stateid) already exists */
3957 status = nfs4_preprocess_seqid_op(cstate,
3958 lock->lk_old_lock_seqid,
3959 &lock->lk_old_lock_stateid,
3960 NFS4_LOCK_STID, &lock_stp);
3963 lock_sop = lockowner(lock_stp->st_stateowner);
3964 fp = lock_stp->st_file;
3966 /* lock_sop and lock_stp have been created or found */
3968 lkflg = setlkflg(lock->lk_type);
3969 status = nfs4_check_openmode(lock_stp, lkflg);
3973 status = nfserr_grace;
3974 if (locks_in_grace() && !lock->lk_reclaim)
3976 status = nfserr_no_grace;
3977 if (!locks_in_grace() && lock->lk_reclaim)
3980 locks_init_lock(&file_lock);
3981 switch (lock->lk_type) {
3984 filp = find_readable_file(lock_stp->st_file);
3986 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
3987 file_lock.fl_type = F_RDLCK;
3990 case NFS4_WRITEW_LT:
3991 filp = find_writeable_file(lock_stp->st_file);
3993 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
3994 file_lock.fl_type = F_WRLCK;
3997 status = nfserr_inval;
4001 status = nfserr_openmode;
4004 file_lock.fl_owner = (fl_owner_t)lock_sop;
4005 file_lock.fl_pid = current->tgid;
4006 file_lock.fl_file = filp;
4007 file_lock.fl_flags = FL_POSIX;
4008 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4010 file_lock.fl_start = lock->lk_offset;
4011 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4012 nfs4_transform_lock_offset(&file_lock);
4015 * Try to lock the file in the VFS.
4016 * Note: locks.c uses the BKL to protect the inode's lock list.
4019 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4021 case 0: /* success! */
4022 update_stateid(&lock_stp->st_stid.sc_stateid);
4023 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4027 case (EAGAIN): /* conflock holds conflicting lock */
4028 status = nfserr_denied;
4029 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4030 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4033 status = nfserr_deadlock;
4036 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4037 status = nfserrno(err);
4041 if (status && lock->lk_is_new && lock_sop)
4042 release_lockowner(lock_sop);
4043 if (!cstate->replay_owner)
4044 nfs4_unlock_state();
4049 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4050 * so we do a temporary open here just to get an open file to pass to
4051 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4054 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4059 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4062 err = vfs_test_lock(file, lock);
4071 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4072 struct nfsd4_lockt *lockt)
4074 struct inode *inode;
4075 struct file_lock file_lock;
4076 struct nfs4_lockowner *lo;
4080 if (locks_in_grace())
4081 return nfserr_grace;
4083 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4084 return nfserr_inval;
4088 status = nfserr_stale_clientid;
4089 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4092 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4095 inode = cstate->current_fh.fh_dentry->d_inode;
4096 locks_init_lock(&file_lock);
4097 switch (lockt->lt_type) {
4100 file_lock.fl_type = F_RDLCK;
4103 case NFS4_WRITEW_LT:
4104 file_lock.fl_type = F_WRLCK;
4107 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4108 status = nfserr_inval;
4112 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4114 file_lock.fl_owner = (fl_owner_t)lo;
4115 file_lock.fl_pid = current->tgid;
4116 file_lock.fl_flags = FL_POSIX;
4118 file_lock.fl_start = lockt->lt_offset;
4119 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4121 nfs4_transform_lock_offset(&file_lock);
4124 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4126 status = nfserrno(error);
4129 if (file_lock.fl_type != F_UNLCK) {
4130 status = nfserr_denied;
4131 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4134 nfs4_unlock_state();
4139 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4140 struct nfsd4_locku *locku)
4142 struct nfs4_ol_stateid *stp;
4143 struct file *filp = NULL;
4144 struct file_lock file_lock;
4148 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4149 (long long) locku->lu_offset,
4150 (long long) locku->lu_length);
4152 if (check_lock_length(locku->lu_offset, locku->lu_length))
4153 return nfserr_inval;
4157 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4158 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4161 filp = find_any_file(stp->st_file);
4163 status = nfserr_lock_range;
4167 locks_init_lock(&file_lock);
4168 file_lock.fl_type = F_UNLCK;
4169 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4170 file_lock.fl_pid = current->tgid;
4171 file_lock.fl_file = filp;
4172 file_lock.fl_flags = FL_POSIX;
4173 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4174 file_lock.fl_start = locku->lu_offset;
4176 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4177 nfs4_transform_lock_offset(&file_lock);
4180 * Try to unlock the file in the VFS.
4182 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4184 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4188 * OK, unlock succeeded; the only thing left to do is update the stateid.
4190 update_stateid(&stp->st_stid.sc_stateid);
4191 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4194 nfs4_unlock_state();
4198 status = nfserrno(err);
4204 * 1: locks held by lockowner
4205 * 0: no locks held by lockowner
4208 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4210 struct file_lock **flpp;
4211 struct inode *inode = filp->fi_inode;
4215 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4216 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4227 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4228 struct nfsd4_compound_state *cstate,
4229 struct nfsd4_release_lockowner *rlockowner)
4231 clientid_t *clid = &rlockowner->rl_clientid;
4232 struct nfs4_stateowner *sop;
4233 struct nfs4_lockowner *lo;
4234 struct nfs4_ol_stateid *stp;
4235 struct xdr_netobj *owner = &rlockowner->rl_owner;
4236 struct list_head matches;
4240 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4241 clid->cl_boot, clid->cl_id);
4243 /* XXX check for lease expiration */
4245 status = nfserr_stale_clientid;
4246 if (STALE_CLIENTID(clid))
4251 status = nfserr_locks_held;
4252 /* XXX: we're doing a linear search through all the lockowners.
4253 * Yipes! For now we'll just hope clients aren't really using
4254 * release_lockowner much, but eventually we have to fix these
4255 * data structures. */
4256 INIT_LIST_HEAD(&matches);
4257 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4258 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
4259 if (!same_owner_str(sop, owner, clid))
4261 list_for_each_entry(stp, &sop->so_stateids,
4263 lo = lockowner(sop);
4264 if (check_for_locks(stp->st_file, lo))
4266 list_add(&lo->lo_list, &matches);
4270 /* Clients probably won't expect us to return with some (but not all)
4271 * of the lockowner state released; so don't release any until all
4272 * have been checked. */
4274 while (!list_empty(&matches)) {
4275 lo = list_entry(matches.next, struct nfs4_lockowner,
4277 /* unhash_stateowner deletes so_perclient only
4278 * for openowners. */
4279 list_del(&lo->lo_list);
4280 release_lockowner(lo);
4283 nfs4_unlock_state();
4287 static inline struct nfs4_client_reclaim *
4290 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4294 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4296 unsigned int strhashval = clientstr_hashval(name);
4297 struct nfs4_client *clp;
4299 clp = find_confirmed_client_by_str(name, strhashval);
4304 * failure => all reset bets are off, nfserr_no_grace...
4307 nfs4_client_to_reclaim(const char *name)
4309 unsigned int strhashval;
4310 struct nfs4_client_reclaim *crp = NULL;
4312 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4313 crp = alloc_reclaim();
4316 strhashval = clientstr_hashval(name);
4317 INIT_LIST_HEAD(&crp->cr_strhash);
4318 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4319 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4320 reclaim_str_hashtbl_size++;
4325 nfs4_release_reclaim(void)
4327 struct nfs4_client_reclaim *crp = NULL;
4330 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4331 while (!list_empty(&reclaim_str_hashtbl[i])) {
4332 crp = list_entry(reclaim_str_hashtbl[i].next,
4333 struct nfs4_client_reclaim, cr_strhash);
4334 list_del(&crp->cr_strhash);
4336 reclaim_str_hashtbl_size--;
4339 BUG_ON(reclaim_str_hashtbl_size);
4343 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4344 static struct nfs4_client_reclaim *
4345 nfs4_find_reclaim_client(clientid_t *clid)
4347 unsigned int strhashval;
4348 struct nfs4_client *clp;
4349 struct nfs4_client_reclaim *crp = NULL;
4352 /* find clientid in conf_id_hashtbl */
4353 clp = find_confirmed_client(clid);
4357 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4358 clp->cl_name.len, clp->cl_name.data,
4361 /* find clp->cl_name in reclaim_str_hashtbl */
4362 strhashval = clientstr_hashval(clp->cl_recdir);
4363 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4364 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4372 * Called from OPEN. Look for clientid in reclaim list.
4375 nfs4_check_open_reclaim(clientid_t *clid)
4377 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4380 /* initialization to perform at module load time: */
4383 nfs4_state_init(void)
4387 status = nfsd4_init_slabs();
4390 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4391 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4392 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4393 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4394 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4395 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4397 for (i = 0; i < SESSION_HASH_SIZE; i++)
4398 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4399 for (i = 0; i < FILE_HASH_SIZE; i++) {
4400 INIT_LIST_HEAD(&file_hashtbl[i]);
4402 for (i = 0; i < OPEN_OWNER_HASH_SIZE; i++) {
4403 INIT_LIST_HEAD(&open_ownerstr_hashtbl[i]);
4404 INIT_LIST_HEAD(&open_ownerid_hashtbl[i]);
4406 for (i = 0; i < STATEID_HASH_SIZE; i++)
4407 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4408 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4409 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4410 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4412 memset(&onestateid, ~0, sizeof(stateid_t));
4413 INIT_LIST_HEAD(&close_lru);
4414 INIT_LIST_HEAD(&client_lru);
4415 INIT_LIST_HEAD(&del_recall_lru);
4416 reclaim_str_hashtbl_size = 0;
4421 nfsd4_load_reboot_recovery_data(void)
4426 nfsd4_init_recdir();
4427 status = nfsd4_recdir_load();
4428 nfs4_unlock_state();
4430 printk("NFSD: Failure reading reboot recovery data\n");
4434 * Since the lifetime of a delegation isn't limited to that of an open, a
4435 * client may quite reasonably hang on to a delegation as long as it has
4436 * the inode cached. This becomes an obvious problem the first time a
4437 * client's inode cache approaches the size of the server's total memory.
4439 * For now we avoid this problem by imposing a hard limit on the number
4440 * of delegations, which varies according to the server's memory size.
4443 set_max_delegations(void)
4446 * Allow at most 4 delegations per megabyte of RAM. Quick
4447 * estimates suggest that in the worst case (where every delegation
4448 * is for a different inode), a delegation could take about 1.5K,
4449 * giving a worst case usage of about 6% of memory.
4451 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4454 /* initialization to perform when the nfsd service is started: */
4457 __nfs4_state_start(void)
4461 boot_time = get_seconds();
4462 locks_start_grace(&nfsd4_manager);
4463 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4465 ret = set_callback_cred();
4468 laundry_wq = create_singlethread_workqueue("nfsd4");
4469 if (laundry_wq == NULL)
4471 ret = nfsd4_create_callback_queue();
4473 goto out_free_laundry;
4474 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4475 set_max_delegations();
4478 destroy_workqueue(laundry_wq);
4483 nfs4_state_start(void)
4485 nfsd4_load_reboot_recovery_data();
4486 return __nfs4_state_start();
4490 __nfs4_state_shutdown(void)
4493 struct nfs4_client *clp = NULL;
4494 struct nfs4_delegation *dp = NULL;
4495 struct list_head *pos, *next, reaplist;
4497 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4498 while (!list_empty(&conf_id_hashtbl[i])) {
4499 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4502 while (!list_empty(&unconf_str_hashtbl[i])) {
4503 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4507 INIT_LIST_HEAD(&reaplist);
4508 spin_lock(&recall_lock);
4509 list_for_each_safe(pos, next, &del_recall_lru) {
4510 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4511 list_move(&dp->dl_recall_lru, &reaplist);
4513 spin_unlock(&recall_lock);
4514 list_for_each_safe(pos, next, &reaplist) {
4515 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4516 list_del_init(&dp->dl_recall_lru);
4517 unhash_delegation(dp);
4520 nfsd4_shutdown_recdir();
4524 nfs4_state_shutdown(void)
4526 cancel_delayed_work_sync(&laundromat_work);
4527 destroy_workqueue(laundry_wq);
4528 locks_end_grace(&nfsd4_manager);
4530 nfs4_release_reclaim();
4531 __nfs4_state_shutdown();
4532 nfs4_unlock_state();
4533 nfsd4_destroy_callback_queue();