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>
36 #include <linux/smp_lock.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease = 90; /* default lease time */
49 time_t nfsd4_grace = 90;
50 static time_t boot_time;
51 static u32 current_ownerid = 1;
52 static u32 current_fileid = 1;
53 static u32 current_delegid = 1;
55 static stateid_t zerostateid; /* bits all 0 */
56 static stateid_t onestateid; /* bits all 1 */
57 static u64 current_sessionid = 1;
59 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
62 /* forward declarations */
63 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
64 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
65 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
66 static void nfs4_set_recdir(char *recdir);
70 /* Currently used for almost all code touching nfsv4 state: */
71 static DEFINE_MUTEX(client_mutex);
74 * Currently used for the del_recall_lru and file hash table. In an
75 * effort to decrease the scope of the client_mutex, this spinlock may
76 * eventually cover more:
78 static DEFINE_SPINLOCK(recall_lock);
80 static struct kmem_cache *stateowner_slab = NULL;
81 static struct kmem_cache *file_slab = NULL;
82 static struct kmem_cache *stateid_slab = NULL;
83 static struct kmem_cache *deleg_slab = NULL;
88 mutex_lock(&client_mutex);
92 nfs4_unlock_state(void)
94 mutex_unlock(&client_mutex);
98 opaque_hashval(const void *ptr, int nbytes)
100 unsigned char *cptr = (unsigned char *) ptr;
110 static struct list_head del_recall_lru;
113 put_nfs4_file(struct nfs4_file *fi)
115 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
116 list_del(&fi->fi_hash);
117 spin_unlock(&recall_lock);
119 kmem_cache_free(file_slab, fi);
124 get_nfs4_file(struct nfs4_file *fi)
126 atomic_inc(&fi->fi_ref);
129 static int num_delegations;
130 unsigned int max_delegations;
133 * Open owner state (share locks)
136 /* hash tables for nfs4_stateowner */
137 #define OWNER_HASH_BITS 8
138 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
139 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
141 #define ownerid_hashval(id) \
142 ((id) & OWNER_HASH_MASK)
143 #define ownerstr_hashval(clientid, ownername) \
144 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
146 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
147 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
149 /* hash table for nfs4_file */
150 #define FILE_HASH_BITS 8
151 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
153 /* hash table for (open)nfs4_stateid */
154 #define STATEID_HASH_BITS 10
155 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
156 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
158 #define file_hashval(x) \
159 hash_ptr(x, FILE_HASH_BITS)
160 #define stateid_hashval(owner_id, file_id) \
161 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
163 static struct list_head file_hashtbl[FILE_HASH_SIZE];
164 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
166 static struct nfs4_delegation *
167 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
169 struct nfs4_delegation *dp;
170 struct nfs4_file *fp = stp->st_file;
171 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
173 dprintk("NFSD alloc_init_deleg\n");
174 if (fp->fi_had_conflict)
176 if (num_delegations > max_delegations)
178 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
182 INIT_LIST_HEAD(&dp->dl_perfile);
183 INIT_LIST_HEAD(&dp->dl_perclnt);
184 INIT_LIST_HEAD(&dp->dl_recall_lru);
189 get_file(stp->st_vfs_file);
190 dp->dl_vfs_file = stp->st_vfs_file;
192 dp->dl_ident = cb->cb_ident;
193 dp->dl_stateid.si_boot = boot_time;
194 dp->dl_stateid.si_stateownerid = current_delegid++;
195 dp->dl_stateid.si_fileid = 0;
196 dp->dl_stateid.si_generation = 0;
197 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
199 atomic_set(&dp->dl_count, 1);
200 list_add(&dp->dl_perfile, &fp->fi_delegations);
201 list_add(&dp->dl_perclnt, &clp->cl_delegations);
202 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
207 nfs4_put_delegation(struct nfs4_delegation *dp)
209 if (atomic_dec_and_test(&dp->dl_count)) {
210 dprintk("NFSD: freeing dp %p\n",dp);
211 put_nfs4_file(dp->dl_file);
212 kmem_cache_free(deleg_slab, dp);
217 /* Remove the associated file_lock first, then remove the delegation.
218 * lease_modify() is called to remove the FS_LEASE file_lock from
219 * the i_flock list, eventually calling nfsd's lock_manager
220 * fl_release_callback.
223 nfs4_close_delegation(struct nfs4_delegation *dp)
225 struct file *filp = dp->dl_vfs_file;
227 dprintk("NFSD: close_delegation dp %p\n",dp);
228 dp->dl_vfs_file = NULL;
229 /* The following nfsd_close may not actually close the file,
230 * but we want to remove the lease in any case. */
232 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
236 /* Called under the state lock. */
238 unhash_delegation(struct nfs4_delegation *dp)
240 list_del_init(&dp->dl_perfile);
241 list_del_init(&dp->dl_perclnt);
242 spin_lock(&recall_lock);
243 list_del_init(&dp->dl_recall_lru);
244 spin_unlock(&recall_lock);
245 nfs4_close_delegation(dp);
246 nfs4_put_delegation(dp);
253 /* client_lock protects the client lru list and session hash table */
254 static DEFINE_SPINLOCK(client_lock);
256 /* Hash tables for nfs4_clientid state */
257 #define CLIENT_HASH_BITS 4
258 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
261 #define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
263 #define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
275 * client_lru holds client queue ordered by nfs4_client.cl_time
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
281 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
282 static int reclaim_str_hashtbl_size = 0;
283 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
284 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
285 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
286 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
287 static struct list_head client_lru;
288 static struct list_head close_lru;
290 static void unhash_generic_stateid(struct nfs4_stateid *stp)
292 list_del(&stp->st_hash);
293 list_del(&stp->st_perfile);
294 list_del(&stp->st_perstateowner);
297 static void free_generic_stateid(struct nfs4_stateid *stp)
299 put_nfs4_file(stp->st_file);
300 kmem_cache_free(stateid_slab, stp);
303 static void release_lock_stateid(struct nfs4_stateid *stp)
305 unhash_generic_stateid(stp);
306 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
307 free_generic_stateid(stp);
310 static void unhash_lockowner(struct nfs4_stateowner *sop)
312 struct nfs4_stateid *stp;
314 list_del(&sop->so_idhash);
315 list_del(&sop->so_strhash);
316 list_del(&sop->so_perstateid);
317 while (!list_empty(&sop->so_stateids)) {
318 stp = list_first_entry(&sop->so_stateids,
319 struct nfs4_stateid, st_perstateowner);
320 release_lock_stateid(stp);
324 static void release_lockowner(struct nfs4_stateowner *sop)
326 unhash_lockowner(sop);
327 nfs4_put_stateowner(sop);
331 release_stateid_lockowners(struct nfs4_stateid *open_stp)
333 struct nfs4_stateowner *lock_sop;
335 while (!list_empty(&open_stp->st_lockowners)) {
336 lock_sop = list_entry(open_stp->st_lockowners.next,
337 struct nfs4_stateowner, so_perstateid);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop->so_is_open_owner);
340 release_lockowner(lock_sop);
344 static void release_open_stateid(struct nfs4_stateid *stp)
346 unhash_generic_stateid(stp);
347 release_stateid_lockowners(stp);
348 nfsd_close(stp->st_vfs_file);
349 free_generic_stateid(stp);
352 static void unhash_openowner(struct nfs4_stateowner *sop)
354 struct nfs4_stateid *stp;
356 list_del(&sop->so_idhash);
357 list_del(&sop->so_strhash);
358 list_del(&sop->so_perclient);
359 list_del(&sop->so_perstateid); /* XXX: necessary? */
360 while (!list_empty(&sop->so_stateids)) {
361 stp = list_first_entry(&sop->so_stateids,
362 struct nfs4_stateid, st_perstateowner);
363 release_open_stateid(stp);
367 static void release_openowner(struct nfs4_stateowner *sop)
369 unhash_openowner(sop);
370 list_del(&sop->so_close_lru);
371 nfs4_put_stateowner(sop);
374 #define SESSION_HASH_SIZE 512
375 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
378 hash_sessionid(struct nfs4_sessionid *sessionid)
380 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
382 return sid->sequence % SESSION_HASH_SIZE;
386 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
388 u32 *ptr = (u32 *)(&sessionid->data[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
393 gen_sessionid(struct nfsd4_session *ses)
395 struct nfs4_client *clp = ses->se_client;
396 struct nfsd4_sessionid *sid;
398 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
399 sid->clientid = clp->cl_clientid;
400 sid->sequence = current_sessionid++;
405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
416 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
428 static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
430 int mem, size = fchan->maxresp_cached;
432 if (fchan->maxreqs < 1)
435 if (size < NFSD_MIN_HDR_SEQ_SZ)
436 size = NFSD_MIN_HDR_SEQ_SZ;
437 size -= NFSD_MIN_HDR_SEQ_SZ;
438 if (size > NFSD_SLOT_CACHE_SIZE)
439 size = NFSD_SLOT_CACHE_SIZE;
441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem = fchan->maxreqs * size;
443 if (mem > NFSD_MAX_MEM_PER_SESSION) {
444 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
445 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
446 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
447 mem = fchan->maxreqs * size;
450 spin_lock(&nfsd_drc_lock);
451 /* bound the total session drc memory ussage */
452 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
453 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
454 mem = fchan->maxreqs * size;
456 nfsd_drc_mem_used += mem;
457 spin_unlock(&nfsd_drc_lock);
459 if (fchan->maxreqs == 0)
460 return nfserr_jukebox;
462 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
467 * fchan holds the client values on input, and the server values on output
468 * sv_max_mesg is the maximum payload plus one page for overhead.
470 static int init_forechannel_attrs(struct svc_rqst *rqstp,
471 struct nfsd4_channel_attrs *session_fchan,
472 struct nfsd4_channel_attrs *fchan)
475 __u32 maxcount = nfsd_serv->sv_max_mesg;
477 /* headerpadsz set to zero in encode routine */
479 /* Use the client's max request and max response size if possible */
480 if (fchan->maxreq_sz > maxcount)
481 fchan->maxreq_sz = maxcount;
482 session_fchan->maxreq_sz = fchan->maxreq_sz;
484 if (fchan->maxresp_sz > maxcount)
485 fchan->maxresp_sz = maxcount;
486 session_fchan->maxresp_sz = fchan->maxresp_sz;
488 /* Use the client's maxops if possible */
489 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
490 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
491 session_fchan->maxops = fchan->maxops;
493 /* FIXME: Error means no more DRC pages so the server should
494 * recover pages from existing sessions. For now fail session
497 status = set_forechannel_drc_size(fchan);
499 session_fchan->maxresp_cached = fchan->maxresp_cached;
500 session_fchan->maxreqs = fchan->maxreqs;
502 dprintk("%s status %d\n", __func__, status);
507 free_session_slots(struct nfsd4_session *ses)
511 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
512 kfree(ses->se_slots[i]);
516 * We don't actually need to cache the rpc and session headers, so we
517 * can allocate a little less for each slot:
519 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
521 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
525 alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
526 struct nfsd4_create_session *cses)
528 struct nfsd4_session *new, tmp;
529 struct nfsd4_slot *sp;
530 int idx, slotsize, cachesize, i;
533 memset(&tmp, 0, sizeof(tmp));
535 /* FIXME: For now, we just accept the client back channel attributes. */
536 tmp.se_bchannel = cses->back_channel;
537 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
538 &cses->fore_channel);
542 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
543 + sizeof(struct nfsd4_session) > PAGE_SIZE);
545 status = nfserr_jukebox;
546 /* allocate struct nfsd4_session and slot table pointers in one piece */
547 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
548 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
552 memcpy(new, &tmp, sizeof(*new));
554 /* allocate each struct nfsd4_slot and data cache in one piece */
555 cachesize = slot_bytes(&new->se_fchannel);
556 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
557 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
560 new->se_slots[i] = sp;
563 new->se_client = clp;
565 idx = hash_sessionid(&new->se_sessionid);
566 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
567 NFS4_MAX_SESSIONID_LEN);
569 new->se_flags = cses->flags;
570 kref_init(&new->se_ref);
571 spin_lock(&client_lock);
572 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
573 list_add(&new->se_perclnt, &clp->cl_sessions);
574 spin_unlock(&client_lock);
580 free_session_slots(new);
585 /* caller must hold client_lock */
586 static struct nfsd4_session *
587 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
589 struct nfsd4_session *elem;
592 dump_sessionid(__func__, sessionid);
593 idx = hash_sessionid(sessionid);
594 /* Search in the appropriate list */
595 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
596 if (!memcmp(elem->se_sessionid.data, sessionid->data,
597 NFS4_MAX_SESSIONID_LEN)) {
602 dprintk("%s: session not found\n", __func__);
606 /* caller must hold client_lock */
608 unhash_session(struct nfsd4_session *ses)
610 list_del(&ses->se_hash);
611 list_del(&ses->se_perclnt);
615 free_session(struct kref *kref)
617 struct nfsd4_session *ses;
620 ses = container_of(kref, struct nfsd4_session, se_ref);
621 spin_lock(&nfsd_drc_lock);
622 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
623 nfsd_drc_mem_used -= mem;
624 spin_unlock(&nfsd_drc_lock);
625 free_session_slots(ses);
629 /* must be called under the client_lock */
631 renew_client_locked(struct nfs4_client *clp)
633 if (is_client_expired(clp)) {
634 dprintk("%s: client (clientid %08x/%08x) already expired\n",
636 clp->cl_clientid.cl_boot,
637 clp->cl_clientid.cl_id);
642 * Move client to the end to the LRU list.
644 dprintk("renewing client (clientid %08x/%08x)\n",
645 clp->cl_clientid.cl_boot,
646 clp->cl_clientid.cl_id);
647 list_move_tail(&clp->cl_lru, &client_lru);
648 clp->cl_time = get_seconds();
652 renew_client(struct nfs4_client *clp)
654 spin_lock(&client_lock);
655 renew_client_locked(clp);
656 spin_unlock(&client_lock);
659 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
661 STALE_CLIENTID(clientid_t *clid)
663 if (clid->cl_boot == boot_time)
665 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
666 clid->cl_boot, clid->cl_id, boot_time);
671 * XXX Should we use a slab cache ?
672 * This type of memory management is somewhat inefficient, but we use it
673 * anyway since SETCLIENTID is not a common operation.
675 static struct nfs4_client *alloc_client(struct xdr_netobj name)
677 struct nfs4_client *clp;
679 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
682 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
683 if (clp->cl_name.data == NULL) {
687 memcpy(clp->cl_name.data, name.data, name.len);
688 clp->cl_name.len = name.len;
693 free_client(struct nfs4_client *clp)
695 if (clp->cl_cred.cr_group_info)
696 put_group_info(clp->cl_cred.cr_group_info);
697 kfree(clp->cl_principal);
698 kfree(clp->cl_name.data);
703 release_session_client(struct nfsd4_session *session)
705 struct nfs4_client *clp = session->se_client;
707 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
709 if (is_client_expired(clp)) {
711 session->se_client = NULL;
713 renew_client_locked(clp);
714 spin_unlock(&client_lock);
717 /* must be called under the client_lock */
719 unhash_client_locked(struct nfs4_client *clp)
721 mark_client_expired(clp);
722 list_del(&clp->cl_lru);
723 while (!list_empty(&clp->cl_sessions)) {
724 struct nfsd4_session *ses;
725 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
728 nfsd4_put_session(ses);
733 expire_client(struct nfs4_client *clp)
735 struct nfs4_stateowner *sop;
736 struct nfs4_delegation *dp;
737 struct list_head reaplist;
739 INIT_LIST_HEAD(&reaplist);
740 spin_lock(&recall_lock);
741 while (!list_empty(&clp->cl_delegations)) {
742 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
743 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
745 list_del_init(&dp->dl_perclnt);
746 list_move(&dp->dl_recall_lru, &reaplist);
748 spin_unlock(&recall_lock);
749 while (!list_empty(&reaplist)) {
750 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
751 list_del_init(&dp->dl_recall_lru);
752 unhash_delegation(dp);
754 while (!list_empty(&clp->cl_openowners)) {
755 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
756 release_openowner(sop);
758 nfsd4_set_callback_client(clp, NULL);
759 if (clp->cl_cb_conn.cb_xprt)
760 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
761 list_del(&clp->cl_idhash);
762 list_del(&clp->cl_strhash);
763 spin_lock(&client_lock);
764 unhash_client_locked(clp);
765 if (atomic_read(&clp->cl_refcount) == 0)
767 spin_unlock(&client_lock);
770 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
772 memcpy(target->cl_verifier.data, source->data,
773 sizeof(target->cl_verifier.data));
776 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
778 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
779 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
782 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
784 target->cr_uid = source->cr_uid;
785 target->cr_gid = source->cr_gid;
786 target->cr_group_info = source->cr_group_info;
787 get_group_info(target->cr_group_info);
790 static int same_name(const char *n1, const char *n2)
792 return 0 == memcmp(n1, n2, HEXDIR_LEN);
796 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
798 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
802 same_clid(clientid_t *cl1, clientid_t *cl2)
804 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
807 /* XXX what about NGROUP */
809 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
811 return cr1->cr_uid == cr2->cr_uid;
814 static void gen_clid(struct nfs4_client *clp)
816 static u32 current_clientid = 1;
818 clp->cl_clientid.cl_boot = boot_time;
819 clp->cl_clientid.cl_id = current_clientid++;
822 static void gen_confirm(struct nfs4_client *clp)
827 p = (u32 *)clp->cl_confirm.data;
828 *p++ = get_seconds();
832 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
833 struct svc_rqst *rqstp, nfs4_verifier *verf)
835 struct nfs4_client *clp;
836 struct sockaddr *sa = svc_addr(rqstp);
839 clp = alloc_client(name);
843 princ = svc_gss_principal(rqstp);
845 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
846 if (clp->cl_principal == NULL) {
852 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
853 atomic_set(&clp->cl_refcount, 0);
854 atomic_set(&clp->cl_cb_set, 0);
855 INIT_LIST_HEAD(&clp->cl_idhash);
856 INIT_LIST_HEAD(&clp->cl_strhash);
857 INIT_LIST_HEAD(&clp->cl_openowners);
858 INIT_LIST_HEAD(&clp->cl_delegations);
859 INIT_LIST_HEAD(&clp->cl_sessions);
860 INIT_LIST_HEAD(&clp->cl_lru);
861 clp->cl_time = get_seconds();
862 clear_bit(0, &clp->cl_cb_slot_busy);
863 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
864 copy_verf(clp, verf);
865 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
866 clp->cl_flavor = rqstp->rq_flavor;
867 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
873 static int check_name(struct xdr_netobj name)
877 if (name.len > NFS4_OPAQUE_LIMIT) {
878 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
885 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
887 unsigned int idhashval;
889 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
890 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
891 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
896 move_to_confirmed(struct nfs4_client *clp)
898 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
899 unsigned int strhashval;
901 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
902 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
903 strhashval = clientstr_hashval(clp->cl_recdir);
904 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
908 static struct nfs4_client *
909 find_confirmed_client(clientid_t *clid)
911 struct nfs4_client *clp;
912 unsigned int idhashval = clientid_hashval(clid->cl_id);
914 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
915 if (same_clid(&clp->cl_clientid, clid))
921 static struct nfs4_client *
922 find_unconfirmed_client(clientid_t *clid)
924 struct nfs4_client *clp;
925 unsigned int idhashval = clientid_hashval(clid->cl_id);
927 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
928 if (same_clid(&clp->cl_clientid, clid))
935 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
936 * parameter. Matching is based on the fact the at least one of the
937 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
939 * FIXME: we need to unify the clientid namespaces for nfsv4.x
940 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
941 * and SET_CLIENTID{,_CONFIRM}
944 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
946 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
947 return use_exchange_id == has_exchange_flags;
950 static struct nfs4_client *
951 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
952 bool use_exchange_id)
954 struct nfs4_client *clp;
956 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
957 if (same_name(clp->cl_recdir, dname) &&
958 match_clientid_establishment(clp, use_exchange_id))
964 static struct nfs4_client *
965 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
966 bool use_exchange_id)
968 struct nfs4_client *clp;
970 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
971 if (same_name(clp->cl_recdir, dname) &&
972 match_clientid_establishment(clp, use_exchange_id))
979 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
981 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
982 unsigned short expected_family;
984 /* Currently, we only support tcp and tcp6 for the callback channel */
985 if (se->se_callback_netid_len == 3 &&
986 !memcmp(se->se_callback_netid_val, "tcp", 3))
987 expected_family = AF_INET;
988 else if (se->se_callback_netid_len == 4 &&
989 !memcmp(se->se_callback_netid_val, "tcp6", 4))
990 expected_family = AF_INET6;
994 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
995 se->se_callback_addr_len,
996 (struct sockaddr *) &cb->cb_addr,
997 sizeof(cb->cb_addr));
999 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1002 if (cb->cb_addr.ss_family == AF_INET6)
1003 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
1005 cb->cb_minorversion = 0;
1006 cb->cb_prog = se->se_callback_prog;
1007 cb->cb_ident = se->se_callback_ident;
1010 cb->cb_addr.ss_family = AF_UNSPEC;
1012 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1013 "will not receive delegations\n",
1014 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1020 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1023 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1025 struct nfsd4_slot *slot = resp->cstate.slot;
1028 dprintk("--> %s slot %p\n", __func__, slot);
1030 slot->sl_opcnt = resp->opcnt;
1031 slot->sl_status = resp->cstate.status;
1033 if (nfsd4_not_cached(resp)) {
1034 slot->sl_datalen = 0;
1037 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1038 base = (char *)resp->cstate.datap -
1039 (char *)resp->xbuf->head[0].iov_base;
1040 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1042 WARN("%s: sessions DRC could not cache compound\n", __func__);
1047 * Encode the replay sequence operation from the slot values.
1048 * If cachethis is FALSE encode the uncached rep error on the next
1049 * operation which sets resp->p and increments resp->opcnt for
1050 * nfs4svc_encode_compoundres.
1054 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1055 struct nfsd4_compoundres *resp)
1057 struct nfsd4_op *op;
1058 struct nfsd4_slot *slot = resp->cstate.slot;
1060 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1061 resp->opcnt, resp->cstate.slot->sl_cachethis);
1063 /* Encode the replayed sequence operation */
1064 op = &args->ops[resp->opcnt - 1];
1065 nfsd4_encode_operation(resp, op);
1067 /* Return nfserr_retry_uncached_rep in next operation. */
1068 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1069 op = &args->ops[resp->opcnt++];
1070 op->status = nfserr_retry_uncached_rep;
1071 nfsd4_encode_operation(resp, op);
1077 * The sequence operation is not cached because we can use the slot and
1081 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1082 struct nfsd4_sequence *seq)
1084 struct nfsd4_slot *slot = resp->cstate.slot;
1087 dprintk("--> %s slot %p\n", __func__, slot);
1089 /* Either returns 0 or nfserr_retry_uncached */
1090 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1091 if (status == nfserr_retry_uncached_rep)
1094 /* The sequence operation has been encoded, cstate->datap set. */
1095 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1097 resp->opcnt = slot->sl_opcnt;
1098 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1099 status = slot->sl_status;
1105 * Set the exchange_id flags returned by the server.
1108 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1110 /* pNFS is not supported */
1111 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1113 /* Referrals are supported, Migration is not. */
1114 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1116 /* set the wire flags to return to client. */
1117 clid->flags = new->cl_exchange_flags;
1121 nfsd4_exchange_id(struct svc_rqst *rqstp,
1122 struct nfsd4_compound_state *cstate,
1123 struct nfsd4_exchange_id *exid)
1125 struct nfs4_client *unconf, *conf, *new;
1127 unsigned int strhashval;
1128 char dname[HEXDIR_LEN];
1129 char addr_str[INET6_ADDRSTRLEN];
1130 nfs4_verifier verf = exid->verifier;
1131 struct sockaddr *sa = svc_addr(rqstp);
1133 rpc_ntop(sa, addr_str, sizeof(addr_str));
1134 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1135 "ip_addr=%s flags %x, spa_how %d\n",
1136 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1137 addr_str, exid->flags, exid->spa_how);
1139 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1140 return nfserr_inval;
1142 /* Currently only support SP4_NONE */
1143 switch (exid->spa_how) {
1147 return nfserr_encr_alg_unsupp;
1149 BUG(); /* checked by xdr code */
1151 return nfserr_serverfault; /* no excuse :-/ */
1154 status = nfs4_make_rec_clidname(dname, &exid->clname);
1159 strhashval = clientstr_hashval(dname);
1164 conf = find_confirmed_client_by_str(dname, strhashval, true);
1166 if (!same_verf(&verf, &conf->cl_verifier)) {
1167 /* 18.35.4 case 8 */
1168 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1169 status = nfserr_not_same;
1172 /* Client reboot: destroy old state */
1173 expire_client(conf);
1176 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1177 /* 18.35.4 case 9 */
1178 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1179 status = nfserr_perm;
1182 expire_client(conf);
1186 * Set bit when the owner id and verifier map to an already
1187 * confirmed client id (18.35.3).
1189 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1192 * Falling into 18.35.4 case 2, possible router replay.
1193 * Leave confirmed record intact and return same result.
1195 copy_verf(conf, &verf);
1200 /* 18.35.4 case 7 */
1201 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1202 status = nfserr_noent;
1206 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1209 * Possible retry or client restart. Per 18.35.4 case 4,
1210 * a new unconfirmed record should be generated regardless
1211 * of whether any properties have changed.
1213 expire_client(unconf);
1218 new = create_client(exid->clname, dname, rqstp, &verf);
1220 status = nfserr_jukebox;
1225 add_to_unconfirmed(new, strhashval);
1227 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1228 exid->clientid.cl_id = new->cl_clientid.cl_id;
1231 nfsd4_set_ex_flags(new, exid);
1233 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1234 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1238 nfs4_unlock_state();
1240 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1245 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1247 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1250 /* The slot is in use, and no response has been sent. */
1252 if (seqid == slot_seqid)
1253 return nfserr_jukebox;
1255 return nfserr_seq_misordered;
1258 if (likely(seqid == slot_seqid + 1))
1261 if (seqid == slot_seqid)
1262 return nfserr_replay_cache;
1264 if (seqid == 1 && (slot_seqid + 1) == 0)
1266 /* Misordered replay or misordered new request */
1267 return nfserr_seq_misordered;
1271 * Cache the create session result into the create session single DRC
1272 * slot cache by saving the xdr structure. sl_seqid has been set.
1273 * Do this for solo or embedded create session operations.
1276 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1277 struct nfsd4_clid_slot *slot, int nfserr)
1279 slot->sl_status = nfserr;
1280 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1284 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1285 struct nfsd4_clid_slot *slot)
1287 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1288 return slot->sl_status;
1292 nfsd4_create_session(struct svc_rqst *rqstp,
1293 struct nfsd4_compound_state *cstate,
1294 struct nfsd4_create_session *cr_ses)
1296 struct sockaddr *sa = svc_addr(rqstp);
1297 struct nfs4_client *conf, *unconf;
1298 struct nfsd4_clid_slot *cs_slot = NULL;
1302 unconf = find_unconfirmed_client(&cr_ses->clientid);
1303 conf = find_confirmed_client(&cr_ses->clientid);
1306 cs_slot = &conf->cl_cs_slot;
1307 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1308 if (status == nfserr_replay_cache) {
1309 dprintk("Got a create_session replay! seqid= %d\n",
1311 /* Return the cached reply status */
1312 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1314 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1315 status = nfserr_seq_misordered;
1316 dprintk("Sequence misordered!\n");
1317 dprintk("Expected seqid= %d but got seqid= %d\n",
1318 cs_slot->sl_seqid, cr_ses->seqid);
1321 cs_slot->sl_seqid++;
1322 } else if (unconf) {
1323 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1324 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1325 status = nfserr_clid_inuse;
1329 cs_slot = &unconf->cl_cs_slot;
1330 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1332 /* an unconfirmed replay returns misordered */
1333 status = nfserr_seq_misordered;
1337 cs_slot->sl_seqid++; /* from 0 to 1 */
1338 move_to_confirmed(unconf);
1340 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1341 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1342 svc_xprt_get(rqstp->rq_xprt);
1344 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1346 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1347 unconf->cl_cb_conn.cb_minorversion =
1348 cstate->minorversion;
1349 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1350 unconf->cl_cb_seq_nr = 1;
1351 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1355 status = nfserr_stale_clientid;
1360 * We do not support RDMA or persistent sessions
1362 cr_ses->flags &= ~SESSION4_PERSIST;
1363 cr_ses->flags &= ~SESSION4_RDMA;
1365 status = alloc_init_session(rqstp, conf, cr_ses);
1369 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1370 NFS4_MAX_SESSIONID_LEN);
1371 cr_ses->seqid = cs_slot->sl_seqid;
1374 /* cache solo and embedded create sessions under the state lock */
1375 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1377 nfs4_unlock_state();
1378 dprintk("%s returns %d\n", __func__, ntohl(status));
1382 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1384 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1385 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1387 return argp->opcnt == resp->opcnt;
1390 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1394 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1398 nfsd4_destroy_session(struct svc_rqst *r,
1399 struct nfsd4_compound_state *cstate,
1400 struct nfsd4_destroy_session *sessionid)
1402 struct nfsd4_session *ses;
1403 u32 status = nfserr_badsession;
1406 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1407 * - Should we return nfserr_back_chan_busy if waiting for
1408 * callbacks on to-be-destroyed session?
1409 * - Do we need to clear any callback info from previous session?
1412 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1413 if (!nfsd4_last_compound_op(r))
1414 return nfserr_not_only_op;
1416 dump_sessionid(__func__, &sessionid->sessionid);
1417 spin_lock(&client_lock);
1418 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1420 spin_unlock(&client_lock);
1424 unhash_session(ses);
1425 spin_unlock(&client_lock);
1428 /* wait for callbacks */
1429 nfsd4_set_callback_client(ses->se_client, NULL);
1430 nfs4_unlock_state();
1431 nfsd4_put_session(ses);
1434 dprintk("%s returns %d\n", __func__, ntohl(status));
1439 nfsd4_sequence(struct svc_rqst *rqstp,
1440 struct nfsd4_compound_state *cstate,
1441 struct nfsd4_sequence *seq)
1443 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1444 struct nfsd4_session *session;
1445 struct nfsd4_slot *slot;
1448 if (resp->opcnt != 1)
1449 return nfserr_sequence_pos;
1451 spin_lock(&client_lock);
1452 status = nfserr_badsession;
1453 session = find_in_sessionid_hashtbl(&seq->sessionid);
1457 status = nfserr_badslot;
1458 if (seq->slotid >= session->se_fchannel.maxreqs)
1461 slot = session->se_slots[seq->slotid];
1462 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1464 /* We do not negotiate the number of slots yet, so set the
1465 * maxslots to the session maxreqs which is used to encode
1466 * sr_highest_slotid and the sr_target_slot id to maxslots */
1467 seq->maxslots = session->se_fchannel.maxreqs;
1469 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1470 if (status == nfserr_replay_cache) {
1471 cstate->slot = slot;
1472 cstate->session = session;
1473 /* Return the cached reply status and set cstate->status
1474 * for nfsd4_proc_compound processing */
1475 status = nfsd4_replay_cache_entry(resp, seq);
1476 cstate->status = nfserr_replay_cache;
1482 /* Success! bump slot seqid */
1483 slot->sl_inuse = true;
1484 slot->sl_seqid = seq->seqid;
1485 slot->sl_cachethis = seq->cachethis;
1487 cstate->slot = slot;
1488 cstate->session = session;
1491 /* Hold a session reference until done processing the compound. */
1492 if (cstate->session) {
1493 nfsd4_get_session(cstate->session);
1494 atomic_inc(&session->se_client->cl_refcount);
1496 spin_unlock(&client_lock);
1497 dprintk("%s: return %d\n", __func__, ntohl(status));
1502 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1504 if (rc->rca_one_fs) {
1505 if (!cstate->current_fh.fh_dentry)
1506 return nfserr_nofilehandle;
1508 * We don't take advantage of the rca_one_fs case.
1509 * That's OK, it's optional, we can safely ignore it.
1514 if (is_client_expired(cstate->session->se_client)) {
1515 nfs4_unlock_state();
1517 * The following error isn't really legal.
1518 * But we only get here if the client just explicitly
1519 * destroyed the client. Surely it no longer cares what
1520 * error it gets back on an operation for the dead
1523 return nfserr_stale_clientid;
1525 nfsd4_create_clid_dir(cstate->session->se_client);
1526 nfs4_unlock_state();
1531 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1532 struct nfsd4_setclientid *setclid)
1534 struct sockaddr *sa = svc_addr(rqstp);
1535 struct xdr_netobj clname = {
1536 .len = setclid->se_namelen,
1537 .data = setclid->se_name,
1539 nfs4_verifier clverifier = setclid->se_verf;
1540 unsigned int strhashval;
1541 struct nfs4_client *conf, *unconf, *new;
1543 char dname[HEXDIR_LEN];
1545 if (!check_name(clname))
1546 return nfserr_inval;
1548 status = nfs4_make_rec_clidname(dname, &clname);
1553 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1554 * We get here on a DRC miss.
1557 strhashval = clientstr_hashval(dname);
1560 conf = find_confirmed_client_by_str(dname, strhashval, false);
1562 /* RFC 3530 14.2.33 CASE 0: */
1563 status = nfserr_clid_inuse;
1564 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1565 char addr_str[INET6_ADDRSTRLEN];
1566 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1568 dprintk("NFSD: setclientid: string in use by client "
1569 "at %s\n", addr_str);
1574 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1575 * has a description of SETCLIENTID request processing consisting
1576 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1578 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1579 status = nfserr_resource;
1582 * RFC 3530 14.2.33 CASE 4:
1583 * placed first, because it is the normal case
1586 expire_client(unconf);
1587 new = create_client(clname, dname, rqstp, &clverifier);
1591 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1593 * RFC 3530 14.2.33 CASE 1:
1594 * probable callback update
1597 /* Note this is removing unconfirmed {*x***},
1598 * which is stronger than RFC recommended {vxc**}.
1599 * This has the advantage that there is at most
1600 * one {*x***} in either list at any time.
1602 expire_client(unconf);
1604 new = create_client(clname, dname, rqstp, &clverifier);
1607 copy_clid(new, conf);
1608 } else if (!unconf) {
1610 * RFC 3530 14.2.33 CASE 2:
1611 * probable client reboot; state will be removed if
1614 new = create_client(clname, dname, rqstp, &clverifier);
1620 * RFC 3530 14.2.33 CASE 3:
1621 * probable client reboot; state will be removed if
1624 expire_client(unconf);
1625 new = create_client(clname, dname, rqstp, &clverifier);
1630 gen_callback(new, setclid, rpc_get_scope_id(sa));
1631 add_to_unconfirmed(new, strhashval);
1632 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1633 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1634 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1637 nfs4_unlock_state();
1643 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1644 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1645 * bullets, labeled as CASE1 - CASE4 below.
1648 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1649 struct nfsd4_compound_state *cstate,
1650 struct nfsd4_setclientid_confirm *setclientid_confirm)
1652 struct sockaddr *sa = svc_addr(rqstp);
1653 struct nfs4_client *conf, *unconf;
1654 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1655 clientid_t * clid = &setclientid_confirm->sc_clientid;
1658 if (STALE_CLIENTID(clid))
1659 return nfserr_stale_clientid;
1661 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1662 * We get here on a DRC miss.
1667 conf = find_confirmed_client(clid);
1668 unconf = find_unconfirmed_client(clid);
1670 status = nfserr_clid_inuse;
1671 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1673 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1677 * section 14.2.34 of RFC 3530 has a description of
1678 * SETCLIENTID_CONFIRM request processing consisting
1679 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1681 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1683 * RFC 3530 14.2.34 CASE 1:
1686 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1687 status = nfserr_clid_inuse;
1689 atomic_set(&conf->cl_cb_set, 0);
1690 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1691 expire_client(unconf);
1695 } else if (conf && !unconf) {
1697 * RFC 3530 14.2.34 CASE 2:
1698 * probable retransmitted request; play it safe and
1701 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1702 status = nfserr_clid_inuse;
1705 } else if (!conf && unconf
1706 && same_verf(&unconf->cl_confirm, &confirm)) {
1708 * RFC 3530 14.2.34 CASE 3:
1709 * Normal case; new or rebooted client:
1711 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1712 status = nfserr_clid_inuse;
1715 clientstr_hashval(unconf->cl_recdir);
1716 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1719 nfsd4_remove_clid_dir(conf);
1720 expire_client(conf);
1722 move_to_confirmed(unconf);
1724 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1727 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1728 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1731 * RFC 3530 14.2.34 CASE 4:
1732 * Client probably hasn't noticed that we rebooted yet.
1734 status = nfserr_stale_clientid;
1736 /* check that we have hit one of the cases...*/
1737 status = nfserr_clid_inuse;
1740 nfs4_unlock_state();
1744 /* OPEN Share state helper functions */
1745 static inline struct nfs4_file *
1746 alloc_init_file(struct inode *ino)
1748 struct nfs4_file *fp;
1749 unsigned int hashval = file_hashval(ino);
1751 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1753 atomic_set(&fp->fi_ref, 1);
1754 INIT_LIST_HEAD(&fp->fi_hash);
1755 INIT_LIST_HEAD(&fp->fi_stateids);
1756 INIT_LIST_HEAD(&fp->fi_delegations);
1757 fp->fi_inode = igrab(ino);
1758 fp->fi_id = current_fileid++;
1759 fp->fi_had_conflict = false;
1760 spin_lock(&recall_lock);
1761 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1762 spin_unlock(&recall_lock);
1769 nfsd4_free_slab(struct kmem_cache **slab)
1773 kmem_cache_destroy(*slab);
1778 nfsd4_free_slabs(void)
1780 nfsd4_free_slab(&stateowner_slab);
1781 nfsd4_free_slab(&file_slab);
1782 nfsd4_free_slab(&stateid_slab);
1783 nfsd4_free_slab(&deleg_slab);
1787 nfsd4_init_slabs(void)
1789 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1790 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1791 if (stateowner_slab == NULL)
1793 file_slab = kmem_cache_create("nfsd4_files",
1794 sizeof(struct nfs4_file), 0, 0, NULL);
1795 if (file_slab == NULL)
1797 stateid_slab = kmem_cache_create("nfsd4_stateids",
1798 sizeof(struct nfs4_stateid), 0, 0, NULL);
1799 if (stateid_slab == NULL)
1801 deleg_slab = kmem_cache_create("nfsd4_delegations",
1802 sizeof(struct nfs4_delegation), 0, 0, NULL);
1803 if (deleg_slab == NULL)
1808 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1813 nfs4_free_stateowner(struct kref *kref)
1815 struct nfs4_stateowner *sop =
1816 container_of(kref, struct nfs4_stateowner, so_ref);
1817 kfree(sop->so_owner.data);
1818 kmem_cache_free(stateowner_slab, sop);
1821 static inline struct nfs4_stateowner *
1822 alloc_stateowner(struct xdr_netobj *owner)
1824 struct nfs4_stateowner *sop;
1826 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1827 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1828 memcpy(sop->so_owner.data, owner->data, owner->len);
1829 sop->so_owner.len = owner->len;
1830 kref_init(&sop->so_ref);
1833 kmem_cache_free(stateowner_slab, sop);
1838 static struct nfs4_stateowner *
1839 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1840 struct nfs4_stateowner *sop;
1841 struct nfs4_replay *rp;
1842 unsigned int idhashval;
1844 if (!(sop = alloc_stateowner(&open->op_owner)))
1846 idhashval = ownerid_hashval(current_ownerid);
1847 INIT_LIST_HEAD(&sop->so_idhash);
1848 INIT_LIST_HEAD(&sop->so_strhash);
1849 INIT_LIST_HEAD(&sop->so_perclient);
1850 INIT_LIST_HEAD(&sop->so_stateids);
1851 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1852 INIT_LIST_HEAD(&sop->so_close_lru);
1854 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1855 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1856 list_add(&sop->so_perclient, &clp->cl_openowners);
1857 sop->so_is_open_owner = 1;
1858 sop->so_id = current_ownerid++;
1859 sop->so_client = clp;
1860 sop->so_seqid = open->op_seqid;
1861 sop->so_confirmed = 0;
1862 rp = &sop->so_replay;
1863 rp->rp_status = nfserr_serverfault;
1865 rp->rp_buf = rp->rp_ibuf;
1870 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1871 struct nfs4_stateowner *sop = open->op_stateowner;
1872 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1874 INIT_LIST_HEAD(&stp->st_hash);
1875 INIT_LIST_HEAD(&stp->st_perstateowner);
1876 INIT_LIST_HEAD(&stp->st_lockowners);
1877 INIT_LIST_HEAD(&stp->st_perfile);
1878 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1879 list_add(&stp->st_perstateowner, &sop->so_stateids);
1880 list_add(&stp->st_perfile, &fp->fi_stateids);
1881 stp->st_stateowner = sop;
1884 stp->st_stateid.si_boot = boot_time;
1885 stp->st_stateid.si_stateownerid = sop->so_id;
1886 stp->st_stateid.si_fileid = fp->fi_id;
1887 stp->st_stateid.si_generation = 0;
1888 stp->st_access_bmap = 0;
1889 stp->st_deny_bmap = 0;
1890 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1891 &stp->st_access_bmap);
1892 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1893 stp->st_openstp = NULL;
1897 move_to_close_lru(struct nfs4_stateowner *sop)
1899 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1901 list_move_tail(&sop->so_close_lru, &close_lru);
1902 sop->so_time = get_seconds();
1906 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1909 return (sop->so_owner.len == owner->len) &&
1910 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1911 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1914 static struct nfs4_stateowner *
1915 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1917 struct nfs4_stateowner *so = NULL;
1919 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1920 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1926 /* search file_hashtbl[] for file */
1927 static struct nfs4_file *
1928 find_file(struct inode *ino)
1930 unsigned int hashval = file_hashval(ino);
1931 struct nfs4_file *fp;
1933 spin_lock(&recall_lock);
1934 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1935 if (fp->fi_inode == ino) {
1937 spin_unlock(&recall_lock);
1941 spin_unlock(&recall_lock);
1945 static inline int access_valid(u32 x, u32 minorversion)
1947 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1949 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1951 x &= ~NFS4_SHARE_ACCESS_MASK;
1952 if (minorversion && x) {
1953 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1955 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1957 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1964 static inline int deny_valid(u32 x)
1966 /* Note: unlike access bits, deny bits may be zero. */
1967 return x <= NFS4_SHARE_DENY_BOTH;
1971 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1972 * st_{access,deny}_bmap field of the stateid, in order to track not
1973 * only what share bits are currently in force, but also what
1974 * combinations of share bits previous opens have used. This allows us
1975 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1976 * return an error if the client attempt to downgrade to a combination
1977 * of share bits not explicable by closing some of its previous opens.
1979 * XXX: This enforcement is actually incomplete, since we don't keep
1980 * track of access/deny bit combinations; so, e.g., we allow:
1982 * OPEN allow read, deny write
1983 * OPEN allow both, deny none
1984 * DOWNGRADE allow read, deny none
1986 * which we should reject.
1989 set_access(unsigned int *access, unsigned long bmap) {
1993 for (i = 1; i < 4; i++) {
1994 if (test_bit(i, &bmap))
2000 set_deny(unsigned int *deny, unsigned long bmap) {
2004 for (i = 0; i < 4; i++) {
2005 if (test_bit(i, &bmap))
2011 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2012 unsigned int access, deny;
2014 set_access(&access, stp->st_access_bmap);
2015 set_deny(&deny, stp->st_deny_bmap);
2016 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2022 * Called to check deny when READ with all zero stateid or
2023 * WRITE with all zero or all one stateid
2026 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2028 struct inode *ino = current_fh->fh_dentry->d_inode;
2029 struct nfs4_file *fp;
2030 struct nfs4_stateid *stp;
2033 dprintk("NFSD: nfs4_share_conflict\n");
2035 fp = find_file(ino);
2038 ret = nfserr_locked;
2039 /* Search for conflicting share reservations */
2040 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2041 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2042 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2052 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2054 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
2055 drop_file_write_access(filp);
2056 spin_lock(&filp->f_lock);
2057 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2058 spin_unlock(&filp->f_lock);
2063 * Spawn a thread to perform a recall on the delegation represented
2064 * by the lease (file_lock)
2066 * Called from break_lease() with lock_kernel() held.
2067 * Note: we assume break_lease will only call this *once* for any given
2071 void nfsd_break_deleg_cb(struct file_lock *fl)
2073 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2075 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2079 /* We're assuming the state code never drops its reference
2080 * without first removing the lease. Since we're in this lease
2081 * callback (and since the lease code is serialized by the kernel
2082 * lock) we know the server hasn't removed the lease yet, we know
2083 * it's safe to take a reference: */
2084 atomic_inc(&dp->dl_count);
2086 spin_lock(&recall_lock);
2087 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2088 spin_unlock(&recall_lock);
2090 /* only place dl_time is set. protected by lock_kernel*/
2091 dp->dl_time = get_seconds();
2094 * We don't want the locks code to timeout the lease for us;
2095 * we'll remove it ourself if the delegation isn't returned
2098 fl->fl_break_time = 0;
2100 dp->dl_file->fi_had_conflict = true;
2101 nfsd4_cb_recall(dp);
2105 * The file_lock is being reapd.
2107 * Called by locks_free_lock() with lock_kernel() held.
2110 void nfsd_release_deleg_cb(struct file_lock *fl)
2112 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2114 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2116 if (!(fl->fl_flags & FL_LEASE) || !dp)
2118 dp->dl_flock = NULL;
2122 * Set the delegation file_lock back pointer.
2124 * Called from setlease() with lock_kernel() held.
2127 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2129 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2131 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2138 * Called from setlease() with lock_kernel() held
2141 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2143 struct nfs4_delegation *onlistd =
2144 (struct nfs4_delegation *)onlist->fl_owner;
2145 struct nfs4_delegation *tryd =
2146 (struct nfs4_delegation *)try->fl_owner;
2148 if (onlist->fl_lmops != try->fl_lmops)
2151 return onlistd->dl_client == tryd->dl_client;
2156 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2159 return lease_modify(onlist, arg);
2164 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2165 .fl_break = nfsd_break_deleg_cb,
2166 .fl_release_private = nfsd_release_deleg_cb,
2167 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2168 .fl_mylease = nfsd_same_client_deleg_cb,
2169 .fl_change = nfsd_change_deleg_cb,
2174 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2175 struct nfsd4_open *open)
2177 clientid_t *clientid = &open->op_clientid;
2178 struct nfs4_client *clp = NULL;
2179 unsigned int strhashval;
2180 struct nfs4_stateowner *sop = NULL;
2182 if (!check_name(open->op_owner))
2183 return nfserr_inval;
2185 if (STALE_CLIENTID(&open->op_clientid))
2186 return nfserr_stale_clientid;
2188 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2189 sop = find_openstateowner_str(strhashval, open);
2190 open->op_stateowner = sop;
2192 /* Make sure the client's lease hasn't expired. */
2193 clp = find_confirmed_client(clientid);
2195 return nfserr_expired;
2198 /* When sessions are used, skip open sequenceid processing */
2199 if (nfsd4_has_session(cstate))
2201 if (!sop->so_confirmed) {
2202 /* Replace unconfirmed owners without checking for replay. */
2203 clp = sop->so_client;
2204 release_openowner(sop);
2205 open->op_stateowner = NULL;
2208 if (open->op_seqid == sop->so_seqid - 1) {
2209 if (sop->so_replay.rp_buflen)
2210 return nfserr_replay_me;
2211 /* The original OPEN failed so spectacularly
2212 * that we don't even have replay data saved!
2213 * Therefore, we have no choice but to continue
2214 * processing this OPEN; presumably, we'll
2215 * fail again for the same reason.
2217 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2220 if (open->op_seqid != sop->so_seqid)
2221 return nfserr_bad_seqid;
2223 if (open->op_stateowner == NULL) {
2224 sop = alloc_init_open_stateowner(strhashval, clp, open);
2226 return nfserr_resource;
2227 open->op_stateowner = sop;
2229 list_del_init(&sop->so_close_lru);
2230 renew_client(sop->so_client);
2234 static inline __be32
2235 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2237 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2238 return nfserr_openmode;
2243 static struct nfs4_delegation *
2244 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2246 struct nfs4_delegation *dp;
2248 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2249 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2255 int share_access_to_flags(u32 share_access)
2257 share_access &= ~NFS4_SHARE_WANT_MASK;
2259 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2263 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2264 struct nfs4_delegation **dp)
2267 __be32 status = nfserr_bad_stateid;
2269 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2272 flags = share_access_to_flags(open->op_share_access);
2273 status = nfs4_check_delegmode(*dp, flags);
2277 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2281 open->op_stateowner->so_confirmed = 1;
2286 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2288 struct nfs4_stateid *local;
2289 __be32 status = nfserr_share_denied;
2290 struct nfs4_stateowner *sop = open->op_stateowner;
2292 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2293 /* ignore lock owners */
2294 if (local->st_stateowner->so_is_open_owner == 0)
2296 /* remember if we have seen this open owner */
2297 if (local->st_stateowner == sop)
2299 /* check for conflicting share reservations */
2300 if (!test_share(local, open))
2308 static inline struct nfs4_stateid *
2309 nfs4_alloc_stateid(void)
2311 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2315 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2316 struct nfs4_delegation *dp,
2317 struct svc_fh *cur_fh, int flags)
2319 struct nfs4_stateid *stp;
2321 stp = nfs4_alloc_stateid();
2323 return nfserr_resource;
2326 get_file(dp->dl_vfs_file);
2327 stp->st_vfs_file = dp->dl_vfs_file;
2330 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2333 if (status == nfserr_dropit)
2334 status = nfserr_jukebox;
2335 kmem_cache_free(stateid_slab, stp);
2343 static inline __be32
2344 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2345 struct nfsd4_open *open)
2347 struct iattr iattr = {
2348 .ia_valid = ATTR_SIZE,
2351 if (!open->op_truncate)
2353 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2354 return nfserr_inval;
2355 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2359 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2361 struct file *filp = stp->st_vfs_file;
2362 struct inode *inode = filp->f_path.dentry->d_inode;
2363 unsigned int share_access, new_writer;
2364 u32 op_share_access;
2367 set_access(&share_access, stp->st_access_bmap);
2368 new_writer = (~share_access) & open->op_share_access
2369 & NFS4_SHARE_ACCESS_WRITE;
2372 int err = get_write_access(inode);
2374 return nfserrno(err);
2375 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2377 return nfserrno(err);
2378 file_take_write(filp);
2380 status = nfsd4_truncate(rqstp, cur_fh, open);
2383 put_write_access(inode);
2386 /* remember the open */
2387 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2388 filp->f_mode |= op_share_access;
2389 __set_bit(op_share_access, &stp->st_access_bmap);
2390 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2397 nfs4_set_claim_prev(struct nfsd4_open *open)
2399 open->op_stateowner->so_confirmed = 1;
2400 open->op_stateowner->so_client->cl_firststate = 1;
2404 * Attempt to hand out a delegation.
2407 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2409 struct nfs4_delegation *dp;
2410 struct nfs4_stateowner *sop = stp->st_stateowner;
2411 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2412 struct file_lock fl, *flp = &fl;
2413 int status, flag = 0;
2415 flag = NFS4_OPEN_DELEGATE_NONE;
2416 open->op_recall = 0;
2417 switch (open->op_claim_type) {
2418 case NFS4_OPEN_CLAIM_PREVIOUS:
2420 open->op_recall = 1;
2421 flag = open->op_delegate_type;
2422 if (flag == NFS4_OPEN_DELEGATE_NONE)
2425 case NFS4_OPEN_CLAIM_NULL:
2426 /* Let's not give out any delegations till everyone's
2427 * had the chance to reclaim theirs.... */
2428 if (locks_in_grace())
2430 if (!cb_up || !sop->so_confirmed)
2432 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2433 flag = NFS4_OPEN_DELEGATE_WRITE;
2435 flag = NFS4_OPEN_DELEGATE_READ;
2441 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2443 flag = NFS4_OPEN_DELEGATE_NONE;
2446 locks_init_lock(&fl);
2447 fl.fl_lmops = &nfsd_lease_mng_ops;
2448 fl.fl_flags = FL_LEASE;
2449 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2450 fl.fl_end = OFFSET_MAX;
2451 fl.fl_owner = (fl_owner_t)dp;
2452 fl.fl_file = stp->st_vfs_file;
2453 fl.fl_pid = current->tgid;
2455 /* vfs_setlease checks to see if delegation should be handed out.
2456 * the lock_manager callbacks fl_mylease and fl_change are used
2458 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2459 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2460 unhash_delegation(dp);
2461 flag = NFS4_OPEN_DELEGATE_NONE;
2465 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2467 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2468 STATEID_VAL(&dp->dl_stateid));
2470 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2471 && flag == NFS4_OPEN_DELEGATE_NONE
2472 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2473 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2474 open->op_delegate_type = flag;
2478 * called with nfs4_lock_state() held.
2481 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2483 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2484 struct nfs4_file *fp = NULL;
2485 struct inode *ino = current_fh->fh_dentry->d_inode;
2486 struct nfs4_stateid *stp = NULL;
2487 struct nfs4_delegation *dp = NULL;
2490 status = nfserr_inval;
2491 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2492 || !deny_valid(open->op_share_deny))
2495 * Lookup file; if found, lookup stateid and check open request,
2496 * and check for delegations in the process of being recalled.
2497 * If not found, create the nfs4_file struct
2499 fp = find_file(ino);
2501 if ((status = nfs4_check_open(fp, open, &stp)))
2503 status = nfs4_check_deleg(fp, open, &dp);
2507 status = nfserr_bad_stateid;
2508 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2510 status = nfserr_resource;
2511 fp = alloc_init_file(ino);
2517 * OPEN the file, or upgrade an existing OPEN.
2518 * If truncate fails, the OPEN fails.
2521 /* Stateid was found, this is an OPEN upgrade */
2522 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2525 update_stateid(&stp->st_stateid);
2527 /* Stateid was not found, this is a new OPEN */
2529 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2530 flags |= NFSD_MAY_READ;
2531 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2532 flags |= NFSD_MAY_WRITE;
2533 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2536 init_stateid(stp, fp, open);
2537 status = nfsd4_truncate(rqstp, current_fh, open);
2539 release_open_stateid(stp);
2542 if (nfsd4_has_session(&resp->cstate))
2543 update_stateid(&stp->st_stateid);
2545 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2547 if (nfsd4_has_session(&resp->cstate))
2548 open->op_stateowner->so_confirmed = 1;
2551 * Attempt to hand out a delegation. No error return, because the
2552 * OPEN succeeds even if we fail.
2554 nfs4_open_delegation(current_fh, open, stp);
2558 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2559 STATEID_VAL(&stp->st_stateid));
2563 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2564 nfs4_set_claim_prev(open);
2566 * To finish the open response, we just need to set the rflags.
2568 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2569 if (!open->op_stateowner->so_confirmed &&
2570 !nfsd4_has_session(&resp->cstate))
2571 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2577 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2580 struct nfs4_client *clp;
2584 dprintk("process_renew(%08x/%08x): starting\n",
2585 clid->cl_boot, clid->cl_id);
2586 status = nfserr_stale_clientid;
2587 if (STALE_CLIENTID(clid))
2589 clp = find_confirmed_client(clid);
2590 status = nfserr_expired;
2592 /* We assume the client took too long to RENEW. */
2593 dprintk("nfsd4_renew: clientid not found!\n");
2597 status = nfserr_cb_path_down;
2598 if (!list_empty(&clp->cl_delegations)
2599 && !atomic_read(&clp->cl_cb_set))
2603 nfs4_unlock_state();
2607 struct lock_manager nfsd4_manager = {
2611 nfsd4_end_grace(void)
2613 dprintk("NFSD: end of grace period\n");
2614 nfsd4_recdir_purge_old();
2615 locks_end_grace(&nfsd4_manager);
2617 * Now that every NFSv4 client has had the chance to recover and
2618 * to see the (possibly new, possibly shorter) lease time, we
2619 * can safely set the next grace time to the current lease time:
2621 nfsd4_grace = nfsd4_lease;
2625 nfs4_laundromat(void)
2627 struct nfs4_client *clp;
2628 struct nfs4_stateowner *sop;
2629 struct nfs4_delegation *dp;
2630 struct list_head *pos, *next, reaplist;
2631 time_t cutoff = get_seconds() - nfsd4_lease;
2632 time_t t, clientid_val = nfsd4_lease;
2633 time_t u, test_val = nfsd4_lease;
2637 dprintk("NFSD: laundromat service - starting\n");
2638 if (locks_in_grace())
2640 INIT_LIST_HEAD(&reaplist);
2641 spin_lock(&client_lock);
2642 list_for_each_safe(pos, next, &client_lru) {
2643 clp = list_entry(pos, struct nfs4_client, cl_lru);
2644 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2645 t = clp->cl_time - cutoff;
2646 if (clientid_val > t)
2650 if (atomic_read(&clp->cl_refcount)) {
2651 dprintk("NFSD: client in use (clientid %08x)\n",
2652 clp->cl_clientid.cl_id);
2655 unhash_client_locked(clp);
2656 list_add(&clp->cl_lru, &reaplist);
2658 spin_unlock(&client_lock);
2659 list_for_each_safe(pos, next, &reaplist) {
2660 clp = list_entry(pos, struct nfs4_client, cl_lru);
2661 dprintk("NFSD: purging unused client (clientid %08x)\n",
2662 clp->cl_clientid.cl_id);
2663 nfsd4_remove_clid_dir(clp);
2666 spin_lock(&recall_lock);
2667 list_for_each_safe(pos, next, &del_recall_lru) {
2668 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2669 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2670 u = dp->dl_time - cutoff;
2675 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2677 list_move(&dp->dl_recall_lru, &reaplist);
2679 spin_unlock(&recall_lock);
2680 list_for_each_safe(pos, next, &reaplist) {
2681 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2682 list_del_init(&dp->dl_recall_lru);
2683 unhash_delegation(dp);
2685 test_val = nfsd4_lease;
2686 list_for_each_safe(pos, next, &close_lru) {
2687 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2688 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2689 u = sop->so_time - cutoff;
2694 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2696 release_openowner(sop);
2698 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2699 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2700 nfs4_unlock_state();
2701 return clientid_val;
2704 static struct workqueue_struct *laundry_wq;
2705 static void laundromat_main(struct work_struct *);
2706 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2709 laundromat_main(struct work_struct *not_used)
2713 t = nfs4_laundromat();
2714 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2715 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2718 static struct nfs4_stateowner *
2719 search_close_lru(u32 st_id, int flags)
2721 struct nfs4_stateowner *local = NULL;
2723 if (flags & CLOSE_STATE) {
2724 list_for_each_entry(local, &close_lru, so_close_lru) {
2725 if (local->so_id == st_id)
2733 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2735 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2739 STALE_STATEID(stateid_t *stateid)
2741 if (stateid->si_boot == boot_time)
2743 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2744 STATEID_VAL(stateid));
2749 access_permit_read(unsigned long access_bmap)
2751 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2752 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2753 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2757 access_permit_write(unsigned long access_bmap)
2759 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2760 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2764 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2766 __be32 status = nfserr_openmode;
2768 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2770 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2777 static inline __be32
2778 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2780 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2782 else if (locks_in_grace()) {
2783 /* Answer in remaining cases depends on existance of
2784 * conflicting state; so we must wait out the grace period. */
2785 return nfserr_grace;
2786 } else if (flags & WR_STATE)
2787 return nfs4_share_conflict(current_fh,
2788 NFS4_SHARE_DENY_WRITE);
2789 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2790 return nfs4_share_conflict(current_fh,
2791 NFS4_SHARE_DENY_READ);
2795 * Allow READ/WRITE during grace period on recovered state only for files
2796 * that are not able to provide mandatory locking.
2799 grace_disallows_io(struct inode *inode)
2801 return locks_in_grace() && mandatory_lock(inode);
2804 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2807 * When sessions are used the stateid generation number is ignored
2810 if ((flags & HAS_SESSION) && in->si_generation == 0)
2813 /* If the client sends us a stateid from the future, it's buggy: */
2814 if (in->si_generation > ref->si_generation)
2815 return nfserr_bad_stateid;
2817 * The following, however, can happen. For example, if the
2818 * client sends an open and some IO at the same time, the open
2819 * may bump si_generation while the IO is still in flight.
2820 * Thanks to hard links and renames, the client never knows what
2821 * file an open will affect. So it could avoid that situation
2822 * only by serializing all opens and IO from the same open
2823 * owner. To recover from the old_stateid error, the client
2824 * will just have to retry the IO:
2826 if (in->si_generation < ref->si_generation)
2827 return nfserr_old_stateid;
2832 static int is_delegation_stateid(stateid_t *stateid)
2834 return stateid->si_fileid == 0;
2838 * Checks for stateid operations
2841 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2842 stateid_t *stateid, int flags, struct file **filpp)
2844 struct nfs4_stateid *stp = NULL;
2845 struct nfs4_delegation *dp = NULL;
2846 struct svc_fh *current_fh = &cstate->current_fh;
2847 struct inode *ino = current_fh->fh_dentry->d_inode;
2853 if (grace_disallows_io(ino))
2854 return nfserr_grace;
2856 if (nfsd4_has_session(cstate))
2857 flags |= HAS_SESSION;
2859 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2860 return check_special_stateids(current_fh, stateid, flags);
2862 status = nfserr_stale_stateid;
2863 if (STALE_STATEID(stateid))
2866 status = nfserr_bad_stateid;
2867 if (is_delegation_stateid(stateid)) {
2868 dp = find_delegation_stateid(ino, stateid);
2871 status = check_stateid_generation(stateid, &dp->dl_stateid,
2875 status = nfs4_check_delegmode(dp, flags);
2878 renew_client(dp->dl_client);
2880 *filpp = dp->dl_vfs_file;
2881 } else { /* open or lock stateid */
2882 stp = find_stateid(stateid, flags);
2885 if (nfs4_check_fh(current_fh, stp))
2887 if (!stp->st_stateowner->so_confirmed)
2889 status = check_stateid_generation(stateid, &stp->st_stateid,
2893 status = nfs4_check_openmode(stp, flags);
2896 renew_client(stp->st_stateowner->so_client);
2898 *filpp = stp->st_vfs_file;
2908 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2909 RD_STATE : WR_STATE;
2913 * Checks for sequence id mutating operations.
2916 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2917 stateid_t *stateid, int flags,
2918 struct nfs4_stateowner **sopp,
2919 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2921 struct nfs4_stateid *stp;
2922 struct nfs4_stateowner *sop;
2923 struct svc_fh *current_fh = &cstate->current_fh;
2926 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2927 seqid, STATEID_VAL(stateid));
2932 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2933 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2934 return nfserr_bad_stateid;
2937 if (STALE_STATEID(stateid))
2938 return nfserr_stale_stateid;
2940 if (nfsd4_has_session(cstate))
2941 flags |= HAS_SESSION;
2944 * We return BAD_STATEID if filehandle doesn't match stateid,
2945 * the confirmed flag is incorrecly set, or the generation
2946 * number is incorrect.
2948 stp = find_stateid(stateid, flags);
2951 * Also, we should make sure this isn't just the result of
2954 sop = search_close_lru(stateid->si_stateownerid, flags);
2956 return nfserr_bad_stateid;
2962 *sopp = sop = stp->st_stateowner;
2965 clientid_t *lockclid = &lock->v.new.clientid;
2966 struct nfs4_client *clp = sop->so_client;
2970 lkflg = setlkflg(lock->lk_type);
2972 if (lock->lk_is_new) {
2973 if (!sop->so_is_open_owner)
2974 return nfserr_bad_stateid;
2975 if (!(flags & HAS_SESSION) &&
2976 !same_clid(&clp->cl_clientid, lockclid))
2977 return nfserr_bad_stateid;
2978 /* stp is the open stateid */
2979 status = nfs4_check_openmode(stp, lkflg);
2983 /* stp is the lock stateid */
2984 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2990 if (nfs4_check_fh(current_fh, stp)) {
2991 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2992 return nfserr_bad_stateid;
2996 * We now validate the seqid and stateid generation numbers.
2997 * For the moment, we ignore the possibility of
2998 * generation number wraparound.
3000 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
3003 if (sop->so_confirmed && flags & CONFIRM) {
3004 dprintk("NFSD: preprocess_seqid_op: expected"
3005 " unconfirmed stateowner!\n");
3006 return nfserr_bad_stateid;
3008 if (!sop->so_confirmed && !(flags & CONFIRM)) {
3009 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3010 " confirmed yet!\n");
3011 return nfserr_bad_stateid;
3013 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
3016 renew_client(sop->so_client);
3020 if (seqid == sop->so_seqid - 1) {
3021 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
3022 /* indicate replay to calling function */
3023 return nfserr_replay_me;
3025 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3026 sop->so_seqid, seqid);
3028 return nfserr_bad_seqid;
3032 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3033 struct nfsd4_open_confirm *oc)
3036 struct nfs4_stateowner *sop;
3037 struct nfs4_stateid *stp;
3039 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3040 (int)cstate->current_fh.fh_dentry->d_name.len,
3041 cstate->current_fh.fh_dentry->d_name.name);
3043 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3049 if ((status = nfs4_preprocess_seqid_op(cstate,
3050 oc->oc_seqid, &oc->oc_req_stateid,
3051 CONFIRM | OPEN_STATE,
3052 &oc->oc_stateowner, &stp, NULL)))
3055 sop = oc->oc_stateowner;
3056 sop->so_confirmed = 1;
3057 update_stateid(&stp->st_stateid);
3058 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3059 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3060 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3062 nfsd4_create_clid_dir(sop->so_client);
3064 if (oc->oc_stateowner) {
3065 nfs4_get_stateowner(oc->oc_stateowner);
3066 cstate->replay_owner = oc->oc_stateowner;
3068 nfs4_unlock_state();
3074 * unset all bits in union bitmap (bmap) that
3075 * do not exist in share (from successful OPEN_DOWNGRADE)
3078 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3081 for (i = 1; i < 4; i++) {
3082 if ((i & access) != i)
3083 __clear_bit(i, bmap);
3088 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3091 for (i = 0; i < 4; i++) {
3092 if ((i & deny) != i)
3093 __clear_bit(i, bmap);
3098 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3099 struct nfsd4_compound_state *cstate,
3100 struct nfsd4_open_downgrade *od)
3103 struct nfs4_stateid *stp;
3104 unsigned int share_access;
3106 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3107 (int)cstate->current_fh.fh_dentry->d_name.len,
3108 cstate->current_fh.fh_dentry->d_name.name);
3110 if (!access_valid(od->od_share_access, cstate->minorversion)
3111 || !deny_valid(od->od_share_deny))
3112 return nfserr_inval;
3115 if ((status = nfs4_preprocess_seqid_op(cstate,
3119 &od->od_stateowner, &stp, NULL)))
3122 status = nfserr_inval;
3123 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3124 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3125 stp->st_access_bmap, od->od_share_access);
3128 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3129 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3130 stp->st_deny_bmap, od->od_share_deny);
3133 set_access(&share_access, stp->st_access_bmap);
3134 nfs4_file_downgrade(stp->st_vfs_file,
3135 share_access & ~od->od_share_access);
3137 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3138 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3140 update_stateid(&stp->st_stateid);
3141 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3144 if (od->od_stateowner) {
3145 nfs4_get_stateowner(od->od_stateowner);
3146 cstate->replay_owner = od->od_stateowner;
3148 nfs4_unlock_state();
3153 * nfs4_unlock_state() called after encode
3156 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3157 struct nfsd4_close *close)
3160 struct nfs4_stateid *stp;
3162 dprintk("NFSD: nfsd4_close on file %.*s\n",
3163 (int)cstate->current_fh.fh_dentry->d_name.len,
3164 cstate->current_fh.fh_dentry->d_name.name);
3167 /* check close_lru for replay */
3168 if ((status = nfs4_preprocess_seqid_op(cstate,
3171 OPEN_STATE | CLOSE_STATE,
3172 &close->cl_stateowner, &stp, NULL)))
3175 update_stateid(&stp->st_stateid);
3176 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3178 /* release_stateid() calls nfsd_close() if needed */
3179 release_open_stateid(stp);
3181 /* place unused nfs4_stateowners on so_close_lru list to be
3182 * released by the laundromat service after the lease period
3183 * to enable us to handle CLOSE replay
3185 if (list_empty(&close->cl_stateowner->so_stateids))
3186 move_to_close_lru(close->cl_stateowner);
3188 if (close->cl_stateowner) {
3189 nfs4_get_stateowner(close->cl_stateowner);
3190 cstate->replay_owner = close->cl_stateowner;
3192 nfs4_unlock_state();
3197 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3198 struct nfsd4_delegreturn *dr)
3200 struct nfs4_delegation *dp;
3201 stateid_t *stateid = &dr->dr_stateid;
3202 struct inode *inode;
3206 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3208 inode = cstate->current_fh.fh_dentry->d_inode;
3210 if (nfsd4_has_session(cstate))
3211 flags |= HAS_SESSION;
3213 status = nfserr_bad_stateid;
3214 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3216 status = nfserr_stale_stateid;
3217 if (STALE_STATEID(stateid))
3219 status = nfserr_bad_stateid;
3220 if (!is_delegation_stateid(stateid))
3222 dp = find_delegation_stateid(inode, stateid);
3225 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3228 renew_client(dp->dl_client);
3230 unhash_delegation(dp);
3232 nfs4_unlock_state();
3239 * Lock owner state (byte-range locks)
3241 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3242 #define LOCK_HASH_BITS 8
3243 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3244 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3247 end_offset(u64 start, u64 len)
3252 return end >= start ? end: NFS4_MAX_UINT64;
3255 /* last octet in a range */
3257 last_byte_offset(u64 start, u64 len)
3263 return end > start ? end - 1: NFS4_MAX_UINT64;
3266 #define lockownerid_hashval(id) \
3267 ((id) & LOCK_HASH_MASK)
3269 static inline unsigned int
3270 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3271 struct xdr_netobj *ownername)
3273 return (file_hashval(inode) + cl_id
3274 + opaque_hashval(ownername->data, ownername->len))
3278 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3279 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3280 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3282 static struct nfs4_stateid *
3283 find_stateid(stateid_t *stid, int flags)
3285 struct nfs4_stateid *local;
3286 u32 st_id = stid->si_stateownerid;
3287 u32 f_id = stid->si_fileid;
3288 unsigned int hashval;
3290 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3291 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3292 hashval = stateid_hashval(st_id, f_id);
3293 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3294 if ((local->st_stateid.si_stateownerid == st_id) &&
3295 (local->st_stateid.si_fileid == f_id))
3300 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3301 hashval = stateid_hashval(st_id, f_id);
3302 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3303 if ((local->st_stateid.si_stateownerid == st_id) &&
3304 (local->st_stateid.si_fileid == f_id))
3311 static struct nfs4_delegation *
3312 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3314 struct nfs4_file *fp;
3315 struct nfs4_delegation *dl;
3317 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3320 fp = find_file(ino);
3323 dl = find_delegation_file(fp, stid);
3329 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3330 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3331 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3332 * locking, this prevents us from being completely protocol-compliant. The
3333 * real solution to this problem is to start using unsigned file offsets in
3334 * the VFS, but this is a very deep change!
3337 nfs4_transform_lock_offset(struct file_lock *lock)
3339 if (lock->fl_start < 0)
3340 lock->fl_start = OFFSET_MAX;
3341 if (lock->fl_end < 0)
3342 lock->fl_end = OFFSET_MAX;
3345 /* Hack!: For now, we're defining this just so we can use a pointer to it
3346 * as a unique cookie to identify our (NFSv4's) posix locks. */
3347 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3351 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3353 struct nfs4_stateowner *sop;
3356 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3357 sop = (struct nfs4_stateowner *) fl->fl_owner;
3358 hval = lockownerid_hashval(sop->so_id);
3359 kref_get(&sop->so_ref);
3361 deny->ld_clientid = sop->so_client->cl_clientid;
3363 deny->ld_sop = NULL;
3364 deny->ld_clientid.cl_boot = 0;
3365 deny->ld_clientid.cl_id = 0;
3367 deny->ld_start = fl->fl_start;
3368 deny->ld_length = NFS4_MAX_UINT64;
3369 if (fl->fl_end != NFS4_MAX_UINT64)
3370 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3371 deny->ld_type = NFS4_READ_LT;
3372 if (fl->fl_type != F_RDLCK)
3373 deny->ld_type = NFS4_WRITE_LT;
3376 static struct nfs4_stateowner *
3377 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3378 struct xdr_netobj *owner)
3380 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3381 struct nfs4_stateowner *op;
3383 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3384 if (same_owner_str(op, owner, clid))
3391 * Alloc a lock owner structure.
3392 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3395 * strhashval = lock_ownerstr_hashval
3398 static struct nfs4_stateowner *
3399 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3400 struct nfs4_stateowner *sop;
3401 struct nfs4_replay *rp;
3402 unsigned int idhashval;
3404 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3406 idhashval = lockownerid_hashval(current_ownerid);
3407 INIT_LIST_HEAD(&sop->so_idhash);
3408 INIT_LIST_HEAD(&sop->so_strhash);
3409 INIT_LIST_HEAD(&sop->so_perclient);
3410 INIT_LIST_HEAD(&sop->so_stateids);
3411 INIT_LIST_HEAD(&sop->so_perstateid);
3412 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3414 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3415 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3416 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3417 sop->so_is_open_owner = 0;
3418 sop->so_id = current_ownerid++;
3419 sop->so_client = clp;
3420 /* It is the openowner seqid that will be incremented in encode in the
3421 * case of new lockowners; so increment the lock seqid manually: */
3422 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3423 sop->so_confirmed = 1;
3424 rp = &sop->so_replay;
3425 rp->rp_status = nfserr_serverfault;
3427 rp->rp_buf = rp->rp_ibuf;
3431 static struct nfs4_stateid *
3432 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3434 struct nfs4_stateid *stp;
3435 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3437 stp = nfs4_alloc_stateid();
3440 INIT_LIST_HEAD(&stp->st_hash);
3441 INIT_LIST_HEAD(&stp->st_perfile);
3442 INIT_LIST_HEAD(&stp->st_perstateowner);
3443 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3444 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3445 list_add(&stp->st_perfile, &fp->fi_stateids);
3446 list_add(&stp->st_perstateowner, &sop->so_stateids);
3447 stp->st_stateowner = sop;
3450 stp->st_stateid.si_boot = boot_time;
3451 stp->st_stateid.si_stateownerid = sop->so_id;
3452 stp->st_stateid.si_fileid = fp->fi_id;
3453 stp->st_stateid.si_generation = 0;
3454 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3455 stp->st_access_bmap = open_stp->st_access_bmap;
3456 stp->st_deny_bmap = open_stp->st_deny_bmap;
3457 stp->st_openstp = open_stp;
3464 check_lock_length(u64 offset, u64 length)
3466 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3467 LOFF_OVERFLOW(offset, length)));
3474 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3475 struct nfsd4_lock *lock)
3477 struct nfs4_stateowner *open_sop = NULL;
3478 struct nfs4_stateowner *lock_sop = NULL;
3479 struct nfs4_stateid *lock_stp;
3481 struct file_lock file_lock;
3482 struct file_lock conflock;
3484 unsigned int strhashval;
3488 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3489 (long long) lock->lk_offset,
3490 (long long) lock->lk_length);
3492 if (check_lock_length(lock->lk_offset, lock->lk_length))
3493 return nfserr_inval;
3495 if ((status = fh_verify(rqstp, &cstate->current_fh,
3496 S_IFREG, NFSD_MAY_LOCK))) {
3497 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3503 if (lock->lk_is_new) {
3505 * Client indicates that this is a new lockowner.
3506 * Use open owner and open stateid to create lock owner and
3509 struct nfs4_stateid *open_stp = NULL;
3510 struct nfs4_file *fp;
3512 status = nfserr_stale_clientid;
3513 if (!nfsd4_has_session(cstate) &&
3514 STALE_CLIENTID(&lock->lk_new_clientid))
3517 /* validate and update open stateid and open seqid */
3518 status = nfs4_preprocess_seqid_op(cstate,
3519 lock->lk_new_open_seqid,
3520 &lock->lk_new_open_stateid,
3522 &lock->lk_replay_owner, &open_stp,
3526 open_sop = lock->lk_replay_owner;
3527 /* create lockowner and lock stateid */
3528 fp = open_stp->st_file;
3529 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3530 open_sop->so_client->cl_clientid.cl_id,
3531 &lock->v.new.owner);
3532 /* XXX: Do we need to check for duplicate stateowners on
3533 * the same file, or should they just be allowed (and
3534 * create new stateids)? */
3535 status = nfserr_resource;
3536 lock_sop = alloc_init_lock_stateowner(strhashval,
3537 open_sop->so_client, open_stp, lock);
3538 if (lock_sop == NULL)
3540 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3541 if (lock_stp == NULL)
3544 /* lock (lock owner + lock stateid) already exists */
3545 status = nfs4_preprocess_seqid_op(cstate,
3546 lock->lk_old_lock_seqid,
3547 &lock->lk_old_lock_stateid,
3549 &lock->lk_replay_owner, &lock_stp, lock);
3552 lock_sop = lock->lk_replay_owner;
3554 /* lock->lk_replay_owner and lock_stp have been created or found */
3555 filp = lock_stp->st_vfs_file;
3557 status = nfserr_grace;
3558 if (locks_in_grace() && !lock->lk_reclaim)
3560 status = nfserr_no_grace;
3561 if (!locks_in_grace() && lock->lk_reclaim)
3564 locks_init_lock(&file_lock);
3565 switch (lock->lk_type) {
3568 file_lock.fl_type = F_RDLCK;
3572 case NFS4_WRITEW_LT:
3573 file_lock.fl_type = F_WRLCK;
3577 status = nfserr_inval;
3580 file_lock.fl_owner = (fl_owner_t)lock_sop;
3581 file_lock.fl_pid = current->tgid;
3582 file_lock.fl_file = filp;
3583 file_lock.fl_flags = FL_POSIX;
3584 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3586 file_lock.fl_start = lock->lk_offset;
3587 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3588 nfs4_transform_lock_offset(&file_lock);
3591 * Try to lock the file in the VFS.
3592 * Note: locks.c uses the BKL to protect the inode's lock list.
3595 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3597 case 0: /* success! */
3598 update_stateid(&lock_stp->st_stateid);
3599 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3603 case (EAGAIN): /* conflock holds conflicting lock */
3604 status = nfserr_denied;
3605 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3606 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3609 status = nfserr_deadlock;
3612 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3613 status = nfserr_resource;
3617 if (status && lock->lk_is_new && lock_sop)
3618 release_lockowner(lock_sop);
3619 if (lock->lk_replay_owner) {
3620 nfs4_get_stateowner(lock->lk_replay_owner);
3621 cstate->replay_owner = lock->lk_replay_owner;
3623 nfs4_unlock_state();
3628 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3629 * so we do a temporary open here just to get an open file to pass to
3630 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3633 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3638 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3641 err = vfs_test_lock(file, lock);
3650 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3651 struct nfsd4_lockt *lockt)
3653 struct inode *inode;
3654 struct file_lock file_lock;
3658 if (locks_in_grace())
3659 return nfserr_grace;
3661 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3662 return nfserr_inval;
3664 lockt->lt_stateowner = NULL;
3667 status = nfserr_stale_clientid;
3668 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3671 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3672 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3673 if (status == nfserr_symlink)
3674 status = nfserr_inval;
3678 inode = cstate->current_fh.fh_dentry->d_inode;
3679 locks_init_lock(&file_lock);
3680 switch (lockt->lt_type) {
3683 file_lock.fl_type = F_RDLCK;
3686 case NFS4_WRITEW_LT:
3687 file_lock.fl_type = F_WRLCK;
3690 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3691 status = nfserr_inval;
3695 lockt->lt_stateowner = find_lockstateowner_str(inode,
3696 &lockt->lt_clientid, &lockt->lt_owner);
3697 if (lockt->lt_stateowner)
3698 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3699 file_lock.fl_pid = current->tgid;
3700 file_lock.fl_flags = FL_POSIX;
3702 file_lock.fl_start = lockt->lt_offset;
3703 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3705 nfs4_transform_lock_offset(&file_lock);
3708 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3710 status = nfserrno(error);
3713 if (file_lock.fl_type != F_UNLCK) {
3714 status = nfserr_denied;
3715 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3718 nfs4_unlock_state();
3723 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3724 struct nfsd4_locku *locku)
3726 struct nfs4_stateid *stp;
3727 struct file *filp = NULL;
3728 struct file_lock file_lock;
3732 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3733 (long long) locku->lu_offset,
3734 (long long) locku->lu_length);
3736 if (check_lock_length(locku->lu_offset, locku->lu_length))
3737 return nfserr_inval;
3741 if ((status = nfs4_preprocess_seqid_op(cstate,
3745 &locku->lu_stateowner, &stp, NULL)))
3748 filp = stp->st_vfs_file;
3750 locks_init_lock(&file_lock);
3751 file_lock.fl_type = F_UNLCK;
3752 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3753 file_lock.fl_pid = current->tgid;
3754 file_lock.fl_file = filp;
3755 file_lock.fl_flags = FL_POSIX;
3756 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3757 file_lock.fl_start = locku->lu_offset;
3759 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3760 nfs4_transform_lock_offset(&file_lock);
3763 * Try to unlock the file in the VFS.
3765 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3767 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3771 * OK, unlock succeeded; the only thing left to do is update the stateid.
3773 update_stateid(&stp->st_stateid);
3774 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3777 if (locku->lu_stateowner) {
3778 nfs4_get_stateowner(locku->lu_stateowner);
3779 cstate->replay_owner = locku->lu_stateowner;
3781 nfs4_unlock_state();
3785 status = nfserrno(err);
3791 * 1: locks held by lockowner
3792 * 0: no locks held by lockowner
3795 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3797 struct file_lock **flpp;
3798 struct inode *inode = filp->f_path.dentry->d_inode;
3802 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3803 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3814 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3815 struct nfsd4_compound_state *cstate,
3816 struct nfsd4_release_lockowner *rlockowner)
3818 clientid_t *clid = &rlockowner->rl_clientid;
3819 struct nfs4_stateowner *sop;
3820 struct nfs4_stateid *stp;
3821 struct xdr_netobj *owner = &rlockowner->rl_owner;
3822 struct list_head matches;
3826 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3827 clid->cl_boot, clid->cl_id);
3829 /* XXX check for lease expiration */
3831 status = nfserr_stale_clientid;
3832 if (STALE_CLIENTID(clid))
3837 status = nfserr_locks_held;
3838 /* XXX: we're doing a linear search through all the lockowners.
3839 * Yipes! For now we'll just hope clients aren't really using
3840 * release_lockowner much, but eventually we have to fix these
3841 * data structures. */
3842 INIT_LIST_HEAD(&matches);
3843 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3844 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3845 if (!same_owner_str(sop, owner, clid))
3847 list_for_each_entry(stp, &sop->so_stateids,
3849 if (check_for_locks(stp->st_vfs_file, sop))
3851 /* Note: so_perclient unused for lockowners,
3852 * so it's OK to fool with here. */
3853 list_add(&sop->so_perclient, &matches);
3857 /* Clients probably won't expect us to return with some (but not all)
3858 * of the lockowner state released; so don't release any until all
3859 * have been checked. */
3861 while (!list_empty(&matches)) {
3862 sop = list_entry(matches.next, struct nfs4_stateowner,
3864 /* unhash_stateowner deletes so_perclient only
3865 * for openowners. */
3866 list_del(&sop->so_perclient);
3867 release_lockowner(sop);
3870 nfs4_unlock_state();
3874 static inline struct nfs4_client_reclaim *
3877 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3881 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3883 unsigned int strhashval = clientstr_hashval(name);
3884 struct nfs4_client *clp;
3886 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3891 * failure => all reset bets are off, nfserr_no_grace...
3894 nfs4_client_to_reclaim(const char *name)
3896 unsigned int strhashval;
3897 struct nfs4_client_reclaim *crp = NULL;
3899 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3900 crp = alloc_reclaim();
3903 strhashval = clientstr_hashval(name);
3904 INIT_LIST_HEAD(&crp->cr_strhash);
3905 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3906 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3907 reclaim_str_hashtbl_size++;
3912 nfs4_release_reclaim(void)
3914 struct nfs4_client_reclaim *crp = NULL;
3917 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3918 while (!list_empty(&reclaim_str_hashtbl[i])) {
3919 crp = list_entry(reclaim_str_hashtbl[i].next,
3920 struct nfs4_client_reclaim, cr_strhash);
3921 list_del(&crp->cr_strhash);
3923 reclaim_str_hashtbl_size--;
3926 BUG_ON(reclaim_str_hashtbl_size);
3930 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3931 static struct nfs4_client_reclaim *
3932 nfs4_find_reclaim_client(clientid_t *clid)
3934 unsigned int strhashval;
3935 struct nfs4_client *clp;
3936 struct nfs4_client_reclaim *crp = NULL;
3939 /* find clientid in conf_id_hashtbl */
3940 clp = find_confirmed_client(clid);
3944 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3945 clp->cl_name.len, clp->cl_name.data,
3948 /* find clp->cl_name in reclaim_str_hashtbl */
3949 strhashval = clientstr_hashval(clp->cl_recdir);
3950 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3951 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3959 * Called from OPEN. Look for clientid in reclaim list.
3962 nfs4_check_open_reclaim(clientid_t *clid)
3964 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3967 /* initialization to perform at module load time: */
3970 nfs4_state_init(void)
3974 status = nfsd4_init_slabs();
3977 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3978 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3979 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3980 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3981 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3982 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3984 for (i = 0; i < SESSION_HASH_SIZE; i++)
3985 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3986 for (i = 0; i < FILE_HASH_SIZE; i++) {
3987 INIT_LIST_HEAD(&file_hashtbl[i]);
3989 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3990 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3991 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3993 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3994 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3995 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3997 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3998 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3999 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4001 memset(&onestateid, ~0, sizeof(stateid_t));
4002 INIT_LIST_HEAD(&close_lru);
4003 INIT_LIST_HEAD(&client_lru);
4004 INIT_LIST_HEAD(&del_recall_lru);
4005 reclaim_str_hashtbl_size = 0;
4010 nfsd4_load_reboot_recovery_data(void)
4015 nfsd4_init_recdir(user_recovery_dirname);
4016 status = nfsd4_recdir_load();
4017 nfs4_unlock_state();
4019 printk("NFSD: Failure reading reboot recovery data\n");
4023 * Since the lifetime of a delegation isn't limited to that of an open, a
4024 * client may quite reasonably hang on to a delegation as long as it has
4025 * the inode cached. This becomes an obvious problem the first time a
4026 * client's inode cache approaches the size of the server's total memory.
4028 * For now we avoid this problem by imposing a hard limit on the number
4029 * of delegations, which varies according to the server's memory size.
4032 set_max_delegations(void)
4035 * Allow at most 4 delegations per megabyte of RAM. Quick
4036 * estimates suggest that in the worst case (where every delegation
4037 * is for a different inode), a delegation could take about 1.5K,
4038 * giving a worst case usage of about 6% of memory.
4040 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4043 /* initialization to perform when the nfsd service is started: */
4046 __nfs4_state_start(void)
4050 boot_time = get_seconds();
4051 locks_start_grace(&nfsd4_manager);
4052 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4054 ret = set_callback_cred();
4057 laundry_wq = create_singlethread_workqueue("nfsd4");
4058 if (laundry_wq == NULL)
4060 ret = nfsd4_create_callback_queue();
4062 goto out_free_laundry;
4063 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4064 set_max_delegations();
4067 destroy_workqueue(laundry_wq);
4072 nfs4_state_start(void)
4078 nfsd4_load_reboot_recovery_data();
4079 ret = __nfs4_state_start();
4087 __nfs4_state_shutdown(void)
4090 struct nfs4_client *clp = NULL;
4091 struct nfs4_delegation *dp = NULL;
4092 struct list_head *pos, *next, reaplist;
4094 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4095 while (!list_empty(&conf_id_hashtbl[i])) {
4096 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4099 while (!list_empty(&unconf_str_hashtbl[i])) {
4100 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4104 INIT_LIST_HEAD(&reaplist);
4105 spin_lock(&recall_lock);
4106 list_for_each_safe(pos, next, &del_recall_lru) {
4107 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4108 list_move(&dp->dl_recall_lru, &reaplist);
4110 spin_unlock(&recall_lock);
4111 list_for_each_safe(pos, next, &reaplist) {
4112 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4113 list_del_init(&dp->dl_recall_lru);
4114 unhash_delegation(dp);
4117 nfsd4_shutdown_recdir();
4122 nfs4_state_shutdown(void)
4124 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4125 destroy_workqueue(laundry_wq);
4126 locks_end_grace(&nfsd4_manager);
4128 nfs4_release_reclaim();
4129 __nfs4_state_shutdown();
4130 nfsd4_destroy_callback_queue();
4131 nfs4_unlock_state();
4135 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4136 * accessed when nfsd is starting.
4139 nfs4_set_recdir(char *recdir)
4141 strcpy(user_recovery_dirname, recdir);
4145 * Change the NFSv4 recovery directory to recdir.
4148 nfs4_reset_recoverydir(char *recdir)
4153 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4157 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4158 nfs4_set_recdir(recdir);
4166 nfs4_recoverydir(void)
4168 return user_recovery_dirname;