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;
53 #define all_ones {{~0,~0},~0}
54 static const stateid_t one_stateid = {
56 .si_opaque = all_ones,
58 static const stateid_t zero_stateid = {
61 static const stateid_t currentstateid = {
65 static u64 current_sessionid = 1;
67 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
68 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
69 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
71 /* forward declarations */
72 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
76 /* Currently used for almost all code touching nfsv4 state: */
77 static DEFINE_MUTEX(client_mutex);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(recall_lock);
86 static struct kmem_cache *openowner_slab = NULL;
87 static struct kmem_cache *lockowner_slab = NULL;
88 static struct kmem_cache *file_slab = NULL;
89 static struct kmem_cache *stateid_slab = NULL;
90 static struct kmem_cache *deleg_slab = NULL;
95 mutex_lock(&client_mutex);
98 static void free_session(struct kref *);
100 /* Must be called under the client_lock */
101 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
103 kref_put(&ses->se_ref, free_session);
106 static void nfsd4_get_session(struct nfsd4_session *ses)
108 kref_get(&ses->se_ref);
112 nfs4_unlock_state(void)
114 mutex_unlock(&client_mutex);
118 opaque_hashval(const void *ptr, int nbytes)
120 unsigned char *cptr = (unsigned char *) ptr;
130 static struct list_head del_recall_lru;
132 static void nfsd4_free_file(struct nfs4_file *f)
134 kmem_cache_free(file_slab, f);
138 put_nfs4_file(struct nfs4_file *fi)
140 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
141 list_del(&fi->fi_hash);
142 spin_unlock(&recall_lock);
149 get_nfs4_file(struct nfs4_file *fi)
151 atomic_inc(&fi->fi_ref);
154 static int num_delegations;
155 unsigned int max_delegations;
158 * Open owner state (share locks)
161 /* hash tables for lock and open owners */
162 #define OWNER_HASH_BITS 8
163 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
164 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
166 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
170 ret = opaque_hashval(ownername->data, ownername->len);
172 return ret & OWNER_HASH_MASK;
175 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
177 /* hash table for nfs4_file */
178 #define FILE_HASH_BITS 8
179 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
181 static unsigned int file_hashval(struct inode *ino)
183 /* XXX: why are we hashing on inode pointer, anyway? */
184 return hash_ptr(ino, FILE_HASH_BITS);
187 static struct list_head file_hashtbl[FILE_HASH_SIZE];
189 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
191 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
192 atomic_inc(&fp->fi_access[oflag]);
195 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
197 if (oflag == O_RDWR) {
198 __nfs4_file_get_access(fp, O_RDONLY);
199 __nfs4_file_get_access(fp, O_WRONLY);
201 __nfs4_file_get_access(fp, oflag);
204 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
206 if (fp->fi_fds[oflag]) {
207 fput(fp->fi_fds[oflag]);
208 fp->fi_fds[oflag] = NULL;
212 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
214 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
215 nfs4_file_put_fd(fp, oflag);
217 * It's also safe to get rid of the RDWR open *if*
218 * we no longer have need of the other kind of access
219 * or if we already have the other kind of open:
221 if (fp->fi_fds[1-oflag]
222 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
223 nfs4_file_put_fd(fp, O_RDWR);
227 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
229 if (oflag == O_RDWR) {
230 __nfs4_file_put_access(fp, O_RDONLY);
231 __nfs4_file_put_access(fp, O_WRONLY);
233 __nfs4_file_put_access(fp, oflag);
236 static inline int get_new_stid(struct nfs4_stid *stid)
238 static int min_stateid = 0;
239 struct idr *stateids = &stid->sc_client->cl_stateids;
243 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
245 * Note: the necessary preallocation was done in
246 * nfs4_alloc_stateid(). The idr code caps the number of
247 * preallocations that can exist at a time, but the state lock
248 * prevents anyone from using ours before we get here:
252 * It shouldn't be a problem to reuse an opaque stateid value.
253 * I don't think it is for 4.1. But with 4.0 I worry that, for
254 * example, a stray write retransmission could be accepted by
255 * the server when it should have been rejected. Therefore,
256 * adopt a trick from the sctp code to attempt to maximize the
257 * amount of time until an id is reused, by ensuring they always
258 * "increase" (mod INT_MAX):
261 min_stateid = new_stid+1;
262 if (min_stateid == INT_MAX)
267 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
269 stateid_t *s = &stid->sc_stateid;
272 stid->sc_type = type;
273 stid->sc_client = cl;
274 s->si_opaque.so_clid = cl->cl_clientid;
275 new_id = get_new_stid(stid);
276 s->si_opaque.so_id = (u32)new_id;
277 /* Will be incremented before return to client: */
278 s->si_generation = 0;
281 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
283 struct idr *stateids = &cl->cl_stateids;
285 if (!idr_pre_get(stateids, GFP_KERNEL))
288 * Note: if we fail here (or any time between now and the time
289 * we actually get the new idr), we won't need to undo the idr
290 * preallocation, since the idr code caps the number of
291 * preallocated entries.
293 return kmem_cache_alloc(slab, GFP_KERNEL);
296 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
298 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
301 static struct nfs4_delegation *
302 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
304 struct nfs4_delegation *dp;
305 struct nfs4_file *fp = stp->st_file;
307 dprintk("NFSD alloc_init_deleg\n");
309 * Major work on the lease subsystem (for example, to support
310 * calbacks on stat) will be required before we can support
311 * write delegations properly.
313 if (type != NFS4_OPEN_DELEGATE_READ)
315 if (fp->fi_had_conflict)
317 if (num_delegations > max_delegations)
319 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
322 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
324 * delegation seqid's are never incremented. The 4.1 special
325 * meaning of seqid 0 isn't meaningful, really, but let's avoid
326 * 0 anyway just for consistency and use 1:
328 dp->dl_stid.sc_stateid.si_generation = 1;
330 INIT_LIST_HEAD(&dp->dl_perfile);
331 INIT_LIST_HEAD(&dp->dl_perclnt);
332 INIT_LIST_HEAD(&dp->dl_recall_lru);
336 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
338 atomic_set(&dp->dl_count, 1);
339 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
344 nfs4_put_delegation(struct nfs4_delegation *dp)
346 if (atomic_dec_and_test(&dp->dl_count)) {
347 dprintk("NFSD: freeing dp %p\n",dp);
348 put_nfs4_file(dp->dl_file);
349 kmem_cache_free(deleg_slab, dp);
354 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
356 if (atomic_dec_and_test(&fp->fi_delegees)) {
357 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
359 fput(fp->fi_deleg_file);
360 fp->fi_deleg_file = NULL;
364 static void unhash_stid(struct nfs4_stid *s)
366 struct idr *stateids = &s->sc_client->cl_stateids;
368 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
371 /* Called under the state lock. */
373 unhash_delegation(struct nfs4_delegation *dp)
375 unhash_stid(&dp->dl_stid);
376 list_del_init(&dp->dl_perclnt);
377 spin_lock(&recall_lock);
378 list_del_init(&dp->dl_perfile);
379 list_del_init(&dp->dl_recall_lru);
380 spin_unlock(&recall_lock);
381 nfs4_put_deleg_lease(dp->dl_file);
382 nfs4_put_delegation(dp);
389 /* client_lock protects the client lru list and session hash table */
390 static DEFINE_SPINLOCK(client_lock);
392 /* Hash tables for nfs4_clientid state */
393 #define CLIENT_HASH_BITS 4
394 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
395 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
397 static unsigned int clientid_hashval(u32 id)
399 return id & CLIENT_HASH_MASK;
402 static unsigned int clientstr_hashval(const char *name)
404 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
408 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
409 * used in reboot/reset lease grace period processing
411 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
412 * setclientid_confirmed info.
414 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
417 * client_lru holds client queue ordered by nfs4_client.cl_time
420 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
421 * for last close replay.
423 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
424 static int reclaim_str_hashtbl_size = 0;
425 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
426 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
427 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
428 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
429 static struct list_head client_lru;
430 static struct list_head close_lru;
433 * We store the NONE, READ, WRITE, and BOTH bits separately in the
434 * st_{access,deny}_bmap field of the stateid, in order to track not
435 * only what share bits are currently in force, but also what
436 * combinations of share bits previous opens have used. This allows us
437 * to enforce the recommendation of rfc 3530 14.2.19 that the server
438 * return an error if the client attempt to downgrade to a combination
439 * of share bits not explicable by closing some of its previous opens.
441 * XXX: This enforcement is actually incomplete, since we don't keep
442 * track of access/deny bit combinations; so, e.g., we allow:
444 * OPEN allow read, deny write
445 * OPEN allow both, deny none
446 * DOWNGRADE allow read, deny none
448 * which we should reject.
451 set_access(unsigned int *access, unsigned long bmap) {
455 for (i = 1; i < 4; i++) {
456 if (test_bit(i, &bmap))
462 set_deny(unsigned int *deny, unsigned long bmap) {
466 for (i = 0; i < 4; i++) {
467 if (test_bit(i, &bmap))
473 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
474 unsigned int access, deny;
476 set_access(&access, stp->st_access_bmap);
477 set_deny(&deny, stp->st_deny_bmap);
478 if ((access & open->op_share_deny) || (deny & open->op_share_access))
483 static int nfs4_access_to_omode(u32 access)
485 switch (access & NFS4_SHARE_ACCESS_BOTH) {
486 case NFS4_SHARE_ACCESS_READ:
488 case NFS4_SHARE_ACCESS_WRITE:
490 case NFS4_SHARE_ACCESS_BOTH:
496 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
498 list_del(&stp->st_perfile);
499 list_del(&stp->st_perstateowner);
502 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
506 if (stp->st_access_bmap) {
507 for (i = 1; i < 4; i++) {
508 if (test_bit(i, &stp->st_access_bmap))
509 nfs4_file_put_access(stp->st_file,
510 nfs4_access_to_omode(i));
511 __clear_bit(i, &stp->st_access_bmap);
514 put_nfs4_file(stp->st_file);
518 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
520 kmem_cache_free(stateid_slab, stp);
523 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
527 unhash_generic_stateid(stp);
528 unhash_stid(&stp->st_stid);
529 file = find_any_file(stp->st_file);
531 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
532 close_generic_stateid(stp);
533 free_generic_stateid(stp);
536 static void unhash_lockowner(struct nfs4_lockowner *lo)
538 struct nfs4_ol_stateid *stp;
540 list_del(&lo->lo_owner.so_strhash);
541 list_del(&lo->lo_perstateid);
542 list_del(&lo->lo_owner_ino_hash);
543 while (!list_empty(&lo->lo_owner.so_stateids)) {
544 stp = list_first_entry(&lo->lo_owner.so_stateids,
545 struct nfs4_ol_stateid, st_perstateowner);
546 release_lock_stateid(stp);
550 static void release_lockowner(struct nfs4_lockowner *lo)
552 unhash_lockowner(lo);
553 nfs4_free_lockowner(lo);
557 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
559 struct nfs4_lockowner *lo;
561 while (!list_empty(&open_stp->st_lockowners)) {
562 lo = list_entry(open_stp->st_lockowners.next,
563 struct nfs4_lockowner, lo_perstateid);
564 release_lockowner(lo);
568 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
570 unhash_generic_stateid(stp);
571 release_stateid_lockowners(stp);
572 close_generic_stateid(stp);
575 static void release_open_stateid(struct nfs4_ol_stateid *stp)
577 unhash_open_stateid(stp);
578 unhash_stid(&stp->st_stid);
579 free_generic_stateid(stp);
582 static void unhash_openowner(struct nfs4_openowner *oo)
584 struct nfs4_ol_stateid *stp;
586 list_del(&oo->oo_owner.so_strhash);
587 list_del(&oo->oo_perclient);
588 while (!list_empty(&oo->oo_owner.so_stateids)) {
589 stp = list_first_entry(&oo->oo_owner.so_stateids,
590 struct nfs4_ol_stateid, st_perstateowner);
591 release_open_stateid(stp);
595 static void release_last_closed_stateid(struct nfs4_openowner *oo)
597 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
600 unhash_stid(&s->st_stid);
601 free_generic_stateid(s);
602 oo->oo_last_closed_stid = NULL;
606 static void release_openowner(struct nfs4_openowner *oo)
608 unhash_openowner(oo);
609 list_del(&oo->oo_close_lru);
610 release_last_closed_stateid(oo);
611 nfs4_free_openowner(oo);
614 #define SESSION_HASH_SIZE 512
615 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
618 hash_sessionid(struct nfs4_sessionid *sessionid)
620 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
622 return sid->sequence % SESSION_HASH_SIZE;
626 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
628 u32 *ptr = (u32 *)(&sessionid->data[0]);
629 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
633 gen_sessionid(struct nfsd4_session *ses)
635 struct nfs4_client *clp = ses->se_client;
636 struct nfsd4_sessionid *sid;
638 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
639 sid->clientid = clp->cl_clientid;
640 sid->sequence = current_sessionid++;
645 * The protocol defines ca_maxresponssize_cached to include the size of
646 * the rpc header, but all we need to cache is the data starting after
647 * the end of the initial SEQUENCE operation--the rest we regenerate
648 * each time. Therefore we can advertise a ca_maxresponssize_cached
649 * value that is the number of bytes in our cache plus a few additional
650 * bytes. In order to stay on the safe side, and not promise more than
651 * we can cache, those additional bytes must be the minimum possible: 24
652 * bytes of rpc header (xid through accept state, with AUTH_NULL
653 * verifier), 12 for the compound header (with zero-length tag), and 44
654 * for the SEQUENCE op response:
656 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
659 free_session_slots(struct nfsd4_session *ses)
663 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
664 kfree(ses->se_slots[i]);
668 * We don't actually need to cache the rpc and session headers, so we
669 * can allocate a little less for each slot:
671 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
673 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
676 static int nfsd4_sanitize_slot_size(u32 size)
678 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
679 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
685 * XXX: If we run out of reserved DRC memory we could (up to a point)
686 * re-negotiate active sessions and reduce their slot usage to make
687 * room for new connections. For now we just fail the create session.
689 static int nfsd4_get_drc_mem(int slotsize, u32 num)
693 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
695 spin_lock(&nfsd_drc_lock);
696 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
697 nfsd_drc_max_mem - nfsd_drc_mem_used);
698 num = min_t(int, num, avail / slotsize);
699 nfsd_drc_mem_used += num * slotsize;
700 spin_unlock(&nfsd_drc_lock);
705 static void nfsd4_put_drc_mem(int slotsize, int num)
707 spin_lock(&nfsd_drc_lock);
708 nfsd_drc_mem_used -= slotsize * num;
709 spin_unlock(&nfsd_drc_lock);
712 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
714 struct nfsd4_session *new;
717 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
718 + sizeof(struct nfsd4_session) > PAGE_SIZE);
719 mem = numslots * sizeof(struct nfsd4_slot *);
721 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
724 /* allocate each struct nfsd4_slot and data cache in one piece */
725 for (i = 0; i < numslots; i++) {
726 mem = sizeof(struct nfsd4_slot) + slotsize;
727 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
728 if (!new->se_slots[i])
734 kfree(new->se_slots[i]);
739 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
741 u32 maxrpc = nfsd_serv->sv_max_mesg;
743 new->maxreqs = numslots;
744 new->maxresp_cached = min_t(u32, req->maxresp_cached,
745 slotsize + NFSD_MIN_HDR_SEQ_SZ);
746 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
747 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
748 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
751 static void free_conn(struct nfsd4_conn *c)
753 svc_xprt_put(c->cn_xprt);
757 static void nfsd4_conn_lost(struct svc_xpt_user *u)
759 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
760 struct nfs4_client *clp = c->cn_session->se_client;
762 spin_lock(&clp->cl_lock);
763 if (!list_empty(&c->cn_persession)) {
764 list_del(&c->cn_persession);
767 spin_unlock(&clp->cl_lock);
768 nfsd4_probe_callback(clp);
771 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
773 struct nfsd4_conn *conn;
775 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
778 svc_xprt_get(rqstp->rq_xprt);
779 conn->cn_xprt = rqstp->rq_xprt;
780 conn->cn_flags = flags;
781 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
785 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
787 conn->cn_session = ses;
788 list_add(&conn->cn_persession, &ses->se_conns);
791 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
793 struct nfs4_client *clp = ses->se_client;
795 spin_lock(&clp->cl_lock);
796 __nfsd4_hash_conn(conn, ses);
797 spin_unlock(&clp->cl_lock);
800 static int nfsd4_register_conn(struct nfsd4_conn *conn)
802 conn->cn_xpt_user.callback = nfsd4_conn_lost;
803 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
806 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
808 struct nfsd4_conn *conn;
811 conn = alloc_conn(rqstp, dir);
813 return nfserr_jukebox;
814 nfsd4_hash_conn(conn, ses);
815 ret = nfsd4_register_conn(conn);
817 /* oops; xprt is already down: */
818 nfsd4_conn_lost(&conn->cn_xpt_user);
822 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
824 u32 dir = NFS4_CDFC4_FORE;
826 if (ses->se_flags & SESSION4_BACK_CHAN)
827 dir |= NFS4_CDFC4_BACK;
829 return nfsd4_new_conn(rqstp, ses, dir);
832 /* must be called under client_lock */
833 static void nfsd4_del_conns(struct nfsd4_session *s)
835 struct nfs4_client *clp = s->se_client;
836 struct nfsd4_conn *c;
838 spin_lock(&clp->cl_lock);
839 while (!list_empty(&s->se_conns)) {
840 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
841 list_del_init(&c->cn_persession);
842 spin_unlock(&clp->cl_lock);
844 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
847 spin_lock(&clp->cl_lock);
849 spin_unlock(&clp->cl_lock);
852 static void free_session(struct kref *kref)
854 struct nfsd4_session *ses;
857 BUG_ON(!spin_is_locked(&client_lock));
858 ses = container_of(kref, struct nfsd4_session, se_ref);
859 nfsd4_del_conns(ses);
860 spin_lock(&nfsd_drc_lock);
861 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
862 nfsd_drc_mem_used -= mem;
863 spin_unlock(&nfsd_drc_lock);
864 free_session_slots(ses);
868 void nfsd4_put_session(struct nfsd4_session *ses)
870 spin_lock(&client_lock);
871 nfsd4_put_session_locked(ses);
872 spin_unlock(&client_lock);
875 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
877 struct nfsd4_session *new;
878 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
879 int numslots, slotsize;
884 * Note decreasing slot size below client's request may
885 * make it difficult for client to function correctly, whereas
886 * decreasing the number of slots will (just?) affect
887 * performance. When short on memory we therefore prefer to
888 * decrease number of slots instead of their size.
890 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
891 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
895 new = alloc_session(slotsize, numslots);
897 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
900 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
902 new->se_client = clp;
905 INIT_LIST_HEAD(&new->se_conns);
907 new->se_cb_seq_nr = 1;
908 new->se_flags = cses->flags;
909 new->se_cb_prog = cses->callback_prog;
910 kref_init(&new->se_ref);
911 idx = hash_sessionid(&new->se_sessionid);
912 spin_lock(&client_lock);
913 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
914 spin_lock(&clp->cl_lock);
915 list_add(&new->se_perclnt, &clp->cl_sessions);
916 spin_unlock(&clp->cl_lock);
917 spin_unlock(&client_lock);
919 status = nfsd4_new_conn_from_crses(rqstp, new);
920 /* whoops: benny points out, status is ignored! (err, or bogus) */
922 spin_lock(&client_lock);
923 free_session(&new->se_ref);
924 spin_unlock(&client_lock);
927 if (cses->flags & SESSION4_BACK_CHAN) {
928 struct sockaddr *sa = svc_addr(rqstp);
930 * This is a little silly; with sessions there's no real
931 * use for the callback address. Use the peer address
932 * as a reasonable default for now, but consider fixing
933 * the rpc client not to require an address in the
936 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
937 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
939 nfsd4_probe_callback(clp);
943 /* caller must hold client_lock */
944 static struct nfsd4_session *
945 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
947 struct nfsd4_session *elem;
950 dump_sessionid(__func__, sessionid);
951 idx = hash_sessionid(sessionid);
952 /* Search in the appropriate list */
953 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
954 if (!memcmp(elem->se_sessionid.data, sessionid->data,
955 NFS4_MAX_SESSIONID_LEN)) {
960 dprintk("%s: session not found\n", __func__);
964 /* caller must hold client_lock */
966 unhash_session(struct nfsd4_session *ses)
968 list_del(&ses->se_hash);
969 spin_lock(&ses->se_client->cl_lock);
970 list_del(&ses->se_perclnt);
971 spin_unlock(&ses->se_client->cl_lock);
974 /* must be called under the client_lock */
976 renew_client_locked(struct nfs4_client *clp)
978 if (is_client_expired(clp)) {
979 dprintk("%s: client (clientid %08x/%08x) already expired\n",
981 clp->cl_clientid.cl_boot,
982 clp->cl_clientid.cl_id);
986 dprintk("renewing client (clientid %08x/%08x)\n",
987 clp->cl_clientid.cl_boot,
988 clp->cl_clientid.cl_id);
989 list_move_tail(&clp->cl_lru, &client_lru);
990 clp->cl_time = get_seconds();
994 renew_client(struct nfs4_client *clp)
996 spin_lock(&client_lock);
997 renew_client_locked(clp);
998 spin_unlock(&client_lock);
1001 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1003 STALE_CLIENTID(clientid_t *clid)
1005 if (clid->cl_boot == boot_time)
1007 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1008 clid->cl_boot, clid->cl_id, boot_time);
1013 * XXX Should we use a slab cache ?
1014 * This type of memory management is somewhat inefficient, but we use it
1015 * anyway since SETCLIENTID is not a common operation.
1017 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1019 struct nfs4_client *clp;
1021 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1024 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1025 if (clp->cl_name.data == NULL) {
1029 clp->cl_name.len = name.len;
1034 free_client(struct nfs4_client *clp)
1036 BUG_ON(!spin_is_locked(&client_lock));
1037 while (!list_empty(&clp->cl_sessions)) {
1038 struct nfsd4_session *ses;
1039 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1041 list_del(&ses->se_perclnt);
1042 nfsd4_put_session_locked(ses);
1044 if (clp->cl_cred.cr_group_info)
1045 put_group_info(clp->cl_cred.cr_group_info);
1046 kfree(clp->cl_principal);
1047 kfree(clp->cl_name.data);
1052 release_session_client(struct nfsd4_session *session)
1054 struct nfs4_client *clp = session->se_client;
1056 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1058 if (is_client_expired(clp)) {
1060 session->se_client = NULL;
1062 renew_client_locked(clp);
1063 spin_unlock(&client_lock);
1066 /* must be called under the client_lock */
1068 unhash_client_locked(struct nfs4_client *clp)
1070 struct nfsd4_session *ses;
1072 mark_client_expired(clp);
1073 list_del(&clp->cl_lru);
1074 spin_lock(&clp->cl_lock);
1075 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1076 list_del_init(&ses->se_hash);
1077 spin_unlock(&clp->cl_lock);
1081 expire_client(struct nfs4_client *clp)
1083 struct nfs4_openowner *oo;
1084 struct nfs4_delegation *dp;
1085 struct list_head reaplist;
1087 INIT_LIST_HEAD(&reaplist);
1088 spin_lock(&recall_lock);
1089 while (!list_empty(&clp->cl_delegations)) {
1090 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1091 list_del_init(&dp->dl_perclnt);
1092 list_move(&dp->dl_recall_lru, &reaplist);
1094 spin_unlock(&recall_lock);
1095 while (!list_empty(&reaplist)) {
1096 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1097 unhash_delegation(dp);
1099 while (!list_empty(&clp->cl_openowners)) {
1100 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1101 release_openowner(oo);
1103 nfsd4_shutdown_callback(clp);
1104 if (clp->cl_cb_conn.cb_xprt)
1105 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1106 list_del(&clp->cl_idhash);
1107 list_del(&clp->cl_strhash);
1108 spin_lock(&client_lock);
1109 unhash_client_locked(clp);
1110 if (atomic_read(&clp->cl_refcount) == 0)
1112 spin_unlock(&client_lock);
1115 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1117 memcpy(target->cl_verifier.data, source->data,
1118 sizeof(target->cl_verifier.data));
1121 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1123 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1124 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1127 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1129 target->cr_uid = source->cr_uid;
1130 target->cr_gid = source->cr_gid;
1131 target->cr_group_info = source->cr_group_info;
1132 get_group_info(target->cr_group_info);
1135 static int same_name(const char *n1, const char *n2)
1137 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1141 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1143 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1147 same_clid(clientid_t *cl1, clientid_t *cl2)
1149 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1152 /* XXX what about NGROUP */
1154 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1156 return cr1->cr_uid == cr2->cr_uid;
1159 static void gen_clid(struct nfs4_client *clp)
1161 static u32 current_clientid = 1;
1163 clp->cl_clientid.cl_boot = boot_time;
1164 clp->cl_clientid.cl_id = current_clientid++;
1167 static void gen_confirm(struct nfs4_client *clp)
1172 p = (u32 *)clp->cl_confirm.data;
1173 *p++ = get_seconds();
1177 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1179 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1182 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1184 struct nfs4_stid *s;
1186 s = find_stateid(cl, t);
1189 if (typemask & s->sc_type)
1194 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1195 struct svc_rqst *rqstp, nfs4_verifier *verf)
1197 struct nfs4_client *clp;
1198 struct sockaddr *sa = svc_addr(rqstp);
1201 clp = alloc_client(name);
1205 INIT_LIST_HEAD(&clp->cl_sessions);
1207 princ = svc_gss_principal(rqstp);
1209 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1210 if (clp->cl_principal == NULL) {
1211 spin_lock(&client_lock);
1213 spin_unlock(&client_lock);
1218 idr_init(&clp->cl_stateids);
1219 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1220 atomic_set(&clp->cl_refcount, 0);
1221 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1222 INIT_LIST_HEAD(&clp->cl_idhash);
1223 INIT_LIST_HEAD(&clp->cl_strhash);
1224 INIT_LIST_HEAD(&clp->cl_openowners);
1225 INIT_LIST_HEAD(&clp->cl_delegations);
1226 INIT_LIST_HEAD(&clp->cl_lru);
1227 INIT_LIST_HEAD(&clp->cl_callbacks);
1228 spin_lock_init(&clp->cl_lock);
1229 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1230 clp->cl_time = get_seconds();
1231 clear_bit(0, &clp->cl_cb_slot_busy);
1232 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1233 copy_verf(clp, verf);
1234 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1235 clp->cl_flavor = rqstp->rq_flavor;
1236 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1238 clp->cl_cb_session = NULL;
1243 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1245 unsigned int idhashval;
1247 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1248 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1249 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1254 move_to_confirmed(struct nfs4_client *clp)
1256 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1257 unsigned int strhashval;
1259 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1260 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1261 strhashval = clientstr_hashval(clp->cl_recdir);
1262 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1266 static struct nfs4_client *
1267 find_confirmed_client(clientid_t *clid)
1269 struct nfs4_client *clp;
1270 unsigned int idhashval = clientid_hashval(clid->cl_id);
1272 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1273 if (same_clid(&clp->cl_clientid, clid)) {
1281 static struct nfs4_client *
1282 find_unconfirmed_client(clientid_t *clid)
1284 struct nfs4_client *clp;
1285 unsigned int idhashval = clientid_hashval(clid->cl_id);
1287 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1288 if (same_clid(&clp->cl_clientid, clid))
1294 static bool clp_used_exchangeid(struct nfs4_client *clp)
1296 return clp->cl_exchange_flags != 0;
1299 static struct nfs4_client *
1300 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1302 struct nfs4_client *clp;
1304 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1305 if (same_name(clp->cl_recdir, dname))
1311 static struct nfs4_client *
1312 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1314 struct nfs4_client *clp;
1316 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1317 if (same_name(clp->cl_recdir, dname))
1324 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1326 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1327 struct sockaddr *sa = svc_addr(rqstp);
1328 u32 scopeid = rpc_get_scope_id(sa);
1329 unsigned short expected_family;
1331 /* Currently, we only support tcp and tcp6 for the callback channel */
1332 if (se->se_callback_netid_len == 3 &&
1333 !memcmp(se->se_callback_netid_val, "tcp", 3))
1334 expected_family = AF_INET;
1335 else if (se->se_callback_netid_len == 4 &&
1336 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1337 expected_family = AF_INET6;
1341 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1342 se->se_callback_addr_len,
1343 (struct sockaddr *)&conn->cb_addr,
1344 sizeof(conn->cb_addr));
1346 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1349 if (conn->cb_addr.ss_family == AF_INET6)
1350 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1352 conn->cb_prog = se->se_callback_prog;
1353 conn->cb_ident = se->se_callback_ident;
1354 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1357 conn->cb_addr.ss_family = AF_UNSPEC;
1358 conn->cb_addrlen = 0;
1359 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1360 "will not receive delegations\n",
1361 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1367 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1370 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1372 struct nfsd4_slot *slot = resp->cstate.slot;
1375 dprintk("--> %s slot %p\n", __func__, slot);
1377 slot->sl_opcnt = resp->opcnt;
1378 slot->sl_status = resp->cstate.status;
1380 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1381 if (nfsd4_not_cached(resp)) {
1382 slot->sl_datalen = 0;
1385 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1386 base = (char *)resp->cstate.datap -
1387 (char *)resp->xbuf->head[0].iov_base;
1388 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1390 WARN("%s: sessions DRC could not cache compound\n", __func__);
1395 * Encode the replay sequence operation from the slot values.
1396 * If cachethis is FALSE encode the uncached rep error on the next
1397 * operation which sets resp->p and increments resp->opcnt for
1398 * nfs4svc_encode_compoundres.
1402 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1403 struct nfsd4_compoundres *resp)
1405 struct nfsd4_op *op;
1406 struct nfsd4_slot *slot = resp->cstate.slot;
1408 /* Encode the replayed sequence operation */
1409 op = &args->ops[resp->opcnt - 1];
1410 nfsd4_encode_operation(resp, op);
1412 /* Return nfserr_retry_uncached_rep in next operation. */
1413 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1414 op = &args->ops[resp->opcnt++];
1415 op->status = nfserr_retry_uncached_rep;
1416 nfsd4_encode_operation(resp, op);
1422 * The sequence operation is not cached because we can use the slot and
1426 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1427 struct nfsd4_sequence *seq)
1429 struct nfsd4_slot *slot = resp->cstate.slot;
1432 dprintk("--> %s slot %p\n", __func__, slot);
1434 /* Either returns 0 or nfserr_retry_uncached */
1435 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1436 if (status == nfserr_retry_uncached_rep)
1439 /* The sequence operation has been encoded, cstate->datap set. */
1440 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1442 resp->opcnt = slot->sl_opcnt;
1443 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1444 status = slot->sl_status;
1450 * Set the exchange_id flags returned by the server.
1453 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1455 /* pNFS is not supported */
1456 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1458 /* Referrals are supported, Migration is not. */
1459 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1461 /* set the wire flags to return to client. */
1462 clid->flags = new->cl_exchange_flags;
1466 nfsd4_exchange_id(struct svc_rqst *rqstp,
1467 struct nfsd4_compound_state *cstate,
1468 struct nfsd4_exchange_id *exid)
1470 struct nfs4_client *unconf, *conf, *new;
1472 unsigned int strhashval;
1473 char dname[HEXDIR_LEN];
1474 char addr_str[INET6_ADDRSTRLEN];
1475 nfs4_verifier verf = exid->verifier;
1476 struct sockaddr *sa = svc_addr(rqstp);
1478 rpc_ntop(sa, addr_str, sizeof(addr_str));
1479 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1480 "ip_addr=%s flags %x, spa_how %d\n",
1481 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1482 addr_str, exid->flags, exid->spa_how);
1484 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1485 return nfserr_inval;
1487 /* Currently only support SP4_NONE */
1488 switch (exid->spa_how) {
1492 return nfserr_serverfault;
1494 BUG(); /* checked by xdr code */
1496 return nfserr_serverfault; /* no excuse :-/ */
1499 status = nfs4_make_rec_clidname(dname, &exid->clname);
1504 strhashval = clientstr_hashval(dname);
1509 conf = find_confirmed_client_by_str(dname, strhashval);
1511 if (!clp_used_exchangeid(conf)) {
1512 status = nfserr_clid_inuse; /* XXX: ? */
1515 if (!same_verf(&verf, &conf->cl_verifier)) {
1516 /* 18.35.4 case 8 */
1517 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1518 status = nfserr_not_same;
1521 /* Client reboot: destroy old state */
1522 expire_client(conf);
1525 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1526 /* 18.35.4 case 9 */
1527 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1528 status = nfserr_perm;
1531 expire_client(conf);
1535 * Set bit when the owner id and verifier map to an already
1536 * confirmed client id (18.35.3).
1538 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1541 * Falling into 18.35.4 case 2, possible router replay.
1542 * Leave confirmed record intact and return same result.
1544 copy_verf(conf, &verf);
1549 /* 18.35.4 case 7 */
1550 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1551 status = nfserr_noent;
1555 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1558 * Possible retry or client restart. Per 18.35.4 case 4,
1559 * a new unconfirmed record should be generated regardless
1560 * of whether any properties have changed.
1562 expire_client(unconf);
1567 new = create_client(exid->clname, dname, rqstp, &verf);
1569 status = nfserr_jukebox;
1574 add_to_unconfirmed(new, strhashval);
1576 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1577 exid->clientid.cl_id = new->cl_clientid.cl_id;
1580 nfsd4_set_ex_flags(new, exid);
1582 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1583 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1587 nfs4_unlock_state();
1589 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1594 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1596 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1599 /* The slot is in use, and no response has been sent. */
1601 if (seqid == slot_seqid)
1602 return nfserr_jukebox;
1604 return nfserr_seq_misordered;
1606 /* Note unsigned 32-bit arithmetic handles wraparound: */
1607 if (likely(seqid == slot_seqid + 1))
1609 if (seqid == slot_seqid)
1610 return nfserr_replay_cache;
1611 return nfserr_seq_misordered;
1615 * Cache the create session result into the create session single DRC
1616 * slot cache by saving the xdr structure. sl_seqid has been set.
1617 * Do this for solo or embedded create session operations.
1620 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1621 struct nfsd4_clid_slot *slot, int nfserr)
1623 slot->sl_status = nfserr;
1624 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1628 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1629 struct nfsd4_clid_slot *slot)
1631 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1632 return slot->sl_status;
1635 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1636 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1637 1 + /* MIN tag is length with zero, only length */ \
1638 3 + /* version, opcount, opcode */ \
1639 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1640 /* seqid, slotID, slotID, cache */ \
1641 4 ) * sizeof(__be32))
1643 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1644 2 + /* verifier: AUTH_NULL, length 0 */\
1646 1 + /* MIN tag is length with zero, only length */ \
1647 3 + /* opcount, opcode, opstatus*/ \
1648 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1649 /* seqid, slotID, slotID, slotID, status */ \
1650 5 ) * sizeof(__be32))
1652 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1654 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1655 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1659 nfsd4_create_session(struct svc_rqst *rqstp,
1660 struct nfsd4_compound_state *cstate,
1661 struct nfsd4_create_session *cr_ses)
1663 struct sockaddr *sa = svc_addr(rqstp);
1664 struct nfs4_client *conf, *unconf;
1665 struct nfsd4_session *new;
1666 struct nfsd4_clid_slot *cs_slot = NULL;
1667 bool confirm_me = false;
1670 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1671 return nfserr_inval;
1674 unconf = find_unconfirmed_client(&cr_ses->clientid);
1675 conf = find_confirmed_client(&cr_ses->clientid);
1678 cs_slot = &conf->cl_cs_slot;
1679 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1680 if (status == nfserr_replay_cache) {
1681 dprintk("Got a create_session replay! seqid= %d\n",
1683 /* Return the cached reply status */
1684 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1686 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1687 status = nfserr_seq_misordered;
1688 dprintk("Sequence misordered!\n");
1689 dprintk("Expected seqid= %d but got seqid= %d\n",
1690 cs_slot->sl_seqid, cr_ses->seqid);
1693 } else if (unconf) {
1694 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1695 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1696 status = nfserr_clid_inuse;
1700 cs_slot = &unconf->cl_cs_slot;
1701 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1703 /* an unconfirmed replay returns misordered */
1704 status = nfserr_seq_misordered;
1711 status = nfserr_stale_clientid;
1716 * XXX: we should probably set this at creation time, and check
1717 * for consistent minorversion use throughout:
1719 conf->cl_minorversion = 1;
1721 * We do not support RDMA or persistent sessions
1723 cr_ses->flags &= ~SESSION4_PERSIST;
1724 cr_ses->flags &= ~SESSION4_RDMA;
1726 status = nfserr_toosmall;
1727 if (check_forechannel_attrs(cr_ses->fore_channel))
1730 status = nfserr_jukebox;
1731 new = alloc_init_session(rqstp, conf, cr_ses);
1735 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1736 NFS4_MAX_SESSIONID_LEN);
1737 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1738 sizeof(struct nfsd4_channel_attrs));
1739 cs_slot->sl_seqid++;
1740 cr_ses->seqid = cs_slot->sl_seqid;
1742 /* cache solo and embedded create sessions under the state lock */
1743 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1745 move_to_confirmed(conf);
1747 nfs4_unlock_state();
1748 dprintk("%s returns %d\n", __func__, ntohl(status));
1752 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1754 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1755 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1757 return argp->opcnt == resp->opcnt;
1760 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1763 case NFS4_CDFC4_FORE:
1764 case NFS4_CDFC4_BACK:
1766 case NFS4_CDFC4_FORE_OR_BOTH:
1767 case NFS4_CDFC4_BACK_OR_BOTH:
1768 *dir = NFS4_CDFC4_BOTH;
1771 return nfserr_inval;
1774 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1775 struct nfsd4_compound_state *cstate,
1776 struct nfsd4_bind_conn_to_session *bcts)
1780 if (!nfsd4_last_compound_op(rqstp))
1781 return nfserr_not_only_op;
1782 spin_lock(&client_lock);
1783 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1784 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1785 * client_lock iself: */
1786 if (cstate->session) {
1787 nfsd4_get_session(cstate->session);
1788 atomic_inc(&cstate->session->se_client->cl_refcount);
1790 spin_unlock(&client_lock);
1791 if (!cstate->session)
1792 return nfserr_badsession;
1794 status = nfsd4_map_bcts_dir(&bcts->dir);
1796 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1800 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1804 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1808 nfsd4_destroy_session(struct svc_rqst *r,
1809 struct nfsd4_compound_state *cstate,
1810 struct nfsd4_destroy_session *sessionid)
1812 struct nfsd4_session *ses;
1813 u32 status = nfserr_badsession;
1816 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1817 * - Should we return nfserr_back_chan_busy if waiting for
1818 * callbacks on to-be-destroyed session?
1819 * - Do we need to clear any callback info from previous session?
1822 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1823 if (!nfsd4_last_compound_op(r))
1824 return nfserr_not_only_op;
1826 dump_sessionid(__func__, &sessionid->sessionid);
1827 spin_lock(&client_lock);
1828 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1830 spin_unlock(&client_lock);
1834 unhash_session(ses);
1835 spin_unlock(&client_lock);
1838 nfsd4_probe_callback_sync(ses->se_client);
1839 nfs4_unlock_state();
1841 spin_lock(&client_lock);
1842 nfsd4_del_conns(ses);
1843 nfsd4_put_session_locked(ses);
1844 spin_unlock(&client_lock);
1847 dprintk("%s returns %d\n", __func__, ntohl(status));
1851 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1853 struct nfsd4_conn *c;
1855 list_for_each_entry(c, &s->se_conns, cn_persession) {
1856 if (c->cn_xprt == xpt) {
1863 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1865 struct nfs4_client *clp = ses->se_client;
1866 struct nfsd4_conn *c;
1869 spin_lock(&clp->cl_lock);
1870 c = __nfsd4_find_conn(new->cn_xprt, ses);
1872 spin_unlock(&clp->cl_lock);
1876 __nfsd4_hash_conn(new, ses);
1877 spin_unlock(&clp->cl_lock);
1878 ret = nfsd4_register_conn(new);
1880 /* oops; xprt is already down: */
1881 nfsd4_conn_lost(&new->cn_xpt_user);
1885 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1887 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1889 return args->opcnt > session->se_fchannel.maxops;
1892 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1893 struct nfsd4_session *session)
1895 struct xdr_buf *xb = &rqstp->rq_arg;
1897 return xb->len > session->se_fchannel.maxreq_sz;
1901 nfsd4_sequence(struct svc_rqst *rqstp,
1902 struct nfsd4_compound_state *cstate,
1903 struct nfsd4_sequence *seq)
1905 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1906 struct nfsd4_session *session;
1907 struct nfsd4_slot *slot;
1908 struct nfsd4_conn *conn;
1911 if (resp->opcnt != 1)
1912 return nfserr_sequence_pos;
1915 * Will be either used or freed by nfsd4_sequence_check_conn
1918 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1920 return nfserr_jukebox;
1922 spin_lock(&client_lock);
1923 status = nfserr_badsession;
1924 session = find_in_sessionid_hashtbl(&seq->sessionid);
1928 status = nfserr_too_many_ops;
1929 if (nfsd4_session_too_many_ops(rqstp, session))
1932 status = nfserr_req_too_big;
1933 if (nfsd4_request_too_big(rqstp, session))
1936 status = nfserr_badslot;
1937 if (seq->slotid >= session->se_fchannel.maxreqs)
1940 slot = session->se_slots[seq->slotid];
1941 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1943 /* We do not negotiate the number of slots yet, so set the
1944 * maxslots to the session maxreqs which is used to encode
1945 * sr_highest_slotid and the sr_target_slot id to maxslots */
1946 seq->maxslots = session->se_fchannel.maxreqs;
1948 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
1949 slot->sl_flags & NFSD4_SLOT_INUSE);
1950 if (status == nfserr_replay_cache) {
1951 status = nfserr_seq_misordered;
1952 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
1954 cstate->slot = slot;
1955 cstate->session = session;
1956 /* Return the cached reply status and set cstate->status
1957 * for nfsd4_proc_compound processing */
1958 status = nfsd4_replay_cache_entry(resp, seq);
1959 cstate->status = nfserr_replay_cache;
1965 nfsd4_sequence_check_conn(conn, session);
1968 /* Success! bump slot seqid */
1969 slot->sl_seqid = seq->seqid;
1970 slot->sl_flags |= NFSD4_SLOT_INUSE;
1972 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
1974 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
1976 cstate->slot = slot;
1977 cstate->session = session;
1980 /* Hold a session reference until done processing the compound. */
1981 if (cstate->session) {
1982 struct nfs4_client *clp = session->se_client;
1984 nfsd4_get_session(cstate->session);
1985 atomic_inc(&clp->cl_refcount);
1986 switch (clp->cl_cb_state) {
1988 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
1990 case NFSD4_CB_FAULT:
1991 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
1994 seq->status_flags = 0;
1998 spin_unlock(&client_lock);
1999 dprintk("%s: return %d\n", __func__, ntohl(status));
2003 static inline bool has_resources(struct nfs4_client *clp)
2005 return !list_empty(&clp->cl_openowners)
2006 || !list_empty(&clp->cl_delegations)
2007 || !list_empty(&clp->cl_sessions);
2011 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2013 struct nfs4_client *conf, *unconf, *clp;
2017 unconf = find_unconfirmed_client(&dc->clientid);
2018 conf = find_confirmed_client(&dc->clientid);
2023 if (!is_client_expired(conf) && has_resources(conf)) {
2024 status = nfserr_clientid_busy;
2028 /* rfc5661 18.50.3 */
2029 if (cstate->session && conf == cstate->session->se_client) {
2030 status = nfserr_clientid_busy;
2036 status = nfserr_stale_clientid;
2042 nfs4_unlock_state();
2043 dprintk("%s return %d\n", __func__, ntohl(status));
2048 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2052 if (rc->rca_one_fs) {
2053 if (!cstate->current_fh.fh_dentry)
2054 return nfserr_nofilehandle;
2056 * We don't take advantage of the rca_one_fs case.
2057 * That's OK, it's optional, we can safely ignore it.
2063 status = nfserr_complete_already;
2064 if (cstate->session->se_client->cl_firststate)
2067 status = nfserr_stale_clientid;
2068 if (is_client_expired(cstate->session->se_client))
2070 * The following error isn't really legal.
2071 * But we only get here if the client just explicitly
2072 * destroyed the client. Surely it no longer cares what
2073 * error it gets back on an operation for the dead
2079 nfsd4_create_clid_dir(cstate->session->se_client);
2081 nfs4_unlock_state();
2086 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2087 struct nfsd4_setclientid *setclid)
2089 struct xdr_netobj clname = setclid->se_name;
2090 nfs4_verifier clverifier = setclid->se_verf;
2091 unsigned int strhashval;
2092 struct nfs4_client *conf, *unconf, *new;
2094 char dname[HEXDIR_LEN];
2096 status = nfs4_make_rec_clidname(dname, &clname);
2101 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2102 * We get here on a DRC miss.
2105 strhashval = clientstr_hashval(dname);
2108 conf = find_confirmed_client_by_str(dname, strhashval);
2110 /* RFC 3530 14.2.33 CASE 0: */
2111 status = nfserr_clid_inuse;
2112 if (clp_used_exchangeid(conf))
2114 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2115 char addr_str[INET6_ADDRSTRLEN];
2116 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2118 dprintk("NFSD: setclientid: string in use by client "
2119 "at %s\n", addr_str);
2124 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2125 * has a description of SETCLIENTID request processing consisting
2126 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2128 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2129 status = nfserr_jukebox;
2132 * RFC 3530 14.2.33 CASE 4:
2133 * placed first, because it is the normal case
2136 expire_client(unconf);
2137 new = create_client(clname, dname, rqstp, &clverifier);
2141 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2143 * RFC 3530 14.2.33 CASE 1:
2144 * probable callback update
2147 /* Note this is removing unconfirmed {*x***},
2148 * which is stronger than RFC recommended {vxc**}.
2149 * This has the advantage that there is at most
2150 * one {*x***} in either list at any time.
2152 expire_client(unconf);
2154 new = create_client(clname, dname, rqstp, &clverifier);
2157 copy_clid(new, conf);
2158 } else if (!unconf) {
2160 * RFC 3530 14.2.33 CASE 2:
2161 * probable client reboot; state will be removed if
2164 new = create_client(clname, dname, rqstp, &clverifier);
2170 * RFC 3530 14.2.33 CASE 3:
2171 * probable client reboot; state will be removed if
2174 expire_client(unconf);
2175 new = create_client(clname, dname, rqstp, &clverifier);
2181 * XXX: we should probably set this at creation time, and check
2182 * for consistent minorversion use throughout:
2184 new->cl_minorversion = 0;
2185 gen_callback(new, setclid, rqstp);
2186 add_to_unconfirmed(new, strhashval);
2187 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2188 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2189 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2192 nfs4_unlock_state();
2198 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2199 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2200 * bullets, labeled as CASE1 - CASE4 below.
2203 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2204 struct nfsd4_compound_state *cstate,
2205 struct nfsd4_setclientid_confirm *setclientid_confirm)
2207 struct sockaddr *sa = svc_addr(rqstp);
2208 struct nfs4_client *conf, *unconf;
2209 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2210 clientid_t * clid = &setclientid_confirm->sc_clientid;
2213 if (STALE_CLIENTID(clid))
2214 return nfserr_stale_clientid;
2216 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2217 * We get here on a DRC miss.
2222 conf = find_confirmed_client(clid);
2223 unconf = find_unconfirmed_client(clid);
2225 status = nfserr_clid_inuse;
2226 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2228 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2232 * section 14.2.34 of RFC 3530 has a description of
2233 * SETCLIENTID_CONFIRM request processing consisting
2234 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2236 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2238 * RFC 3530 14.2.34 CASE 1:
2241 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2242 status = nfserr_clid_inuse;
2244 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2245 nfsd4_probe_callback(conf);
2246 expire_client(unconf);
2250 } else if (conf && !unconf) {
2252 * RFC 3530 14.2.34 CASE 2:
2253 * probable retransmitted request; play it safe and
2256 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2257 status = nfserr_clid_inuse;
2260 } else if (!conf && unconf
2261 && same_verf(&unconf->cl_confirm, &confirm)) {
2263 * RFC 3530 14.2.34 CASE 3:
2264 * Normal case; new or rebooted client:
2266 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2267 status = nfserr_clid_inuse;
2270 clientstr_hashval(unconf->cl_recdir);
2271 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2274 nfsd4_remove_clid_dir(conf);
2275 expire_client(conf);
2277 move_to_confirmed(unconf);
2279 nfsd4_probe_callback(conf);
2282 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2283 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2286 * RFC 3530 14.2.34 CASE 4:
2287 * Client probably hasn't noticed that we rebooted yet.
2289 status = nfserr_stale_clientid;
2291 /* check that we have hit one of the cases...*/
2292 status = nfserr_clid_inuse;
2295 nfs4_unlock_state();
2299 static struct nfs4_file *nfsd4_alloc_file(void)
2301 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2304 /* OPEN Share state helper functions */
2305 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2307 unsigned int hashval = file_hashval(ino);
2309 atomic_set(&fp->fi_ref, 1);
2310 INIT_LIST_HEAD(&fp->fi_hash);
2311 INIT_LIST_HEAD(&fp->fi_stateids);
2312 INIT_LIST_HEAD(&fp->fi_delegations);
2313 fp->fi_inode = igrab(ino);
2314 fp->fi_had_conflict = false;
2315 fp->fi_lease = NULL;
2316 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2317 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2318 spin_lock(&recall_lock);
2319 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2320 spin_unlock(&recall_lock);
2324 nfsd4_free_slab(struct kmem_cache **slab)
2328 kmem_cache_destroy(*slab);
2333 nfsd4_free_slabs(void)
2335 nfsd4_free_slab(&openowner_slab);
2336 nfsd4_free_slab(&lockowner_slab);
2337 nfsd4_free_slab(&file_slab);
2338 nfsd4_free_slab(&stateid_slab);
2339 nfsd4_free_slab(&deleg_slab);
2343 nfsd4_init_slabs(void)
2345 openowner_slab = kmem_cache_create("nfsd4_openowners",
2346 sizeof(struct nfs4_openowner), 0, 0, NULL);
2347 if (openowner_slab == NULL)
2349 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2350 sizeof(struct nfs4_openowner), 0, 0, NULL);
2351 if (lockowner_slab == NULL)
2353 file_slab = kmem_cache_create("nfsd4_files",
2354 sizeof(struct nfs4_file), 0, 0, NULL);
2355 if (file_slab == NULL)
2357 stateid_slab = kmem_cache_create("nfsd4_stateids",
2358 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2359 if (stateid_slab == NULL)
2361 deleg_slab = kmem_cache_create("nfsd4_delegations",
2362 sizeof(struct nfs4_delegation), 0, 0, NULL);
2363 if (deleg_slab == NULL)
2368 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2372 void nfs4_free_openowner(struct nfs4_openowner *oo)
2374 kfree(oo->oo_owner.so_owner.data);
2375 kmem_cache_free(openowner_slab, oo);
2378 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2380 kfree(lo->lo_owner.so_owner.data);
2381 kmem_cache_free(lockowner_slab, lo);
2384 static void init_nfs4_replay(struct nfs4_replay *rp)
2386 rp->rp_status = nfserr_serverfault;
2388 rp->rp_buf = rp->rp_ibuf;
2391 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2393 struct nfs4_stateowner *sop;
2395 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2399 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2400 if (!sop->so_owner.data) {
2401 kmem_cache_free(slab, sop);
2404 sop->so_owner.len = owner->len;
2406 INIT_LIST_HEAD(&sop->so_stateids);
2407 sop->so_client = clp;
2408 init_nfs4_replay(&sop->so_replay);
2412 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2414 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2415 list_add(&oo->oo_perclient, &clp->cl_openowners);
2418 static struct nfs4_openowner *
2419 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2420 struct nfs4_openowner *oo;
2422 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2425 oo->oo_owner.so_is_open_owner = 1;
2426 oo->oo_owner.so_seqid = open->op_seqid;
2427 oo->oo_flags = NFS4_OO_NEW;
2429 oo->oo_last_closed_stid = NULL;
2430 INIT_LIST_HEAD(&oo->oo_close_lru);
2431 hash_openowner(oo, clp, strhashval);
2435 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2436 struct nfs4_openowner *oo = open->op_openowner;
2437 struct nfs4_client *clp = oo->oo_owner.so_client;
2439 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2440 INIT_LIST_HEAD(&stp->st_lockowners);
2441 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2442 list_add(&stp->st_perfile, &fp->fi_stateids);
2443 stp->st_stateowner = &oo->oo_owner;
2446 stp->st_access_bmap = 0;
2447 stp->st_deny_bmap = 0;
2448 __set_bit(open->op_share_access, &stp->st_access_bmap);
2449 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2450 stp->st_openstp = NULL;
2454 move_to_close_lru(struct nfs4_openowner *oo)
2456 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2458 list_move_tail(&oo->oo_close_lru, &close_lru);
2459 oo->oo_time = get_seconds();
2463 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2466 return (sop->so_owner.len == owner->len) &&
2467 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2468 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2471 static struct nfs4_openowner *
2472 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2474 struct nfs4_stateowner *so;
2475 struct nfs4_openowner *oo;
2477 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2478 if (!so->so_is_open_owner)
2480 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2482 renew_client(oo->oo_owner.so_client);
2489 /* search file_hashtbl[] for file */
2490 static struct nfs4_file *
2491 find_file(struct inode *ino)
2493 unsigned int hashval = file_hashval(ino);
2494 struct nfs4_file *fp;
2496 spin_lock(&recall_lock);
2497 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2498 if (fp->fi_inode == ino) {
2500 spin_unlock(&recall_lock);
2504 spin_unlock(&recall_lock);
2509 * Called to check deny when READ with all zero stateid or
2510 * WRITE with all zero or all one stateid
2513 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2515 struct inode *ino = current_fh->fh_dentry->d_inode;
2516 struct nfs4_file *fp;
2517 struct nfs4_ol_stateid *stp;
2520 dprintk("NFSD: nfs4_share_conflict\n");
2522 fp = find_file(ino);
2525 ret = nfserr_locked;
2526 /* Search for conflicting share reservations */
2527 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2528 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2529 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2538 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2540 /* We're assuming the state code never drops its reference
2541 * without first removing the lease. Since we're in this lease
2542 * callback (and since the lease code is serialized by the kernel
2543 * lock) we know the server hasn't removed the lease yet, we know
2544 * it's safe to take a reference: */
2545 atomic_inc(&dp->dl_count);
2547 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2549 /* only place dl_time is set. protected by lock_flocks*/
2550 dp->dl_time = get_seconds();
2552 nfsd4_cb_recall(dp);
2555 /* Called from break_lease() with lock_flocks() held. */
2556 static void nfsd_break_deleg_cb(struct file_lock *fl)
2558 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2559 struct nfs4_delegation *dp;
2562 /* We assume break_lease is only called once per lease: */
2563 BUG_ON(fp->fi_had_conflict);
2565 * We don't want the locks code to timeout the lease for us;
2566 * we'll remove it ourself if a delegation isn't returned
2569 fl->fl_break_time = 0;
2571 spin_lock(&recall_lock);
2572 fp->fi_had_conflict = true;
2573 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2574 nfsd_break_one_deleg(dp);
2575 spin_unlock(&recall_lock);
2579 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2582 return lease_modify(onlist, arg);
2587 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2588 .lm_break = nfsd_break_deleg_cb,
2589 .lm_change = nfsd_change_deleg_cb,
2592 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2594 if (nfsd4_has_session(cstate))
2596 if (seqid == so->so_seqid - 1)
2597 return nfserr_replay_me;
2598 if (seqid == so->so_seqid)
2600 return nfserr_bad_seqid;
2604 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2605 struct nfsd4_open *open)
2607 clientid_t *clientid = &open->op_clientid;
2608 struct nfs4_client *clp = NULL;
2609 unsigned int strhashval;
2610 struct nfs4_openowner *oo = NULL;
2613 if (STALE_CLIENTID(&open->op_clientid))
2614 return nfserr_stale_clientid;
2616 * In case we need it later, after we've already created the
2617 * file and don't want to risk a further failure:
2619 open->op_file = nfsd4_alloc_file();
2620 if (open->op_file == NULL)
2621 return nfserr_jukebox;
2623 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2624 oo = find_openstateowner_str(strhashval, open);
2625 open->op_openowner = oo;
2627 clp = find_confirmed_client(clientid);
2629 return nfserr_expired;
2632 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2633 /* Replace unconfirmed owners without checking for replay. */
2634 clp = oo->oo_owner.so_client;
2635 release_openowner(oo);
2636 open->op_openowner = NULL;
2639 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2642 clp = oo->oo_owner.so_client;
2645 oo = alloc_init_open_stateowner(strhashval, clp, open);
2647 return nfserr_jukebox;
2648 open->op_openowner = oo;
2650 open->op_stp = nfs4_alloc_stateid(clp);
2652 return nfserr_jukebox;
2656 static inline __be32
2657 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2659 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2660 return nfserr_openmode;
2665 static int share_access_to_flags(u32 share_access)
2667 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2670 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2672 struct nfs4_stid *ret;
2674 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2677 return delegstateid(ret);
2680 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2682 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2683 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2687 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2688 struct nfs4_delegation **dp)
2691 __be32 status = nfserr_bad_stateid;
2693 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2696 flags = share_access_to_flags(open->op_share_access);
2697 status = nfs4_check_delegmode(*dp, flags);
2701 if (!nfsd4_is_deleg_cur(open))
2705 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2710 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2712 struct nfs4_ol_stateid *local;
2713 struct nfs4_openowner *oo = open->op_openowner;
2715 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2716 /* ignore lock owners */
2717 if (local->st_stateowner->so_is_open_owner == 0)
2719 /* remember if we have seen this open owner */
2720 if (local->st_stateowner == &oo->oo_owner)
2722 /* check for conflicting share reservations */
2723 if (!test_share(local, open))
2724 return nfserr_share_denied;
2729 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2731 kmem_cache_free(stateid_slab, s);
2734 static inline int nfs4_access_to_access(u32 nfs4_access)
2738 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2739 flags |= NFSD_MAY_READ;
2740 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2741 flags |= NFSD_MAY_WRITE;
2745 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2746 struct svc_fh *cur_fh, struct nfsd4_open *open)
2749 int oflag = nfs4_access_to_omode(open->op_share_access);
2750 int access = nfs4_access_to_access(open->op_share_access);
2752 if (!fp->fi_fds[oflag]) {
2753 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2754 &fp->fi_fds[oflag]);
2758 nfs4_file_get_access(fp, oflag);
2763 static inline __be32
2764 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2765 struct nfsd4_open *open)
2767 struct iattr iattr = {
2768 .ia_valid = ATTR_SIZE,
2771 if (!open->op_truncate)
2773 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2774 return nfserr_inval;
2775 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2779 nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
2781 u32 op_share_access = open->op_share_access;
2785 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2787 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2791 status = nfsd4_truncate(rqstp, cur_fh, open);
2794 int oflag = nfs4_access_to_omode(op_share_access);
2795 nfs4_file_put_access(fp, oflag);
2799 /* remember the open */
2800 __set_bit(op_share_access, &stp->st_access_bmap);
2801 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2808 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2810 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2812 * On a 4.1+ client, we don't create a state record for a client
2813 * until it performs RECLAIM_COMPLETE:
2816 open->op_openowner->oo_owner.so_client->cl_firststate = 1;
2819 /* Should we give out recallable state?: */
2820 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2822 if (clp->cl_cb_state == NFSD4_CB_UP)
2825 * In the sessions case, since we don't have to establish a
2826 * separate connection for callbacks, we assume it's OK
2827 * until we hear otherwise:
2829 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2832 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2834 struct file_lock *fl;
2836 fl = locks_alloc_lock();
2839 locks_init_lock(fl);
2840 fl->fl_lmops = &nfsd_lease_mng_ops;
2841 fl->fl_flags = FL_LEASE;
2842 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2843 fl->fl_end = OFFSET_MAX;
2844 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2845 fl->fl_pid = current->tgid;
2849 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2851 struct nfs4_file *fp = dp->dl_file;
2852 struct file_lock *fl;
2855 fl = nfs4_alloc_init_lease(dp, flag);
2858 fl->fl_file = find_readable_file(fp);
2859 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2860 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2862 list_del_init(&dp->dl_perclnt);
2863 locks_free_lock(fl);
2867 fp->fi_deleg_file = fl->fl_file;
2868 get_file(fp->fi_deleg_file);
2869 atomic_set(&fp->fi_delegees, 1);
2870 list_add(&dp->dl_perfile, &fp->fi_delegations);
2874 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2876 struct nfs4_file *fp = dp->dl_file;
2879 return nfs4_setlease(dp, flag);
2880 spin_lock(&recall_lock);
2881 if (fp->fi_had_conflict) {
2882 spin_unlock(&recall_lock);
2885 atomic_inc(&fp->fi_delegees);
2886 list_add(&dp->dl_perfile, &fp->fi_delegations);
2887 spin_unlock(&recall_lock);
2888 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2892 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2894 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2895 if (status == -EAGAIN)
2896 open->op_why_no_deleg = WND4_CONTENTION;
2898 open->op_why_no_deleg = WND4_RESOURCE;
2899 switch (open->op_deleg_want) {
2900 case NFS4_SHARE_WANT_READ_DELEG:
2901 case NFS4_SHARE_WANT_WRITE_DELEG:
2902 case NFS4_SHARE_WANT_ANY_DELEG:
2904 case NFS4_SHARE_WANT_CANCEL:
2905 open->op_why_no_deleg = WND4_CANCELLED;
2907 case NFS4_SHARE_WANT_NO_DELEG:
2908 BUG(); /* not supposed to get here */
2914 * Attempt to hand out a delegation.
2917 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2919 struct nfs4_delegation *dp;
2920 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2922 int status = 0, flag = 0;
2924 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2925 flag = NFS4_OPEN_DELEGATE_NONE;
2926 open->op_recall = 0;
2927 switch (open->op_claim_type) {
2928 case NFS4_OPEN_CLAIM_PREVIOUS:
2930 open->op_recall = 1;
2931 flag = open->op_delegate_type;
2932 if (flag == NFS4_OPEN_DELEGATE_NONE)
2935 case NFS4_OPEN_CLAIM_NULL:
2936 /* Let's not give out any delegations till everyone's
2937 * had the chance to reclaim theirs.... */
2938 if (locks_in_grace())
2940 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2942 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2943 flag = NFS4_OPEN_DELEGATE_WRITE;
2945 flag = NFS4_OPEN_DELEGATE_READ;
2951 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2954 status = nfs4_set_delegation(dp, flag);
2958 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2960 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2961 STATEID_VAL(&dp->dl_stid.sc_stateid));
2963 open->op_delegate_type = flag;
2964 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2965 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2966 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2967 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2969 /* 4.1 client asking for a delegation? */
2970 if (open->op_deleg_want)
2971 nfsd4_open_deleg_none_ext(open, status);
2975 nfs4_put_delegation(dp);
2977 flag = NFS4_OPEN_DELEGATE_NONE;
2981 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2982 struct nfs4_delegation *dp)
2984 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2985 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2986 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2987 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2988 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2989 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2990 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2991 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2993 /* Otherwise the client must be confused wanting a delegation
2994 * it already has, therefore we don't return
2995 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3000 * called with nfs4_lock_state() held.
3003 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3005 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3006 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3007 struct nfs4_file *fp = NULL;
3008 struct inode *ino = current_fh->fh_dentry->d_inode;
3009 struct nfs4_ol_stateid *stp = NULL;
3010 struct nfs4_delegation *dp = NULL;
3014 * Lookup file; if found, lookup stateid and check open request,
3015 * and check for delegations in the process of being recalled.
3016 * If not found, create the nfs4_file struct
3018 fp = find_file(ino);
3020 if ((status = nfs4_check_open(fp, open, &stp)))
3022 status = nfs4_check_deleg(cl, fp, open, &dp);
3026 status = nfserr_bad_stateid;
3027 if (nfsd4_is_deleg_cur(open))
3029 status = nfserr_jukebox;
3031 open->op_file = NULL;
3032 nfsd4_init_file(fp, ino);
3036 * OPEN the file, or upgrade an existing OPEN.
3037 * If truncate fails, the OPEN fails.
3040 /* Stateid was found, this is an OPEN upgrade */
3041 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3045 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3049 open->op_stp = NULL;
3050 init_open_stateid(stp, fp, open);
3051 status = nfsd4_truncate(rqstp, current_fh, open);
3053 release_open_stateid(stp);
3057 update_stateid(&stp->st_stid.sc_stateid);
3058 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3060 if (nfsd4_has_session(&resp->cstate)) {
3061 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3063 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3064 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3065 open->op_why_no_deleg = WND4_NOT_WANTED;
3071 * Attempt to hand out a delegation. No error return, because the
3072 * OPEN succeeds even if we fail.
3074 nfs4_open_delegation(current_fh, open, stp);
3078 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3079 STATEID_VAL(&stp->st_stid.sc_stateid));
3081 /* 4.1 client trying to upgrade/downgrade delegation? */
3082 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3083 open->op_deleg_want)
3084 nfsd4_deleg_xgrade_none_ext(open, dp);
3088 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3089 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3091 * To finish the open response, we just need to set the rflags.
3093 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3094 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3095 !nfsd4_has_session(&resp->cstate))
3096 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3101 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3103 if (open->op_openowner) {
3104 struct nfs4_openowner *oo = open->op_openowner;
3106 if (!list_empty(&oo->oo_owner.so_stateids))
3107 list_del_init(&oo->oo_close_lru);
3108 if (oo->oo_flags & NFS4_OO_NEW) {
3110 release_openowner(oo);
3111 open->op_openowner = NULL;
3113 oo->oo_flags &= ~NFS4_OO_NEW;
3117 nfsd4_free_file(open->op_file);
3119 nfs4_free_stateid(open->op_stp);
3123 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3126 struct nfs4_client *clp;
3130 dprintk("process_renew(%08x/%08x): starting\n",
3131 clid->cl_boot, clid->cl_id);
3132 status = nfserr_stale_clientid;
3133 if (STALE_CLIENTID(clid))
3135 clp = find_confirmed_client(clid);
3136 status = nfserr_expired;
3138 /* We assume the client took too long to RENEW. */
3139 dprintk("nfsd4_renew: clientid not found!\n");
3142 status = nfserr_cb_path_down;
3143 if (!list_empty(&clp->cl_delegations)
3144 && clp->cl_cb_state != NFSD4_CB_UP)
3148 nfs4_unlock_state();
3152 static struct lock_manager nfsd4_manager = {
3156 nfsd4_end_grace(void)
3158 dprintk("NFSD: end of grace period\n");
3159 nfsd4_recdir_purge_old();
3160 locks_end_grace(&nfsd4_manager);
3162 * Now that every NFSv4 client has had the chance to recover and
3163 * to see the (possibly new, possibly shorter) lease time, we
3164 * can safely set the next grace time to the current lease time:
3166 nfsd4_grace = nfsd4_lease;
3170 nfs4_laundromat(void)
3172 struct nfs4_client *clp;
3173 struct nfs4_openowner *oo;
3174 struct nfs4_delegation *dp;
3175 struct list_head *pos, *next, reaplist;
3176 time_t cutoff = get_seconds() - nfsd4_lease;
3177 time_t t, clientid_val = nfsd4_lease;
3178 time_t u, test_val = nfsd4_lease;
3182 dprintk("NFSD: laundromat service - starting\n");
3183 if (locks_in_grace())
3185 INIT_LIST_HEAD(&reaplist);
3186 spin_lock(&client_lock);
3187 list_for_each_safe(pos, next, &client_lru) {
3188 clp = list_entry(pos, struct nfs4_client, cl_lru);
3189 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3190 t = clp->cl_time - cutoff;
3191 if (clientid_val > t)
3195 if (atomic_read(&clp->cl_refcount)) {
3196 dprintk("NFSD: client in use (clientid %08x)\n",
3197 clp->cl_clientid.cl_id);
3200 unhash_client_locked(clp);
3201 list_add(&clp->cl_lru, &reaplist);
3203 spin_unlock(&client_lock);
3204 list_for_each_safe(pos, next, &reaplist) {
3205 clp = list_entry(pos, struct nfs4_client, cl_lru);
3206 dprintk("NFSD: purging unused client (clientid %08x)\n",
3207 clp->cl_clientid.cl_id);
3208 nfsd4_remove_clid_dir(clp);
3211 spin_lock(&recall_lock);
3212 list_for_each_safe(pos, next, &del_recall_lru) {
3213 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3214 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3215 u = dp->dl_time - cutoff;
3220 list_move(&dp->dl_recall_lru, &reaplist);
3222 spin_unlock(&recall_lock);
3223 list_for_each_safe(pos, next, &reaplist) {
3224 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3225 unhash_delegation(dp);
3227 test_val = nfsd4_lease;
3228 list_for_each_safe(pos, next, &close_lru) {
3229 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3230 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3231 u = oo->oo_time - cutoff;
3236 release_openowner(oo);
3238 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3239 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3240 nfs4_unlock_state();
3241 return clientid_val;
3244 static struct workqueue_struct *laundry_wq;
3245 static void laundromat_main(struct work_struct *);
3246 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3249 laundromat_main(struct work_struct *not_used)
3253 t = nfs4_laundromat();
3254 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3255 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3258 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3260 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3261 return nfserr_bad_stateid;
3266 STALE_STATEID(stateid_t *stateid)
3268 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3270 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3271 STATEID_VAL(stateid));
3276 access_permit_read(unsigned long access_bmap)
3278 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3279 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3280 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3284 access_permit_write(unsigned long access_bmap)
3286 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3287 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3291 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3293 __be32 status = nfserr_openmode;
3295 /* For lock stateid's, we test the parent open, not the lock: */
3296 if (stp->st_openstp)
3297 stp = stp->st_openstp;
3298 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3300 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3307 static inline __be32
3308 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3310 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3312 else if (locks_in_grace()) {
3313 /* Answer in remaining cases depends on existence of
3314 * conflicting state; so we must wait out the grace period. */
3315 return nfserr_grace;
3316 } else if (flags & WR_STATE)
3317 return nfs4_share_conflict(current_fh,
3318 NFS4_SHARE_DENY_WRITE);
3319 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3320 return nfs4_share_conflict(current_fh,
3321 NFS4_SHARE_DENY_READ);
3325 * Allow READ/WRITE during grace period on recovered state only for files
3326 * that are not able to provide mandatory locking.
3329 grace_disallows_io(struct inode *inode)
3331 return locks_in_grace() && mandatory_lock(inode);
3334 /* Returns true iff a is later than b: */
3335 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3337 return (s32)a->si_generation - (s32)b->si_generation > 0;
3340 static int check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3343 * When sessions are used the stateid generation number is ignored
3346 if (has_session && in->si_generation == 0)
3349 if (in->si_generation == ref->si_generation)
3352 /* If the client sends us a stateid from the future, it's buggy: */
3353 if (stateid_generation_after(in, ref))
3354 return nfserr_bad_stateid;
3356 * However, we could see a stateid from the past, even from a
3357 * non-buggy client. For example, if the client sends a lock
3358 * while some IO is outstanding, the lock may bump si_generation
3359 * while the IO is still in flight. The client could avoid that
3360 * situation by waiting for responses on all the IO requests,
3361 * but better performance may result in retrying IO that
3362 * receives an old_stateid error if requests are rarely
3363 * reordered in flight:
3365 return nfserr_old_stateid;
3368 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3370 struct nfs4_stid *s;
3371 struct nfs4_ol_stateid *ols;
3374 if (STALE_STATEID(stateid))
3375 return nfserr_stale_stateid;
3377 s = find_stateid(cl, stateid);
3379 return nfserr_stale_stateid;
3380 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3383 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3385 ols = openlockstateid(s);
3386 if (ols->st_stateowner->so_is_open_owner
3387 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3388 return nfserr_bad_stateid;
3392 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3394 struct nfs4_client *cl;
3396 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3397 return nfserr_bad_stateid;
3398 if (STALE_STATEID(stateid))
3399 return nfserr_stale_stateid;
3400 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3402 return nfserr_expired;
3403 *s = find_stateid_by_type(cl, stateid, typemask);
3405 return nfserr_bad_stateid;
3411 * Checks for stateid operations
3414 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3415 stateid_t *stateid, int flags, struct file **filpp)
3417 struct nfs4_stid *s;
3418 struct nfs4_ol_stateid *stp = NULL;
3419 struct nfs4_delegation *dp = NULL;
3420 struct svc_fh *current_fh = &cstate->current_fh;
3421 struct inode *ino = current_fh->fh_dentry->d_inode;
3427 if (grace_disallows_io(ino))
3428 return nfserr_grace;
3430 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3431 return check_special_stateids(current_fh, stateid, flags);
3433 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3436 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3439 switch (s->sc_type) {
3440 case NFS4_DELEG_STID:
3441 dp = delegstateid(s);
3442 status = nfs4_check_delegmode(dp, flags);
3446 *filpp = dp->dl_file->fi_deleg_file;
3450 case NFS4_OPEN_STID:
3451 case NFS4_LOCK_STID:
3452 stp = openlockstateid(s);
3453 status = nfs4_check_fh(current_fh, stp);
3456 if (stp->st_stateowner->so_is_open_owner
3457 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3459 status = nfs4_check_openmode(stp, flags);
3463 if (flags & RD_STATE)
3464 *filpp = find_readable_file(stp->st_file);
3466 *filpp = find_writeable_file(stp->st_file);
3470 return nfserr_bad_stateid;
3478 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3480 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3481 return nfserr_locks_held;
3482 release_lock_stateid(stp);
3487 * Test if the stateid is valid
3490 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3491 struct nfsd4_test_stateid *test_stateid)
3493 struct nfsd4_test_stateid_id *stateid;
3494 struct nfs4_client *cl = cstate->session->se_client;
3497 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3498 stateid->ts_id_status = nfs4_validate_stateid(cl, &stateid->ts_id_stateid);
3499 nfs4_unlock_state();
3505 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3506 struct nfsd4_free_stateid *free_stateid)
3508 stateid_t *stateid = &free_stateid->fr_stateid;
3509 struct nfs4_stid *s;
3510 struct nfs4_client *cl = cstate->session->se_client;
3511 __be32 ret = nfserr_bad_stateid;
3514 s = find_stateid(cl, stateid);
3517 switch (s->sc_type) {
3518 case NFS4_DELEG_STID:
3519 ret = nfserr_locks_held;
3521 case NFS4_OPEN_STID:
3522 case NFS4_LOCK_STID:
3523 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3526 if (s->sc_type == NFS4_LOCK_STID)
3527 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3529 ret = nfserr_locks_held;
3532 ret = nfserr_bad_stateid;
3535 nfs4_unlock_state();
3542 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3543 RD_STATE : WR_STATE;
3546 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3548 struct svc_fh *current_fh = &cstate->current_fh;
3549 struct nfs4_stateowner *sop = stp->st_stateowner;
3552 status = nfsd4_check_seqid(cstate, sop, seqid);
3555 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3557 * "Closed" stateid's exist *only* to return
3558 * nfserr_replay_me from the previous step.
3560 return nfserr_bad_stateid;
3561 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3564 return nfs4_check_fh(current_fh, stp);
3568 * Checks for sequence id mutating operations.
3571 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3572 stateid_t *stateid, char typemask,
3573 struct nfs4_ol_stateid **stpp)
3576 struct nfs4_stid *s;
3578 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3579 seqid, STATEID_VAL(stateid));
3582 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3585 *stpp = openlockstateid(s);
3586 cstate->replay_owner = (*stpp)->st_stateowner;
3588 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3591 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3594 struct nfs4_openowner *oo;
3596 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3597 NFS4_OPEN_STID, stpp);
3600 oo = openowner((*stpp)->st_stateowner);
3601 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3602 return nfserr_bad_stateid;
3607 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3608 struct nfsd4_open_confirm *oc)
3611 struct nfs4_openowner *oo;
3612 struct nfs4_ol_stateid *stp;
3614 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3615 (int)cstate->current_fh.fh_dentry->d_name.len,
3616 cstate->current_fh.fh_dentry->d_name.name);
3618 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3624 status = nfs4_preprocess_seqid_op(cstate,
3625 oc->oc_seqid, &oc->oc_req_stateid,
3626 NFS4_OPEN_STID, &stp);
3629 oo = openowner(stp->st_stateowner);
3630 status = nfserr_bad_stateid;
3631 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3633 oo->oo_flags |= NFS4_OO_CONFIRMED;
3634 update_stateid(&stp->st_stid.sc_stateid);
3635 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3636 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3637 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3639 nfsd4_create_clid_dir(oo->oo_owner.so_client);
3642 if (!cstate->replay_owner)
3643 nfs4_unlock_state();
3647 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3649 if (!test_bit(access, &stp->st_access_bmap))
3651 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3652 __clear_bit(access, &stp->st_access_bmap);
3655 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3657 switch (to_access) {
3658 case NFS4_SHARE_ACCESS_READ:
3659 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3660 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3662 case NFS4_SHARE_ACCESS_WRITE:
3663 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3664 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3666 case NFS4_SHARE_ACCESS_BOTH:
3674 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3677 for (i = 0; i < 4; i++) {
3678 if ((i & deny) != i)
3679 __clear_bit(i, bmap);
3684 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3685 struct nfsd4_compound_state *cstate,
3686 struct nfsd4_open_downgrade *od)
3689 struct nfs4_ol_stateid *stp;
3691 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3692 (int)cstate->current_fh.fh_dentry->d_name.len,
3693 cstate->current_fh.fh_dentry->d_name.name);
3695 /* We don't yet support WANT bits: */
3696 if (od->od_deleg_want)
3697 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3701 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3702 &od->od_stateid, &stp);
3705 status = nfserr_inval;
3706 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3707 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3708 stp->st_access_bmap, od->od_share_access);
3711 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3712 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3713 stp->st_deny_bmap, od->od_share_deny);
3716 nfs4_stateid_downgrade(stp, od->od_share_access);
3718 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3720 update_stateid(&stp->st_stid.sc_stateid);
3721 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3724 if (!cstate->replay_owner)
3725 nfs4_unlock_state();
3729 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3731 struct nfs4_openowner *oo;
3732 struct nfs4_ol_stateid *s;
3734 if (!so->so_is_open_owner)
3737 s = oo->oo_last_closed_stid;
3740 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3741 /* Release the last_closed_stid on the next seqid bump: */
3742 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3745 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3746 release_last_closed_stateid(oo);
3749 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3751 unhash_open_stateid(s);
3752 s->st_stid.sc_type = NFS4_CLOSED_STID;
3756 * nfs4_unlock_state() called after encode
3759 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3760 struct nfsd4_close *close)
3763 struct nfs4_openowner *oo;
3764 struct nfs4_ol_stateid *stp;
3766 dprintk("NFSD: nfsd4_close on file %.*s\n",
3767 (int)cstate->current_fh.fh_dentry->d_name.len,
3768 cstate->current_fh.fh_dentry->d_name.name);
3771 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3773 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3777 oo = openowner(stp->st_stateowner);
3779 update_stateid(&stp->st_stid.sc_stateid);
3780 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3782 nfsd4_close_open_stateid(stp);
3783 oo->oo_last_closed_stid = stp;
3785 /* place unused nfs4_stateowners on so_close_lru list to be
3786 * released by the laundromat service after the lease period
3787 * to enable us to handle CLOSE replay
3789 if (list_empty(&oo->oo_owner.so_stateids))
3790 move_to_close_lru(oo);
3792 if (!cstate->replay_owner)
3793 nfs4_unlock_state();
3798 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3799 struct nfsd4_delegreturn *dr)
3801 struct nfs4_delegation *dp;
3802 stateid_t *stateid = &dr->dr_stateid;
3803 struct nfs4_stid *s;
3804 struct inode *inode;
3807 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3809 inode = cstate->current_fh.fh_dentry->d_inode;
3812 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3815 dp = delegstateid(s);
3816 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3820 unhash_delegation(dp);
3822 nfs4_unlock_state();
3828 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3830 #define LOCKOWNER_INO_HASH_BITS 8
3831 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3832 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3835 end_offset(u64 start, u64 len)
3840 return end >= start ? end: NFS4_MAX_UINT64;
3843 /* last octet in a range */
3845 last_byte_offset(u64 start, u64 len)
3851 return end > start ? end - 1: NFS4_MAX_UINT64;
3854 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3856 return (file_hashval(inode) + cl_id
3857 + opaque_hashval(ownername->data, ownername->len))
3858 & LOCKOWNER_INO_HASH_MASK;
3861 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3864 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3865 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3866 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3867 * locking, this prevents us from being completely protocol-compliant. The
3868 * real solution to this problem is to start using unsigned file offsets in
3869 * the VFS, but this is a very deep change!
3872 nfs4_transform_lock_offset(struct file_lock *lock)
3874 if (lock->fl_start < 0)
3875 lock->fl_start = OFFSET_MAX;
3876 if (lock->fl_end < 0)
3877 lock->fl_end = OFFSET_MAX;
3880 /* Hack!: For now, we're defining this just so we can use a pointer to it
3881 * as a unique cookie to identify our (NFSv4's) posix locks. */
3882 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3886 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3888 struct nfs4_lockowner *lo;
3890 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3891 lo = (struct nfs4_lockowner *) fl->fl_owner;
3892 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3893 lo->lo_owner.so_owner.len, GFP_KERNEL);
3894 if (!deny->ld_owner.data)
3895 /* We just don't care that much */
3897 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3898 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3901 deny->ld_owner.len = 0;
3902 deny->ld_owner.data = NULL;
3903 deny->ld_clientid.cl_boot = 0;
3904 deny->ld_clientid.cl_id = 0;
3906 deny->ld_start = fl->fl_start;
3907 deny->ld_length = NFS4_MAX_UINT64;
3908 if (fl->fl_end != NFS4_MAX_UINT64)
3909 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3910 deny->ld_type = NFS4_READ_LT;
3911 if (fl->fl_type != F_RDLCK)
3912 deny->ld_type = NFS4_WRITE_LT;
3915 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3917 struct nfs4_ol_stateid *lst;
3919 if (!same_owner_str(&lo->lo_owner, owner, clid))
3921 lst = list_first_entry(&lo->lo_owner.so_stateids,
3922 struct nfs4_ol_stateid, st_perstateowner);
3923 return lst->st_file->fi_inode == inode;
3926 static struct nfs4_lockowner *
3927 find_lockowner_str(struct inode *inode, clientid_t *clid,
3928 struct xdr_netobj *owner)
3930 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3931 struct nfs4_lockowner *lo;
3933 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3934 if (same_lockowner_ino(lo, inode, clid, owner))
3940 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3942 struct inode *inode = open_stp->st_file->fi_inode;
3943 unsigned int inohash = lockowner_ino_hashval(inode,
3944 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3946 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3947 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3948 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3952 * Alloc a lock owner structure.
3953 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3956 * strhashval = ownerstr_hashval
3959 static struct nfs4_lockowner *
3960 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3961 struct nfs4_lockowner *lo;
3963 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3966 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3967 lo->lo_owner.so_is_open_owner = 0;
3968 /* It is the openowner seqid that will be incremented in encode in the
3969 * case of new lockowners; so increment the lock seqid manually: */
3970 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3971 hash_lockowner(lo, strhashval, clp, open_stp);
3975 static struct nfs4_ol_stateid *
3976 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3978 struct nfs4_ol_stateid *stp;
3979 struct nfs4_client *clp = lo->lo_owner.so_client;
3981 stp = nfs4_alloc_stateid(clp);
3984 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3985 list_add(&stp->st_perfile, &fp->fi_stateids);
3986 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3987 stp->st_stateowner = &lo->lo_owner;
3990 stp->st_access_bmap = 0;
3991 stp->st_deny_bmap = open_stp->st_deny_bmap;
3992 stp->st_openstp = open_stp;
3997 check_lock_length(u64 offset, u64 length)
3999 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4000 LOFF_OVERFLOW(offset, length)));
4003 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4005 struct nfs4_file *fp = lock_stp->st_file;
4006 int oflag = nfs4_access_to_omode(access);
4008 if (test_bit(access, &lock_stp->st_access_bmap))
4010 nfs4_file_get_access(fp, oflag);
4011 __set_bit(access, &lock_stp->st_access_bmap);
4014 __be32 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate, struct nfs4_ol_stateid *ost, struct nfsd4_lock *lock, struct nfs4_ol_stateid **lst, bool *new)
4016 struct nfs4_file *fi = ost->st_file;
4017 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4018 struct nfs4_client *cl = oo->oo_owner.so_client;
4019 struct nfs4_lockowner *lo;
4020 unsigned int strhashval;
4022 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4024 if (!cstate->minorversion)
4025 return nfserr_bad_seqid;
4026 /* XXX: a lockowner always has exactly one stateid: */
4027 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4028 struct nfs4_ol_stateid, st_perstateowner);
4031 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4032 &lock->v.new.owner);
4033 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4035 return nfserr_jukebox;
4036 *lst = alloc_init_lock_stateid(lo, fi, ost);
4038 release_lockowner(lo);
4039 return nfserr_jukebox;
4049 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4050 struct nfsd4_lock *lock)
4052 struct nfs4_openowner *open_sop = NULL;
4053 struct nfs4_lockowner *lock_sop = NULL;
4054 struct nfs4_ol_stateid *lock_stp;
4055 struct nfs4_file *fp;
4056 struct file *filp = NULL;
4057 struct file_lock file_lock;
4058 struct file_lock conflock;
4060 bool new_state = false;
4064 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4065 (long long) lock->lk_offset,
4066 (long long) lock->lk_length);
4068 if (check_lock_length(lock->lk_offset, lock->lk_length))
4069 return nfserr_inval;
4071 if ((status = fh_verify(rqstp, &cstate->current_fh,
4072 S_IFREG, NFSD_MAY_LOCK))) {
4073 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4079 if (lock->lk_is_new) {
4081 * Client indicates that this is a new lockowner.
4082 * Use open owner and open stateid to create lock owner and
4085 struct nfs4_ol_stateid *open_stp = NULL;
4087 if (nfsd4_has_session(cstate))
4088 /* See rfc 5661 18.10.3: given clientid is ignored: */
4089 memcpy(&lock->v.new.clientid,
4090 &cstate->session->se_client->cl_clientid,
4091 sizeof(clientid_t));
4093 status = nfserr_stale_clientid;
4094 if (STALE_CLIENTID(&lock->lk_new_clientid))
4097 /* validate and update open stateid and open seqid */
4098 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4099 lock->lk_new_open_seqid,
4100 &lock->lk_new_open_stateid,
4104 open_sop = openowner(open_stp->st_stateowner);
4105 status = nfserr_bad_stateid;
4106 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4107 &lock->v.new.clientid))
4109 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4110 &lock_stp, &new_state);
4114 /* lock (lock owner + lock stateid) already exists */
4115 status = nfs4_preprocess_seqid_op(cstate,
4116 lock->lk_old_lock_seqid,
4117 &lock->lk_old_lock_stateid,
4118 NFS4_LOCK_STID, &lock_stp);
4122 lock_sop = lockowner(lock_stp->st_stateowner);
4123 fp = lock_stp->st_file;
4125 lkflg = setlkflg(lock->lk_type);
4126 status = nfs4_check_openmode(lock_stp, lkflg);
4130 status = nfserr_grace;
4131 if (locks_in_grace() && !lock->lk_reclaim)
4133 status = nfserr_no_grace;
4134 if (!locks_in_grace() && lock->lk_reclaim)
4137 locks_init_lock(&file_lock);
4138 switch (lock->lk_type) {
4141 filp = find_readable_file(lock_stp->st_file);
4143 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4144 file_lock.fl_type = F_RDLCK;
4147 case NFS4_WRITEW_LT:
4148 filp = find_writeable_file(lock_stp->st_file);
4150 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4151 file_lock.fl_type = F_WRLCK;
4154 status = nfserr_inval;
4158 status = nfserr_openmode;
4161 file_lock.fl_owner = (fl_owner_t)lock_sop;
4162 file_lock.fl_pid = current->tgid;
4163 file_lock.fl_file = filp;
4164 file_lock.fl_flags = FL_POSIX;
4165 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4167 file_lock.fl_start = lock->lk_offset;
4168 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4169 nfs4_transform_lock_offset(&file_lock);
4172 * Try to lock the file in the VFS.
4173 * Note: locks.c uses the BKL to protect the inode's lock list.
4176 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4178 case 0: /* success! */
4179 update_stateid(&lock_stp->st_stid.sc_stateid);
4180 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4184 case (EAGAIN): /* conflock holds conflicting lock */
4185 status = nfserr_denied;
4186 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4187 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4190 status = nfserr_deadlock;
4193 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4194 status = nfserrno(err);
4198 if (status && new_state)
4199 release_lockowner(lock_sop);
4200 if (!cstate->replay_owner)
4201 nfs4_unlock_state();
4206 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4207 * so we do a temporary open here just to get an open file to pass to
4208 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4211 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4216 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4219 err = vfs_test_lock(file, lock);
4228 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4229 struct nfsd4_lockt *lockt)
4231 struct inode *inode;
4232 struct file_lock file_lock;
4233 struct nfs4_lockowner *lo;
4237 if (locks_in_grace())
4238 return nfserr_grace;
4240 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4241 return nfserr_inval;
4245 status = nfserr_stale_clientid;
4246 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4249 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4252 inode = cstate->current_fh.fh_dentry->d_inode;
4253 locks_init_lock(&file_lock);
4254 switch (lockt->lt_type) {
4257 file_lock.fl_type = F_RDLCK;
4260 case NFS4_WRITEW_LT:
4261 file_lock.fl_type = F_WRLCK;
4264 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4265 status = nfserr_inval;
4269 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4271 file_lock.fl_owner = (fl_owner_t)lo;
4272 file_lock.fl_pid = current->tgid;
4273 file_lock.fl_flags = FL_POSIX;
4275 file_lock.fl_start = lockt->lt_offset;
4276 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4278 nfs4_transform_lock_offset(&file_lock);
4281 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4283 status = nfserrno(error);
4286 if (file_lock.fl_type != F_UNLCK) {
4287 status = nfserr_denied;
4288 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4291 nfs4_unlock_state();
4296 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4297 struct nfsd4_locku *locku)
4299 struct nfs4_ol_stateid *stp;
4300 struct file *filp = NULL;
4301 struct file_lock file_lock;
4305 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4306 (long long) locku->lu_offset,
4307 (long long) locku->lu_length);
4309 if (check_lock_length(locku->lu_offset, locku->lu_length))
4310 return nfserr_inval;
4314 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4315 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4318 filp = find_any_file(stp->st_file);
4320 status = nfserr_lock_range;
4324 locks_init_lock(&file_lock);
4325 file_lock.fl_type = F_UNLCK;
4326 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4327 file_lock.fl_pid = current->tgid;
4328 file_lock.fl_file = filp;
4329 file_lock.fl_flags = FL_POSIX;
4330 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4331 file_lock.fl_start = locku->lu_offset;
4333 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4334 nfs4_transform_lock_offset(&file_lock);
4337 * Try to unlock the file in the VFS.
4339 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4341 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4345 * OK, unlock succeeded; the only thing left to do is update the stateid.
4347 update_stateid(&stp->st_stid.sc_stateid);
4348 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4351 if (!cstate->replay_owner)
4352 nfs4_unlock_state();
4356 status = nfserrno(err);
4362 * 1: locks held by lockowner
4363 * 0: no locks held by lockowner
4366 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4368 struct file_lock **flpp;
4369 struct inode *inode = filp->fi_inode;
4373 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4374 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4385 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4386 struct nfsd4_compound_state *cstate,
4387 struct nfsd4_release_lockowner *rlockowner)
4389 clientid_t *clid = &rlockowner->rl_clientid;
4390 struct nfs4_stateowner *sop;
4391 struct nfs4_lockowner *lo;
4392 struct nfs4_ol_stateid *stp;
4393 struct xdr_netobj *owner = &rlockowner->rl_owner;
4394 struct list_head matches;
4395 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4398 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4399 clid->cl_boot, clid->cl_id);
4401 /* XXX check for lease expiration */
4403 status = nfserr_stale_clientid;
4404 if (STALE_CLIENTID(clid))
4409 status = nfserr_locks_held;
4410 INIT_LIST_HEAD(&matches);
4412 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4413 if (sop->so_is_open_owner)
4415 if (!same_owner_str(sop, owner, clid))
4417 list_for_each_entry(stp, &sop->so_stateids,
4419 lo = lockowner(sop);
4420 if (check_for_locks(stp->st_file, lo))
4422 list_add(&lo->lo_list, &matches);
4425 /* Clients probably won't expect us to return with some (but not all)
4426 * of the lockowner state released; so don't release any until all
4427 * have been checked. */
4429 while (!list_empty(&matches)) {
4430 lo = list_entry(matches.next, struct nfs4_lockowner,
4432 /* unhash_stateowner deletes so_perclient only
4433 * for openowners. */
4434 list_del(&lo->lo_list);
4435 release_lockowner(lo);
4438 nfs4_unlock_state();
4442 static inline struct nfs4_client_reclaim *
4445 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4449 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4451 unsigned int strhashval = clientstr_hashval(name);
4452 struct nfs4_client *clp;
4454 clp = find_confirmed_client_by_str(name, strhashval);
4457 return clp->cl_firststate;
4461 * failure => all reset bets are off, nfserr_no_grace...
4464 nfs4_client_to_reclaim(const char *name)
4466 unsigned int strhashval;
4467 struct nfs4_client_reclaim *crp = NULL;
4469 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4470 crp = alloc_reclaim();
4473 strhashval = clientstr_hashval(name);
4474 INIT_LIST_HEAD(&crp->cr_strhash);
4475 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4476 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4477 reclaim_str_hashtbl_size++;
4482 nfs4_release_reclaim(void)
4484 struct nfs4_client_reclaim *crp = NULL;
4487 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4488 while (!list_empty(&reclaim_str_hashtbl[i])) {
4489 crp = list_entry(reclaim_str_hashtbl[i].next,
4490 struct nfs4_client_reclaim, cr_strhash);
4491 list_del(&crp->cr_strhash);
4493 reclaim_str_hashtbl_size--;
4496 BUG_ON(reclaim_str_hashtbl_size);
4500 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4501 static struct nfs4_client_reclaim *
4502 nfs4_find_reclaim_client(clientid_t *clid)
4504 unsigned int strhashval;
4505 struct nfs4_client *clp;
4506 struct nfs4_client_reclaim *crp = NULL;
4509 /* find clientid in conf_id_hashtbl */
4510 clp = find_confirmed_client(clid);
4514 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4515 clp->cl_name.len, clp->cl_name.data,
4518 /* find clp->cl_name in reclaim_str_hashtbl */
4519 strhashval = clientstr_hashval(clp->cl_recdir);
4520 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4521 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4529 * Called from OPEN. Look for clientid in reclaim list.
4532 nfs4_check_open_reclaim(clientid_t *clid)
4534 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4537 #ifdef CONFIG_NFSD_FAULT_INJECTION
4539 void nfsd_forget_clients(u64 num)
4541 struct nfs4_client *clp, *next;
4545 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4546 nfsd4_remove_clid_dir(clp);
4551 nfs4_unlock_state();
4553 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4556 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4558 release_lockowner(lockowner(sop));
4561 static void release_openowner_sop(struct nfs4_stateowner *sop)
4563 release_openowner(openowner(sop));
4566 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4567 void (*release_sop)(struct nfs4_stateowner *))
4570 struct nfs4_stateowner *sop, *next;
4572 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4573 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4574 if (sop->so_is_open_owner != is_open_owner)
4584 void nfsd_forget_locks(u64 num)
4589 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4590 nfs4_unlock_state();
4592 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4595 void nfsd_forget_openowners(u64 num)
4600 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4601 nfs4_unlock_state();
4603 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4606 int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegation *))
4609 struct nfs4_file *fp, *fnext;
4610 struct nfs4_delegation *dp, *dnext;
4612 for (i = 0; i < FILE_HASH_SIZE; i++) {
4613 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4614 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4625 void nfsd_forget_delegations(u64 num)
4630 count = nfsd_process_n_delegations(num, unhash_delegation);
4631 nfs4_unlock_state();
4633 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4636 void nfsd_recall_delegations(u64 num)
4641 spin_lock(&recall_lock);
4642 count = nfsd_process_n_delegations(num, nfsd_break_one_deleg);
4643 spin_unlock(&recall_lock);
4644 nfs4_unlock_state();
4646 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4649 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4651 /* initialization to perform at module load time: */
4654 nfs4_state_init(void)
4658 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4659 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4660 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4661 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4662 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4663 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4665 for (i = 0; i < SESSION_HASH_SIZE; i++)
4666 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4667 for (i = 0; i < FILE_HASH_SIZE; i++) {
4668 INIT_LIST_HEAD(&file_hashtbl[i]);
4670 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4671 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4673 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4674 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4675 INIT_LIST_HEAD(&close_lru);
4676 INIT_LIST_HEAD(&client_lru);
4677 INIT_LIST_HEAD(&del_recall_lru);
4678 reclaim_str_hashtbl_size = 0;
4682 nfsd4_load_reboot_recovery_data(void)
4687 nfsd4_init_recdir();
4688 status = nfsd4_recdir_load();
4689 nfs4_unlock_state();
4691 printk("NFSD: Failure reading reboot recovery data\n");
4695 * Since the lifetime of a delegation isn't limited to that of an open, a
4696 * client may quite reasonably hang on to a delegation as long as it has
4697 * the inode cached. This becomes an obvious problem the first time a
4698 * client's inode cache approaches the size of the server's total memory.
4700 * For now we avoid this problem by imposing a hard limit on the number
4701 * of delegations, which varies according to the server's memory size.
4704 set_max_delegations(void)
4707 * Allow at most 4 delegations per megabyte of RAM. Quick
4708 * estimates suggest that in the worst case (where every delegation
4709 * is for a different inode), a delegation could take about 1.5K,
4710 * giving a worst case usage of about 6% of memory.
4712 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4715 /* initialization to perform when the nfsd service is started: */
4718 nfs4_state_start(void)
4722 nfsd4_load_reboot_recovery_data();
4723 boot_time = get_seconds();
4724 locks_start_grace(&nfsd4_manager);
4725 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4727 ret = set_callback_cred();
4732 laundry_wq = create_singlethread_workqueue("nfsd4");
4733 if (laundry_wq == NULL) {
4737 ret = nfsd4_create_callback_queue();
4739 goto out_free_laundry;
4740 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4741 set_max_delegations();
4744 destroy_workqueue(laundry_wq);
4746 nfs4_release_reclaim();
4747 nfsd4_shutdown_recdir();
4752 __nfs4_state_shutdown(void)
4755 struct nfs4_client *clp = NULL;
4756 struct nfs4_delegation *dp = NULL;
4757 struct list_head *pos, *next, reaplist;
4759 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4760 while (!list_empty(&conf_id_hashtbl[i])) {
4761 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4764 while (!list_empty(&unconf_str_hashtbl[i])) {
4765 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4769 INIT_LIST_HEAD(&reaplist);
4770 spin_lock(&recall_lock);
4771 list_for_each_safe(pos, next, &del_recall_lru) {
4772 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4773 list_move(&dp->dl_recall_lru, &reaplist);
4775 spin_unlock(&recall_lock);
4776 list_for_each_safe(pos, next, &reaplist) {
4777 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4778 unhash_delegation(dp);
4781 nfsd4_shutdown_recdir();
4785 nfs4_state_shutdown(void)
4787 cancel_delayed_work_sync(&laundromat_work);
4788 destroy_workqueue(laundry_wq);
4789 locks_end_grace(&nfsd4_manager);
4791 nfs4_release_reclaim();
4792 __nfs4_state_shutdown();
4793 nfs4_unlock_state();
4794 nfsd4_destroy_callback_queue();
4798 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4800 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4801 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4805 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4807 if (cstate->minorversion) {
4808 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4809 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4814 clear_current_stateid(struct nfsd4_compound_state *cstate)
4816 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4820 * functions to set current state id
4823 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4825 put_stateid(cstate, &odp->od_stateid);
4829 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4831 put_stateid(cstate, &open->op_stateid);
4835 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4837 put_stateid(cstate, &close->cl_stateid);
4841 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4843 put_stateid(cstate, &lock->lk_resp_stateid);
4847 * functions to consume current state id
4851 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4853 get_stateid(cstate, &odp->od_stateid);
4857 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4859 get_stateid(cstate, &drp->dr_stateid);
4863 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4865 get_stateid(cstate, &fsp->fr_stateid);
4869 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4871 get_stateid(cstate, &setattr->sa_stateid);
4875 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4877 get_stateid(cstate, &close->cl_stateid);
4881 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4883 get_stateid(cstate, &locku->lu_stateid);
4887 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4889 get_stateid(cstate, &read->rd_stateid);
4893 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4895 get_stateid(cstate, &write->wr_stateid);