2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
46 #define NFSDDBG_FACILITY NFSDDBG_PROC
49 time_t nfsd4_lease = 90; /* default lease time */
50 time_t nfsd4_grace = 90;
51 static time_t boot_time;
52 static u32 current_ownerid = 1;
53 static u32 current_fileid = 1;
54 static u32 current_delegid = 1;
55 static stateid_t zerostateid; /* bits all 0 */
56 static stateid_t onestateid; /* bits all 1 */
57 static u64 current_sessionid = 1;
59 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
62 /* forward declarations */
63 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
64 static struct nfs4_stateid * search_for_stateid(stateid_t *stid);
65 static struct nfs4_delegation * search_for_delegation(stateid_t *stid);
66 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
67 static int check_for_locks(struct nfs4_file *filp, struct nfs4_stateowner *lowner);
71 /* Currently used for almost all code touching nfsv4 state: */
72 static DEFINE_MUTEX(client_mutex);
75 * Currently used for the del_recall_lru and file hash table. In an
76 * effort to decrease the scope of the client_mutex, this spinlock may
77 * eventually cover more:
79 static DEFINE_SPINLOCK(recall_lock);
81 static struct kmem_cache *stateowner_slab = NULL;
82 static struct kmem_cache *file_slab = NULL;
83 static struct kmem_cache *stateid_slab = NULL;
84 static struct kmem_cache *deleg_slab = NULL;
89 mutex_lock(&client_mutex);
93 nfs4_unlock_state(void)
95 mutex_unlock(&client_mutex);
99 opaque_hashval(const void *ptr, int nbytes)
101 unsigned char *cptr = (unsigned char *) ptr;
111 static struct list_head del_recall_lru;
114 put_nfs4_file(struct nfs4_file *fi)
116 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
117 list_del(&fi->fi_hash);
118 spin_unlock(&recall_lock);
120 kmem_cache_free(file_slab, fi);
125 get_nfs4_file(struct nfs4_file *fi)
127 atomic_inc(&fi->fi_ref);
130 static int num_delegations;
131 unsigned int max_delegations;
134 * Open owner state (share locks)
137 /* hash tables for nfs4_stateowner */
138 #define OWNER_HASH_BITS 8
139 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
140 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
142 static unsigned int ownerid_hashval(const u32 id)
144 return id & OWNER_HASH_MASK;
147 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
151 ret = opaque_hashval(ownername->data, ownername->len);
153 return ret & OWNER_HASH_MASK;
156 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
157 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
159 /* hash table for nfs4_file */
160 #define FILE_HASH_BITS 8
161 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
163 /* hash table for (open)nfs4_stateid */
164 #define STATEID_HASH_BITS 10
165 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
166 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
168 static unsigned int file_hashval(struct inode *ino)
170 /* XXX: why are we hashing on inode pointer, anyway? */
171 return hash_ptr(ino, FILE_HASH_BITS);
174 static unsigned int stateid_hashval(u32 owner_id, u32 file_id)
176 return (owner_id + file_id) & STATEID_HASH_MASK;
179 static struct list_head file_hashtbl[FILE_HASH_SIZE];
180 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
182 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
184 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
185 atomic_inc(&fp->fi_access[oflag]);
188 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
190 if (oflag == O_RDWR) {
191 __nfs4_file_get_access(fp, O_RDONLY);
192 __nfs4_file_get_access(fp, O_WRONLY);
194 __nfs4_file_get_access(fp, oflag);
197 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
199 if (fp->fi_fds[oflag]) {
200 fput(fp->fi_fds[oflag]);
201 fp->fi_fds[oflag] = NULL;
205 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
207 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
208 nfs4_file_put_fd(fp, O_RDWR);
209 nfs4_file_put_fd(fp, oflag);
213 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
215 if (oflag == O_RDWR) {
216 __nfs4_file_put_access(fp, O_RDONLY);
217 __nfs4_file_put_access(fp, O_WRONLY);
219 __nfs4_file_put_access(fp, oflag);
222 static struct nfs4_delegation *
223 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
225 struct nfs4_delegation *dp;
226 struct nfs4_file *fp = stp->st_file;
228 dprintk("NFSD alloc_init_deleg\n");
230 * Major work on the lease subsystem (for example, to support
231 * calbacks on stat) will be required before we can support
232 * write delegations properly.
234 if (type != NFS4_OPEN_DELEGATE_READ)
236 if (fp->fi_had_conflict)
238 if (num_delegations > max_delegations)
240 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
244 INIT_LIST_HEAD(&dp->dl_perfile);
245 INIT_LIST_HEAD(&dp->dl_perclnt);
246 INIT_LIST_HEAD(&dp->dl_recall_lru);
251 dp->dl_stateid.si_boot = boot_time;
252 dp->dl_stateid.si_stateownerid = current_delegid++;
253 dp->dl_stateid.si_fileid = 0;
254 dp->dl_stateid.si_generation = 0;
255 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
257 atomic_set(&dp->dl_count, 1);
258 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
263 nfs4_put_delegation(struct nfs4_delegation *dp)
265 if (atomic_dec_and_test(&dp->dl_count)) {
266 dprintk("NFSD: freeing dp %p\n",dp);
267 put_nfs4_file(dp->dl_file);
268 kmem_cache_free(deleg_slab, dp);
273 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
275 if (atomic_dec_and_test(&fp->fi_delegees)) {
276 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
278 fput(fp->fi_deleg_file);
279 fp->fi_deleg_file = NULL;
283 /* Called under the state lock. */
285 unhash_delegation(struct nfs4_delegation *dp)
287 list_del_init(&dp->dl_perclnt);
288 spin_lock(&recall_lock);
289 list_del_init(&dp->dl_perfile);
290 list_del_init(&dp->dl_recall_lru);
291 spin_unlock(&recall_lock);
292 nfs4_put_deleg_lease(dp->dl_file);
293 nfs4_put_delegation(dp);
300 /* client_lock protects the client lru list and session hash table */
301 static DEFINE_SPINLOCK(client_lock);
303 /* Hash tables for nfs4_clientid state */
304 #define CLIENT_HASH_BITS 4
305 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
306 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
308 static unsigned int clientid_hashval(u32 id)
310 return id & CLIENT_HASH_MASK;
313 static unsigned int clientstr_hashval(const char *name)
315 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
319 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
320 * used in reboot/reset lease grace period processing
322 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
323 * setclientid_confirmed info.
325 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
328 * client_lru holds client queue ordered by nfs4_client.cl_time
331 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
332 * for last close replay.
334 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
335 static int reclaim_str_hashtbl_size = 0;
336 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
337 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
338 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
339 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
340 static struct list_head client_lru;
341 static struct list_head close_lru;
344 * We store the NONE, READ, WRITE, and BOTH bits separately in the
345 * st_{access,deny}_bmap field of the stateid, in order to track not
346 * only what share bits are currently in force, but also what
347 * combinations of share bits previous opens have used. This allows us
348 * to enforce the recommendation of rfc 3530 14.2.19 that the server
349 * return an error if the client attempt to downgrade to a combination
350 * of share bits not explicable by closing some of its previous opens.
352 * XXX: This enforcement is actually incomplete, since we don't keep
353 * track of access/deny bit combinations; so, e.g., we allow:
355 * OPEN allow read, deny write
356 * OPEN allow both, deny none
357 * DOWNGRADE allow read, deny none
359 * which we should reject.
362 set_access(unsigned int *access, unsigned long bmap) {
366 for (i = 1; i < 4; i++) {
367 if (test_bit(i, &bmap))
373 set_deny(unsigned int *deny, unsigned long bmap) {
377 for (i = 0; i < 4; i++) {
378 if (test_bit(i, &bmap))
384 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
385 unsigned int access, deny;
387 set_access(&access, stp->st_access_bmap);
388 set_deny(&deny, stp->st_deny_bmap);
389 if ((access & open->op_share_deny) || (deny & open->op_share_access))
394 static int nfs4_access_to_omode(u32 access)
396 switch (access & NFS4_SHARE_ACCESS_BOTH) {
397 case NFS4_SHARE_ACCESS_READ:
399 case NFS4_SHARE_ACCESS_WRITE:
401 case NFS4_SHARE_ACCESS_BOTH:
407 static void unhash_generic_stateid(struct nfs4_stateid *stp)
409 list_del(&stp->st_hash);
410 list_del(&stp->st_perfile);
411 list_del(&stp->st_perstateowner);
414 static void free_generic_stateid(struct nfs4_stateid *stp)
418 if (stp->st_access_bmap) {
419 for (i = 1; i < 4; i++) {
420 if (test_bit(i, &stp->st_access_bmap))
421 nfs4_file_put_access(stp->st_file,
422 nfs4_access_to_omode(i));
425 put_nfs4_file(stp->st_file);
426 kmem_cache_free(stateid_slab, stp);
429 static void release_lock_stateid(struct nfs4_stateid *stp)
433 unhash_generic_stateid(stp);
434 file = find_any_file(stp->st_file);
436 locks_remove_posix(file, (fl_owner_t)stp->st_stateowner);
437 free_generic_stateid(stp);
440 static void unhash_lockowner(struct nfs4_stateowner *sop)
442 struct nfs4_stateid *stp;
444 list_del(&sop->so_idhash);
445 list_del(&sop->so_strhash);
446 list_del(&sop->so_perstateid);
447 while (!list_empty(&sop->so_stateids)) {
448 stp = list_first_entry(&sop->so_stateids,
449 struct nfs4_stateid, st_perstateowner);
450 release_lock_stateid(stp);
454 static void release_lockowner(struct nfs4_stateowner *sop)
456 unhash_lockowner(sop);
457 nfs4_put_stateowner(sop);
461 release_stateid_lockowners(struct nfs4_stateid *open_stp)
463 struct nfs4_stateowner *lock_sop;
465 while (!list_empty(&open_stp->st_lockowners)) {
466 lock_sop = list_entry(open_stp->st_lockowners.next,
467 struct nfs4_stateowner, so_perstateid);
468 /* list_del(&open_stp->st_lockowners); */
469 BUG_ON(lock_sop->so_is_open_owner);
470 release_lockowner(lock_sop);
474 static void release_open_stateid(struct nfs4_stateid *stp)
476 unhash_generic_stateid(stp);
477 release_stateid_lockowners(stp);
478 free_generic_stateid(stp);
481 static void unhash_openowner(struct nfs4_stateowner *sop)
483 struct nfs4_stateid *stp;
485 list_del(&sop->so_idhash);
486 list_del(&sop->so_strhash);
487 list_del(&sop->so_perclient);
488 list_del(&sop->so_perstateid); /* XXX: necessary? */
489 while (!list_empty(&sop->so_stateids)) {
490 stp = list_first_entry(&sop->so_stateids,
491 struct nfs4_stateid, st_perstateowner);
492 release_open_stateid(stp);
496 static void release_openowner(struct nfs4_stateowner *sop)
498 unhash_openowner(sop);
499 list_del(&sop->so_close_lru);
500 nfs4_put_stateowner(sop);
503 #define SESSION_HASH_SIZE 512
504 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
507 hash_sessionid(struct nfs4_sessionid *sessionid)
509 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
511 return sid->sequence % SESSION_HASH_SIZE;
515 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
517 u32 *ptr = (u32 *)(&sessionid->data[0]);
518 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
522 gen_sessionid(struct nfsd4_session *ses)
524 struct nfs4_client *clp = ses->se_client;
525 struct nfsd4_sessionid *sid;
527 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
528 sid->clientid = clp->cl_clientid;
529 sid->sequence = current_sessionid++;
534 * The protocol defines ca_maxresponssize_cached to include the size of
535 * the rpc header, but all we need to cache is the data starting after
536 * the end of the initial SEQUENCE operation--the rest we regenerate
537 * each time. Therefore we can advertise a ca_maxresponssize_cached
538 * value that is the number of bytes in our cache plus a few additional
539 * bytes. In order to stay on the safe side, and not promise more than
540 * we can cache, those additional bytes must be the minimum possible: 24
541 * bytes of rpc header (xid through accept state, with AUTH_NULL
542 * verifier), 12 for the compound header (with zero-length tag), and 44
543 * for the SEQUENCE op response:
545 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
548 free_session_slots(struct nfsd4_session *ses)
552 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
553 kfree(ses->se_slots[i]);
557 * We don't actually need to cache the rpc and session headers, so we
558 * can allocate a little less for each slot:
560 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
562 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
565 static int nfsd4_sanitize_slot_size(u32 size)
567 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
568 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
574 * XXX: If we run out of reserved DRC memory we could (up to a point)
575 * re-negotiate active sessions and reduce their slot usage to make
576 * rooom for new connections. For now we just fail the create session.
578 static int nfsd4_get_drc_mem(int slotsize, u32 num)
582 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
584 spin_lock(&nfsd_drc_lock);
585 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
586 nfsd_drc_max_mem - nfsd_drc_mem_used);
587 num = min_t(int, num, avail / slotsize);
588 nfsd_drc_mem_used += num * slotsize;
589 spin_unlock(&nfsd_drc_lock);
594 static void nfsd4_put_drc_mem(int slotsize, int num)
596 spin_lock(&nfsd_drc_lock);
597 nfsd_drc_mem_used -= slotsize * num;
598 spin_unlock(&nfsd_drc_lock);
601 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
603 struct nfsd4_session *new;
606 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
607 + sizeof(struct nfsd4_session) > PAGE_SIZE);
608 mem = numslots * sizeof(struct nfsd4_slot *);
610 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
613 /* allocate each struct nfsd4_slot and data cache in one piece */
614 for (i = 0; i < numslots; i++) {
615 mem = sizeof(struct nfsd4_slot) + slotsize;
616 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
617 if (!new->se_slots[i])
623 kfree(new->se_slots[i]);
628 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
630 u32 maxrpc = nfsd_serv->sv_max_mesg;
632 new->maxreqs = numslots;
633 new->maxresp_cached = min_t(u32, req->maxresp_cached,
634 slotsize + NFSD_MIN_HDR_SEQ_SZ);
635 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
636 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
637 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
640 static void free_conn(struct nfsd4_conn *c)
642 svc_xprt_put(c->cn_xprt);
646 static void nfsd4_conn_lost(struct svc_xpt_user *u)
648 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
649 struct nfs4_client *clp = c->cn_session->se_client;
651 spin_lock(&clp->cl_lock);
652 if (!list_empty(&c->cn_persession)) {
653 list_del(&c->cn_persession);
656 spin_unlock(&clp->cl_lock);
657 nfsd4_probe_callback(clp);
660 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
662 struct nfsd4_conn *conn;
664 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
667 svc_xprt_get(rqstp->rq_xprt);
668 conn->cn_xprt = rqstp->rq_xprt;
669 conn->cn_flags = flags;
670 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
674 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
676 conn->cn_session = ses;
677 list_add(&conn->cn_persession, &ses->se_conns);
680 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
682 struct nfs4_client *clp = ses->se_client;
684 spin_lock(&clp->cl_lock);
685 __nfsd4_hash_conn(conn, ses);
686 spin_unlock(&clp->cl_lock);
689 static int nfsd4_register_conn(struct nfsd4_conn *conn)
691 conn->cn_xpt_user.callback = nfsd4_conn_lost;
692 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
695 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
697 struct nfsd4_conn *conn;
700 conn = alloc_conn(rqstp, dir);
702 return nfserr_jukebox;
703 nfsd4_hash_conn(conn, ses);
704 ret = nfsd4_register_conn(conn);
706 /* oops; xprt is already down: */
707 nfsd4_conn_lost(&conn->cn_xpt_user);
711 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
713 u32 dir = NFS4_CDFC4_FORE;
715 if (ses->se_flags & SESSION4_BACK_CHAN)
716 dir |= NFS4_CDFC4_BACK;
718 return nfsd4_new_conn(rqstp, ses, dir);
721 /* must be called under client_lock */
722 static void nfsd4_del_conns(struct nfsd4_session *s)
724 struct nfs4_client *clp = s->se_client;
725 struct nfsd4_conn *c;
727 spin_lock(&clp->cl_lock);
728 while (!list_empty(&s->se_conns)) {
729 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
730 list_del_init(&c->cn_persession);
731 spin_unlock(&clp->cl_lock);
733 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
736 spin_lock(&clp->cl_lock);
738 spin_unlock(&clp->cl_lock);
741 void free_session(struct kref *kref)
743 struct nfsd4_session *ses;
746 ses = container_of(kref, struct nfsd4_session, se_ref);
747 nfsd4_del_conns(ses);
748 spin_lock(&nfsd_drc_lock);
749 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
750 nfsd_drc_mem_used -= mem;
751 spin_unlock(&nfsd_drc_lock);
752 free_session_slots(ses);
756 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
758 struct nfsd4_session *new;
759 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
760 int numslots, slotsize;
765 * Note decreasing slot size below client's request may
766 * make it difficult for client to function correctly, whereas
767 * decreasing the number of slots will (just?) affect
768 * performance. When short on memory we therefore prefer to
769 * decrease number of slots instead of their size.
771 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
772 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
776 new = alloc_session(slotsize, numslots);
778 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
781 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
783 new->se_client = clp;
786 INIT_LIST_HEAD(&new->se_conns);
788 new->se_cb_seq_nr = 1;
789 new->se_flags = cses->flags;
790 new->se_cb_prog = cses->callback_prog;
791 kref_init(&new->se_ref);
792 idx = hash_sessionid(&new->se_sessionid);
793 spin_lock(&client_lock);
794 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
795 spin_lock(&clp->cl_lock);
796 list_add(&new->se_perclnt, &clp->cl_sessions);
797 spin_unlock(&clp->cl_lock);
798 spin_unlock(&client_lock);
800 status = nfsd4_new_conn_from_crses(rqstp, new);
801 /* whoops: benny points out, status is ignored! (err, or bogus) */
803 free_session(&new->se_ref);
806 if (cses->flags & SESSION4_BACK_CHAN) {
807 struct sockaddr *sa = svc_addr(rqstp);
809 * This is a little silly; with sessions there's no real
810 * use for the callback address. Use the peer address
811 * as a reasonable default for now, but consider fixing
812 * the rpc client not to require an address in the
815 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
816 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
818 nfsd4_probe_callback(clp);
822 /* caller must hold client_lock */
823 static struct nfsd4_session *
824 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
826 struct nfsd4_session *elem;
829 dump_sessionid(__func__, sessionid);
830 idx = hash_sessionid(sessionid);
831 /* Search in the appropriate list */
832 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
833 if (!memcmp(elem->se_sessionid.data, sessionid->data,
834 NFS4_MAX_SESSIONID_LEN)) {
839 dprintk("%s: session not found\n", __func__);
843 /* caller must hold client_lock */
845 unhash_session(struct nfsd4_session *ses)
847 list_del(&ses->se_hash);
848 spin_lock(&ses->se_client->cl_lock);
849 list_del(&ses->se_perclnt);
850 spin_unlock(&ses->se_client->cl_lock);
853 /* must be called under the client_lock */
855 renew_client_locked(struct nfs4_client *clp)
857 if (is_client_expired(clp)) {
858 dprintk("%s: client (clientid %08x/%08x) already expired\n",
860 clp->cl_clientid.cl_boot,
861 clp->cl_clientid.cl_id);
866 * Move client to the end to the LRU list.
868 dprintk("renewing client (clientid %08x/%08x)\n",
869 clp->cl_clientid.cl_boot,
870 clp->cl_clientid.cl_id);
871 list_move_tail(&clp->cl_lru, &client_lru);
872 clp->cl_time = get_seconds();
876 renew_client(struct nfs4_client *clp)
878 spin_lock(&client_lock);
879 renew_client_locked(clp);
880 spin_unlock(&client_lock);
883 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
885 STALE_CLIENTID(clientid_t *clid)
887 if (clid->cl_boot == boot_time)
889 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
890 clid->cl_boot, clid->cl_id, boot_time);
895 * XXX Should we use a slab cache ?
896 * This type of memory management is somewhat inefficient, but we use it
897 * anyway since SETCLIENTID is not a common operation.
899 static struct nfs4_client *alloc_client(struct xdr_netobj name)
901 struct nfs4_client *clp;
903 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
906 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
907 if (clp->cl_name.data == NULL) {
911 memcpy(clp->cl_name.data, name.data, name.len);
912 clp->cl_name.len = name.len;
917 free_client(struct nfs4_client *clp)
919 while (!list_empty(&clp->cl_sessions)) {
920 struct nfsd4_session *ses;
921 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
923 list_del(&ses->se_perclnt);
924 nfsd4_put_session(ses);
926 if (clp->cl_cred.cr_group_info)
927 put_group_info(clp->cl_cred.cr_group_info);
928 kfree(clp->cl_principal);
929 kfree(clp->cl_name.data);
934 release_session_client(struct nfsd4_session *session)
936 struct nfs4_client *clp = session->se_client;
938 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
940 if (is_client_expired(clp)) {
942 session->se_client = NULL;
944 renew_client_locked(clp);
945 spin_unlock(&client_lock);
948 /* must be called under the client_lock */
950 unhash_client_locked(struct nfs4_client *clp)
952 struct nfsd4_session *ses;
954 mark_client_expired(clp);
955 list_del(&clp->cl_lru);
956 spin_lock(&clp->cl_lock);
957 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
958 list_del_init(&ses->se_hash);
959 spin_unlock(&clp->cl_lock);
963 expire_client(struct nfs4_client *clp)
965 struct nfs4_stateowner *sop;
966 struct nfs4_delegation *dp;
967 struct list_head reaplist;
969 INIT_LIST_HEAD(&reaplist);
970 spin_lock(&recall_lock);
971 while (!list_empty(&clp->cl_delegations)) {
972 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
973 list_del_init(&dp->dl_perclnt);
974 list_move(&dp->dl_recall_lru, &reaplist);
976 spin_unlock(&recall_lock);
977 while (!list_empty(&reaplist)) {
978 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
979 list_del_init(&dp->dl_recall_lru);
980 unhash_delegation(dp);
982 while (!list_empty(&clp->cl_openowners)) {
983 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
984 release_openowner(sop);
986 nfsd4_shutdown_callback(clp);
987 if (clp->cl_cb_conn.cb_xprt)
988 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
989 list_del(&clp->cl_idhash);
990 list_del(&clp->cl_strhash);
991 spin_lock(&client_lock);
992 unhash_client_locked(clp);
993 if (atomic_read(&clp->cl_refcount) == 0)
995 spin_unlock(&client_lock);
998 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1000 memcpy(target->cl_verifier.data, source->data,
1001 sizeof(target->cl_verifier.data));
1004 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1006 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1007 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1010 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1012 target->cr_uid = source->cr_uid;
1013 target->cr_gid = source->cr_gid;
1014 target->cr_group_info = source->cr_group_info;
1015 get_group_info(target->cr_group_info);
1018 static int same_name(const char *n1, const char *n2)
1020 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1024 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1026 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1030 same_clid(clientid_t *cl1, clientid_t *cl2)
1032 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1035 /* XXX what about NGROUP */
1037 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1039 return cr1->cr_uid == cr2->cr_uid;
1042 static void gen_clid(struct nfs4_client *clp)
1044 static u32 current_clientid = 1;
1046 clp->cl_clientid.cl_boot = boot_time;
1047 clp->cl_clientid.cl_id = current_clientid++;
1050 static void gen_confirm(struct nfs4_client *clp)
1055 p = (u32 *)clp->cl_confirm.data;
1056 *p++ = get_seconds();
1060 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1061 struct svc_rqst *rqstp, nfs4_verifier *verf)
1063 struct nfs4_client *clp;
1064 struct sockaddr *sa = svc_addr(rqstp);
1067 clp = alloc_client(name);
1071 INIT_LIST_HEAD(&clp->cl_sessions);
1073 princ = svc_gss_principal(rqstp);
1075 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1076 if (clp->cl_principal == NULL) {
1082 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1083 atomic_set(&clp->cl_refcount, 0);
1084 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1085 INIT_LIST_HEAD(&clp->cl_idhash);
1086 INIT_LIST_HEAD(&clp->cl_strhash);
1087 INIT_LIST_HEAD(&clp->cl_openowners);
1088 INIT_LIST_HEAD(&clp->cl_delegations);
1089 INIT_LIST_HEAD(&clp->cl_lru);
1090 INIT_LIST_HEAD(&clp->cl_callbacks);
1091 spin_lock_init(&clp->cl_lock);
1092 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1093 clp->cl_time = get_seconds();
1094 clear_bit(0, &clp->cl_cb_slot_busy);
1095 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1096 copy_verf(clp, verf);
1097 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1098 clp->cl_flavor = rqstp->rq_flavor;
1099 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1101 clp->cl_cb_session = NULL;
1105 static int check_name(struct xdr_netobj name)
1109 if (name.len > NFS4_OPAQUE_LIMIT) {
1110 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1117 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1119 unsigned int idhashval;
1121 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1122 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1123 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1128 move_to_confirmed(struct nfs4_client *clp)
1130 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1131 unsigned int strhashval;
1133 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1134 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1135 strhashval = clientstr_hashval(clp->cl_recdir);
1136 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1140 static struct nfs4_client *
1141 find_confirmed_client(clientid_t *clid)
1143 struct nfs4_client *clp;
1144 unsigned int idhashval = clientid_hashval(clid->cl_id);
1146 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1147 if (same_clid(&clp->cl_clientid, clid))
1153 static struct nfs4_client *
1154 find_unconfirmed_client(clientid_t *clid)
1156 struct nfs4_client *clp;
1157 unsigned int idhashval = clientid_hashval(clid->cl_id);
1159 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1160 if (same_clid(&clp->cl_clientid, clid))
1166 static bool clp_used_exchangeid(struct nfs4_client *clp)
1168 return clp->cl_exchange_flags != 0;
1171 static struct nfs4_client *
1172 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1174 struct nfs4_client *clp;
1176 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1177 if (same_name(clp->cl_recdir, dname))
1183 static struct nfs4_client *
1184 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1186 struct nfs4_client *clp;
1188 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1189 if (same_name(clp->cl_recdir, dname))
1195 static void rpc_svcaddr2sockaddr(struct sockaddr *sa, unsigned short family, union svc_addr_u *svcaddr)
1199 ((struct sockaddr_in *)sa)->sin_family = AF_INET;
1200 ((struct sockaddr_in *)sa)->sin_addr = svcaddr->addr;
1203 ((struct sockaddr_in6 *)sa)->sin6_family = AF_INET6;
1204 ((struct sockaddr_in6 *)sa)->sin6_addr = svcaddr->addr6;
1210 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1212 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1213 struct sockaddr *sa = svc_addr(rqstp);
1214 u32 scopeid = rpc_get_scope_id(sa);
1215 unsigned short expected_family;
1217 /* Currently, we only support tcp and tcp6 for the callback channel */
1218 if (se->se_callback_netid_len == 3 &&
1219 !memcmp(se->se_callback_netid_val, "tcp", 3))
1220 expected_family = AF_INET;
1221 else if (se->se_callback_netid_len == 4 &&
1222 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1223 expected_family = AF_INET6;
1227 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1228 se->se_callback_addr_len,
1229 (struct sockaddr *)&conn->cb_addr,
1230 sizeof(conn->cb_addr));
1232 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1235 if (conn->cb_addr.ss_family == AF_INET6)
1236 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1238 conn->cb_prog = se->se_callback_prog;
1239 conn->cb_ident = se->se_callback_ident;
1240 rpc_svcaddr2sockaddr((struct sockaddr *)&conn->cb_saddr, expected_family, &rqstp->rq_daddr);
1243 conn->cb_addr.ss_family = AF_UNSPEC;
1244 conn->cb_addrlen = 0;
1245 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1246 "will not receive delegations\n",
1247 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1253 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1256 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1258 struct nfsd4_slot *slot = resp->cstate.slot;
1261 dprintk("--> %s slot %p\n", __func__, slot);
1263 slot->sl_opcnt = resp->opcnt;
1264 slot->sl_status = resp->cstate.status;
1266 if (nfsd4_not_cached(resp)) {
1267 slot->sl_datalen = 0;
1270 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1271 base = (char *)resp->cstate.datap -
1272 (char *)resp->xbuf->head[0].iov_base;
1273 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1275 WARN("%s: sessions DRC could not cache compound\n", __func__);
1280 * Encode the replay sequence operation from the slot values.
1281 * If cachethis is FALSE encode the uncached rep error on the next
1282 * operation which sets resp->p and increments resp->opcnt for
1283 * nfs4svc_encode_compoundres.
1287 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1288 struct nfsd4_compoundres *resp)
1290 struct nfsd4_op *op;
1291 struct nfsd4_slot *slot = resp->cstate.slot;
1293 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1294 resp->opcnt, resp->cstate.slot->sl_cachethis);
1296 /* Encode the replayed sequence operation */
1297 op = &args->ops[resp->opcnt - 1];
1298 nfsd4_encode_operation(resp, op);
1300 /* Return nfserr_retry_uncached_rep in next operation. */
1301 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1302 op = &args->ops[resp->opcnt++];
1303 op->status = nfserr_retry_uncached_rep;
1304 nfsd4_encode_operation(resp, op);
1310 * The sequence operation is not cached because we can use the slot and
1314 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1315 struct nfsd4_sequence *seq)
1317 struct nfsd4_slot *slot = resp->cstate.slot;
1320 dprintk("--> %s slot %p\n", __func__, slot);
1322 /* Either returns 0 or nfserr_retry_uncached */
1323 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1324 if (status == nfserr_retry_uncached_rep)
1327 /* The sequence operation has been encoded, cstate->datap set. */
1328 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1330 resp->opcnt = slot->sl_opcnt;
1331 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1332 status = slot->sl_status;
1338 * Set the exchange_id flags returned by the server.
1341 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1343 /* pNFS is not supported */
1344 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1346 /* Referrals are supported, Migration is not. */
1347 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1349 /* set the wire flags to return to client. */
1350 clid->flags = new->cl_exchange_flags;
1354 nfsd4_exchange_id(struct svc_rqst *rqstp,
1355 struct nfsd4_compound_state *cstate,
1356 struct nfsd4_exchange_id *exid)
1358 struct nfs4_client *unconf, *conf, *new;
1360 unsigned int strhashval;
1361 char dname[HEXDIR_LEN];
1362 char addr_str[INET6_ADDRSTRLEN];
1363 nfs4_verifier verf = exid->verifier;
1364 struct sockaddr *sa = svc_addr(rqstp);
1366 rpc_ntop(sa, addr_str, sizeof(addr_str));
1367 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1368 "ip_addr=%s flags %x, spa_how %d\n",
1369 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1370 addr_str, exid->flags, exid->spa_how);
1372 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1373 return nfserr_inval;
1375 /* Currently only support SP4_NONE */
1376 switch (exid->spa_how) {
1380 return nfserr_serverfault;
1382 BUG(); /* checked by xdr code */
1384 return nfserr_serverfault; /* no excuse :-/ */
1387 status = nfs4_make_rec_clidname(dname, &exid->clname);
1392 strhashval = clientstr_hashval(dname);
1397 conf = find_confirmed_client_by_str(dname, strhashval);
1399 if (!clp_used_exchangeid(conf)) {
1400 status = nfserr_clid_inuse; /* XXX: ? */
1403 if (!same_verf(&verf, &conf->cl_verifier)) {
1404 /* 18.35.4 case 8 */
1405 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1406 status = nfserr_not_same;
1409 /* Client reboot: destroy old state */
1410 expire_client(conf);
1413 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1414 /* 18.35.4 case 9 */
1415 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1416 status = nfserr_perm;
1419 expire_client(conf);
1423 * Set bit when the owner id and verifier map to an already
1424 * confirmed client id (18.35.3).
1426 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1429 * Falling into 18.35.4 case 2, possible router replay.
1430 * Leave confirmed record intact and return same result.
1432 copy_verf(conf, &verf);
1437 /* 18.35.4 case 7 */
1438 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1439 status = nfserr_noent;
1443 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1446 * Possible retry or client restart. Per 18.35.4 case 4,
1447 * a new unconfirmed record should be generated regardless
1448 * of whether any properties have changed.
1450 expire_client(unconf);
1455 new = create_client(exid->clname, dname, rqstp, &verf);
1457 status = nfserr_jukebox;
1462 add_to_unconfirmed(new, strhashval);
1464 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1465 exid->clientid.cl_id = new->cl_clientid.cl_id;
1468 nfsd4_set_ex_flags(new, exid);
1470 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1471 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1475 nfs4_unlock_state();
1477 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1482 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1484 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1487 /* The slot is in use, and no response has been sent. */
1489 if (seqid == slot_seqid)
1490 return nfserr_jukebox;
1492 return nfserr_seq_misordered;
1495 if (likely(seqid == slot_seqid + 1))
1498 if (seqid == slot_seqid)
1499 return nfserr_replay_cache;
1501 if (seqid == 1 && (slot_seqid + 1) == 0)
1503 /* Misordered replay or misordered new request */
1504 return nfserr_seq_misordered;
1508 * Cache the create session result into the create session single DRC
1509 * slot cache by saving the xdr structure. sl_seqid has been set.
1510 * Do this for solo or embedded create session operations.
1513 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1514 struct nfsd4_clid_slot *slot, int nfserr)
1516 slot->sl_status = nfserr;
1517 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1521 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1522 struct nfsd4_clid_slot *slot)
1524 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1525 return slot->sl_status;
1528 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1529 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1530 1 + /* MIN tag is length with zero, only length */ \
1531 3 + /* version, opcount, opcode */ \
1532 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1533 /* seqid, slotID, slotID, cache */ \
1534 4 ) * sizeof(__be32))
1536 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1537 2 + /* verifier: AUTH_NULL, length 0 */\
1539 1 + /* MIN tag is length with zero, only length */ \
1540 3 + /* opcount, opcode, opstatus*/ \
1541 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1542 /* seqid, slotID, slotID, slotID, status */ \
1543 5 ) * sizeof(__be32))
1545 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1547 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1548 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1552 nfsd4_create_session(struct svc_rqst *rqstp,
1553 struct nfsd4_compound_state *cstate,
1554 struct nfsd4_create_session *cr_ses)
1556 struct sockaddr *sa = svc_addr(rqstp);
1557 struct nfs4_client *conf, *unconf;
1558 struct nfsd4_session *new;
1559 struct nfsd4_clid_slot *cs_slot = NULL;
1560 bool confirm_me = false;
1563 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1564 return nfserr_inval;
1567 unconf = find_unconfirmed_client(&cr_ses->clientid);
1568 conf = find_confirmed_client(&cr_ses->clientid);
1571 cs_slot = &conf->cl_cs_slot;
1572 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1573 if (status == nfserr_replay_cache) {
1574 dprintk("Got a create_session replay! seqid= %d\n",
1576 /* Return the cached reply status */
1577 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1579 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1580 status = nfserr_seq_misordered;
1581 dprintk("Sequence misordered!\n");
1582 dprintk("Expected seqid= %d but got seqid= %d\n",
1583 cs_slot->sl_seqid, cr_ses->seqid);
1586 } else if (unconf) {
1587 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1588 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1589 status = nfserr_clid_inuse;
1593 cs_slot = &unconf->cl_cs_slot;
1594 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1596 /* an unconfirmed replay returns misordered */
1597 status = nfserr_seq_misordered;
1604 status = nfserr_stale_clientid;
1609 * XXX: we should probably set this at creation time, and check
1610 * for consistent minorversion use throughout:
1612 conf->cl_minorversion = 1;
1614 * We do not support RDMA or persistent sessions
1616 cr_ses->flags &= ~SESSION4_PERSIST;
1617 cr_ses->flags &= ~SESSION4_RDMA;
1619 status = nfserr_toosmall;
1620 if (check_forechannel_attrs(cr_ses->fore_channel))
1623 status = nfserr_jukebox;
1624 new = alloc_init_session(rqstp, conf, cr_ses);
1628 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1629 NFS4_MAX_SESSIONID_LEN);
1630 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1631 sizeof(struct nfsd4_channel_attrs));
1632 cs_slot->sl_seqid++;
1633 cr_ses->seqid = cs_slot->sl_seqid;
1635 /* cache solo and embedded create sessions under the state lock */
1636 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1638 move_to_confirmed(conf);
1640 nfs4_unlock_state();
1641 dprintk("%s returns %d\n", __func__, ntohl(status));
1645 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1647 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1648 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1650 return argp->opcnt == resp->opcnt;
1653 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1656 case NFS4_CDFC4_FORE:
1657 case NFS4_CDFC4_BACK:
1659 case NFS4_CDFC4_FORE_OR_BOTH:
1660 case NFS4_CDFC4_BACK_OR_BOTH:
1661 *dir = NFS4_CDFC4_BOTH;
1664 return nfserr_inval;
1667 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1668 struct nfsd4_compound_state *cstate,
1669 struct nfsd4_bind_conn_to_session *bcts)
1673 if (!nfsd4_last_compound_op(rqstp))
1674 return nfserr_not_only_op;
1675 spin_lock(&client_lock);
1676 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1677 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1678 * client_lock iself: */
1679 if (cstate->session) {
1680 nfsd4_get_session(cstate->session);
1681 atomic_inc(&cstate->session->se_client->cl_refcount);
1683 spin_unlock(&client_lock);
1684 if (!cstate->session)
1685 return nfserr_badsession;
1687 status = nfsd4_map_bcts_dir(&bcts->dir);
1689 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1693 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1697 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1701 nfsd4_destroy_session(struct svc_rqst *r,
1702 struct nfsd4_compound_state *cstate,
1703 struct nfsd4_destroy_session *sessionid)
1705 struct nfsd4_session *ses;
1706 u32 status = nfserr_badsession;
1709 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1710 * - Should we return nfserr_back_chan_busy if waiting for
1711 * callbacks on to-be-destroyed session?
1712 * - Do we need to clear any callback info from previous session?
1715 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1716 if (!nfsd4_last_compound_op(r))
1717 return nfserr_not_only_op;
1719 dump_sessionid(__func__, &sessionid->sessionid);
1720 spin_lock(&client_lock);
1721 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1723 spin_unlock(&client_lock);
1727 unhash_session(ses);
1728 spin_unlock(&client_lock);
1731 nfsd4_probe_callback_sync(ses->se_client);
1732 nfs4_unlock_state();
1734 nfsd4_del_conns(ses);
1736 nfsd4_put_session(ses);
1739 dprintk("%s returns %d\n", __func__, ntohl(status));
1743 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1745 struct nfsd4_conn *c;
1747 list_for_each_entry(c, &s->se_conns, cn_persession) {
1748 if (c->cn_xprt == xpt) {
1755 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1757 struct nfs4_client *clp = ses->se_client;
1758 struct nfsd4_conn *c;
1761 spin_lock(&clp->cl_lock);
1762 c = __nfsd4_find_conn(new->cn_xprt, ses);
1764 spin_unlock(&clp->cl_lock);
1768 __nfsd4_hash_conn(new, ses);
1769 spin_unlock(&clp->cl_lock);
1770 ret = nfsd4_register_conn(new);
1772 /* oops; xprt is already down: */
1773 nfsd4_conn_lost(&new->cn_xpt_user);
1777 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1779 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1781 return args->opcnt > session->se_fchannel.maxops;
1784 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1785 struct nfsd4_session *session)
1787 struct xdr_buf *xb = &rqstp->rq_arg;
1789 return xb->len > session->se_fchannel.maxreq_sz;
1793 nfsd4_sequence(struct svc_rqst *rqstp,
1794 struct nfsd4_compound_state *cstate,
1795 struct nfsd4_sequence *seq)
1797 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1798 struct nfsd4_session *session;
1799 struct nfsd4_slot *slot;
1800 struct nfsd4_conn *conn;
1803 if (resp->opcnt != 1)
1804 return nfserr_sequence_pos;
1807 * Will be either used or freed by nfsd4_sequence_check_conn
1810 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1812 return nfserr_jukebox;
1814 spin_lock(&client_lock);
1815 status = nfserr_badsession;
1816 session = find_in_sessionid_hashtbl(&seq->sessionid);
1820 status = nfserr_too_many_ops;
1821 if (nfsd4_session_too_many_ops(rqstp, session))
1824 status = nfserr_req_too_big;
1825 if (nfsd4_request_too_big(rqstp, session))
1828 status = nfserr_badslot;
1829 if (seq->slotid >= session->se_fchannel.maxreqs)
1832 slot = session->se_slots[seq->slotid];
1833 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1835 /* We do not negotiate the number of slots yet, so set the
1836 * maxslots to the session maxreqs which is used to encode
1837 * sr_highest_slotid and the sr_target_slot id to maxslots */
1838 seq->maxslots = session->se_fchannel.maxreqs;
1840 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1841 if (status == nfserr_replay_cache) {
1842 cstate->slot = slot;
1843 cstate->session = session;
1844 /* Return the cached reply status and set cstate->status
1845 * for nfsd4_proc_compound processing */
1846 status = nfsd4_replay_cache_entry(resp, seq);
1847 cstate->status = nfserr_replay_cache;
1853 nfsd4_sequence_check_conn(conn, session);
1856 /* Success! bump slot seqid */
1857 slot->sl_inuse = true;
1858 slot->sl_seqid = seq->seqid;
1859 slot->sl_cachethis = seq->cachethis;
1861 cstate->slot = slot;
1862 cstate->session = session;
1865 /* Hold a session reference until done processing the compound. */
1866 if (cstate->session) {
1867 struct nfs4_client *clp = session->se_client;
1869 nfsd4_get_session(cstate->session);
1870 atomic_inc(&clp->cl_refcount);
1871 if (clp->cl_cb_state == NFSD4_CB_DOWN)
1872 seq->status_flags |= SEQ4_STATUS_CB_PATH_DOWN;
1875 spin_unlock(&client_lock);
1876 dprintk("%s: return %d\n", __func__, ntohl(status));
1881 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1885 if (rc->rca_one_fs) {
1886 if (!cstate->current_fh.fh_dentry)
1887 return nfserr_nofilehandle;
1889 * We don't take advantage of the rca_one_fs case.
1890 * That's OK, it's optional, we can safely ignore it.
1896 status = nfserr_complete_already;
1897 if (cstate->session->se_client->cl_firststate)
1900 status = nfserr_stale_clientid;
1901 if (is_client_expired(cstate->session->se_client))
1903 * The following error isn't really legal.
1904 * But we only get here if the client just explicitly
1905 * destroyed the client. Surely it no longer cares what
1906 * error it gets back on an operation for the dead
1912 nfsd4_create_clid_dir(cstate->session->se_client);
1914 nfs4_unlock_state();
1919 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1920 struct nfsd4_setclientid *setclid)
1922 struct xdr_netobj clname = {
1923 .len = setclid->se_namelen,
1924 .data = setclid->se_name,
1926 nfs4_verifier clverifier = setclid->se_verf;
1927 unsigned int strhashval;
1928 struct nfs4_client *conf, *unconf, *new;
1930 char dname[HEXDIR_LEN];
1932 if (!check_name(clname))
1933 return nfserr_inval;
1935 status = nfs4_make_rec_clidname(dname, &clname);
1940 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1941 * We get here on a DRC miss.
1944 strhashval = clientstr_hashval(dname);
1947 conf = find_confirmed_client_by_str(dname, strhashval);
1949 /* RFC 3530 14.2.33 CASE 0: */
1950 status = nfserr_clid_inuse;
1951 if (clp_used_exchangeid(conf))
1953 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1954 char addr_str[INET6_ADDRSTRLEN];
1955 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1957 dprintk("NFSD: setclientid: string in use by client "
1958 "at %s\n", addr_str);
1963 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1964 * has a description of SETCLIENTID request processing consisting
1965 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1967 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1968 status = nfserr_jukebox;
1971 * RFC 3530 14.2.33 CASE 4:
1972 * placed first, because it is the normal case
1975 expire_client(unconf);
1976 new = create_client(clname, dname, rqstp, &clverifier);
1980 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1982 * RFC 3530 14.2.33 CASE 1:
1983 * probable callback update
1986 /* Note this is removing unconfirmed {*x***},
1987 * which is stronger than RFC recommended {vxc**}.
1988 * This has the advantage that there is at most
1989 * one {*x***} in either list at any time.
1991 expire_client(unconf);
1993 new = create_client(clname, dname, rqstp, &clverifier);
1996 copy_clid(new, conf);
1997 } else if (!unconf) {
1999 * RFC 3530 14.2.33 CASE 2:
2000 * probable client reboot; state will be removed if
2003 new = create_client(clname, dname, rqstp, &clverifier);
2009 * RFC 3530 14.2.33 CASE 3:
2010 * probable client reboot; state will be removed if
2013 expire_client(unconf);
2014 new = create_client(clname, dname, rqstp, &clverifier);
2020 * XXX: we should probably set this at creation time, and check
2021 * for consistent minorversion use throughout:
2023 new->cl_minorversion = 0;
2024 gen_callback(new, setclid, rqstp);
2025 add_to_unconfirmed(new, strhashval);
2026 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2027 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2028 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2031 nfs4_unlock_state();
2037 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2038 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2039 * bullets, labeled as CASE1 - CASE4 below.
2042 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2043 struct nfsd4_compound_state *cstate,
2044 struct nfsd4_setclientid_confirm *setclientid_confirm)
2046 struct sockaddr *sa = svc_addr(rqstp);
2047 struct nfs4_client *conf, *unconf;
2048 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2049 clientid_t * clid = &setclientid_confirm->sc_clientid;
2052 if (STALE_CLIENTID(clid))
2053 return nfserr_stale_clientid;
2055 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2056 * We get here on a DRC miss.
2061 conf = find_confirmed_client(clid);
2062 unconf = find_unconfirmed_client(clid);
2064 status = nfserr_clid_inuse;
2065 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2067 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2071 * section 14.2.34 of RFC 3530 has a description of
2072 * SETCLIENTID_CONFIRM request processing consisting
2073 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2075 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2077 * RFC 3530 14.2.34 CASE 1:
2080 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2081 status = nfserr_clid_inuse;
2083 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2084 nfsd4_probe_callback(conf);
2085 expire_client(unconf);
2089 } else if (conf && !unconf) {
2091 * RFC 3530 14.2.34 CASE 2:
2092 * probable retransmitted request; play it safe and
2095 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2096 status = nfserr_clid_inuse;
2099 } else if (!conf && unconf
2100 && same_verf(&unconf->cl_confirm, &confirm)) {
2102 * RFC 3530 14.2.34 CASE 3:
2103 * Normal case; new or rebooted client:
2105 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2106 status = nfserr_clid_inuse;
2109 clientstr_hashval(unconf->cl_recdir);
2110 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2113 nfsd4_remove_clid_dir(conf);
2114 expire_client(conf);
2116 move_to_confirmed(unconf);
2118 nfsd4_probe_callback(conf);
2121 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2122 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2125 * RFC 3530 14.2.34 CASE 4:
2126 * Client probably hasn't noticed that we rebooted yet.
2128 status = nfserr_stale_clientid;
2130 /* check that we have hit one of the cases...*/
2131 status = nfserr_clid_inuse;
2134 nfs4_unlock_state();
2138 /* OPEN Share state helper functions */
2139 static inline struct nfs4_file *
2140 alloc_init_file(struct inode *ino)
2142 struct nfs4_file *fp;
2143 unsigned int hashval = file_hashval(ino);
2145 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
2147 atomic_set(&fp->fi_ref, 1);
2148 INIT_LIST_HEAD(&fp->fi_hash);
2149 INIT_LIST_HEAD(&fp->fi_stateids);
2150 INIT_LIST_HEAD(&fp->fi_delegations);
2151 fp->fi_inode = igrab(ino);
2152 fp->fi_id = current_fileid++;
2153 fp->fi_had_conflict = false;
2154 fp->fi_lease = NULL;
2155 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2156 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2157 spin_lock(&recall_lock);
2158 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2159 spin_unlock(&recall_lock);
2166 nfsd4_free_slab(struct kmem_cache **slab)
2170 kmem_cache_destroy(*slab);
2175 nfsd4_free_slabs(void)
2177 nfsd4_free_slab(&stateowner_slab);
2178 nfsd4_free_slab(&file_slab);
2179 nfsd4_free_slab(&stateid_slab);
2180 nfsd4_free_slab(&deleg_slab);
2184 nfsd4_init_slabs(void)
2186 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
2187 sizeof(struct nfs4_stateowner), 0, 0, NULL);
2188 if (stateowner_slab == NULL)
2190 file_slab = kmem_cache_create("nfsd4_files",
2191 sizeof(struct nfs4_file), 0, 0, NULL);
2192 if (file_slab == NULL)
2194 stateid_slab = kmem_cache_create("nfsd4_stateids",
2195 sizeof(struct nfs4_stateid), 0, 0, NULL);
2196 if (stateid_slab == NULL)
2198 deleg_slab = kmem_cache_create("nfsd4_delegations",
2199 sizeof(struct nfs4_delegation), 0, 0, NULL);
2200 if (deleg_slab == NULL)
2205 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2210 nfs4_free_stateowner(struct kref *kref)
2212 struct nfs4_stateowner *sop =
2213 container_of(kref, struct nfs4_stateowner, so_ref);
2214 kfree(sop->so_owner.data);
2215 kmem_cache_free(stateowner_slab, sop);
2218 static inline struct nfs4_stateowner *
2219 alloc_stateowner(struct xdr_netobj *owner)
2221 struct nfs4_stateowner *sop;
2223 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
2224 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
2225 memcpy(sop->so_owner.data, owner->data, owner->len);
2226 sop->so_owner.len = owner->len;
2227 kref_init(&sop->so_ref);
2230 kmem_cache_free(stateowner_slab, sop);
2235 static struct nfs4_stateowner *
2236 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2237 struct nfs4_stateowner *sop;
2238 struct nfs4_replay *rp;
2239 unsigned int idhashval;
2241 if (!(sop = alloc_stateowner(&open->op_owner)))
2243 idhashval = ownerid_hashval(current_ownerid);
2244 INIT_LIST_HEAD(&sop->so_idhash);
2245 INIT_LIST_HEAD(&sop->so_strhash);
2246 INIT_LIST_HEAD(&sop->so_perclient);
2247 INIT_LIST_HEAD(&sop->so_stateids);
2248 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
2249 INIT_LIST_HEAD(&sop->so_close_lru);
2251 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
2252 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
2253 list_add(&sop->so_perclient, &clp->cl_openowners);
2254 sop->so_is_open_owner = 1;
2255 sop->so_id = current_ownerid++;
2256 sop->so_client = clp;
2257 sop->so_seqid = open->op_seqid;
2258 sop->so_confirmed = 0;
2259 rp = &sop->so_replay;
2260 rp->rp_status = nfserr_serverfault;
2262 rp->rp_buf = rp->rp_ibuf;
2267 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2268 struct nfs4_stateowner *sop = open->op_stateowner;
2269 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2271 INIT_LIST_HEAD(&stp->st_hash);
2272 INIT_LIST_HEAD(&stp->st_perstateowner);
2273 INIT_LIST_HEAD(&stp->st_lockowners);
2274 INIT_LIST_HEAD(&stp->st_perfile);
2275 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
2276 list_add(&stp->st_perstateowner, &sop->so_stateids);
2277 list_add(&stp->st_perfile, &fp->fi_stateids);
2278 stp->st_stateowner = sop;
2281 stp->st_stateid.si_boot = boot_time;
2282 stp->st_stateid.si_stateownerid = sop->so_id;
2283 stp->st_stateid.si_fileid = fp->fi_id;
2284 stp->st_stateid.si_generation = 0;
2285 stp->st_access_bmap = 0;
2286 stp->st_deny_bmap = 0;
2287 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
2288 &stp->st_access_bmap);
2289 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2290 stp->st_openstp = NULL;
2294 move_to_close_lru(struct nfs4_stateowner *sop)
2296 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
2298 list_move_tail(&sop->so_close_lru, &close_lru);
2299 sop->so_time = get_seconds();
2303 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2306 return (sop->so_owner.len == owner->len) &&
2307 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2308 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2311 static struct nfs4_stateowner *
2312 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2314 struct nfs4_stateowner *so = NULL;
2316 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2317 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
2323 /* search file_hashtbl[] for file */
2324 static struct nfs4_file *
2325 find_file(struct inode *ino)
2327 unsigned int hashval = file_hashval(ino);
2328 struct nfs4_file *fp;
2330 spin_lock(&recall_lock);
2331 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2332 if (fp->fi_inode == ino) {
2334 spin_unlock(&recall_lock);
2338 spin_unlock(&recall_lock);
2342 static inline int access_valid(u32 x, u32 minorversion)
2344 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
2346 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
2348 x &= ~NFS4_SHARE_ACCESS_MASK;
2349 if (minorversion && x) {
2350 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
2352 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
2354 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
2361 static inline int deny_valid(u32 x)
2363 /* Note: unlike access bits, deny bits may be zero. */
2364 return x <= NFS4_SHARE_DENY_BOTH;
2368 * Called to check deny when READ with all zero stateid or
2369 * WRITE with all zero or all one stateid
2372 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2374 struct inode *ino = current_fh->fh_dentry->d_inode;
2375 struct nfs4_file *fp;
2376 struct nfs4_stateid *stp;
2379 dprintk("NFSD: nfs4_share_conflict\n");
2381 fp = find_file(ino);
2384 ret = nfserr_locked;
2385 /* Search for conflicting share reservations */
2386 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2387 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2388 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2397 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2399 /* We're assuming the state code never drops its reference
2400 * without first removing the lease. Since we're in this lease
2401 * callback (and since the lease code is serialized by the kernel
2402 * lock) we know the server hasn't removed the lease yet, we know
2403 * it's safe to take a reference: */
2404 atomic_inc(&dp->dl_count);
2406 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2408 /* only place dl_time is set. protected by lock_flocks*/
2409 dp->dl_time = get_seconds();
2411 nfsd4_cb_recall(dp);
2414 /* Called from break_lease() with lock_flocks() held. */
2415 static void nfsd_break_deleg_cb(struct file_lock *fl)
2417 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2418 struct nfs4_delegation *dp;
2421 /* We assume break_lease is only called once per lease: */
2422 BUG_ON(fp->fi_had_conflict);
2424 * We don't want the locks code to timeout the lease for us;
2425 * we'll remove it ourself if a delegation isn't returned
2428 fl->fl_break_time = 0;
2430 spin_lock(&recall_lock);
2431 fp->fi_had_conflict = true;
2432 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2433 nfsd_break_one_deleg(dp);
2434 spin_unlock(&recall_lock);
2438 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2441 return lease_modify(onlist, arg);
2446 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2447 .lm_break = nfsd_break_deleg_cb,
2448 .lm_change = nfsd_change_deleg_cb,
2453 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2454 struct nfsd4_open *open)
2456 clientid_t *clientid = &open->op_clientid;
2457 struct nfs4_client *clp = NULL;
2458 unsigned int strhashval;
2459 struct nfs4_stateowner *sop = NULL;
2461 if (!check_name(open->op_owner))
2462 return nfserr_inval;
2464 if (STALE_CLIENTID(&open->op_clientid))
2465 return nfserr_stale_clientid;
2467 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2468 sop = find_openstateowner_str(strhashval, open);
2469 open->op_stateowner = sop;
2471 /* Make sure the client's lease hasn't expired. */
2472 clp = find_confirmed_client(clientid);
2474 return nfserr_expired;
2477 /* When sessions are used, skip open sequenceid processing */
2478 if (nfsd4_has_session(cstate))
2480 if (!sop->so_confirmed) {
2481 /* Replace unconfirmed owners without checking for replay. */
2482 clp = sop->so_client;
2483 release_openowner(sop);
2484 open->op_stateowner = NULL;
2487 if (open->op_seqid == sop->so_seqid - 1) {
2488 if (sop->so_replay.rp_buflen)
2489 return nfserr_replay_me;
2490 /* The original OPEN failed so spectacularly
2491 * that we don't even have replay data saved!
2492 * Therefore, we have no choice but to continue
2493 * processing this OPEN; presumably, we'll
2494 * fail again for the same reason.
2496 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2499 if (open->op_seqid != sop->so_seqid)
2500 return nfserr_bad_seqid;
2502 if (open->op_stateowner == NULL) {
2503 sop = alloc_init_open_stateowner(strhashval, clp, open);
2505 return nfserr_jukebox;
2506 open->op_stateowner = sop;
2508 list_del_init(&sop->so_close_lru);
2509 renew_client(sop->so_client);
2513 static inline __be32
2514 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2516 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2517 return nfserr_openmode;
2522 static struct nfs4_delegation *
2523 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2525 struct nfs4_delegation *dp;
2527 spin_lock(&recall_lock);
2528 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2529 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid) {
2530 spin_unlock(&recall_lock);
2533 spin_unlock(&recall_lock);
2537 static int share_access_to_flags(u32 share_access)
2539 share_access &= ~NFS4_SHARE_WANT_MASK;
2541 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2545 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2546 struct nfs4_delegation **dp)
2549 __be32 status = nfserr_bad_stateid;
2551 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2554 flags = share_access_to_flags(open->op_share_access);
2555 status = nfs4_check_delegmode(*dp, flags);
2559 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2563 open->op_stateowner->so_confirmed = 1;
2568 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2570 struct nfs4_stateid *local;
2571 __be32 status = nfserr_share_denied;
2572 struct nfs4_stateowner *sop = open->op_stateowner;
2574 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2575 /* ignore lock owners */
2576 if (local->st_stateowner->so_is_open_owner == 0)
2578 /* remember if we have seen this open owner */
2579 if (local->st_stateowner == sop)
2581 /* check for conflicting share reservations */
2582 if (!test_share(local, open))
2590 static inline struct nfs4_stateid *
2591 nfs4_alloc_stateid(void)
2593 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2596 static inline int nfs4_access_to_access(u32 nfs4_access)
2600 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2601 flags |= NFSD_MAY_READ;
2602 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2603 flags |= NFSD_MAY_WRITE;
2607 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2608 struct svc_fh *cur_fh, struct nfsd4_open *open)
2611 int oflag = nfs4_access_to_omode(open->op_share_access);
2612 int access = nfs4_access_to_access(open->op_share_access);
2614 /* CLAIM_DELEGATE_CUR is used in response to a broken lease;
2615 * allowing it to break the lease and return EAGAIN leaves the
2616 * client unable to make progress in returning the delegation */
2617 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2618 access |= NFSD_MAY_NOT_BREAK_LEASE;
2620 if (!fp->fi_fds[oflag]) {
2621 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2622 &fp->fi_fds[oflag]);
2626 nfs4_file_get_access(fp, oflag);
2632 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2633 struct nfs4_file *fp, struct svc_fh *cur_fh,
2634 struct nfsd4_open *open)
2636 struct nfs4_stateid *stp;
2639 stp = nfs4_alloc_stateid();
2641 return nfserr_jukebox;
2643 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2645 kmem_cache_free(stateid_slab, stp);
2652 static inline __be32
2653 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2654 struct nfsd4_open *open)
2656 struct iattr iattr = {
2657 .ia_valid = ATTR_SIZE,
2660 if (!open->op_truncate)
2662 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2663 return nfserr_inval;
2664 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2668 nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2670 u32 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2674 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2676 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2680 status = nfsd4_truncate(rqstp, cur_fh, open);
2683 int oflag = nfs4_access_to_omode(op_share_access);
2684 nfs4_file_put_access(fp, oflag);
2688 /* remember the open */
2689 __set_bit(op_share_access, &stp->st_access_bmap);
2690 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2697 nfs4_set_claim_prev(struct nfsd4_open *open)
2699 open->op_stateowner->so_confirmed = 1;
2700 open->op_stateowner->so_client->cl_firststate = 1;
2703 /* Should we give out recallable state?: */
2704 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2706 if (clp->cl_cb_state == NFSD4_CB_UP)
2709 * In the sessions case, since we don't have to establish a
2710 * separate connection for callbacks, we assume it's OK
2711 * until we hear otherwise:
2713 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2716 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2718 struct file_lock *fl;
2720 fl = locks_alloc_lock();
2723 locks_init_lock(fl);
2724 fl->fl_lmops = &nfsd_lease_mng_ops;
2725 fl->fl_flags = FL_LEASE;
2726 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2727 fl->fl_end = OFFSET_MAX;
2728 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2729 fl->fl_pid = current->tgid;
2733 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2735 struct nfs4_file *fp = dp->dl_file;
2736 struct file_lock *fl;
2739 fl = nfs4_alloc_init_lease(dp, flag);
2742 fl->fl_file = find_readable_file(fp);
2743 list_add(&dp->dl_perclnt, &dp->dl_client->cl_delegations);
2744 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2746 list_del_init(&dp->dl_perclnt);
2747 locks_free_lock(fl);
2751 fp->fi_deleg_file = fl->fl_file;
2752 get_file(fp->fi_deleg_file);
2753 atomic_set(&fp->fi_delegees, 1);
2754 list_add(&dp->dl_perfile, &fp->fi_delegations);
2758 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2760 struct nfs4_file *fp = dp->dl_file;
2763 return nfs4_setlease(dp, flag);
2764 spin_lock(&recall_lock);
2765 if (fp->fi_had_conflict) {
2766 spin_unlock(&recall_lock);
2769 atomic_inc(&fp->fi_delegees);
2770 list_add(&dp->dl_perfile, &fp->fi_delegations);
2771 spin_unlock(&recall_lock);
2772 list_add(&dp->dl_perclnt, &dp->dl_client->cl_delegations);
2777 * Attempt to hand out a delegation.
2780 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2782 struct nfs4_delegation *dp;
2783 struct nfs4_stateowner *sop = stp->st_stateowner;
2785 int status, flag = 0;
2787 cb_up = nfsd4_cb_channel_good(sop->so_client);
2788 flag = NFS4_OPEN_DELEGATE_NONE;
2789 open->op_recall = 0;
2790 switch (open->op_claim_type) {
2791 case NFS4_OPEN_CLAIM_PREVIOUS:
2793 open->op_recall = 1;
2794 flag = open->op_delegate_type;
2795 if (flag == NFS4_OPEN_DELEGATE_NONE)
2798 case NFS4_OPEN_CLAIM_NULL:
2799 /* Let's not give out any delegations till everyone's
2800 * had the chance to reclaim theirs.... */
2801 if (locks_in_grace())
2803 if (!cb_up || !sop->so_confirmed)
2805 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2806 flag = NFS4_OPEN_DELEGATE_WRITE;
2808 flag = NFS4_OPEN_DELEGATE_READ;
2814 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2817 status = nfs4_set_delegation(dp, flag);
2821 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2823 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2824 STATEID_VAL(&dp->dl_stateid));
2826 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2827 && flag == NFS4_OPEN_DELEGATE_NONE
2828 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2829 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2830 open->op_delegate_type = flag;
2833 nfs4_put_delegation(dp);
2835 flag = NFS4_OPEN_DELEGATE_NONE;
2840 * called with nfs4_lock_state() held.
2843 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2845 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2846 struct nfs4_file *fp = NULL;
2847 struct inode *ino = current_fh->fh_dentry->d_inode;
2848 struct nfs4_stateid *stp = NULL;
2849 struct nfs4_delegation *dp = NULL;
2852 status = nfserr_inval;
2853 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2854 || !deny_valid(open->op_share_deny))
2857 * Lookup file; if found, lookup stateid and check open request,
2858 * and check for delegations in the process of being recalled.
2859 * If not found, create the nfs4_file struct
2861 fp = find_file(ino);
2863 if ((status = nfs4_check_open(fp, open, &stp)))
2865 status = nfs4_check_deleg(fp, open, &dp);
2869 status = nfserr_bad_stateid;
2870 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2872 status = nfserr_jukebox;
2873 fp = alloc_init_file(ino);
2879 * OPEN the file, or upgrade an existing OPEN.
2880 * If truncate fails, the OPEN fails.
2883 /* Stateid was found, this is an OPEN upgrade */
2884 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
2887 update_stateid(&stp->st_stateid);
2889 status = nfs4_new_open(rqstp, &stp, fp, current_fh, open);
2892 init_stateid(stp, fp, open);
2893 status = nfsd4_truncate(rqstp, current_fh, open);
2895 release_open_stateid(stp);
2898 if (nfsd4_has_session(&resp->cstate))
2899 update_stateid(&stp->st_stateid);
2901 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2903 if (nfsd4_has_session(&resp->cstate))
2904 open->op_stateowner->so_confirmed = 1;
2907 * Attempt to hand out a delegation. No error return, because the
2908 * OPEN succeeds even if we fail.
2910 nfs4_open_delegation(current_fh, open, stp);
2914 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2915 STATEID_VAL(&stp->st_stateid));
2919 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2920 nfs4_set_claim_prev(open);
2922 * To finish the open response, we just need to set the rflags.
2924 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2925 if (!open->op_stateowner->so_confirmed &&
2926 !nfsd4_has_session(&resp->cstate))
2927 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2933 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2936 struct nfs4_client *clp;
2940 dprintk("process_renew(%08x/%08x): starting\n",
2941 clid->cl_boot, clid->cl_id);
2942 status = nfserr_stale_clientid;
2943 if (STALE_CLIENTID(clid))
2945 clp = find_confirmed_client(clid);
2946 status = nfserr_expired;
2948 /* We assume the client took too long to RENEW. */
2949 dprintk("nfsd4_renew: clientid not found!\n");
2953 status = nfserr_cb_path_down;
2954 if (!list_empty(&clp->cl_delegations)
2955 && clp->cl_cb_state != NFSD4_CB_UP)
2959 nfs4_unlock_state();
2963 static struct lock_manager nfsd4_manager = {
2967 nfsd4_end_grace(void)
2969 dprintk("NFSD: end of grace period\n");
2970 nfsd4_recdir_purge_old();
2971 locks_end_grace(&nfsd4_manager);
2973 * Now that every NFSv4 client has had the chance to recover and
2974 * to see the (possibly new, possibly shorter) lease time, we
2975 * can safely set the next grace time to the current lease time:
2977 nfsd4_grace = nfsd4_lease;
2981 nfs4_laundromat(void)
2983 struct nfs4_client *clp;
2984 struct nfs4_stateowner *sop;
2985 struct nfs4_delegation *dp;
2986 struct list_head *pos, *next, reaplist;
2987 time_t cutoff = get_seconds() - nfsd4_lease;
2988 time_t t, clientid_val = nfsd4_lease;
2989 time_t u, test_val = nfsd4_lease;
2993 dprintk("NFSD: laundromat service - starting\n");
2994 if (locks_in_grace())
2996 INIT_LIST_HEAD(&reaplist);
2997 spin_lock(&client_lock);
2998 list_for_each_safe(pos, next, &client_lru) {
2999 clp = list_entry(pos, struct nfs4_client, cl_lru);
3000 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3001 t = clp->cl_time - cutoff;
3002 if (clientid_val > t)
3006 if (atomic_read(&clp->cl_refcount)) {
3007 dprintk("NFSD: client in use (clientid %08x)\n",
3008 clp->cl_clientid.cl_id);
3011 unhash_client_locked(clp);
3012 list_add(&clp->cl_lru, &reaplist);
3014 spin_unlock(&client_lock);
3015 list_for_each_safe(pos, next, &reaplist) {
3016 clp = list_entry(pos, struct nfs4_client, cl_lru);
3017 dprintk("NFSD: purging unused client (clientid %08x)\n",
3018 clp->cl_clientid.cl_id);
3019 nfsd4_remove_clid_dir(clp);
3022 spin_lock(&recall_lock);
3023 list_for_each_safe(pos, next, &del_recall_lru) {
3024 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3025 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3026 u = dp->dl_time - cutoff;
3031 list_move(&dp->dl_recall_lru, &reaplist);
3033 spin_unlock(&recall_lock);
3034 list_for_each_safe(pos, next, &reaplist) {
3035 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3036 list_del_init(&dp->dl_recall_lru);
3037 unhash_delegation(dp);
3039 test_val = nfsd4_lease;
3040 list_for_each_safe(pos, next, &close_lru) {
3041 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
3042 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
3043 u = sop->so_time - cutoff;
3048 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
3050 release_openowner(sop);
3052 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3053 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3054 nfs4_unlock_state();
3055 return clientid_val;
3058 static struct workqueue_struct *laundry_wq;
3059 static void laundromat_main(struct work_struct *);
3060 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3063 laundromat_main(struct work_struct *not_used)
3067 t = nfs4_laundromat();
3068 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3069 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3072 static struct nfs4_stateowner *
3073 search_close_lru(u32 st_id, int flags)
3075 struct nfs4_stateowner *local = NULL;
3077 if (flags & CLOSE_STATE) {
3078 list_for_each_entry(local, &close_lru, so_close_lru) {
3079 if (local->so_id == st_id)
3087 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
3089 return fhp->fh_dentry->d_inode != stp->st_file->fi_inode;
3093 STALE_STATEID(stateid_t *stateid)
3095 if (stateid->si_boot == boot_time)
3097 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3098 STATEID_VAL(stateid));
3103 access_permit_read(unsigned long access_bmap)
3105 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3106 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3107 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3111 access_permit_write(unsigned long access_bmap)
3113 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3114 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3118 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
3120 __be32 status = nfserr_openmode;
3122 /* For lock stateid's, we test the parent open, not the lock: */
3123 if (stp->st_openstp)
3124 stp = stp->st_openstp;
3125 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3127 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3134 static inline __be32
3135 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3137 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3139 else if (locks_in_grace()) {
3140 /* Answer in remaining cases depends on existence of
3141 * conflicting state; so we must wait out the grace period. */
3142 return nfserr_grace;
3143 } else if (flags & WR_STATE)
3144 return nfs4_share_conflict(current_fh,
3145 NFS4_SHARE_DENY_WRITE);
3146 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3147 return nfs4_share_conflict(current_fh,
3148 NFS4_SHARE_DENY_READ);
3152 * Allow READ/WRITE during grace period on recovered state only for files
3153 * that are not able to provide mandatory locking.
3156 grace_disallows_io(struct inode *inode)
3158 return locks_in_grace() && mandatory_lock(inode);
3161 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
3164 * When sessions are used the stateid generation number is ignored
3167 if ((flags & HAS_SESSION) && in->si_generation == 0)
3170 /* If the client sends us a stateid from the future, it's buggy: */
3171 if (in->si_generation > ref->si_generation)
3172 return nfserr_bad_stateid;
3174 * The following, however, can happen. For example, if the
3175 * client sends an open and some IO at the same time, the open
3176 * may bump si_generation while the IO is still in flight.
3177 * Thanks to hard links and renames, the client never knows what
3178 * file an open will affect. So it could avoid that situation
3179 * only by serializing all opens and IO from the same open
3180 * owner. To recover from the old_stateid error, the client
3181 * will just have to retry the IO:
3183 if (in->si_generation < ref->si_generation)
3184 return nfserr_old_stateid;
3189 static int is_delegation_stateid(stateid_t *stateid)
3191 return stateid->si_fileid == 0;
3194 static int is_open_stateid(struct nfs4_stateid *stateid)
3196 return stateid->st_openstp == NULL;
3199 __be32 nfs4_validate_stateid(stateid_t *stateid, int flags)
3201 struct nfs4_stateid *stp = NULL;
3202 __be32 status = nfserr_stale_stateid;
3204 if (STALE_STATEID(stateid))
3207 status = nfserr_expired;
3208 stp = search_for_stateid(stateid);
3211 status = nfserr_bad_stateid;
3213 if (!stp->st_stateowner->so_confirmed)
3216 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
3226 * Checks for stateid operations
3229 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3230 stateid_t *stateid, int flags, struct file **filpp)
3232 struct nfs4_stateid *stp = NULL;
3233 struct nfs4_delegation *dp = NULL;
3234 struct svc_fh *current_fh = &cstate->current_fh;
3235 struct inode *ino = current_fh->fh_dentry->d_inode;
3241 if (grace_disallows_io(ino))
3242 return nfserr_grace;
3244 if (nfsd4_has_session(cstate))
3245 flags |= HAS_SESSION;
3247 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3248 return check_special_stateids(current_fh, stateid, flags);
3250 status = nfserr_stale_stateid;
3251 if (STALE_STATEID(stateid))
3255 * We assume that any stateid that has the current boot time,
3256 * but that we can't find, is expired:
3258 status = nfserr_expired;
3259 if (is_delegation_stateid(stateid)) {
3260 dp = find_delegation_stateid(ino, stateid);
3263 status = check_stateid_generation(stateid, &dp->dl_stateid,
3267 status = nfs4_check_delegmode(dp, flags);
3270 renew_client(dp->dl_client);
3272 *filpp = dp->dl_file->fi_deleg_file;
3275 } else { /* open or lock stateid */
3276 stp = find_stateid(stateid, flags);
3279 status = nfserr_bad_stateid;
3280 if (nfs4_check_fh(current_fh, stp))
3282 if (!stp->st_stateowner->so_confirmed)
3284 status = check_stateid_generation(stateid, &stp->st_stateid,
3288 status = nfs4_check_openmode(stp, flags);
3291 renew_client(stp->st_stateowner->so_client);
3293 if (flags & RD_STATE)
3294 *filpp = find_readable_file(stp->st_file);
3296 *filpp = find_writeable_file(stp->st_file);
3305 nfsd4_free_delegation_stateid(stateid_t *stateid)
3307 struct nfs4_delegation *dp = search_for_delegation(stateid);
3309 return nfserr_locks_held;
3310 return nfserr_bad_stateid;
3314 nfsd4_free_lock_stateid(struct nfs4_stateid *stp)
3316 if (check_for_locks(stp->st_file, stp->st_stateowner))
3317 return nfserr_locks_held;
3318 release_lock_stateid(stp);
3323 * Test if the stateid is valid
3326 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3327 struct nfsd4_test_stateid *test_stateid)
3329 test_stateid->ts_has_session = nfsd4_has_session(cstate);
3337 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3338 struct nfsd4_free_stateid *free_stateid)
3340 stateid_t *stateid = &free_stateid->fr_stateid;
3341 struct nfs4_stateid *stp;
3345 if (is_delegation_stateid(stateid)) {
3346 ret = nfsd4_free_delegation_stateid(stateid);
3350 stp = search_for_stateid(stateid);
3352 ret = nfserr_bad_stateid;
3355 if (stateid->si_generation != 0) {
3356 if (stateid->si_generation < stp->st_stateid.si_generation) {
3357 ret = nfserr_old_stateid;
3360 if (stateid->si_generation > stp->st_stateid.si_generation) {
3361 ret = nfserr_bad_stateid;
3366 if (is_open_stateid(stp)) {
3367 ret = nfserr_locks_held;
3370 ret = nfsd4_free_lock_stateid(stp);
3375 nfs4_unlock_state();
3382 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3383 RD_STATE : WR_STATE;
3387 * Checks for sequence id mutating operations.
3390 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3391 stateid_t *stateid, int flags,
3392 struct nfs4_stateowner **sopp,
3393 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
3395 struct nfs4_stateid *stp;
3396 struct nfs4_stateowner *sop;
3397 struct svc_fh *current_fh = &cstate->current_fh;
3400 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3401 seqid, STATEID_VAL(stateid));
3406 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
3407 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3408 return nfserr_bad_stateid;
3411 if (STALE_STATEID(stateid))
3412 return nfserr_stale_stateid;
3414 if (nfsd4_has_session(cstate))
3415 flags |= HAS_SESSION;
3418 * We return BAD_STATEID if filehandle doesn't match stateid,
3419 * the confirmed flag is incorrecly set, or the generation
3420 * number is incorrect.
3422 stp = find_stateid(stateid, flags);
3425 * Also, we should make sure this isn't just the result of
3428 sop = search_close_lru(stateid->si_stateownerid, flags);
3429 /* It's not stale; let's assume it's expired: */
3431 return nfserr_expired;
3437 *sopp = sop = stp->st_stateowner;
3440 clientid_t *lockclid = &lock->v.new.clientid;
3441 struct nfs4_client *clp = sop->so_client;
3445 lkflg = setlkflg(lock->lk_type);
3447 if (lock->lk_is_new) {
3448 if (!sop->so_is_open_owner)
3449 return nfserr_bad_stateid;
3450 if (!(flags & HAS_SESSION) &&
3451 !same_clid(&clp->cl_clientid, lockclid))
3452 return nfserr_bad_stateid;
3453 /* stp is the open stateid */
3454 status = nfs4_check_openmode(stp, lkflg);
3458 /* stp is the lock stateid */
3459 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3465 if (nfs4_check_fh(current_fh, stp)) {
3466 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3467 return nfserr_bad_stateid;
3471 * We now validate the seqid and stateid generation numbers.
3472 * For the moment, we ignore the possibility of
3473 * generation number wraparound.
3475 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
3478 if (sop->so_confirmed && flags & CONFIRM) {
3479 dprintk("NFSD: preprocess_seqid_op: expected"
3480 " unconfirmed stateowner!\n");
3481 return nfserr_bad_stateid;
3483 if (!sop->so_confirmed && !(flags & CONFIRM)) {
3484 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3485 " confirmed yet!\n");
3486 return nfserr_bad_stateid;
3488 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
3491 renew_client(sop->so_client);
3495 if (seqid == sop->so_seqid - 1) {
3496 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3497 /* indicate replay to calling function */
3498 return nfserr_replay_me;
3500 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3501 sop->so_seqid, seqid);
3503 return nfserr_bad_seqid;
3507 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3508 struct nfsd4_open_confirm *oc)
3511 struct nfs4_stateowner *sop;
3512 struct nfs4_stateid *stp;
3514 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3515 (int)cstate->current_fh.fh_dentry->d_name.len,
3516 cstate->current_fh.fh_dentry->d_name.name);
3518 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3524 if ((status = nfs4_preprocess_seqid_op(cstate,
3525 oc->oc_seqid, &oc->oc_req_stateid,
3526 CONFIRM | OPEN_STATE,
3527 &oc->oc_stateowner, &stp, NULL)))
3530 sop = oc->oc_stateowner;
3531 sop->so_confirmed = 1;
3532 update_stateid(&stp->st_stateid);
3533 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3534 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3535 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3537 nfsd4_create_clid_dir(sop->so_client);
3539 if (oc->oc_stateowner) {
3540 nfs4_get_stateowner(oc->oc_stateowner);
3541 cstate->replay_owner = oc->oc_stateowner;
3543 nfs4_unlock_state();
3547 static inline void nfs4_file_downgrade(struct nfs4_stateid *stp, unsigned int to_access)
3551 for (i = 1; i < 4; i++) {
3552 if (test_bit(i, &stp->st_access_bmap) && !(i & to_access)) {
3553 nfs4_file_put_access(stp->st_file, i);
3554 __clear_bit(i, &stp->st_access_bmap);
3560 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3563 for (i = 0; i < 4; i++) {
3564 if ((i & deny) != i)
3565 __clear_bit(i, bmap);
3570 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3571 struct nfsd4_compound_state *cstate,
3572 struct nfsd4_open_downgrade *od)
3575 struct nfs4_stateid *stp;
3577 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3578 (int)cstate->current_fh.fh_dentry->d_name.len,
3579 cstate->current_fh.fh_dentry->d_name.name);
3581 if (!access_valid(od->od_share_access, cstate->minorversion)
3582 || !deny_valid(od->od_share_deny))
3583 return nfserr_inval;
3586 if ((status = nfs4_preprocess_seqid_op(cstate,
3590 &od->od_stateowner, &stp, NULL)))
3593 status = nfserr_inval;
3594 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3595 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3596 stp->st_access_bmap, od->od_share_access);
3599 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3600 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3601 stp->st_deny_bmap, od->od_share_deny);
3604 nfs4_file_downgrade(stp, od->od_share_access);
3606 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3608 update_stateid(&stp->st_stateid);
3609 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3612 if (od->od_stateowner) {
3613 nfs4_get_stateowner(od->od_stateowner);
3614 cstate->replay_owner = od->od_stateowner;
3616 nfs4_unlock_state();
3621 * nfs4_unlock_state() called after encode
3624 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3625 struct nfsd4_close *close)
3628 struct nfs4_stateid *stp;
3630 dprintk("NFSD: nfsd4_close on file %.*s\n",
3631 (int)cstate->current_fh.fh_dentry->d_name.len,
3632 cstate->current_fh.fh_dentry->d_name.name);
3635 /* check close_lru for replay */
3636 if ((status = nfs4_preprocess_seqid_op(cstate,
3639 OPEN_STATE | CLOSE_STATE,
3640 &close->cl_stateowner, &stp, NULL)))
3643 update_stateid(&stp->st_stateid);
3644 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3646 /* release_stateid() calls nfsd_close() if needed */
3647 release_open_stateid(stp);
3649 /* place unused nfs4_stateowners on so_close_lru list to be
3650 * released by the laundromat service after the lease period
3651 * to enable us to handle CLOSE replay
3653 if (list_empty(&close->cl_stateowner->so_stateids))
3654 move_to_close_lru(close->cl_stateowner);
3656 if (close->cl_stateowner) {
3657 nfs4_get_stateowner(close->cl_stateowner);
3658 cstate->replay_owner = close->cl_stateowner;
3660 nfs4_unlock_state();
3665 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3666 struct nfsd4_delegreturn *dr)
3668 struct nfs4_delegation *dp;
3669 stateid_t *stateid = &dr->dr_stateid;
3670 struct inode *inode;
3674 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3676 inode = cstate->current_fh.fh_dentry->d_inode;
3678 if (nfsd4_has_session(cstate))
3679 flags |= HAS_SESSION;
3681 status = nfserr_bad_stateid;
3682 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3684 status = nfserr_stale_stateid;
3685 if (STALE_STATEID(stateid))
3687 status = nfserr_bad_stateid;
3688 if (!is_delegation_stateid(stateid))
3690 status = nfserr_expired;
3691 dp = find_delegation_stateid(inode, stateid);
3694 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3697 renew_client(dp->dl_client);
3699 unhash_delegation(dp);
3701 nfs4_unlock_state();
3708 * Lock owner state (byte-range locks)
3710 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3711 #define LOCK_HASH_BITS 8
3712 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3713 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3716 end_offset(u64 start, u64 len)
3721 return end >= start ? end: NFS4_MAX_UINT64;
3724 /* last octet in a range */
3726 last_byte_offset(u64 start, u64 len)
3732 return end > start ? end - 1: NFS4_MAX_UINT64;
3735 static unsigned int lockownerid_hashval(u32 id)
3737 return id & LOCK_HASH_MASK;
3740 static inline unsigned int
3741 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3742 struct xdr_netobj *ownername)
3744 return (file_hashval(inode) + cl_id
3745 + opaque_hashval(ownername->data, ownername->len))
3749 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3750 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3751 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3754 same_stateid(stateid_t *id_one, stateid_t *id_two)
3756 if (id_one->si_stateownerid != id_two->si_stateownerid)
3758 return id_one->si_fileid == id_two->si_fileid;
3761 static struct nfs4_stateid *
3762 find_stateid(stateid_t *stid, int flags)
3764 struct nfs4_stateid *local;
3765 u32 st_id = stid->si_stateownerid;
3766 u32 f_id = stid->si_fileid;
3767 unsigned int hashval;
3769 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3770 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3771 hashval = stateid_hashval(st_id, f_id);
3772 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3773 if ((local->st_stateid.si_stateownerid == st_id) &&
3774 (local->st_stateid.si_fileid == f_id))
3779 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3780 hashval = stateid_hashval(st_id, f_id);
3781 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3782 if ((local->st_stateid.si_stateownerid == st_id) &&
3783 (local->st_stateid.si_fileid == f_id))
3790 static struct nfs4_stateid *
3791 search_for_stateid(stateid_t *stid)
3793 struct nfs4_stateid *local;
3794 unsigned int hashval = stateid_hashval(stid->si_stateownerid, stid->si_fileid);
3796 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3797 if (same_stateid(&local->st_stateid, stid))
3801 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3802 if (same_stateid(&local->st_stateid, stid))
3808 static struct nfs4_delegation *
3809 search_for_delegation(stateid_t *stid)
3811 struct nfs4_file *fp;
3812 struct nfs4_delegation *dp;
3813 struct list_head *pos;
3816 for (i = 0; i < FILE_HASH_SIZE; i++) {
3817 list_for_each_entry(fp, &file_hashtbl[i], fi_hash) {
3818 list_for_each(pos, &fp->fi_delegations) {
3819 dp = list_entry(pos, struct nfs4_delegation, dl_perfile);
3820 if (same_stateid(&dp->dl_stateid, stid))
3828 static struct nfs4_delegation *
3829 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3831 struct nfs4_file *fp;
3832 struct nfs4_delegation *dl;
3834 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3837 fp = find_file(ino);
3840 dl = find_delegation_file(fp, stid);
3846 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3847 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3848 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3849 * locking, this prevents us from being completely protocol-compliant. The
3850 * real solution to this problem is to start using unsigned file offsets in
3851 * the VFS, but this is a very deep change!
3854 nfs4_transform_lock_offset(struct file_lock *lock)
3856 if (lock->fl_start < 0)
3857 lock->fl_start = OFFSET_MAX;
3858 if (lock->fl_end < 0)
3859 lock->fl_end = OFFSET_MAX;
3862 /* Hack!: For now, we're defining this just so we can use a pointer to it
3863 * as a unique cookie to identify our (NFSv4's) posix locks. */
3864 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3868 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3870 struct nfs4_stateowner *sop;
3872 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3873 sop = (struct nfs4_stateowner *) fl->fl_owner;
3874 kref_get(&sop->so_ref);
3876 deny->ld_clientid = sop->so_client->cl_clientid;
3878 deny->ld_sop = NULL;
3879 deny->ld_clientid.cl_boot = 0;
3880 deny->ld_clientid.cl_id = 0;
3882 deny->ld_start = fl->fl_start;
3883 deny->ld_length = NFS4_MAX_UINT64;
3884 if (fl->fl_end != NFS4_MAX_UINT64)
3885 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3886 deny->ld_type = NFS4_READ_LT;
3887 if (fl->fl_type != F_RDLCK)
3888 deny->ld_type = NFS4_WRITE_LT;
3891 static struct nfs4_stateowner *
3892 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3893 struct xdr_netobj *owner)
3895 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3896 struct nfs4_stateowner *op;
3898 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3899 if (same_owner_str(op, owner, clid))
3906 * Alloc a lock owner structure.
3907 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3910 * strhashval = lock_ownerstr_hashval
3913 static struct nfs4_stateowner *
3914 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3915 struct nfs4_stateowner *sop;
3916 struct nfs4_replay *rp;
3917 unsigned int idhashval;
3919 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3921 idhashval = lockownerid_hashval(current_ownerid);
3922 INIT_LIST_HEAD(&sop->so_idhash);
3923 INIT_LIST_HEAD(&sop->so_strhash);
3924 INIT_LIST_HEAD(&sop->so_perclient);
3925 INIT_LIST_HEAD(&sop->so_stateids);
3926 INIT_LIST_HEAD(&sop->so_perstateid);
3927 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3929 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3930 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3931 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3932 sop->so_is_open_owner = 0;
3933 sop->so_id = current_ownerid++;
3934 sop->so_client = clp;
3935 /* It is the openowner seqid that will be incremented in encode in the
3936 * case of new lockowners; so increment the lock seqid manually: */
3937 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3938 sop->so_confirmed = 1;
3939 rp = &sop->so_replay;
3940 rp->rp_status = nfserr_serverfault;
3942 rp->rp_buf = rp->rp_ibuf;
3946 static struct nfs4_stateid *
3947 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3949 struct nfs4_stateid *stp;
3950 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3952 stp = nfs4_alloc_stateid();
3955 INIT_LIST_HEAD(&stp->st_hash);
3956 INIT_LIST_HEAD(&stp->st_perfile);
3957 INIT_LIST_HEAD(&stp->st_perstateowner);
3958 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3959 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3960 list_add(&stp->st_perfile, &fp->fi_stateids);
3961 list_add(&stp->st_perstateowner, &sop->so_stateids);
3962 stp->st_stateowner = sop;
3965 stp->st_stateid.si_boot = boot_time;
3966 stp->st_stateid.si_stateownerid = sop->so_id;
3967 stp->st_stateid.si_fileid = fp->fi_id;
3968 stp->st_stateid.si_generation = 0;
3969 stp->st_access_bmap = 0;
3970 stp->st_deny_bmap = open_stp->st_deny_bmap;
3971 stp->st_openstp = open_stp;
3978 check_lock_length(u64 offset, u64 length)
3980 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3981 LOFF_OVERFLOW(offset, length)));
3984 static void get_lock_access(struct nfs4_stateid *lock_stp, u32 access)
3986 struct nfs4_file *fp = lock_stp->st_file;
3987 int oflag = nfs4_access_to_omode(access);
3989 if (test_bit(access, &lock_stp->st_access_bmap))
3991 nfs4_file_get_access(fp, oflag);
3992 __set_bit(access, &lock_stp->st_access_bmap);
3999 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4000 struct nfsd4_lock *lock)
4002 struct nfs4_stateowner *open_sop = NULL;
4003 struct nfs4_stateowner *lock_sop = NULL;
4004 struct nfs4_stateid *lock_stp;
4005 struct nfs4_file *fp;
4006 struct file *filp = NULL;
4007 struct file_lock file_lock;
4008 struct file_lock conflock;
4010 unsigned int strhashval;
4013 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4014 (long long) lock->lk_offset,
4015 (long long) lock->lk_length);
4017 if (check_lock_length(lock->lk_offset, lock->lk_length))
4018 return nfserr_inval;
4020 if ((status = fh_verify(rqstp, &cstate->current_fh,
4021 S_IFREG, NFSD_MAY_LOCK))) {
4022 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4028 if (lock->lk_is_new) {
4030 * Client indicates that this is a new lockowner.
4031 * Use open owner and open stateid to create lock owner and
4034 struct nfs4_stateid *open_stp = NULL;
4036 status = nfserr_stale_clientid;
4037 if (!nfsd4_has_session(cstate) &&
4038 STALE_CLIENTID(&lock->lk_new_clientid))
4041 /* validate and update open stateid and open seqid */
4042 status = nfs4_preprocess_seqid_op(cstate,
4043 lock->lk_new_open_seqid,
4044 &lock->lk_new_open_stateid,
4046 &lock->lk_replay_owner, &open_stp,
4050 open_sop = lock->lk_replay_owner;
4051 /* create lockowner and lock stateid */
4052 fp = open_stp->st_file;
4053 strhashval = lock_ownerstr_hashval(fp->fi_inode,
4054 open_sop->so_client->cl_clientid.cl_id,
4055 &lock->v.new.owner);
4056 /* XXX: Do we need to check for duplicate stateowners on
4057 * the same file, or should they just be allowed (and
4058 * create new stateids)? */
4059 status = nfserr_jukebox;
4060 lock_sop = alloc_init_lock_stateowner(strhashval,
4061 open_sop->so_client, open_stp, lock);
4062 if (lock_sop == NULL)
4064 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
4065 if (lock_stp == NULL)
4068 /* lock (lock owner + lock stateid) already exists */
4069 status = nfs4_preprocess_seqid_op(cstate,
4070 lock->lk_old_lock_seqid,
4071 &lock->lk_old_lock_stateid,
4073 &lock->lk_replay_owner, &lock_stp, lock);
4076 lock_sop = lock->lk_replay_owner;
4077 fp = lock_stp->st_file;
4079 /* lock->lk_replay_owner and lock_stp have been created or found */
4081 status = nfserr_grace;
4082 if (locks_in_grace() && !lock->lk_reclaim)
4084 status = nfserr_no_grace;
4085 if (!locks_in_grace() && lock->lk_reclaim)
4088 locks_init_lock(&file_lock);
4089 switch (lock->lk_type) {
4092 filp = find_readable_file(lock_stp->st_file);
4094 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4095 file_lock.fl_type = F_RDLCK;
4098 case NFS4_WRITEW_LT:
4099 filp = find_writeable_file(lock_stp->st_file);
4101 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4102 file_lock.fl_type = F_WRLCK;
4105 status = nfserr_inval;
4109 status = nfserr_openmode;
4112 file_lock.fl_owner = (fl_owner_t)lock_sop;
4113 file_lock.fl_pid = current->tgid;
4114 file_lock.fl_file = filp;
4115 file_lock.fl_flags = FL_POSIX;
4116 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4118 file_lock.fl_start = lock->lk_offset;
4119 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4120 nfs4_transform_lock_offset(&file_lock);
4123 * Try to lock the file in the VFS.
4124 * Note: locks.c uses the BKL to protect the inode's lock list.
4127 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4129 case 0: /* success! */
4130 update_stateid(&lock_stp->st_stateid);
4131 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
4135 case (EAGAIN): /* conflock holds conflicting lock */
4136 status = nfserr_denied;
4137 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4138 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4141 status = nfserr_deadlock;
4144 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4145 status = nfserrno(err);
4149 if (status && lock->lk_is_new && lock_sop)
4150 release_lockowner(lock_sop);
4151 if (lock->lk_replay_owner) {
4152 nfs4_get_stateowner(lock->lk_replay_owner);
4153 cstate->replay_owner = lock->lk_replay_owner;
4155 nfs4_unlock_state();
4160 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4161 * so we do a temporary open here just to get an open file to pass to
4162 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4165 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4170 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4173 err = vfs_test_lock(file, lock);
4182 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4183 struct nfsd4_lockt *lockt)
4185 struct inode *inode;
4186 struct file_lock file_lock;
4190 if (locks_in_grace())
4191 return nfserr_grace;
4193 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4194 return nfserr_inval;
4196 lockt->lt_stateowner = NULL;
4199 status = nfserr_stale_clientid;
4200 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4203 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4206 inode = cstate->current_fh.fh_dentry->d_inode;
4207 locks_init_lock(&file_lock);
4208 switch (lockt->lt_type) {
4211 file_lock.fl_type = F_RDLCK;
4214 case NFS4_WRITEW_LT:
4215 file_lock.fl_type = F_WRLCK;
4218 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4219 status = nfserr_inval;
4223 lockt->lt_stateowner = find_lockstateowner_str(inode,
4224 &lockt->lt_clientid, &lockt->lt_owner);
4225 if (lockt->lt_stateowner)
4226 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
4227 file_lock.fl_pid = current->tgid;
4228 file_lock.fl_flags = FL_POSIX;
4230 file_lock.fl_start = lockt->lt_offset;
4231 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4233 nfs4_transform_lock_offset(&file_lock);
4236 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4238 status = nfserrno(error);
4241 if (file_lock.fl_type != F_UNLCK) {
4242 status = nfserr_denied;
4243 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4246 nfs4_unlock_state();
4251 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4252 struct nfsd4_locku *locku)
4254 struct nfs4_stateid *stp;
4255 struct file *filp = NULL;
4256 struct file_lock file_lock;
4260 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4261 (long long) locku->lu_offset,
4262 (long long) locku->lu_length);
4264 if (check_lock_length(locku->lu_offset, locku->lu_length))
4265 return nfserr_inval;
4269 if ((status = nfs4_preprocess_seqid_op(cstate,
4273 &locku->lu_stateowner, &stp, NULL)))
4276 filp = find_any_file(stp->st_file);
4278 status = nfserr_lock_range;
4282 locks_init_lock(&file_lock);
4283 file_lock.fl_type = F_UNLCK;
4284 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
4285 file_lock.fl_pid = current->tgid;
4286 file_lock.fl_file = filp;
4287 file_lock.fl_flags = FL_POSIX;
4288 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4289 file_lock.fl_start = locku->lu_offset;
4291 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4292 nfs4_transform_lock_offset(&file_lock);
4295 * Try to unlock the file in the VFS.
4297 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4299 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4303 * OK, unlock succeeded; the only thing left to do is update the stateid.
4305 update_stateid(&stp->st_stateid);
4306 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
4309 if (locku->lu_stateowner) {
4310 nfs4_get_stateowner(locku->lu_stateowner);
4311 cstate->replay_owner = locku->lu_stateowner;
4313 nfs4_unlock_state();
4317 status = nfserrno(err);
4323 * 1: locks held by lockowner
4324 * 0: no locks held by lockowner
4327 check_for_locks(struct nfs4_file *filp, struct nfs4_stateowner *lowner)
4329 struct file_lock **flpp;
4330 struct inode *inode = filp->fi_inode;
4334 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4335 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4346 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4347 struct nfsd4_compound_state *cstate,
4348 struct nfsd4_release_lockowner *rlockowner)
4350 clientid_t *clid = &rlockowner->rl_clientid;
4351 struct nfs4_stateowner *sop;
4352 struct nfs4_stateid *stp;
4353 struct xdr_netobj *owner = &rlockowner->rl_owner;
4354 struct list_head matches;
4358 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4359 clid->cl_boot, clid->cl_id);
4361 /* XXX check for lease expiration */
4363 status = nfserr_stale_clientid;
4364 if (STALE_CLIENTID(clid))
4369 status = nfserr_locks_held;
4370 /* XXX: we're doing a linear search through all the lockowners.
4371 * Yipes! For now we'll just hope clients aren't really using
4372 * release_lockowner much, but eventually we have to fix these
4373 * data structures. */
4374 INIT_LIST_HEAD(&matches);
4375 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4376 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
4377 if (!same_owner_str(sop, owner, clid))
4379 list_for_each_entry(stp, &sop->so_stateids,
4381 if (check_for_locks(stp->st_file, sop))
4383 /* Note: so_perclient unused for lockowners,
4384 * so it's OK to fool with here. */
4385 list_add(&sop->so_perclient, &matches);
4389 /* Clients probably won't expect us to return with some (but not all)
4390 * of the lockowner state released; so don't release any until all
4391 * have been checked. */
4393 while (!list_empty(&matches)) {
4394 sop = list_entry(matches.next, struct nfs4_stateowner,
4396 /* unhash_stateowner deletes so_perclient only
4397 * for openowners. */
4398 list_del(&sop->so_perclient);
4399 release_lockowner(sop);
4402 nfs4_unlock_state();
4406 static inline struct nfs4_client_reclaim *
4409 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4413 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4415 unsigned int strhashval = clientstr_hashval(name);
4416 struct nfs4_client *clp;
4418 clp = find_confirmed_client_by_str(name, strhashval);
4423 * failure => all reset bets are off, nfserr_no_grace...
4426 nfs4_client_to_reclaim(const char *name)
4428 unsigned int strhashval;
4429 struct nfs4_client_reclaim *crp = NULL;
4431 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4432 crp = alloc_reclaim();
4435 strhashval = clientstr_hashval(name);
4436 INIT_LIST_HEAD(&crp->cr_strhash);
4437 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4438 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4439 reclaim_str_hashtbl_size++;
4444 nfs4_release_reclaim(void)
4446 struct nfs4_client_reclaim *crp = NULL;
4449 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4450 while (!list_empty(&reclaim_str_hashtbl[i])) {
4451 crp = list_entry(reclaim_str_hashtbl[i].next,
4452 struct nfs4_client_reclaim, cr_strhash);
4453 list_del(&crp->cr_strhash);
4455 reclaim_str_hashtbl_size--;
4458 BUG_ON(reclaim_str_hashtbl_size);
4462 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4463 static struct nfs4_client_reclaim *
4464 nfs4_find_reclaim_client(clientid_t *clid)
4466 unsigned int strhashval;
4467 struct nfs4_client *clp;
4468 struct nfs4_client_reclaim *crp = NULL;
4471 /* find clientid in conf_id_hashtbl */
4472 clp = find_confirmed_client(clid);
4476 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4477 clp->cl_name.len, clp->cl_name.data,
4480 /* find clp->cl_name in reclaim_str_hashtbl */
4481 strhashval = clientstr_hashval(clp->cl_recdir);
4482 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4483 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4491 * Called from OPEN. Look for clientid in reclaim list.
4494 nfs4_check_open_reclaim(clientid_t *clid)
4496 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4499 /* initialization to perform at module load time: */
4502 nfs4_state_init(void)
4506 status = nfsd4_init_slabs();
4509 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4510 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4511 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4512 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4513 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4514 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4516 for (i = 0; i < SESSION_HASH_SIZE; i++)
4517 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4518 for (i = 0; i < FILE_HASH_SIZE; i++) {
4519 INIT_LIST_HEAD(&file_hashtbl[i]);
4521 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4522 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4523 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4525 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4526 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4527 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4529 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4530 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4531 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4533 memset(&onestateid, ~0, sizeof(stateid_t));
4534 INIT_LIST_HEAD(&close_lru);
4535 INIT_LIST_HEAD(&client_lru);
4536 INIT_LIST_HEAD(&del_recall_lru);
4537 reclaim_str_hashtbl_size = 0;
4542 nfsd4_load_reboot_recovery_data(void)
4547 nfsd4_init_recdir();
4548 status = nfsd4_recdir_load();
4549 nfs4_unlock_state();
4551 printk("NFSD: Failure reading reboot recovery data\n");
4555 * Since the lifetime of a delegation isn't limited to that of an open, a
4556 * client may quite reasonably hang on to a delegation as long as it has
4557 * the inode cached. This becomes an obvious problem the first time a
4558 * client's inode cache approaches the size of the server's total memory.
4560 * For now we avoid this problem by imposing a hard limit on the number
4561 * of delegations, which varies according to the server's memory size.
4564 set_max_delegations(void)
4567 * Allow at most 4 delegations per megabyte of RAM. Quick
4568 * estimates suggest that in the worst case (where every delegation
4569 * is for a different inode), a delegation could take about 1.5K,
4570 * giving a worst case usage of about 6% of memory.
4572 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4575 /* initialization to perform when the nfsd service is started: */
4578 __nfs4_state_start(void)
4582 boot_time = get_seconds();
4583 locks_start_grace(&nfsd4_manager);
4584 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4586 ret = set_callback_cred();
4589 laundry_wq = create_singlethread_workqueue("nfsd4");
4590 if (laundry_wq == NULL)
4592 ret = nfsd4_create_callback_queue();
4594 goto out_free_laundry;
4595 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4596 set_max_delegations();
4599 destroy_workqueue(laundry_wq);
4604 nfs4_state_start(void)
4606 nfsd4_load_reboot_recovery_data();
4607 return __nfs4_state_start();
4611 __nfs4_state_shutdown(void)
4614 struct nfs4_client *clp = NULL;
4615 struct nfs4_delegation *dp = NULL;
4616 struct list_head *pos, *next, reaplist;
4618 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4619 while (!list_empty(&conf_id_hashtbl[i])) {
4620 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4623 while (!list_empty(&unconf_str_hashtbl[i])) {
4624 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4628 INIT_LIST_HEAD(&reaplist);
4629 spin_lock(&recall_lock);
4630 list_for_each_safe(pos, next, &del_recall_lru) {
4631 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4632 list_move(&dp->dl_recall_lru, &reaplist);
4634 spin_unlock(&recall_lock);
4635 list_for_each_safe(pos, next, &reaplist) {
4636 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4637 list_del_init(&dp->dl_recall_lru);
4638 unhash_delegation(dp);
4641 nfsd4_shutdown_recdir();
4645 nfs4_state_shutdown(void)
4647 cancel_delayed_work_sync(&laundromat_work);
4648 destroy_workqueue(laundry_wq);
4649 locks_end_grace(&nfsd4_manager);
4651 nfs4_release_reclaim();
4652 __nfs4_state_shutdown();
4653 nfs4_unlock_state();
4654 nfsd4_destroy_callback_queue();