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 stateid_t zerostateid; /* bits all 0 */
53 static stateid_t onestateid; /* bits all 1 */
54 static u64 current_sessionid = 1;
56 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
57 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
59 /* forward declarations */
60 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
64 /* Currently used for almost all code touching nfsv4 state: */
65 static DEFINE_MUTEX(client_mutex);
68 * Currently used for the del_recall_lru and file hash table. In an
69 * effort to decrease the scope of the client_mutex, this spinlock may
70 * eventually cover more:
72 static DEFINE_SPINLOCK(recall_lock);
74 static struct kmem_cache *openowner_slab = NULL;
75 static struct kmem_cache *lockowner_slab = NULL;
76 static struct kmem_cache *file_slab = NULL;
77 static struct kmem_cache *stateid_slab = NULL;
78 static struct kmem_cache *deleg_slab = NULL;
83 mutex_lock(&client_mutex);
87 nfs4_unlock_state(void)
89 mutex_unlock(&client_mutex);
93 opaque_hashval(const void *ptr, int nbytes)
95 unsigned char *cptr = (unsigned char *) ptr;
105 static struct list_head del_recall_lru;
108 put_nfs4_file(struct nfs4_file *fi)
110 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
111 list_del(&fi->fi_hash);
112 spin_unlock(&recall_lock);
114 kmem_cache_free(file_slab, fi);
119 get_nfs4_file(struct nfs4_file *fi)
121 atomic_inc(&fi->fi_ref);
124 static int num_delegations;
125 unsigned int max_delegations;
128 * Open owner state (share locks)
131 /* hash tables for open owners */
132 #define OPEN_OWNER_HASH_BITS 8
133 #define OPEN_OWNER_HASH_SIZE (1 << OPEN_OWNER_HASH_BITS)
134 #define OPEN_OWNER_HASH_MASK (OPEN_OWNER_HASH_SIZE - 1)
136 static unsigned int open_ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
140 ret = opaque_hashval(ownername->data, ownername->len);
142 return ret & OPEN_OWNER_HASH_MASK;
145 static struct list_head open_ownerstr_hashtbl[OPEN_OWNER_HASH_SIZE];
147 /* hash table for nfs4_file */
148 #define FILE_HASH_BITS 8
149 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151 static unsigned int file_hashval(struct inode *ino)
153 /* XXX: why are we hashing on inode pointer, anyway? */
154 return hash_ptr(ino, FILE_HASH_BITS);
157 static struct list_head file_hashtbl[FILE_HASH_SIZE];
159 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
161 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
162 atomic_inc(&fp->fi_access[oflag]);
165 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
167 if (oflag == O_RDWR) {
168 __nfs4_file_get_access(fp, O_RDONLY);
169 __nfs4_file_get_access(fp, O_WRONLY);
171 __nfs4_file_get_access(fp, oflag);
174 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
176 if (fp->fi_fds[oflag]) {
177 fput(fp->fi_fds[oflag]);
178 fp->fi_fds[oflag] = NULL;
182 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
184 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
185 nfs4_file_put_fd(fp, oflag);
187 * It's also safe to get rid of the RDWR open *if*
188 * we no longer have need of the other kind of access
189 * or if we already have the other kind of open:
191 if (fp->fi_fds[1-oflag]
192 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
193 nfs4_file_put_fd(fp, O_RDWR);
197 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
199 if (oflag == O_RDWR) {
200 __nfs4_file_put_access(fp, O_RDONLY);
201 __nfs4_file_put_access(fp, O_WRONLY);
203 __nfs4_file_put_access(fp, oflag);
206 static inline int get_new_stid(struct nfs4_stid *stid)
208 static int min_stateid = 0;
209 struct idr *stateids = &stid->sc_client->cl_stateids;
213 if (!idr_pre_get(stateids, GFP_KERNEL))
216 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
218 * All this code is currently serialized; the preallocation
219 * above should still be ours:
223 * It shouldn't be a problem to reuse an opaque stateid value.
224 * I don't think it is for 4.1. But with 4.0 I worry that, for
225 * example, a stray write retransmission could be accepted by
226 * the server when it should have been rejected. Therefore,
227 * adopt a trick from the sctp code to attempt to maximize the
228 * amount of time until an id is reused, by ensuring they always
229 * "increase" (mod INT_MAX):
232 min_stateid = new_stid+1;
233 if (min_stateid == INT_MAX)
238 static inline __be32 init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
240 stateid_t *s = &stid->sc_stateid;
243 stid->sc_type = type;
244 stid->sc_client = cl;
245 s->si_opaque.so_clid = cl->cl_clientid;
246 new_id = get_new_stid(stid);
248 return nfserr_jukebox;
249 s->si_opaque.so_id = (u32)new_id;
250 /* Will be incremented before return to client: */
251 s->si_generation = 0;
255 static struct nfs4_delegation *
256 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
258 struct nfs4_delegation *dp;
259 struct nfs4_file *fp = stp->st_file;
262 dprintk("NFSD alloc_init_deleg\n");
264 * Major work on the lease subsystem (for example, to support
265 * calbacks on stat) will be required before we can support
266 * write delegations properly.
268 if (type != NFS4_OPEN_DELEGATE_READ)
270 if (fp->fi_had_conflict)
272 if (num_delegations > max_delegations)
274 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
277 status = init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
279 kmem_cache_free(deleg_slab, dp);
283 * delegation seqid's are never incremented. The 4.1 special
284 * meaning of seqid 0 isn't meaningful, really, but let's avoid
285 * 0 anyway just for consistency and use 1:
287 dp->dl_stid.sc_stateid.si_generation = 1;
289 INIT_LIST_HEAD(&dp->dl_perfile);
290 INIT_LIST_HEAD(&dp->dl_perclnt);
291 INIT_LIST_HEAD(&dp->dl_recall_lru);
295 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
297 atomic_set(&dp->dl_count, 1);
298 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
303 nfs4_put_delegation(struct nfs4_delegation *dp)
305 if (atomic_dec_and_test(&dp->dl_count)) {
306 dprintk("NFSD: freeing dp %p\n",dp);
307 put_nfs4_file(dp->dl_file);
308 kmem_cache_free(deleg_slab, dp);
313 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
315 if (atomic_dec_and_test(&fp->fi_delegees)) {
316 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
318 fput(fp->fi_deleg_file);
319 fp->fi_deleg_file = NULL;
323 static void unhash_stid(struct nfs4_stid *s)
325 struct idr *stateids = &s->sc_client->cl_stateids;
327 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
330 /* Called under the state lock. */
332 unhash_delegation(struct nfs4_delegation *dp)
334 unhash_stid(&dp->dl_stid);
335 list_del_init(&dp->dl_perclnt);
336 spin_lock(&recall_lock);
337 list_del_init(&dp->dl_perfile);
338 list_del_init(&dp->dl_recall_lru);
339 spin_unlock(&recall_lock);
340 nfs4_put_deleg_lease(dp->dl_file);
341 nfs4_put_delegation(dp);
348 /* client_lock protects the client lru list and session hash table */
349 static DEFINE_SPINLOCK(client_lock);
351 /* Hash tables for nfs4_clientid state */
352 #define CLIENT_HASH_BITS 4
353 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
354 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
356 static unsigned int clientid_hashval(u32 id)
358 return id & CLIENT_HASH_MASK;
361 static unsigned int clientstr_hashval(const char *name)
363 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
367 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
368 * used in reboot/reset lease grace period processing
370 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
371 * setclientid_confirmed info.
373 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
376 * client_lru holds client queue ordered by nfs4_client.cl_time
379 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
380 * for last close replay.
382 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
383 static int reclaim_str_hashtbl_size = 0;
384 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
385 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
386 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
387 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
388 static struct list_head client_lru;
389 static struct list_head close_lru;
392 * We store the NONE, READ, WRITE, and BOTH bits separately in the
393 * st_{access,deny}_bmap field of the stateid, in order to track not
394 * only what share bits are currently in force, but also what
395 * combinations of share bits previous opens have used. This allows us
396 * to enforce the recommendation of rfc 3530 14.2.19 that the server
397 * return an error if the client attempt to downgrade to a combination
398 * of share bits not explicable by closing some of its previous opens.
400 * XXX: This enforcement is actually incomplete, since we don't keep
401 * track of access/deny bit combinations; so, e.g., we allow:
403 * OPEN allow read, deny write
404 * OPEN allow both, deny none
405 * DOWNGRADE allow read, deny none
407 * which we should reject.
410 set_access(unsigned int *access, unsigned long bmap) {
414 for (i = 1; i < 4; i++) {
415 if (test_bit(i, &bmap))
421 set_deny(unsigned int *deny, unsigned long bmap) {
425 for (i = 0; i < 4; i++) {
426 if (test_bit(i, &bmap))
432 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
433 unsigned int access, deny;
435 set_access(&access, stp->st_access_bmap);
436 set_deny(&deny, stp->st_deny_bmap);
437 if ((access & open->op_share_deny) || (deny & open->op_share_access))
442 static int nfs4_access_to_omode(u32 access)
444 switch (access & NFS4_SHARE_ACCESS_BOTH) {
445 case NFS4_SHARE_ACCESS_READ:
447 case NFS4_SHARE_ACCESS_WRITE:
449 case NFS4_SHARE_ACCESS_BOTH:
455 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
457 list_del(&stp->st_perfile);
458 list_del(&stp->st_perstateowner);
461 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
465 if (stp->st_access_bmap) {
466 for (i = 1; i < 4; i++) {
467 if (test_bit(i, &stp->st_access_bmap))
468 nfs4_file_put_access(stp->st_file,
469 nfs4_access_to_omode(i));
470 __clear_bit(i, &stp->st_access_bmap);
473 put_nfs4_file(stp->st_file);
477 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
479 kmem_cache_free(stateid_slab, stp);
482 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
486 unhash_generic_stateid(stp);
487 unhash_stid(&stp->st_stid);
488 file = find_any_file(stp->st_file);
490 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
491 close_generic_stateid(stp);
492 free_generic_stateid(stp);
495 static void unhash_lockowner(struct nfs4_lockowner *lo)
497 struct nfs4_ol_stateid *stp;
499 list_del(&lo->lo_owner.so_strhash);
500 list_del(&lo->lo_perstateid);
501 while (!list_empty(&lo->lo_owner.so_stateids)) {
502 stp = list_first_entry(&lo->lo_owner.so_stateids,
503 struct nfs4_ol_stateid, st_perstateowner);
504 release_lock_stateid(stp);
508 static void release_lockowner(struct nfs4_lockowner *lo)
510 unhash_lockowner(lo);
511 nfs4_free_lockowner(lo);
515 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
517 struct nfs4_lockowner *lo;
519 while (!list_empty(&open_stp->st_lockowners)) {
520 lo = list_entry(open_stp->st_lockowners.next,
521 struct nfs4_lockowner, lo_perstateid);
522 release_lockowner(lo);
526 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
528 unhash_generic_stateid(stp);
529 release_stateid_lockowners(stp);
530 close_generic_stateid(stp);
533 static void release_open_stateid(struct nfs4_ol_stateid *stp)
535 unhash_open_stateid(stp);
536 unhash_stid(&stp->st_stid);
537 free_generic_stateid(stp);
540 static void unhash_openowner(struct nfs4_openowner *oo)
542 struct nfs4_ol_stateid *stp;
544 list_del(&oo->oo_owner.so_strhash);
545 list_del(&oo->oo_perclient);
546 while (!list_empty(&oo->oo_owner.so_stateids)) {
547 stp = list_first_entry(&oo->oo_owner.so_stateids,
548 struct nfs4_ol_stateid, st_perstateowner);
549 release_open_stateid(stp);
553 static void release_last_closed_stateid(struct nfs4_openowner *oo)
555 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
558 unhash_stid(&s->st_stid);
559 free_generic_stateid(s);
560 oo->oo_last_closed_stid = NULL;
564 static void release_openowner(struct nfs4_openowner *oo)
566 unhash_openowner(oo);
567 list_del(&oo->oo_close_lru);
568 release_last_closed_stateid(oo);
569 nfs4_free_openowner(oo);
572 #define SESSION_HASH_SIZE 512
573 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
576 hash_sessionid(struct nfs4_sessionid *sessionid)
578 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
580 return sid->sequence % SESSION_HASH_SIZE;
584 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
586 u32 *ptr = (u32 *)(&sessionid->data[0]);
587 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
591 gen_sessionid(struct nfsd4_session *ses)
593 struct nfs4_client *clp = ses->se_client;
594 struct nfsd4_sessionid *sid;
596 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
597 sid->clientid = clp->cl_clientid;
598 sid->sequence = current_sessionid++;
603 * The protocol defines ca_maxresponssize_cached to include the size of
604 * the rpc header, but all we need to cache is the data starting after
605 * the end of the initial SEQUENCE operation--the rest we regenerate
606 * each time. Therefore we can advertise a ca_maxresponssize_cached
607 * value that is the number of bytes in our cache plus a few additional
608 * bytes. In order to stay on the safe side, and not promise more than
609 * we can cache, those additional bytes must be the minimum possible: 24
610 * bytes of rpc header (xid through accept state, with AUTH_NULL
611 * verifier), 12 for the compound header (with zero-length tag), and 44
612 * for the SEQUENCE op response:
614 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
617 free_session_slots(struct nfsd4_session *ses)
621 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
622 kfree(ses->se_slots[i]);
626 * We don't actually need to cache the rpc and session headers, so we
627 * can allocate a little less for each slot:
629 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
631 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
634 static int nfsd4_sanitize_slot_size(u32 size)
636 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
637 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
643 * XXX: If we run out of reserved DRC memory we could (up to a point)
644 * re-negotiate active sessions and reduce their slot usage to make
645 * rooom for new connections. For now we just fail the create session.
647 static int nfsd4_get_drc_mem(int slotsize, u32 num)
651 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
653 spin_lock(&nfsd_drc_lock);
654 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
655 nfsd_drc_max_mem - nfsd_drc_mem_used);
656 num = min_t(int, num, avail / slotsize);
657 nfsd_drc_mem_used += num * slotsize;
658 spin_unlock(&nfsd_drc_lock);
663 static void nfsd4_put_drc_mem(int slotsize, int num)
665 spin_lock(&nfsd_drc_lock);
666 nfsd_drc_mem_used -= slotsize * num;
667 spin_unlock(&nfsd_drc_lock);
670 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
672 struct nfsd4_session *new;
675 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
676 + sizeof(struct nfsd4_session) > PAGE_SIZE);
677 mem = numslots * sizeof(struct nfsd4_slot *);
679 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
682 /* allocate each struct nfsd4_slot and data cache in one piece */
683 for (i = 0; i < numslots; i++) {
684 mem = sizeof(struct nfsd4_slot) + slotsize;
685 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
686 if (!new->se_slots[i])
692 kfree(new->se_slots[i]);
697 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
699 u32 maxrpc = nfsd_serv->sv_max_mesg;
701 new->maxreqs = numslots;
702 new->maxresp_cached = min_t(u32, req->maxresp_cached,
703 slotsize + NFSD_MIN_HDR_SEQ_SZ);
704 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
705 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
706 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
709 static void free_conn(struct nfsd4_conn *c)
711 svc_xprt_put(c->cn_xprt);
715 static void nfsd4_conn_lost(struct svc_xpt_user *u)
717 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
718 struct nfs4_client *clp = c->cn_session->se_client;
720 spin_lock(&clp->cl_lock);
721 if (!list_empty(&c->cn_persession)) {
722 list_del(&c->cn_persession);
725 spin_unlock(&clp->cl_lock);
726 nfsd4_probe_callback(clp);
729 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
731 struct nfsd4_conn *conn;
733 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
736 svc_xprt_get(rqstp->rq_xprt);
737 conn->cn_xprt = rqstp->rq_xprt;
738 conn->cn_flags = flags;
739 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
743 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
745 conn->cn_session = ses;
746 list_add(&conn->cn_persession, &ses->se_conns);
749 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
751 struct nfs4_client *clp = ses->se_client;
753 spin_lock(&clp->cl_lock);
754 __nfsd4_hash_conn(conn, ses);
755 spin_unlock(&clp->cl_lock);
758 static int nfsd4_register_conn(struct nfsd4_conn *conn)
760 conn->cn_xpt_user.callback = nfsd4_conn_lost;
761 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
764 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
766 struct nfsd4_conn *conn;
769 conn = alloc_conn(rqstp, dir);
771 return nfserr_jukebox;
772 nfsd4_hash_conn(conn, ses);
773 ret = nfsd4_register_conn(conn);
775 /* oops; xprt is already down: */
776 nfsd4_conn_lost(&conn->cn_xpt_user);
780 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
782 u32 dir = NFS4_CDFC4_FORE;
784 if (ses->se_flags & SESSION4_BACK_CHAN)
785 dir |= NFS4_CDFC4_BACK;
787 return nfsd4_new_conn(rqstp, ses, dir);
790 /* must be called under client_lock */
791 static void nfsd4_del_conns(struct nfsd4_session *s)
793 struct nfs4_client *clp = s->se_client;
794 struct nfsd4_conn *c;
796 spin_lock(&clp->cl_lock);
797 while (!list_empty(&s->se_conns)) {
798 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
799 list_del_init(&c->cn_persession);
800 spin_unlock(&clp->cl_lock);
802 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
805 spin_lock(&clp->cl_lock);
807 spin_unlock(&clp->cl_lock);
810 void free_session(struct kref *kref)
812 struct nfsd4_session *ses;
815 ses = container_of(kref, struct nfsd4_session, se_ref);
816 nfsd4_del_conns(ses);
817 spin_lock(&nfsd_drc_lock);
818 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
819 nfsd_drc_mem_used -= mem;
820 spin_unlock(&nfsd_drc_lock);
821 free_session_slots(ses);
825 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
827 struct nfsd4_session *new;
828 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
829 int numslots, slotsize;
834 * Note decreasing slot size below client's request may
835 * make it difficult for client to function correctly, whereas
836 * decreasing the number of slots will (just?) affect
837 * performance. When short on memory we therefore prefer to
838 * decrease number of slots instead of their size.
840 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
841 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
845 new = alloc_session(slotsize, numslots);
847 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
850 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
852 new->se_client = clp;
855 INIT_LIST_HEAD(&new->se_conns);
857 new->se_cb_seq_nr = 1;
858 new->se_flags = cses->flags;
859 new->se_cb_prog = cses->callback_prog;
860 kref_init(&new->se_ref);
861 idx = hash_sessionid(&new->se_sessionid);
862 spin_lock(&client_lock);
863 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
864 spin_lock(&clp->cl_lock);
865 list_add(&new->se_perclnt, &clp->cl_sessions);
866 spin_unlock(&clp->cl_lock);
867 spin_unlock(&client_lock);
869 status = nfsd4_new_conn_from_crses(rqstp, new);
870 /* whoops: benny points out, status is ignored! (err, or bogus) */
872 free_session(&new->se_ref);
875 if (cses->flags & SESSION4_BACK_CHAN) {
876 struct sockaddr *sa = svc_addr(rqstp);
878 * This is a little silly; with sessions there's no real
879 * use for the callback address. Use the peer address
880 * as a reasonable default for now, but consider fixing
881 * the rpc client not to require an address in the
884 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
885 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
887 nfsd4_probe_callback(clp);
891 /* caller must hold client_lock */
892 static struct nfsd4_session *
893 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
895 struct nfsd4_session *elem;
898 dump_sessionid(__func__, sessionid);
899 idx = hash_sessionid(sessionid);
900 /* Search in the appropriate list */
901 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
902 if (!memcmp(elem->se_sessionid.data, sessionid->data,
903 NFS4_MAX_SESSIONID_LEN)) {
908 dprintk("%s: session not found\n", __func__);
912 /* caller must hold client_lock */
914 unhash_session(struct nfsd4_session *ses)
916 list_del(&ses->se_hash);
917 spin_lock(&ses->se_client->cl_lock);
918 list_del(&ses->se_perclnt);
919 spin_unlock(&ses->se_client->cl_lock);
922 /* must be called under the client_lock */
924 renew_client_locked(struct nfs4_client *clp)
926 if (is_client_expired(clp)) {
927 dprintk("%s: client (clientid %08x/%08x) already expired\n",
929 clp->cl_clientid.cl_boot,
930 clp->cl_clientid.cl_id);
935 * Move client to the end to the LRU list.
937 dprintk("renewing client (clientid %08x/%08x)\n",
938 clp->cl_clientid.cl_boot,
939 clp->cl_clientid.cl_id);
940 list_move_tail(&clp->cl_lru, &client_lru);
941 clp->cl_time = get_seconds();
945 renew_client(struct nfs4_client *clp)
947 spin_lock(&client_lock);
948 renew_client_locked(clp);
949 spin_unlock(&client_lock);
952 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
954 STALE_CLIENTID(clientid_t *clid)
956 if (clid->cl_boot == boot_time)
958 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
959 clid->cl_boot, clid->cl_id, boot_time);
964 * XXX Should we use a slab cache ?
965 * This type of memory management is somewhat inefficient, but we use it
966 * anyway since SETCLIENTID is not a common operation.
968 static struct nfs4_client *alloc_client(struct xdr_netobj name)
970 struct nfs4_client *clp;
972 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
975 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
976 if (clp->cl_name.data == NULL) {
980 memcpy(clp->cl_name.data, name.data, name.len);
981 clp->cl_name.len = name.len;
986 free_client(struct nfs4_client *clp)
988 while (!list_empty(&clp->cl_sessions)) {
989 struct nfsd4_session *ses;
990 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
992 list_del(&ses->se_perclnt);
993 nfsd4_put_session(ses);
995 if (clp->cl_cred.cr_group_info)
996 put_group_info(clp->cl_cred.cr_group_info);
997 kfree(clp->cl_principal);
998 kfree(clp->cl_name.data);
1003 release_session_client(struct nfsd4_session *session)
1005 struct nfs4_client *clp = session->se_client;
1007 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1009 if (is_client_expired(clp)) {
1011 session->se_client = NULL;
1013 renew_client_locked(clp);
1014 spin_unlock(&client_lock);
1017 /* must be called under the client_lock */
1019 unhash_client_locked(struct nfs4_client *clp)
1021 struct nfsd4_session *ses;
1023 mark_client_expired(clp);
1024 list_del(&clp->cl_lru);
1025 spin_lock(&clp->cl_lock);
1026 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1027 list_del_init(&ses->se_hash);
1028 spin_unlock(&clp->cl_lock);
1032 expire_client(struct nfs4_client *clp)
1034 struct nfs4_openowner *oo;
1035 struct nfs4_delegation *dp;
1036 struct list_head reaplist;
1038 INIT_LIST_HEAD(&reaplist);
1039 spin_lock(&recall_lock);
1040 while (!list_empty(&clp->cl_delegations)) {
1041 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1042 list_del_init(&dp->dl_perclnt);
1043 list_move(&dp->dl_recall_lru, &reaplist);
1045 spin_unlock(&recall_lock);
1046 while (!list_empty(&reaplist)) {
1047 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1048 list_del_init(&dp->dl_recall_lru);
1049 unhash_delegation(dp);
1051 while (!list_empty(&clp->cl_openowners)) {
1052 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1053 release_openowner(oo);
1055 nfsd4_shutdown_callback(clp);
1056 if (clp->cl_cb_conn.cb_xprt)
1057 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1058 list_del(&clp->cl_idhash);
1059 list_del(&clp->cl_strhash);
1060 spin_lock(&client_lock);
1061 unhash_client_locked(clp);
1062 if (atomic_read(&clp->cl_refcount) == 0)
1064 spin_unlock(&client_lock);
1067 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1069 memcpy(target->cl_verifier.data, source->data,
1070 sizeof(target->cl_verifier.data));
1073 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1075 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1076 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1079 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1081 target->cr_uid = source->cr_uid;
1082 target->cr_gid = source->cr_gid;
1083 target->cr_group_info = source->cr_group_info;
1084 get_group_info(target->cr_group_info);
1087 static int same_name(const char *n1, const char *n2)
1089 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1093 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1095 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1099 same_clid(clientid_t *cl1, clientid_t *cl2)
1101 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1104 /* XXX what about NGROUP */
1106 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1108 return cr1->cr_uid == cr2->cr_uid;
1111 static void gen_clid(struct nfs4_client *clp)
1113 static u32 current_clientid = 1;
1115 clp->cl_clientid.cl_boot = boot_time;
1116 clp->cl_clientid.cl_id = current_clientid++;
1119 static void gen_confirm(struct nfs4_client *clp)
1124 p = (u32 *)clp->cl_confirm.data;
1125 *p++ = get_seconds();
1129 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1131 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1134 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1136 struct nfs4_stid *s;
1138 s = find_stateid(cl, t);
1141 if (typemask & s->sc_type)
1146 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1147 struct svc_rqst *rqstp, nfs4_verifier *verf)
1149 struct nfs4_client *clp;
1150 struct sockaddr *sa = svc_addr(rqstp);
1153 clp = alloc_client(name);
1157 INIT_LIST_HEAD(&clp->cl_sessions);
1159 princ = svc_gss_principal(rqstp);
1161 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1162 if (clp->cl_principal == NULL) {
1168 idr_init(&clp->cl_stateids);
1169 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1170 atomic_set(&clp->cl_refcount, 0);
1171 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1172 INIT_LIST_HEAD(&clp->cl_idhash);
1173 INIT_LIST_HEAD(&clp->cl_strhash);
1174 INIT_LIST_HEAD(&clp->cl_openowners);
1175 INIT_LIST_HEAD(&clp->cl_delegations);
1176 INIT_LIST_HEAD(&clp->cl_lru);
1177 INIT_LIST_HEAD(&clp->cl_callbacks);
1178 spin_lock_init(&clp->cl_lock);
1179 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1180 clp->cl_time = get_seconds();
1181 clear_bit(0, &clp->cl_cb_slot_busy);
1182 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1183 copy_verf(clp, verf);
1184 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1185 clp->cl_flavor = rqstp->rq_flavor;
1186 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1188 clp->cl_cb_session = NULL;
1192 static int check_name(struct xdr_netobj name)
1196 if (name.len > NFS4_OPAQUE_LIMIT) {
1197 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1204 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1206 unsigned int idhashval;
1208 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1209 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1210 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1215 move_to_confirmed(struct nfs4_client *clp)
1217 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1218 unsigned int strhashval;
1220 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1221 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1222 strhashval = clientstr_hashval(clp->cl_recdir);
1223 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1227 static struct nfs4_client *
1228 find_confirmed_client(clientid_t *clid)
1230 struct nfs4_client *clp;
1231 unsigned int idhashval = clientid_hashval(clid->cl_id);
1233 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1234 if (same_clid(&clp->cl_clientid, clid))
1240 static struct nfs4_client *
1241 find_unconfirmed_client(clientid_t *clid)
1243 struct nfs4_client *clp;
1244 unsigned int idhashval = clientid_hashval(clid->cl_id);
1246 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1247 if (same_clid(&clp->cl_clientid, clid))
1253 static bool clp_used_exchangeid(struct nfs4_client *clp)
1255 return clp->cl_exchange_flags != 0;
1258 static struct nfs4_client *
1259 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1261 struct nfs4_client *clp;
1263 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1264 if (same_name(clp->cl_recdir, dname))
1270 static struct nfs4_client *
1271 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1273 struct nfs4_client *clp;
1275 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1276 if (same_name(clp->cl_recdir, dname))
1283 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1285 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1286 struct sockaddr *sa = svc_addr(rqstp);
1287 u32 scopeid = rpc_get_scope_id(sa);
1288 unsigned short expected_family;
1290 /* Currently, we only support tcp and tcp6 for the callback channel */
1291 if (se->se_callback_netid_len == 3 &&
1292 !memcmp(se->se_callback_netid_val, "tcp", 3))
1293 expected_family = AF_INET;
1294 else if (se->se_callback_netid_len == 4 &&
1295 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1296 expected_family = AF_INET6;
1300 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1301 se->se_callback_addr_len,
1302 (struct sockaddr *)&conn->cb_addr,
1303 sizeof(conn->cb_addr));
1305 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1308 if (conn->cb_addr.ss_family == AF_INET6)
1309 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1311 conn->cb_prog = se->se_callback_prog;
1312 conn->cb_ident = se->se_callback_ident;
1313 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1316 conn->cb_addr.ss_family = AF_UNSPEC;
1317 conn->cb_addrlen = 0;
1318 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1319 "will not receive delegations\n",
1320 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1326 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1329 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1331 struct nfsd4_slot *slot = resp->cstate.slot;
1334 dprintk("--> %s slot %p\n", __func__, slot);
1336 slot->sl_opcnt = resp->opcnt;
1337 slot->sl_status = resp->cstate.status;
1339 if (nfsd4_not_cached(resp)) {
1340 slot->sl_datalen = 0;
1343 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1344 base = (char *)resp->cstate.datap -
1345 (char *)resp->xbuf->head[0].iov_base;
1346 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1348 WARN("%s: sessions DRC could not cache compound\n", __func__);
1353 * Encode the replay sequence operation from the slot values.
1354 * If cachethis is FALSE encode the uncached rep error on the next
1355 * operation which sets resp->p and increments resp->opcnt for
1356 * nfs4svc_encode_compoundres.
1360 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1361 struct nfsd4_compoundres *resp)
1363 struct nfsd4_op *op;
1364 struct nfsd4_slot *slot = resp->cstate.slot;
1366 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1367 resp->opcnt, resp->cstate.slot->sl_cachethis);
1369 /* Encode the replayed sequence operation */
1370 op = &args->ops[resp->opcnt - 1];
1371 nfsd4_encode_operation(resp, op);
1373 /* Return nfserr_retry_uncached_rep in next operation. */
1374 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1375 op = &args->ops[resp->opcnt++];
1376 op->status = nfserr_retry_uncached_rep;
1377 nfsd4_encode_operation(resp, op);
1383 * The sequence operation is not cached because we can use the slot and
1387 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1388 struct nfsd4_sequence *seq)
1390 struct nfsd4_slot *slot = resp->cstate.slot;
1393 dprintk("--> %s slot %p\n", __func__, slot);
1395 /* Either returns 0 or nfserr_retry_uncached */
1396 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1397 if (status == nfserr_retry_uncached_rep)
1400 /* The sequence operation has been encoded, cstate->datap set. */
1401 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1403 resp->opcnt = slot->sl_opcnt;
1404 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1405 status = slot->sl_status;
1411 * Set the exchange_id flags returned by the server.
1414 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1416 /* pNFS is not supported */
1417 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1419 /* Referrals are supported, Migration is not. */
1420 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1422 /* set the wire flags to return to client. */
1423 clid->flags = new->cl_exchange_flags;
1427 nfsd4_exchange_id(struct svc_rqst *rqstp,
1428 struct nfsd4_compound_state *cstate,
1429 struct nfsd4_exchange_id *exid)
1431 struct nfs4_client *unconf, *conf, *new;
1433 unsigned int strhashval;
1434 char dname[HEXDIR_LEN];
1435 char addr_str[INET6_ADDRSTRLEN];
1436 nfs4_verifier verf = exid->verifier;
1437 struct sockaddr *sa = svc_addr(rqstp);
1439 rpc_ntop(sa, addr_str, sizeof(addr_str));
1440 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1441 "ip_addr=%s flags %x, spa_how %d\n",
1442 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1443 addr_str, exid->flags, exid->spa_how);
1445 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1446 return nfserr_inval;
1448 /* Currently only support SP4_NONE */
1449 switch (exid->spa_how) {
1453 return nfserr_serverfault;
1455 BUG(); /* checked by xdr code */
1457 return nfserr_serverfault; /* no excuse :-/ */
1460 status = nfs4_make_rec_clidname(dname, &exid->clname);
1465 strhashval = clientstr_hashval(dname);
1470 conf = find_confirmed_client_by_str(dname, strhashval);
1472 if (!clp_used_exchangeid(conf)) {
1473 status = nfserr_clid_inuse; /* XXX: ? */
1476 if (!same_verf(&verf, &conf->cl_verifier)) {
1477 /* 18.35.4 case 8 */
1478 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1479 status = nfserr_not_same;
1482 /* Client reboot: destroy old state */
1483 expire_client(conf);
1486 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1487 /* 18.35.4 case 9 */
1488 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1489 status = nfserr_perm;
1492 expire_client(conf);
1496 * Set bit when the owner id and verifier map to an already
1497 * confirmed client id (18.35.3).
1499 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1502 * Falling into 18.35.4 case 2, possible router replay.
1503 * Leave confirmed record intact and return same result.
1505 copy_verf(conf, &verf);
1510 /* 18.35.4 case 7 */
1511 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1512 status = nfserr_noent;
1516 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1519 * Possible retry or client restart. Per 18.35.4 case 4,
1520 * a new unconfirmed record should be generated regardless
1521 * of whether any properties have changed.
1523 expire_client(unconf);
1528 new = create_client(exid->clname, dname, rqstp, &verf);
1530 status = nfserr_jukebox;
1535 add_to_unconfirmed(new, strhashval);
1537 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1538 exid->clientid.cl_id = new->cl_clientid.cl_id;
1541 nfsd4_set_ex_flags(new, exid);
1543 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1544 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1548 nfs4_unlock_state();
1550 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1555 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1557 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1560 /* The slot is in use, and no response has been sent. */
1562 if (seqid == slot_seqid)
1563 return nfserr_jukebox;
1565 return nfserr_seq_misordered;
1568 if (likely(seqid == slot_seqid + 1))
1571 if (seqid == slot_seqid)
1572 return nfserr_replay_cache;
1574 if (seqid == 1 && (slot_seqid + 1) == 0)
1576 /* Misordered replay or misordered new request */
1577 return nfserr_seq_misordered;
1581 * Cache the create session result into the create session single DRC
1582 * slot cache by saving the xdr structure. sl_seqid has been set.
1583 * Do this for solo or embedded create session operations.
1586 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1587 struct nfsd4_clid_slot *slot, int nfserr)
1589 slot->sl_status = nfserr;
1590 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1594 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1595 struct nfsd4_clid_slot *slot)
1597 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1598 return slot->sl_status;
1601 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1602 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1603 1 + /* MIN tag is length with zero, only length */ \
1604 3 + /* version, opcount, opcode */ \
1605 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1606 /* seqid, slotID, slotID, cache */ \
1607 4 ) * sizeof(__be32))
1609 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1610 2 + /* verifier: AUTH_NULL, length 0 */\
1612 1 + /* MIN tag is length with zero, only length */ \
1613 3 + /* opcount, opcode, opstatus*/ \
1614 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1615 /* seqid, slotID, slotID, slotID, status */ \
1616 5 ) * sizeof(__be32))
1618 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1620 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1621 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1625 nfsd4_create_session(struct svc_rqst *rqstp,
1626 struct nfsd4_compound_state *cstate,
1627 struct nfsd4_create_session *cr_ses)
1629 struct sockaddr *sa = svc_addr(rqstp);
1630 struct nfs4_client *conf, *unconf;
1631 struct nfsd4_session *new;
1632 struct nfsd4_clid_slot *cs_slot = NULL;
1633 bool confirm_me = false;
1636 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1637 return nfserr_inval;
1640 unconf = find_unconfirmed_client(&cr_ses->clientid);
1641 conf = find_confirmed_client(&cr_ses->clientid);
1644 cs_slot = &conf->cl_cs_slot;
1645 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1646 if (status == nfserr_replay_cache) {
1647 dprintk("Got a create_session replay! seqid= %d\n",
1649 /* Return the cached reply status */
1650 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1652 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1653 status = nfserr_seq_misordered;
1654 dprintk("Sequence misordered!\n");
1655 dprintk("Expected seqid= %d but got seqid= %d\n",
1656 cs_slot->sl_seqid, cr_ses->seqid);
1659 } else if (unconf) {
1660 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1661 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1662 status = nfserr_clid_inuse;
1666 cs_slot = &unconf->cl_cs_slot;
1667 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1669 /* an unconfirmed replay returns misordered */
1670 status = nfserr_seq_misordered;
1677 status = nfserr_stale_clientid;
1682 * XXX: we should probably set this at creation time, and check
1683 * for consistent minorversion use throughout:
1685 conf->cl_minorversion = 1;
1687 * We do not support RDMA or persistent sessions
1689 cr_ses->flags &= ~SESSION4_PERSIST;
1690 cr_ses->flags &= ~SESSION4_RDMA;
1692 status = nfserr_toosmall;
1693 if (check_forechannel_attrs(cr_ses->fore_channel))
1696 status = nfserr_jukebox;
1697 new = alloc_init_session(rqstp, conf, cr_ses);
1701 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1702 NFS4_MAX_SESSIONID_LEN);
1703 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1704 sizeof(struct nfsd4_channel_attrs));
1705 cs_slot->sl_seqid++;
1706 cr_ses->seqid = cs_slot->sl_seqid;
1708 /* cache solo and embedded create sessions under the state lock */
1709 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1711 move_to_confirmed(conf);
1713 nfs4_unlock_state();
1714 dprintk("%s returns %d\n", __func__, ntohl(status));
1718 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1720 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1721 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1723 return argp->opcnt == resp->opcnt;
1726 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1729 case NFS4_CDFC4_FORE:
1730 case NFS4_CDFC4_BACK:
1732 case NFS4_CDFC4_FORE_OR_BOTH:
1733 case NFS4_CDFC4_BACK_OR_BOTH:
1734 *dir = NFS4_CDFC4_BOTH;
1737 return nfserr_inval;
1740 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1741 struct nfsd4_compound_state *cstate,
1742 struct nfsd4_bind_conn_to_session *bcts)
1746 if (!nfsd4_last_compound_op(rqstp))
1747 return nfserr_not_only_op;
1748 spin_lock(&client_lock);
1749 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1750 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1751 * client_lock iself: */
1752 if (cstate->session) {
1753 nfsd4_get_session(cstate->session);
1754 atomic_inc(&cstate->session->se_client->cl_refcount);
1756 spin_unlock(&client_lock);
1757 if (!cstate->session)
1758 return nfserr_badsession;
1760 status = nfsd4_map_bcts_dir(&bcts->dir);
1762 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1766 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1770 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1774 nfsd4_destroy_session(struct svc_rqst *r,
1775 struct nfsd4_compound_state *cstate,
1776 struct nfsd4_destroy_session *sessionid)
1778 struct nfsd4_session *ses;
1779 u32 status = nfserr_badsession;
1782 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1783 * - Should we return nfserr_back_chan_busy if waiting for
1784 * callbacks on to-be-destroyed session?
1785 * - Do we need to clear any callback info from previous session?
1788 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1789 if (!nfsd4_last_compound_op(r))
1790 return nfserr_not_only_op;
1792 dump_sessionid(__func__, &sessionid->sessionid);
1793 spin_lock(&client_lock);
1794 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1796 spin_unlock(&client_lock);
1800 unhash_session(ses);
1801 spin_unlock(&client_lock);
1804 nfsd4_probe_callback_sync(ses->se_client);
1805 nfs4_unlock_state();
1807 nfsd4_del_conns(ses);
1809 nfsd4_put_session(ses);
1812 dprintk("%s returns %d\n", __func__, ntohl(status));
1816 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1818 struct nfsd4_conn *c;
1820 list_for_each_entry(c, &s->se_conns, cn_persession) {
1821 if (c->cn_xprt == xpt) {
1828 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1830 struct nfs4_client *clp = ses->se_client;
1831 struct nfsd4_conn *c;
1834 spin_lock(&clp->cl_lock);
1835 c = __nfsd4_find_conn(new->cn_xprt, ses);
1837 spin_unlock(&clp->cl_lock);
1841 __nfsd4_hash_conn(new, ses);
1842 spin_unlock(&clp->cl_lock);
1843 ret = nfsd4_register_conn(new);
1845 /* oops; xprt is already down: */
1846 nfsd4_conn_lost(&new->cn_xpt_user);
1850 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1852 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1854 return args->opcnt > session->se_fchannel.maxops;
1857 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1858 struct nfsd4_session *session)
1860 struct xdr_buf *xb = &rqstp->rq_arg;
1862 return xb->len > session->se_fchannel.maxreq_sz;
1866 nfsd4_sequence(struct svc_rqst *rqstp,
1867 struct nfsd4_compound_state *cstate,
1868 struct nfsd4_sequence *seq)
1870 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1871 struct nfsd4_session *session;
1872 struct nfsd4_slot *slot;
1873 struct nfsd4_conn *conn;
1876 if (resp->opcnt != 1)
1877 return nfserr_sequence_pos;
1880 * Will be either used or freed by nfsd4_sequence_check_conn
1883 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1885 return nfserr_jukebox;
1887 spin_lock(&client_lock);
1888 status = nfserr_badsession;
1889 session = find_in_sessionid_hashtbl(&seq->sessionid);
1893 status = nfserr_too_many_ops;
1894 if (nfsd4_session_too_many_ops(rqstp, session))
1897 status = nfserr_req_too_big;
1898 if (nfsd4_request_too_big(rqstp, session))
1901 status = nfserr_badslot;
1902 if (seq->slotid >= session->se_fchannel.maxreqs)
1905 slot = session->se_slots[seq->slotid];
1906 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1908 /* We do not negotiate the number of slots yet, so set the
1909 * maxslots to the session maxreqs which is used to encode
1910 * sr_highest_slotid and the sr_target_slot id to maxslots */
1911 seq->maxslots = session->se_fchannel.maxreqs;
1913 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1914 if (status == nfserr_replay_cache) {
1915 cstate->slot = slot;
1916 cstate->session = session;
1917 /* Return the cached reply status and set cstate->status
1918 * for nfsd4_proc_compound processing */
1919 status = nfsd4_replay_cache_entry(resp, seq);
1920 cstate->status = nfserr_replay_cache;
1926 nfsd4_sequence_check_conn(conn, session);
1929 /* Success! bump slot seqid */
1930 slot->sl_inuse = true;
1931 slot->sl_seqid = seq->seqid;
1932 slot->sl_cachethis = seq->cachethis;
1934 cstate->slot = slot;
1935 cstate->session = session;
1938 /* Hold a session reference until done processing the compound. */
1939 if (cstate->session) {
1940 struct nfs4_client *clp = session->se_client;
1942 nfsd4_get_session(cstate->session);
1943 atomic_inc(&clp->cl_refcount);
1944 if (clp->cl_cb_state == NFSD4_CB_DOWN)
1945 seq->status_flags |= SEQ4_STATUS_CB_PATH_DOWN;
1948 spin_unlock(&client_lock);
1949 dprintk("%s: return %d\n", __func__, ntohl(status));
1954 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1958 if (rc->rca_one_fs) {
1959 if (!cstate->current_fh.fh_dentry)
1960 return nfserr_nofilehandle;
1962 * We don't take advantage of the rca_one_fs case.
1963 * That's OK, it's optional, we can safely ignore it.
1969 status = nfserr_complete_already;
1970 if (cstate->session->se_client->cl_firststate)
1973 status = nfserr_stale_clientid;
1974 if (is_client_expired(cstate->session->se_client))
1976 * The following error isn't really legal.
1977 * But we only get here if the client just explicitly
1978 * destroyed the client. Surely it no longer cares what
1979 * error it gets back on an operation for the dead
1985 nfsd4_create_clid_dir(cstate->session->se_client);
1987 nfs4_unlock_state();
1992 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1993 struct nfsd4_setclientid *setclid)
1995 struct xdr_netobj clname = {
1996 .len = setclid->se_namelen,
1997 .data = setclid->se_name,
1999 nfs4_verifier clverifier = setclid->se_verf;
2000 unsigned int strhashval;
2001 struct nfs4_client *conf, *unconf, *new;
2003 char dname[HEXDIR_LEN];
2005 if (!check_name(clname))
2006 return nfserr_inval;
2008 status = nfs4_make_rec_clidname(dname, &clname);
2013 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2014 * We get here on a DRC miss.
2017 strhashval = clientstr_hashval(dname);
2020 conf = find_confirmed_client_by_str(dname, strhashval);
2022 /* RFC 3530 14.2.33 CASE 0: */
2023 status = nfserr_clid_inuse;
2024 if (clp_used_exchangeid(conf))
2026 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2027 char addr_str[INET6_ADDRSTRLEN];
2028 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2030 dprintk("NFSD: setclientid: string in use by client "
2031 "at %s\n", addr_str);
2036 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2037 * has a description of SETCLIENTID request processing consisting
2038 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2040 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2041 status = nfserr_jukebox;
2044 * RFC 3530 14.2.33 CASE 4:
2045 * placed first, because it is the normal case
2048 expire_client(unconf);
2049 new = create_client(clname, dname, rqstp, &clverifier);
2053 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2055 * RFC 3530 14.2.33 CASE 1:
2056 * probable callback update
2059 /* Note this is removing unconfirmed {*x***},
2060 * which is stronger than RFC recommended {vxc**}.
2061 * This has the advantage that there is at most
2062 * one {*x***} in either list at any time.
2064 expire_client(unconf);
2066 new = create_client(clname, dname, rqstp, &clverifier);
2069 copy_clid(new, conf);
2070 } else if (!unconf) {
2072 * RFC 3530 14.2.33 CASE 2:
2073 * probable client reboot; state will be removed if
2076 new = create_client(clname, dname, rqstp, &clverifier);
2082 * RFC 3530 14.2.33 CASE 3:
2083 * probable client reboot; state will be removed if
2086 expire_client(unconf);
2087 new = create_client(clname, dname, rqstp, &clverifier);
2093 * XXX: we should probably set this at creation time, and check
2094 * for consistent minorversion use throughout:
2096 new->cl_minorversion = 0;
2097 gen_callback(new, setclid, rqstp);
2098 add_to_unconfirmed(new, strhashval);
2099 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2100 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2101 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2104 nfs4_unlock_state();
2110 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2111 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2112 * bullets, labeled as CASE1 - CASE4 below.
2115 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2116 struct nfsd4_compound_state *cstate,
2117 struct nfsd4_setclientid_confirm *setclientid_confirm)
2119 struct sockaddr *sa = svc_addr(rqstp);
2120 struct nfs4_client *conf, *unconf;
2121 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2122 clientid_t * clid = &setclientid_confirm->sc_clientid;
2125 if (STALE_CLIENTID(clid))
2126 return nfserr_stale_clientid;
2128 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2129 * We get here on a DRC miss.
2134 conf = find_confirmed_client(clid);
2135 unconf = find_unconfirmed_client(clid);
2137 status = nfserr_clid_inuse;
2138 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2140 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2144 * section 14.2.34 of RFC 3530 has a description of
2145 * SETCLIENTID_CONFIRM request processing consisting
2146 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2148 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2150 * RFC 3530 14.2.34 CASE 1:
2153 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2154 status = nfserr_clid_inuse;
2156 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2157 nfsd4_probe_callback(conf);
2158 expire_client(unconf);
2162 } else if (conf && !unconf) {
2164 * RFC 3530 14.2.34 CASE 2:
2165 * probable retransmitted request; play it safe and
2168 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2169 status = nfserr_clid_inuse;
2172 } else if (!conf && unconf
2173 && same_verf(&unconf->cl_confirm, &confirm)) {
2175 * RFC 3530 14.2.34 CASE 3:
2176 * Normal case; new or rebooted client:
2178 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2179 status = nfserr_clid_inuse;
2182 clientstr_hashval(unconf->cl_recdir);
2183 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2186 nfsd4_remove_clid_dir(conf);
2187 expire_client(conf);
2189 move_to_confirmed(unconf);
2191 nfsd4_probe_callback(conf);
2194 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2195 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2198 * RFC 3530 14.2.34 CASE 4:
2199 * Client probably hasn't noticed that we rebooted yet.
2201 status = nfserr_stale_clientid;
2203 /* check that we have hit one of the cases...*/
2204 status = nfserr_clid_inuse;
2207 nfs4_unlock_state();
2211 /* OPEN Share state helper functions */
2212 static inline struct nfs4_file *
2213 alloc_init_file(struct inode *ino)
2215 struct nfs4_file *fp;
2216 unsigned int hashval = file_hashval(ino);
2218 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
2220 atomic_set(&fp->fi_ref, 1);
2221 INIT_LIST_HEAD(&fp->fi_hash);
2222 INIT_LIST_HEAD(&fp->fi_stateids);
2223 INIT_LIST_HEAD(&fp->fi_delegations);
2224 fp->fi_inode = igrab(ino);
2225 fp->fi_had_conflict = false;
2226 fp->fi_lease = NULL;
2227 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2228 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2229 spin_lock(&recall_lock);
2230 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2231 spin_unlock(&recall_lock);
2238 nfsd4_free_slab(struct kmem_cache **slab)
2242 kmem_cache_destroy(*slab);
2247 nfsd4_free_slabs(void)
2249 nfsd4_free_slab(&openowner_slab);
2250 nfsd4_free_slab(&lockowner_slab);
2251 nfsd4_free_slab(&file_slab);
2252 nfsd4_free_slab(&stateid_slab);
2253 nfsd4_free_slab(&deleg_slab);
2257 nfsd4_init_slabs(void)
2259 openowner_slab = kmem_cache_create("nfsd4_openowners",
2260 sizeof(struct nfs4_openowner), 0, 0, NULL);
2261 if (openowner_slab == NULL)
2263 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2264 sizeof(struct nfs4_openowner), 0, 0, NULL);
2265 if (lockowner_slab == NULL)
2267 file_slab = kmem_cache_create("nfsd4_files",
2268 sizeof(struct nfs4_file), 0, 0, NULL);
2269 if (file_slab == NULL)
2271 stateid_slab = kmem_cache_create("nfsd4_stateids",
2272 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2273 if (stateid_slab == NULL)
2275 deleg_slab = kmem_cache_create("nfsd4_delegations",
2276 sizeof(struct nfs4_delegation), 0, 0, NULL);
2277 if (deleg_slab == NULL)
2282 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2286 void nfs4_free_openowner(struct nfs4_openowner *oo)
2288 kfree(oo->oo_owner.so_owner.data);
2289 kmem_cache_free(openowner_slab, oo);
2292 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2294 kfree(lo->lo_owner.so_owner.data);
2295 kmem_cache_free(lockowner_slab, lo);
2298 static void init_nfs4_replay(struct nfs4_replay *rp)
2300 rp->rp_status = nfserr_serverfault;
2302 rp->rp_buf = rp->rp_ibuf;
2305 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2307 struct nfs4_stateowner *sop;
2309 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2313 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2314 if (!sop->so_owner.data) {
2315 kmem_cache_free(slab, sop);
2318 sop->so_owner.len = owner->len;
2320 INIT_LIST_HEAD(&sop->so_stateids);
2321 sop->so_client = clp;
2322 init_nfs4_replay(&sop->so_replay);
2326 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2328 list_add(&oo->oo_owner.so_strhash, &open_ownerstr_hashtbl[strhashval]);
2329 list_add(&oo->oo_perclient, &clp->cl_openowners);
2332 static struct nfs4_openowner *
2333 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2334 struct nfs4_openowner *oo;
2336 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2339 oo->oo_owner.so_is_open_owner = 1;
2340 oo->oo_owner.so_seqid = open->op_seqid;
2343 oo->oo_last_closed_stid = NULL;
2344 INIT_LIST_HEAD(&oo->oo_close_lru);
2345 hash_openowner(oo, clp, strhashval);
2349 static inline __be32 init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2350 struct nfs4_openowner *oo = open->op_openowner;
2351 struct nfs4_client *clp = oo->oo_owner.so_client;
2354 status = init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2357 INIT_LIST_HEAD(&stp->st_lockowners);
2358 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2359 list_add(&stp->st_perfile, &fp->fi_stateids);
2360 stp->st_stateowner = &oo->oo_owner;
2363 stp->st_access_bmap = 0;
2364 stp->st_deny_bmap = 0;
2365 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
2366 &stp->st_access_bmap);
2367 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2368 stp->st_openstp = NULL;
2373 move_to_close_lru(struct nfs4_openowner *oo)
2375 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2377 list_move_tail(&oo->oo_close_lru, &close_lru);
2378 oo->oo_time = get_seconds();
2382 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2385 return (sop->so_owner.len == owner->len) &&
2386 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2387 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2390 static struct nfs4_openowner *
2391 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2393 struct nfs4_stateowner *so = NULL;
2395 list_for_each_entry(so, &open_ownerstr_hashtbl[hashval], so_strhash) {
2396 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
2397 return container_of(so, struct nfs4_openowner, oo_owner);
2402 /* search file_hashtbl[] for file */
2403 static struct nfs4_file *
2404 find_file(struct inode *ino)
2406 unsigned int hashval = file_hashval(ino);
2407 struct nfs4_file *fp;
2409 spin_lock(&recall_lock);
2410 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2411 if (fp->fi_inode == ino) {
2413 spin_unlock(&recall_lock);
2417 spin_unlock(&recall_lock);
2421 static inline int access_valid(u32 x, u32 minorversion)
2423 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
2425 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
2427 x &= ~NFS4_SHARE_ACCESS_MASK;
2428 if (minorversion && x) {
2429 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
2431 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
2433 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
2440 static inline int deny_valid(u32 x)
2442 /* Note: unlike access bits, deny bits may be zero. */
2443 return x <= NFS4_SHARE_DENY_BOTH;
2447 * Called to check deny when READ with all zero stateid or
2448 * WRITE with all zero or all one stateid
2451 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2453 struct inode *ino = current_fh->fh_dentry->d_inode;
2454 struct nfs4_file *fp;
2455 struct nfs4_ol_stateid *stp;
2458 dprintk("NFSD: nfs4_share_conflict\n");
2460 fp = find_file(ino);
2463 ret = nfserr_locked;
2464 /* Search for conflicting share reservations */
2465 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2466 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2467 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2476 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2478 /* We're assuming the state code never drops its reference
2479 * without first removing the lease. Since we're in this lease
2480 * callback (and since the lease code is serialized by the kernel
2481 * lock) we know the server hasn't removed the lease yet, we know
2482 * it's safe to take a reference: */
2483 atomic_inc(&dp->dl_count);
2485 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2487 /* only place dl_time is set. protected by lock_flocks*/
2488 dp->dl_time = get_seconds();
2490 nfsd4_cb_recall(dp);
2493 /* Called from break_lease() with lock_flocks() held. */
2494 static void nfsd_break_deleg_cb(struct file_lock *fl)
2496 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2497 struct nfs4_delegation *dp;
2500 /* We assume break_lease is only called once per lease: */
2501 BUG_ON(fp->fi_had_conflict);
2503 * We don't want the locks code to timeout the lease for us;
2504 * we'll remove it ourself if a delegation isn't returned
2507 fl->fl_break_time = 0;
2509 spin_lock(&recall_lock);
2510 fp->fi_had_conflict = true;
2511 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2512 nfsd_break_one_deleg(dp);
2513 spin_unlock(&recall_lock);
2517 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2520 return lease_modify(onlist, arg);
2525 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2526 .lm_break = nfsd_break_deleg_cb,
2527 .lm_change = nfsd_change_deleg_cb,
2530 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2532 if (nfsd4_has_session(cstate))
2534 if (seqid == so->so_seqid - 1)
2535 return nfserr_replay_me;
2536 if (seqid == so->so_seqid)
2538 return nfserr_bad_seqid;
2542 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2543 struct nfsd4_open *open)
2545 clientid_t *clientid = &open->op_clientid;
2546 struct nfs4_client *clp = NULL;
2547 unsigned int strhashval;
2548 struct nfs4_openowner *oo = NULL;
2551 if (!check_name(open->op_owner))
2552 return nfserr_inval;
2554 if (STALE_CLIENTID(&open->op_clientid))
2555 return nfserr_stale_clientid;
2557 strhashval = open_ownerstr_hashval(clientid->cl_id, &open->op_owner);
2558 oo = find_openstateowner_str(strhashval, open);
2559 open->op_openowner = oo;
2561 /* Make sure the client's lease hasn't expired. */
2562 clp = find_confirmed_client(clientid);
2564 return nfserr_expired;
2567 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2568 /* Replace unconfirmed owners without checking for replay. */
2569 clp = oo->oo_owner.so_client;
2570 release_openowner(oo);
2571 open->op_openowner = NULL;
2574 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2578 if (open->op_openowner == NULL) {
2579 oo = alloc_init_open_stateowner(strhashval, clp, open);
2581 return nfserr_jukebox;
2582 open->op_openowner = oo;
2584 list_del_init(&oo->oo_close_lru);
2585 renew_client(oo->oo_owner.so_client);
2589 static inline __be32
2590 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2592 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2593 return nfserr_openmode;
2598 static int share_access_to_flags(u32 share_access)
2600 share_access &= ~NFS4_SHARE_WANT_MASK;
2602 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2605 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2607 struct nfs4_stid *ret;
2609 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2612 return delegstateid(ret);
2616 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2617 struct nfs4_delegation **dp)
2620 __be32 status = nfserr_bad_stateid;
2622 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2625 flags = share_access_to_flags(open->op_share_access);
2626 status = nfs4_check_delegmode(*dp, flags);
2630 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2634 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2639 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2641 struct nfs4_ol_stateid *local;
2642 struct nfs4_openowner *oo = open->op_openowner;
2644 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2645 /* ignore lock owners */
2646 if (local->st_stateowner->so_is_open_owner == 0)
2648 /* remember if we have seen this open owner */
2649 if (local->st_stateowner == &oo->oo_owner)
2651 /* check for conflicting share reservations */
2652 if (!test_share(local, open))
2653 return nfserr_share_denied;
2658 static inline struct nfs4_ol_stateid *
2659 nfs4_alloc_stateid(void)
2661 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2664 static inline int nfs4_access_to_access(u32 nfs4_access)
2668 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2669 flags |= NFSD_MAY_READ;
2670 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2671 flags |= NFSD_MAY_WRITE;
2675 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2676 struct svc_fh *cur_fh, struct nfsd4_open *open)
2679 int oflag = nfs4_access_to_omode(open->op_share_access);
2680 int access = nfs4_access_to_access(open->op_share_access);
2682 /* CLAIM_DELEGATE_CUR is used in response to a broken lease;
2683 * allowing it to break the lease and return EAGAIN leaves the
2684 * client unable to make progress in returning the delegation */
2685 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2686 access |= NFSD_MAY_NOT_BREAK_LEASE;
2688 if (!fp->fi_fds[oflag]) {
2689 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2690 &fp->fi_fds[oflag]);
2694 nfs4_file_get_access(fp, oflag);
2700 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_ol_stateid **stpp,
2701 struct nfs4_file *fp, struct svc_fh *cur_fh,
2702 struct nfsd4_open *open)
2704 struct nfs4_ol_stateid *stp;
2707 stp = nfs4_alloc_stateid();
2709 return nfserr_jukebox;
2711 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2713 kmem_cache_free(stateid_slab, stp);
2720 static inline __be32
2721 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2722 struct nfsd4_open *open)
2724 struct iattr iattr = {
2725 .ia_valid = ATTR_SIZE,
2728 if (!open->op_truncate)
2730 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2731 return nfserr_inval;
2732 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2736 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)
2738 u32 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2742 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2744 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2748 status = nfsd4_truncate(rqstp, cur_fh, open);
2751 int oflag = nfs4_access_to_omode(op_share_access);
2752 nfs4_file_put_access(fp, oflag);
2756 /* remember the open */
2757 __set_bit(op_share_access, &stp->st_access_bmap);
2758 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2765 nfs4_set_claim_prev(struct nfsd4_open *open)
2767 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2768 open->op_openowner->oo_owner.so_client->cl_firststate = 1;
2771 /* Should we give out recallable state?: */
2772 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2774 if (clp->cl_cb_state == NFSD4_CB_UP)
2777 * In the sessions case, since we don't have to establish a
2778 * separate connection for callbacks, we assume it's OK
2779 * until we hear otherwise:
2781 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2784 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2786 struct file_lock *fl;
2788 fl = locks_alloc_lock();
2791 locks_init_lock(fl);
2792 fl->fl_lmops = &nfsd_lease_mng_ops;
2793 fl->fl_flags = FL_LEASE;
2794 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2795 fl->fl_end = OFFSET_MAX;
2796 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2797 fl->fl_pid = current->tgid;
2801 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2803 struct nfs4_file *fp = dp->dl_file;
2804 struct file_lock *fl;
2807 fl = nfs4_alloc_init_lease(dp, flag);
2810 fl->fl_file = find_readable_file(fp);
2811 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2812 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2814 list_del_init(&dp->dl_perclnt);
2815 locks_free_lock(fl);
2819 fp->fi_deleg_file = fl->fl_file;
2820 get_file(fp->fi_deleg_file);
2821 atomic_set(&fp->fi_delegees, 1);
2822 list_add(&dp->dl_perfile, &fp->fi_delegations);
2826 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2828 struct nfs4_file *fp = dp->dl_file;
2831 return nfs4_setlease(dp, flag);
2832 spin_lock(&recall_lock);
2833 if (fp->fi_had_conflict) {
2834 spin_unlock(&recall_lock);
2837 atomic_inc(&fp->fi_delegees);
2838 list_add(&dp->dl_perfile, &fp->fi_delegations);
2839 spin_unlock(&recall_lock);
2840 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2845 * Attempt to hand out a delegation.
2848 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2850 struct nfs4_delegation *dp;
2851 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2853 int status, flag = 0;
2855 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2856 flag = NFS4_OPEN_DELEGATE_NONE;
2857 open->op_recall = 0;
2858 switch (open->op_claim_type) {
2859 case NFS4_OPEN_CLAIM_PREVIOUS:
2861 open->op_recall = 1;
2862 flag = open->op_delegate_type;
2863 if (flag == NFS4_OPEN_DELEGATE_NONE)
2866 case NFS4_OPEN_CLAIM_NULL:
2867 /* Let's not give out any delegations till everyone's
2868 * had the chance to reclaim theirs.... */
2869 if (locks_in_grace())
2871 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2873 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2874 flag = NFS4_OPEN_DELEGATE_WRITE;
2876 flag = NFS4_OPEN_DELEGATE_READ;
2882 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2885 status = nfs4_set_delegation(dp, flag);
2889 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2891 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2892 STATEID_VAL(&dp->dl_stid.sc_stateid));
2894 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2895 && flag == NFS4_OPEN_DELEGATE_NONE
2896 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2897 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2898 open->op_delegate_type = flag;
2901 nfs4_put_delegation(dp);
2903 flag = NFS4_OPEN_DELEGATE_NONE;
2908 * called with nfs4_lock_state() held.
2911 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2913 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2914 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2915 struct nfs4_file *fp = NULL;
2916 struct inode *ino = current_fh->fh_dentry->d_inode;
2917 struct nfs4_ol_stateid *stp = NULL;
2918 struct nfs4_delegation *dp = NULL;
2921 status = nfserr_inval;
2922 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2923 || !deny_valid(open->op_share_deny))
2926 * Lookup file; if found, lookup stateid and check open request,
2927 * and check for delegations in the process of being recalled.
2928 * If not found, create the nfs4_file struct
2930 fp = find_file(ino);
2932 if ((status = nfs4_check_open(fp, open, &stp)))
2934 status = nfs4_check_deleg(cl, fp, open, &dp);
2938 status = nfserr_bad_stateid;
2939 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2941 status = nfserr_jukebox;
2942 fp = alloc_init_file(ino);
2948 * OPEN the file, or upgrade an existing OPEN.
2949 * If truncate fails, the OPEN fails.
2952 /* Stateid was found, this is an OPEN upgrade */
2953 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
2957 status = nfs4_new_open(rqstp, &stp, fp, current_fh, open);
2960 status = init_open_stateid(stp, fp, open);
2962 release_open_stateid(stp);
2965 status = nfsd4_truncate(rqstp, current_fh, open);
2967 release_open_stateid(stp);
2971 update_stateid(&stp->st_stid.sc_stateid);
2972 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
2974 if (nfsd4_has_session(&resp->cstate))
2975 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2978 * Attempt to hand out a delegation. No error return, because the
2979 * OPEN succeeds even if we fail.
2981 nfs4_open_delegation(current_fh, open, stp);
2985 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2986 STATEID_VAL(&stp->st_stid.sc_stateid));
2990 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2991 nfs4_set_claim_prev(open);
2993 * To finish the open response, we just need to set the rflags.
2995 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2996 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
2997 !nfsd4_has_session(&resp->cstate))
2998 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3004 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3007 struct nfs4_client *clp;
3011 dprintk("process_renew(%08x/%08x): starting\n",
3012 clid->cl_boot, clid->cl_id);
3013 status = nfserr_stale_clientid;
3014 if (STALE_CLIENTID(clid))
3016 clp = find_confirmed_client(clid);
3017 status = nfserr_expired;
3019 /* We assume the client took too long to RENEW. */
3020 dprintk("nfsd4_renew: clientid not found!\n");
3024 status = nfserr_cb_path_down;
3025 if (!list_empty(&clp->cl_delegations)
3026 && clp->cl_cb_state != NFSD4_CB_UP)
3030 nfs4_unlock_state();
3034 static struct lock_manager nfsd4_manager = {
3038 nfsd4_end_grace(void)
3040 dprintk("NFSD: end of grace period\n");
3041 nfsd4_recdir_purge_old();
3042 locks_end_grace(&nfsd4_manager);
3044 * Now that every NFSv4 client has had the chance to recover and
3045 * to see the (possibly new, possibly shorter) lease time, we
3046 * can safely set the next grace time to the current lease time:
3048 nfsd4_grace = nfsd4_lease;
3052 nfs4_laundromat(void)
3054 struct nfs4_client *clp;
3055 struct nfs4_openowner *oo;
3056 struct nfs4_delegation *dp;
3057 struct list_head *pos, *next, reaplist;
3058 time_t cutoff = get_seconds() - nfsd4_lease;
3059 time_t t, clientid_val = nfsd4_lease;
3060 time_t u, test_val = nfsd4_lease;
3064 dprintk("NFSD: laundromat service - starting\n");
3065 if (locks_in_grace())
3067 INIT_LIST_HEAD(&reaplist);
3068 spin_lock(&client_lock);
3069 list_for_each_safe(pos, next, &client_lru) {
3070 clp = list_entry(pos, struct nfs4_client, cl_lru);
3071 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3072 t = clp->cl_time - cutoff;
3073 if (clientid_val > t)
3077 if (atomic_read(&clp->cl_refcount)) {
3078 dprintk("NFSD: client in use (clientid %08x)\n",
3079 clp->cl_clientid.cl_id);
3082 unhash_client_locked(clp);
3083 list_add(&clp->cl_lru, &reaplist);
3085 spin_unlock(&client_lock);
3086 list_for_each_safe(pos, next, &reaplist) {
3087 clp = list_entry(pos, struct nfs4_client, cl_lru);
3088 dprintk("NFSD: purging unused client (clientid %08x)\n",
3089 clp->cl_clientid.cl_id);
3090 nfsd4_remove_clid_dir(clp);
3093 spin_lock(&recall_lock);
3094 list_for_each_safe(pos, next, &del_recall_lru) {
3095 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3096 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3097 u = dp->dl_time - cutoff;
3102 list_move(&dp->dl_recall_lru, &reaplist);
3104 spin_unlock(&recall_lock);
3105 list_for_each_safe(pos, next, &reaplist) {
3106 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3107 list_del_init(&dp->dl_recall_lru);
3108 unhash_delegation(dp);
3110 test_val = nfsd4_lease;
3111 list_for_each_safe(pos, next, &close_lru) {
3112 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3113 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3114 u = oo->oo_time - cutoff;
3119 release_openowner(oo);
3121 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3122 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3123 nfs4_unlock_state();
3124 return clientid_val;
3127 static struct workqueue_struct *laundry_wq;
3128 static void laundromat_main(struct work_struct *);
3129 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3132 laundromat_main(struct work_struct *not_used)
3136 t = nfs4_laundromat();
3137 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3138 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3141 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3143 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3144 return nfserr_bad_stateid;
3149 STALE_STATEID(stateid_t *stateid)
3151 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3153 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3154 STATEID_VAL(stateid));
3159 access_permit_read(unsigned long access_bmap)
3161 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3162 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3163 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3167 access_permit_write(unsigned long access_bmap)
3169 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3170 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3174 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3176 __be32 status = nfserr_openmode;
3178 /* For lock stateid's, we test the parent open, not the lock: */
3179 if (stp->st_openstp)
3180 stp = stp->st_openstp;
3181 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3183 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3190 static inline __be32
3191 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3193 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3195 else if (locks_in_grace()) {
3196 /* Answer in remaining cases depends on existence of
3197 * conflicting state; so we must wait out the grace period. */
3198 return nfserr_grace;
3199 } else if (flags & WR_STATE)
3200 return nfs4_share_conflict(current_fh,
3201 NFS4_SHARE_DENY_WRITE);
3202 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3203 return nfs4_share_conflict(current_fh,
3204 NFS4_SHARE_DENY_READ);
3208 * Allow READ/WRITE during grace period on recovered state only for files
3209 * that are not able to provide mandatory locking.
3212 grace_disallows_io(struct inode *inode)
3214 return locks_in_grace() && mandatory_lock(inode);
3217 /* Returns true iff a is later than b: */
3218 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3220 return (s32)a->si_generation - (s32)b->si_generation > 0;
3223 static int check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3226 * When sessions are used the stateid generation number is ignored
3229 if (has_session && in->si_generation == 0)
3232 if (in->si_generation == ref->si_generation)
3235 /* If the client sends us a stateid from the future, it's buggy: */
3236 if (stateid_generation_after(in, ref))
3237 return nfserr_bad_stateid;
3239 * However, we could see a stateid from the past, even from a
3240 * non-buggy client. For example, if the client sends a lock
3241 * while some IO is outstanding, the lock may bump si_generation
3242 * while the IO is still in flight. The client could avoid that
3243 * situation by waiting for responses on all the IO requests,
3244 * but better performance may result in retrying IO that
3245 * receives an old_stateid error if requests are rarely
3246 * reordered in flight:
3248 return nfserr_old_stateid;
3251 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3253 struct nfs4_stid *s;
3254 struct nfs4_ol_stateid *ols;
3257 if (STALE_STATEID(stateid))
3258 return nfserr_stale_stateid;
3260 s = find_stateid(cl, stateid);
3262 return nfserr_stale_stateid;
3263 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3266 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3268 ols = openlockstateid(s);
3269 if (ols->st_stateowner->so_is_open_owner
3270 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3271 return nfserr_bad_stateid;
3275 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3277 struct nfs4_client *cl;
3279 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3280 return nfserr_bad_stateid;
3281 if (STALE_STATEID(stateid))
3282 return nfserr_stale_stateid;
3283 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3285 return nfserr_expired;
3286 *s = find_stateid_by_type(cl, stateid, typemask);
3288 return nfserr_bad_stateid;
3294 * Checks for stateid operations
3297 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3298 stateid_t *stateid, int flags, struct file **filpp)
3300 struct nfs4_stid *s;
3301 struct nfs4_ol_stateid *stp = NULL;
3302 struct nfs4_delegation *dp = NULL;
3303 struct svc_fh *current_fh = &cstate->current_fh;
3304 struct inode *ino = current_fh->fh_dentry->d_inode;
3310 if (grace_disallows_io(ino))
3311 return nfserr_grace;
3313 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3314 return check_special_stateids(current_fh, stateid, flags);
3316 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3319 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3322 switch (s->sc_type) {
3323 case NFS4_DELEG_STID:
3324 dp = delegstateid(s);
3325 status = nfs4_check_delegmode(dp, flags);
3328 renew_client(dp->dl_stid.sc_client);
3330 *filpp = dp->dl_file->fi_deleg_file;
3334 case NFS4_OPEN_STID:
3335 case NFS4_LOCK_STID:
3336 stp = openlockstateid(s);
3337 status = nfs4_check_fh(current_fh, stp);
3340 if (stp->st_stateowner->so_is_open_owner
3341 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3343 status = nfs4_check_openmode(stp, flags);
3346 renew_client(stp->st_stateowner->so_client);
3348 if (flags & RD_STATE)
3349 *filpp = find_readable_file(stp->st_file);
3351 *filpp = find_writeable_file(stp->st_file);
3355 return nfserr_bad_stateid;
3363 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3365 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3366 return nfserr_locks_held;
3367 release_lock_stateid(stp);
3372 * Test if the stateid is valid
3375 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3376 struct nfsd4_test_stateid *test_stateid)
3378 /* real work is done during encoding */
3383 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3384 struct nfsd4_free_stateid *free_stateid)
3386 stateid_t *stateid = &free_stateid->fr_stateid;
3387 struct nfs4_stid *s;
3388 struct nfs4_client *cl = cstate->session->se_client;
3389 __be32 ret = nfserr_bad_stateid;
3392 s = find_stateid(cl, stateid);
3395 switch (s->sc_type) {
3396 case NFS4_DELEG_STID:
3397 ret = nfserr_locks_held;
3399 case NFS4_OPEN_STID:
3400 case NFS4_LOCK_STID:
3401 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3404 if (s->sc_type == NFS4_LOCK_STID)
3405 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3407 ret = nfserr_locks_held;
3410 ret = nfserr_bad_stateid;
3413 nfs4_unlock_state();
3420 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3421 RD_STATE : WR_STATE;
3424 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3426 struct svc_fh *current_fh = &cstate->current_fh;
3427 struct nfs4_stateowner *sop = stp->st_stateowner;
3430 status = nfsd4_check_seqid(cstate, sop, seqid);
3433 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3435 * "Closed" stateid's exist *only* to return
3436 * nfserr_replay_me from the previous step.
3438 return nfserr_bad_stateid;
3439 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3442 return nfs4_check_fh(current_fh, stp);
3446 * Checks for sequence id mutating operations.
3449 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3450 stateid_t *stateid, char typemask,
3451 struct nfs4_ol_stateid **stpp)
3454 struct nfs4_stid *s;
3456 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3457 seqid, STATEID_VAL(stateid));
3460 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3463 *stpp = openlockstateid(s);
3464 cstate->replay_owner = (*stpp)->st_stateowner;
3465 renew_client((*stpp)->st_stateowner->so_client);
3467 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3470 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3473 struct nfs4_openowner *oo;
3475 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3476 NFS4_OPEN_STID, stpp);
3479 oo = openowner((*stpp)->st_stateowner);
3480 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3481 return nfserr_bad_stateid;
3486 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3487 struct nfsd4_open_confirm *oc)
3490 struct nfs4_openowner *oo;
3491 struct nfs4_ol_stateid *stp;
3493 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3494 (int)cstate->current_fh.fh_dentry->d_name.len,
3495 cstate->current_fh.fh_dentry->d_name.name);
3497 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3503 status = nfs4_preprocess_seqid_op(cstate,
3504 oc->oc_seqid, &oc->oc_req_stateid,
3505 NFS4_OPEN_STID, &stp);
3508 oo = openowner(stp->st_stateowner);
3509 status = nfserr_bad_stateid;
3510 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3512 oo->oo_flags |= NFS4_OO_CONFIRMED;
3513 update_stateid(&stp->st_stid.sc_stateid);
3514 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3515 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3516 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3518 nfsd4_create_clid_dir(oo->oo_owner.so_client);
3521 if (!cstate->replay_owner)
3522 nfs4_unlock_state();
3526 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3528 if (!test_bit(access, &stp->st_access_bmap))
3530 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3531 __clear_bit(access, &stp->st_access_bmap);
3534 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3536 switch (to_access) {
3537 case NFS4_SHARE_ACCESS_READ:
3538 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3539 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3541 case NFS4_SHARE_ACCESS_WRITE:
3542 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3543 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3545 case NFS4_SHARE_ACCESS_BOTH:
3553 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3556 for (i = 0; i < 4; i++) {
3557 if ((i & deny) != i)
3558 __clear_bit(i, bmap);
3563 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3564 struct nfsd4_compound_state *cstate,
3565 struct nfsd4_open_downgrade *od)
3568 struct nfs4_ol_stateid *stp;
3570 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3571 (int)cstate->current_fh.fh_dentry->d_name.len,
3572 cstate->current_fh.fh_dentry->d_name.name);
3574 if (!access_valid(od->od_share_access, cstate->minorversion)
3575 || !deny_valid(od->od_share_deny))
3576 return nfserr_inval;
3577 /* We don't yet support WANT bits: */
3578 od->od_share_access &= NFS4_SHARE_ACCESS_MASK;
3581 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3582 &od->od_stateid, &stp);
3585 status = nfserr_inval;
3586 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3587 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3588 stp->st_access_bmap, od->od_share_access);
3591 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3592 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3593 stp->st_deny_bmap, od->od_share_deny);
3596 nfs4_stateid_downgrade(stp, od->od_share_access);
3598 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3600 update_stateid(&stp->st_stid.sc_stateid);
3601 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3604 if (!cstate->replay_owner)
3605 nfs4_unlock_state();
3609 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3611 struct nfs4_openowner *oo;
3612 struct nfs4_ol_stateid *s;
3614 if (!so->so_is_open_owner)
3617 s = oo->oo_last_closed_stid;
3620 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3621 /* Release the last_closed_stid on the next seqid bump: */
3622 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3625 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3626 release_last_closed_stateid(oo);
3629 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3631 unhash_open_stateid(s);
3632 s->st_stid.sc_type = NFS4_CLOSED_STID;
3636 * nfs4_unlock_state() called after encode
3639 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3640 struct nfsd4_close *close)
3643 struct nfs4_openowner *oo;
3644 struct nfs4_ol_stateid *stp;
3646 dprintk("NFSD: nfsd4_close on file %.*s\n",
3647 (int)cstate->current_fh.fh_dentry->d_name.len,
3648 cstate->current_fh.fh_dentry->d_name.name);
3651 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3653 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3657 oo = openowner(stp->st_stateowner);
3659 update_stateid(&stp->st_stid.sc_stateid);
3660 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3662 nfsd4_close_open_stateid(stp);
3663 oo->oo_last_closed_stid = stp;
3665 /* place unused nfs4_stateowners on so_close_lru list to be
3666 * released by the laundromat service after the lease period
3667 * to enable us to handle CLOSE replay
3669 if (list_empty(&oo->oo_owner.so_stateids))
3670 move_to_close_lru(oo);
3672 if (!cstate->replay_owner)
3673 nfs4_unlock_state();
3678 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3679 struct nfsd4_delegreturn *dr)
3681 struct nfs4_delegation *dp;
3682 stateid_t *stateid = &dr->dr_stateid;
3683 struct nfs4_stid *s;
3684 struct inode *inode;
3687 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3689 inode = cstate->current_fh.fh_dentry->d_inode;
3692 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3695 dp = delegstateid(s);
3696 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3699 renew_client(dp->dl_stid.sc_client);
3701 unhash_delegation(dp);
3703 nfs4_unlock_state();
3710 * Lock owner state (byte-range locks)
3712 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3713 #define LOCK_HASH_BITS 8
3714 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3715 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3718 end_offset(u64 start, u64 len)
3723 return end >= start ? end: NFS4_MAX_UINT64;
3726 /* last octet in a range */
3728 last_byte_offset(u64 start, u64 len)
3734 return end > start ? end - 1: NFS4_MAX_UINT64;
3737 static inline unsigned int
3738 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3739 struct xdr_netobj *ownername)
3741 return (file_hashval(inode) + cl_id
3742 + opaque_hashval(ownername->data, ownername->len))
3746 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3749 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3750 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3751 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3752 * locking, this prevents us from being completely protocol-compliant. The
3753 * real solution to this problem is to start using unsigned file offsets in
3754 * the VFS, but this is a very deep change!
3757 nfs4_transform_lock_offset(struct file_lock *lock)
3759 if (lock->fl_start < 0)
3760 lock->fl_start = OFFSET_MAX;
3761 if (lock->fl_end < 0)
3762 lock->fl_end = OFFSET_MAX;
3765 /* Hack!: For now, we're defining this just so we can use a pointer to it
3766 * as a unique cookie to identify our (NFSv4's) posix locks. */
3767 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3771 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3773 struct nfs4_lockowner *lo;
3775 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3776 lo = (struct nfs4_lockowner *) fl->fl_owner;
3777 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3778 lo->lo_owner.so_owner.len, GFP_KERNEL);
3779 if (!deny->ld_owner.data)
3780 /* We just don't care that much */
3782 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3783 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3786 deny->ld_owner.len = 0;
3787 deny->ld_owner.data = NULL;
3788 deny->ld_clientid.cl_boot = 0;
3789 deny->ld_clientid.cl_id = 0;
3791 deny->ld_start = fl->fl_start;
3792 deny->ld_length = NFS4_MAX_UINT64;
3793 if (fl->fl_end != NFS4_MAX_UINT64)
3794 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3795 deny->ld_type = NFS4_READ_LT;
3796 if (fl->fl_type != F_RDLCK)
3797 deny->ld_type = NFS4_WRITE_LT;
3800 static struct nfs4_lockowner *
3801 find_lockowner_str(struct inode *inode, clientid_t *clid,
3802 struct xdr_netobj *owner)
3804 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3805 struct nfs4_stateowner *op;
3807 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3808 if (same_owner_str(op, owner, clid))
3809 return lockowner(op);
3814 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3816 list_add(&lo->lo_owner.so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3817 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3821 * Alloc a lock owner structure.
3822 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3825 * strhashval = lock_ownerstr_hashval
3828 static struct nfs4_lockowner *
3829 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3830 struct nfs4_lockowner *lo;
3832 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3835 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3836 lo->lo_owner.so_is_open_owner = 0;
3837 /* It is the openowner seqid that will be incremented in encode in the
3838 * case of new lockowners; so increment the lock seqid manually: */
3839 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3840 hash_lockowner(lo, strhashval, clp, open_stp);
3844 static struct nfs4_ol_stateid *
3845 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3847 struct nfs4_ol_stateid *stp;
3848 struct nfs4_client *clp = lo->lo_owner.so_client;
3851 stp = nfs4_alloc_stateid();
3854 status = init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3856 free_generic_stateid(stp);
3859 list_add(&stp->st_perfile, &fp->fi_stateids);
3860 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3861 stp->st_stateowner = &lo->lo_owner;
3864 stp->st_access_bmap = 0;
3865 stp->st_deny_bmap = open_stp->st_deny_bmap;
3866 stp->st_openstp = open_stp;
3871 check_lock_length(u64 offset, u64 length)
3873 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3874 LOFF_OVERFLOW(offset, length)));
3877 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
3879 struct nfs4_file *fp = lock_stp->st_file;
3880 int oflag = nfs4_access_to_omode(access);
3882 if (test_bit(access, &lock_stp->st_access_bmap))
3884 nfs4_file_get_access(fp, oflag);
3885 __set_bit(access, &lock_stp->st_access_bmap);
3892 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3893 struct nfsd4_lock *lock)
3895 struct nfs4_openowner *open_sop = NULL;
3896 struct nfs4_lockowner *lock_sop = NULL;
3897 struct nfs4_ol_stateid *lock_stp;
3898 struct nfs4_file *fp;
3899 struct file *filp = NULL;
3900 struct file_lock file_lock;
3901 struct file_lock conflock;
3903 unsigned int strhashval;
3907 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3908 (long long) lock->lk_offset,
3909 (long long) lock->lk_length);
3911 if (check_lock_length(lock->lk_offset, lock->lk_length))
3912 return nfserr_inval;
3914 if ((status = fh_verify(rqstp, &cstate->current_fh,
3915 S_IFREG, NFSD_MAY_LOCK))) {
3916 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3922 if (lock->lk_is_new) {
3924 * Client indicates that this is a new lockowner.
3925 * Use open owner and open stateid to create lock owner and
3928 struct nfs4_ol_stateid *open_stp = NULL;
3930 status = nfserr_stale_clientid;
3931 if (!nfsd4_has_session(cstate) &&
3932 STALE_CLIENTID(&lock->lk_new_clientid))
3935 /* validate and update open stateid and open seqid */
3936 status = nfs4_preprocess_confirmed_seqid_op(cstate,
3937 lock->lk_new_open_seqid,
3938 &lock->lk_new_open_stateid,
3942 open_sop = openowner(open_stp->st_stateowner);
3943 status = nfserr_bad_stateid;
3944 if (!nfsd4_has_session(cstate) &&
3945 !same_clid(&open_sop->oo_owner.so_client->cl_clientid,
3946 &lock->v.new.clientid))
3948 /* create lockowner and lock stateid */
3949 fp = open_stp->st_file;
3950 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3951 open_sop->oo_owner.so_client->cl_clientid.cl_id,
3952 &lock->v.new.owner);
3953 /* XXX: Do we need to check for duplicate stateowners on
3954 * the same file, or should they just be allowed (and
3955 * create new stateids)? */
3956 status = nfserr_jukebox;
3957 lock_sop = alloc_init_lock_stateowner(strhashval,
3958 open_sop->oo_owner.so_client, open_stp, lock);
3959 if (lock_sop == NULL)
3961 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3962 if (lock_stp == NULL)
3965 /* lock (lock owner + lock stateid) already exists */
3966 status = nfs4_preprocess_seqid_op(cstate,
3967 lock->lk_old_lock_seqid,
3968 &lock->lk_old_lock_stateid,
3969 NFS4_LOCK_STID, &lock_stp);
3972 lock_sop = lockowner(lock_stp->st_stateowner);
3973 fp = lock_stp->st_file;
3975 /* lock_sop and lock_stp have been created or found */
3977 lkflg = setlkflg(lock->lk_type);
3978 status = nfs4_check_openmode(lock_stp, lkflg);
3982 status = nfserr_grace;
3983 if (locks_in_grace() && !lock->lk_reclaim)
3985 status = nfserr_no_grace;
3986 if (!locks_in_grace() && lock->lk_reclaim)
3989 locks_init_lock(&file_lock);
3990 switch (lock->lk_type) {
3993 filp = find_readable_file(lock_stp->st_file);
3995 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
3996 file_lock.fl_type = F_RDLCK;
3999 case NFS4_WRITEW_LT:
4000 filp = find_writeable_file(lock_stp->st_file);
4002 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4003 file_lock.fl_type = F_WRLCK;
4006 status = nfserr_inval;
4010 status = nfserr_openmode;
4013 file_lock.fl_owner = (fl_owner_t)lock_sop;
4014 file_lock.fl_pid = current->tgid;
4015 file_lock.fl_file = filp;
4016 file_lock.fl_flags = FL_POSIX;
4017 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4019 file_lock.fl_start = lock->lk_offset;
4020 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4021 nfs4_transform_lock_offset(&file_lock);
4024 * Try to lock the file in the VFS.
4025 * Note: locks.c uses the BKL to protect the inode's lock list.
4028 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4030 case 0: /* success! */
4031 update_stateid(&lock_stp->st_stid.sc_stateid);
4032 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4036 case (EAGAIN): /* conflock holds conflicting lock */
4037 status = nfserr_denied;
4038 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4039 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4042 status = nfserr_deadlock;
4045 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4046 status = nfserrno(err);
4050 if (status && lock->lk_is_new && lock_sop)
4051 release_lockowner(lock_sop);
4052 if (!cstate->replay_owner)
4053 nfs4_unlock_state();
4058 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4059 * so we do a temporary open here just to get an open file to pass to
4060 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4063 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4068 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4071 err = vfs_test_lock(file, lock);
4080 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4081 struct nfsd4_lockt *lockt)
4083 struct inode *inode;
4084 struct file_lock file_lock;
4085 struct nfs4_lockowner *lo;
4089 if (locks_in_grace())
4090 return nfserr_grace;
4092 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4093 return nfserr_inval;
4097 status = nfserr_stale_clientid;
4098 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4101 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4104 inode = cstate->current_fh.fh_dentry->d_inode;
4105 locks_init_lock(&file_lock);
4106 switch (lockt->lt_type) {
4109 file_lock.fl_type = F_RDLCK;
4112 case NFS4_WRITEW_LT:
4113 file_lock.fl_type = F_WRLCK;
4116 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4117 status = nfserr_inval;
4121 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4123 file_lock.fl_owner = (fl_owner_t)lo;
4124 file_lock.fl_pid = current->tgid;
4125 file_lock.fl_flags = FL_POSIX;
4127 file_lock.fl_start = lockt->lt_offset;
4128 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4130 nfs4_transform_lock_offset(&file_lock);
4133 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4135 status = nfserrno(error);
4138 if (file_lock.fl_type != F_UNLCK) {
4139 status = nfserr_denied;
4140 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4143 nfs4_unlock_state();
4148 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4149 struct nfsd4_locku *locku)
4151 struct nfs4_ol_stateid *stp;
4152 struct file *filp = NULL;
4153 struct file_lock file_lock;
4157 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4158 (long long) locku->lu_offset,
4159 (long long) locku->lu_length);
4161 if (check_lock_length(locku->lu_offset, locku->lu_length))
4162 return nfserr_inval;
4166 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4167 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4170 filp = find_any_file(stp->st_file);
4172 status = nfserr_lock_range;
4176 locks_init_lock(&file_lock);
4177 file_lock.fl_type = F_UNLCK;
4178 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4179 file_lock.fl_pid = current->tgid;
4180 file_lock.fl_file = filp;
4181 file_lock.fl_flags = FL_POSIX;
4182 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4183 file_lock.fl_start = locku->lu_offset;
4185 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4186 nfs4_transform_lock_offset(&file_lock);
4189 * Try to unlock the file in the VFS.
4191 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4193 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4197 * OK, unlock succeeded; the only thing left to do is update the stateid.
4199 update_stateid(&stp->st_stid.sc_stateid);
4200 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4203 if (!cstate->replay_owner)
4204 nfs4_unlock_state();
4208 status = nfserrno(err);
4214 * 1: locks held by lockowner
4215 * 0: no locks held by lockowner
4218 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4220 struct file_lock **flpp;
4221 struct inode *inode = filp->fi_inode;
4225 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4226 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4237 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4238 struct nfsd4_compound_state *cstate,
4239 struct nfsd4_release_lockowner *rlockowner)
4241 clientid_t *clid = &rlockowner->rl_clientid;
4242 struct nfs4_stateowner *sop;
4243 struct nfs4_lockowner *lo;
4244 struct nfs4_ol_stateid *stp;
4245 struct xdr_netobj *owner = &rlockowner->rl_owner;
4246 struct list_head matches;
4250 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4251 clid->cl_boot, clid->cl_id);
4253 /* XXX check for lease expiration */
4255 status = nfserr_stale_clientid;
4256 if (STALE_CLIENTID(clid))
4261 status = nfserr_locks_held;
4262 /* XXX: we're doing a linear search through all the lockowners.
4263 * Yipes! For now we'll just hope clients aren't really using
4264 * release_lockowner much, but eventually we have to fix these
4265 * data structures. */
4266 INIT_LIST_HEAD(&matches);
4267 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4268 list_for_each_entry(sop, &lock_ownerstr_hashtbl[i], so_strhash) {
4269 if (!same_owner_str(sop, owner, clid))
4271 list_for_each_entry(stp, &sop->so_stateids,
4273 lo = lockowner(sop);
4274 if (check_for_locks(stp->st_file, lo))
4276 list_add(&lo->lo_list, &matches);
4280 /* Clients probably won't expect us to return with some (but not all)
4281 * of the lockowner state released; so don't release any until all
4282 * have been checked. */
4284 while (!list_empty(&matches)) {
4285 lo = list_entry(matches.next, struct nfs4_lockowner,
4287 /* unhash_stateowner deletes so_perclient only
4288 * for openowners. */
4289 list_del(&lo->lo_list);
4290 release_lockowner(lo);
4293 nfs4_unlock_state();
4297 static inline struct nfs4_client_reclaim *
4300 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4304 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4306 unsigned int strhashval = clientstr_hashval(name);
4307 struct nfs4_client *clp;
4309 clp = find_confirmed_client_by_str(name, strhashval);
4314 * failure => all reset bets are off, nfserr_no_grace...
4317 nfs4_client_to_reclaim(const char *name)
4319 unsigned int strhashval;
4320 struct nfs4_client_reclaim *crp = NULL;
4322 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4323 crp = alloc_reclaim();
4326 strhashval = clientstr_hashval(name);
4327 INIT_LIST_HEAD(&crp->cr_strhash);
4328 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4329 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4330 reclaim_str_hashtbl_size++;
4335 nfs4_release_reclaim(void)
4337 struct nfs4_client_reclaim *crp = NULL;
4340 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4341 while (!list_empty(&reclaim_str_hashtbl[i])) {
4342 crp = list_entry(reclaim_str_hashtbl[i].next,
4343 struct nfs4_client_reclaim, cr_strhash);
4344 list_del(&crp->cr_strhash);
4346 reclaim_str_hashtbl_size--;
4349 BUG_ON(reclaim_str_hashtbl_size);
4353 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4354 static struct nfs4_client_reclaim *
4355 nfs4_find_reclaim_client(clientid_t *clid)
4357 unsigned int strhashval;
4358 struct nfs4_client *clp;
4359 struct nfs4_client_reclaim *crp = NULL;
4362 /* find clientid in conf_id_hashtbl */
4363 clp = find_confirmed_client(clid);
4367 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4368 clp->cl_name.len, clp->cl_name.data,
4371 /* find clp->cl_name in reclaim_str_hashtbl */
4372 strhashval = clientstr_hashval(clp->cl_recdir);
4373 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4374 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4382 * Called from OPEN. Look for clientid in reclaim list.
4385 nfs4_check_open_reclaim(clientid_t *clid)
4387 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4390 /* initialization to perform at module load time: */
4393 nfs4_state_init(void)
4397 status = nfsd4_init_slabs();
4400 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4401 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4402 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4403 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4404 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4405 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4407 for (i = 0; i < SESSION_HASH_SIZE; i++)
4408 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4409 for (i = 0; i < FILE_HASH_SIZE; i++) {
4410 INIT_LIST_HEAD(&file_hashtbl[i]);
4412 for (i = 0; i < OPEN_OWNER_HASH_SIZE; i++) {
4413 INIT_LIST_HEAD(&open_ownerstr_hashtbl[i]);
4415 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4416 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4418 memset(&onestateid, ~0, sizeof(stateid_t));
4419 INIT_LIST_HEAD(&close_lru);
4420 INIT_LIST_HEAD(&client_lru);
4421 INIT_LIST_HEAD(&del_recall_lru);
4422 reclaim_str_hashtbl_size = 0;
4427 nfsd4_load_reboot_recovery_data(void)
4432 nfsd4_init_recdir();
4433 status = nfsd4_recdir_load();
4434 nfs4_unlock_state();
4436 printk("NFSD: Failure reading reboot recovery data\n");
4440 * Since the lifetime of a delegation isn't limited to that of an open, a
4441 * client may quite reasonably hang on to a delegation as long as it has
4442 * the inode cached. This becomes an obvious problem the first time a
4443 * client's inode cache approaches the size of the server's total memory.
4445 * For now we avoid this problem by imposing a hard limit on the number
4446 * of delegations, which varies according to the server's memory size.
4449 set_max_delegations(void)
4452 * Allow at most 4 delegations per megabyte of RAM. Quick
4453 * estimates suggest that in the worst case (where every delegation
4454 * is for a different inode), a delegation could take about 1.5K,
4455 * giving a worst case usage of about 6% of memory.
4457 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4460 /* initialization to perform when the nfsd service is started: */
4463 __nfs4_state_start(void)
4467 boot_time = get_seconds();
4468 locks_start_grace(&nfsd4_manager);
4469 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4471 ret = set_callback_cred();
4474 laundry_wq = create_singlethread_workqueue("nfsd4");
4475 if (laundry_wq == NULL)
4477 ret = nfsd4_create_callback_queue();
4479 goto out_free_laundry;
4480 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4481 set_max_delegations();
4484 destroy_workqueue(laundry_wq);
4489 nfs4_state_start(void)
4491 nfsd4_load_reboot_recovery_data();
4492 return __nfs4_state_start();
4496 __nfs4_state_shutdown(void)
4499 struct nfs4_client *clp = NULL;
4500 struct nfs4_delegation *dp = NULL;
4501 struct list_head *pos, *next, reaplist;
4503 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4504 while (!list_empty(&conf_id_hashtbl[i])) {
4505 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4508 while (!list_empty(&unconf_str_hashtbl[i])) {
4509 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4513 INIT_LIST_HEAD(&reaplist);
4514 spin_lock(&recall_lock);
4515 list_for_each_safe(pos, next, &del_recall_lru) {
4516 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4517 list_move(&dp->dl_recall_lru, &reaplist);
4519 spin_unlock(&recall_lock);
4520 list_for_each_safe(pos, next, &reaplist) {
4521 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4522 list_del_init(&dp->dl_recall_lru);
4523 unhash_delegation(dp);
4526 nfsd4_shutdown_recdir();
4530 nfs4_state_shutdown(void)
4532 cancel_delayed_work_sync(&laundromat_work);
4533 destroy_workqueue(laundry_wq);
4534 locks_end_grace(&nfsd4_manager);
4536 nfs4_release_reclaim();
4537 __nfs4_state_shutdown();
4538 nfs4_unlock_state();
4539 nfsd4_destroy_callback_queue();