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 = get_seconds();
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 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_serverfault;
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_serverfault;
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 dprintk("%s: idx is %d\n", __func__, idx);
595 /* Search in the appropriate list */
596 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
597 dump_sessionid("list traversal", &elem->se_sessionid);
598 if (!memcmp(elem->se_sessionid.data, sessionid->data,
599 NFS4_MAX_SESSIONID_LEN)) {
604 dprintk("%s: session not found\n", __func__);
608 /* caller must hold client_lock */
610 unhash_session(struct nfsd4_session *ses)
612 list_del(&ses->se_hash);
613 list_del(&ses->se_perclnt);
617 free_session(struct kref *kref)
619 struct nfsd4_session *ses;
622 ses = container_of(kref, struct nfsd4_session, se_ref);
623 spin_lock(&nfsd_drc_lock);
624 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
625 nfsd_drc_mem_used -= mem;
626 spin_unlock(&nfsd_drc_lock);
627 free_session_slots(ses);
632 renew_client(struct nfs4_client *clp)
635 * Move client to the end to the LRU list.
637 dprintk("renewing client (clientid %08x/%08x)\n",
638 clp->cl_clientid.cl_boot,
639 clp->cl_clientid.cl_id);
640 list_move_tail(&clp->cl_lru, &client_lru);
641 clp->cl_time = get_seconds();
644 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
646 STALE_CLIENTID(clientid_t *clid)
648 if (clid->cl_boot == boot_time)
650 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
651 clid->cl_boot, clid->cl_id, boot_time);
656 * XXX Should we use a slab cache ?
657 * This type of memory management is somewhat inefficient, but we use it
658 * anyway since SETCLIENTID is not a common operation.
660 static struct nfs4_client *alloc_client(struct xdr_netobj name)
662 struct nfs4_client *clp;
664 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
667 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
668 if (clp->cl_name.data == NULL) {
672 memcpy(clp->cl_name.data, name.data, name.len);
673 clp->cl_name.len = name.len;
678 free_client(struct nfs4_client *clp)
680 if (clp->cl_cred.cr_group_info)
681 put_group_info(clp->cl_cred.cr_group_info);
682 kfree(clp->cl_principal);
683 kfree(clp->cl_name.data);
688 expire_client(struct nfs4_client *clp)
690 struct nfs4_stateowner *sop;
691 struct nfs4_delegation *dp;
692 struct list_head reaplist;
694 INIT_LIST_HEAD(&reaplist);
695 spin_lock(&recall_lock);
696 while (!list_empty(&clp->cl_delegations)) {
697 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
698 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
700 list_del_init(&dp->dl_perclnt);
701 list_move(&dp->dl_recall_lru, &reaplist);
703 spin_unlock(&recall_lock);
704 while (!list_empty(&reaplist)) {
705 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
706 list_del_init(&dp->dl_recall_lru);
707 unhash_delegation(dp);
709 list_del(&clp->cl_idhash);
710 list_del(&clp->cl_strhash);
711 list_del(&clp->cl_lru);
712 while (!list_empty(&clp->cl_openowners)) {
713 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
714 release_openowner(sop);
716 spin_lock(&client_lock);
717 while (!list_empty(&clp->cl_sessions)) {
718 struct nfsd4_session *ses;
719 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
722 nfsd4_put_session(ses);
724 spin_unlock(&client_lock);
725 nfsd4_set_callback_client(clp, NULL);
726 if (clp->cl_cb_conn.cb_xprt)
727 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
731 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
733 memcpy(target->cl_verifier.data, source->data,
734 sizeof(target->cl_verifier.data));
737 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
739 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
740 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
743 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
745 target->cr_uid = source->cr_uid;
746 target->cr_gid = source->cr_gid;
747 target->cr_group_info = source->cr_group_info;
748 get_group_info(target->cr_group_info);
751 static int same_name(const char *n1, const char *n2)
753 return 0 == memcmp(n1, n2, HEXDIR_LEN);
757 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
759 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
763 same_clid(clientid_t *cl1, clientid_t *cl2)
765 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
768 /* XXX what about NGROUP */
770 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
772 return cr1->cr_uid == cr2->cr_uid;
775 static void gen_clid(struct nfs4_client *clp)
777 static u32 current_clientid = 1;
779 clp->cl_clientid.cl_boot = boot_time;
780 clp->cl_clientid.cl_id = current_clientid++;
783 static void gen_confirm(struct nfs4_client *clp)
788 p = (u32 *)clp->cl_confirm.data;
789 *p++ = get_seconds();
793 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
794 struct svc_rqst *rqstp, nfs4_verifier *verf)
796 struct nfs4_client *clp;
797 struct sockaddr *sa = svc_addr(rqstp);
800 clp = alloc_client(name);
804 princ = svc_gss_principal(rqstp);
806 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
807 if (clp->cl_principal == NULL) {
813 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
814 atomic_set(&clp->cl_cb_set, 0);
815 INIT_LIST_HEAD(&clp->cl_idhash);
816 INIT_LIST_HEAD(&clp->cl_strhash);
817 INIT_LIST_HEAD(&clp->cl_openowners);
818 INIT_LIST_HEAD(&clp->cl_delegations);
819 INIT_LIST_HEAD(&clp->cl_sessions);
820 INIT_LIST_HEAD(&clp->cl_lru);
821 clear_bit(0, &clp->cl_cb_slot_busy);
822 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
823 copy_verf(clp, verf);
824 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
825 clp->cl_flavor = rqstp->rq_flavor;
826 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
832 static int check_name(struct xdr_netobj name)
836 if (name.len > NFS4_OPAQUE_LIMIT) {
837 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
844 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
846 unsigned int idhashval;
848 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
849 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
850 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
851 list_add_tail(&clp->cl_lru, &client_lru);
852 clp->cl_time = get_seconds();
856 move_to_confirmed(struct nfs4_client *clp)
858 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
859 unsigned int strhashval;
861 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
862 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
863 strhashval = clientstr_hashval(clp->cl_recdir);
864 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
868 static struct nfs4_client *
869 find_confirmed_client(clientid_t *clid)
871 struct nfs4_client *clp;
872 unsigned int idhashval = clientid_hashval(clid->cl_id);
874 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
875 if (same_clid(&clp->cl_clientid, clid))
881 static struct nfs4_client *
882 find_unconfirmed_client(clientid_t *clid)
884 struct nfs4_client *clp;
885 unsigned int idhashval = clientid_hashval(clid->cl_id);
887 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
888 if (same_clid(&clp->cl_clientid, clid))
895 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
896 * parameter. Matching is based on the fact the at least one of the
897 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
899 * FIXME: we need to unify the clientid namespaces for nfsv4.x
900 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
901 * and SET_CLIENTID{,_CONFIRM}
904 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
906 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
907 return use_exchange_id == has_exchange_flags;
910 static struct nfs4_client *
911 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
912 bool use_exchange_id)
914 struct nfs4_client *clp;
916 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
917 if (same_name(clp->cl_recdir, dname) &&
918 match_clientid_establishment(clp, use_exchange_id))
924 static struct nfs4_client *
925 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
926 bool use_exchange_id)
928 struct nfs4_client *clp;
930 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
931 if (same_name(clp->cl_recdir, dname) &&
932 match_clientid_establishment(clp, use_exchange_id))
939 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
941 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
942 unsigned short expected_family;
944 /* Currently, we only support tcp and tcp6 for the callback channel */
945 if (se->se_callback_netid_len == 3 &&
946 !memcmp(se->se_callback_netid_val, "tcp", 3))
947 expected_family = AF_INET;
948 else if (se->se_callback_netid_len == 4 &&
949 !memcmp(se->se_callback_netid_val, "tcp6", 4))
950 expected_family = AF_INET6;
954 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
955 se->se_callback_addr_len,
956 (struct sockaddr *) &cb->cb_addr,
957 sizeof(cb->cb_addr));
959 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
962 if (cb->cb_addr.ss_family == AF_INET6)
963 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
965 cb->cb_minorversion = 0;
966 cb->cb_prog = se->se_callback_prog;
967 cb->cb_ident = se->se_callback_ident;
970 cb->cb_addr.ss_family = AF_UNSPEC;
972 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
973 "will not receive delegations\n",
974 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
980 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
983 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
985 struct nfsd4_slot *slot = resp->cstate.slot;
988 dprintk("--> %s slot %p\n", __func__, slot);
990 slot->sl_opcnt = resp->opcnt;
991 slot->sl_status = resp->cstate.status;
993 if (nfsd4_not_cached(resp)) {
994 slot->sl_datalen = 0;
997 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
998 base = (char *)resp->cstate.datap -
999 (char *)resp->xbuf->head[0].iov_base;
1000 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1002 WARN("%s: sessions DRC could not cache compound\n", __func__);
1007 * Encode the replay sequence operation from the slot values.
1008 * If cachethis is FALSE encode the uncached rep error on the next
1009 * operation which sets resp->p and increments resp->opcnt for
1010 * nfs4svc_encode_compoundres.
1014 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1015 struct nfsd4_compoundres *resp)
1017 struct nfsd4_op *op;
1018 struct nfsd4_slot *slot = resp->cstate.slot;
1020 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1021 resp->opcnt, resp->cstate.slot->sl_cachethis);
1023 /* Encode the replayed sequence operation */
1024 op = &args->ops[resp->opcnt - 1];
1025 nfsd4_encode_operation(resp, op);
1027 /* Return nfserr_retry_uncached_rep in next operation. */
1028 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1029 op = &args->ops[resp->opcnt++];
1030 op->status = nfserr_retry_uncached_rep;
1031 nfsd4_encode_operation(resp, op);
1037 * The sequence operation is not cached because we can use the slot and
1041 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1042 struct nfsd4_sequence *seq)
1044 struct nfsd4_slot *slot = resp->cstate.slot;
1047 dprintk("--> %s slot %p\n", __func__, slot);
1049 /* Either returns 0 or nfserr_retry_uncached */
1050 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1051 if (status == nfserr_retry_uncached_rep)
1054 /* The sequence operation has been encoded, cstate->datap set. */
1055 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1057 resp->opcnt = slot->sl_opcnt;
1058 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1059 status = slot->sl_status;
1065 * Set the exchange_id flags returned by the server.
1068 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1070 /* pNFS is not supported */
1071 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1073 /* Referrals are supported, Migration is not. */
1074 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1076 /* set the wire flags to return to client. */
1077 clid->flags = new->cl_exchange_flags;
1081 nfsd4_exchange_id(struct svc_rqst *rqstp,
1082 struct nfsd4_compound_state *cstate,
1083 struct nfsd4_exchange_id *exid)
1085 struct nfs4_client *unconf, *conf, *new;
1087 unsigned int strhashval;
1088 char dname[HEXDIR_LEN];
1089 char addr_str[INET6_ADDRSTRLEN];
1090 nfs4_verifier verf = exid->verifier;
1091 struct sockaddr *sa = svc_addr(rqstp);
1093 rpc_ntop(sa, addr_str, sizeof(addr_str));
1094 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1095 "ip_addr=%s flags %x, spa_how %d\n",
1096 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1097 addr_str, exid->flags, exid->spa_how);
1099 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1100 return nfserr_inval;
1102 /* Currently only support SP4_NONE */
1103 switch (exid->spa_how) {
1107 return nfserr_encr_alg_unsupp;
1109 BUG(); /* checked by xdr code */
1111 return nfserr_serverfault; /* no excuse :-/ */
1114 status = nfs4_make_rec_clidname(dname, &exid->clname);
1119 strhashval = clientstr_hashval(dname);
1124 conf = find_confirmed_client_by_str(dname, strhashval, true);
1126 if (!same_verf(&verf, &conf->cl_verifier)) {
1127 /* 18.35.4 case 8 */
1128 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1129 status = nfserr_not_same;
1132 /* Client reboot: destroy old state */
1133 expire_client(conf);
1136 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1137 /* 18.35.4 case 9 */
1138 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1139 status = nfserr_perm;
1142 expire_client(conf);
1146 * Set bit when the owner id and verifier map to an already
1147 * confirmed client id (18.35.3).
1149 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1152 * Falling into 18.35.4 case 2, possible router replay.
1153 * Leave confirmed record intact and return same result.
1155 copy_verf(conf, &verf);
1160 /* 18.35.4 case 7 */
1161 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1162 status = nfserr_noent;
1166 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1169 * Possible retry or client restart. Per 18.35.4 case 4,
1170 * a new unconfirmed record should be generated regardless
1171 * of whether any properties have changed.
1173 expire_client(unconf);
1178 new = create_client(exid->clname, dname, rqstp, &verf);
1180 status = nfserr_serverfault;
1185 add_to_unconfirmed(new, strhashval);
1187 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1188 exid->clientid.cl_id = new->cl_clientid.cl_id;
1191 nfsd4_set_ex_flags(new, exid);
1193 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1194 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1198 nfs4_unlock_state();
1200 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1205 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1207 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1210 /* The slot is in use, and no response has been sent. */
1212 if (seqid == slot_seqid)
1213 return nfserr_jukebox;
1215 return nfserr_seq_misordered;
1218 if (likely(seqid == slot_seqid + 1))
1221 if (seqid == slot_seqid)
1222 return nfserr_replay_cache;
1224 if (seqid == 1 && (slot_seqid + 1) == 0)
1226 /* Misordered replay or misordered new request */
1227 return nfserr_seq_misordered;
1231 * Cache the create session result into the create session single DRC
1232 * slot cache by saving the xdr structure. sl_seqid has been set.
1233 * Do this for solo or embedded create session operations.
1236 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1237 struct nfsd4_clid_slot *slot, int nfserr)
1239 slot->sl_status = nfserr;
1240 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1244 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1245 struct nfsd4_clid_slot *slot)
1247 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1248 return slot->sl_status;
1252 nfsd4_create_session(struct svc_rqst *rqstp,
1253 struct nfsd4_compound_state *cstate,
1254 struct nfsd4_create_session *cr_ses)
1256 struct sockaddr *sa = svc_addr(rqstp);
1257 struct nfs4_client *conf, *unconf;
1258 struct nfsd4_clid_slot *cs_slot = NULL;
1262 unconf = find_unconfirmed_client(&cr_ses->clientid);
1263 conf = find_confirmed_client(&cr_ses->clientid);
1266 cs_slot = &conf->cl_cs_slot;
1267 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1268 if (status == nfserr_replay_cache) {
1269 dprintk("Got a create_session replay! seqid= %d\n",
1271 /* Return the cached reply status */
1272 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1274 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1275 status = nfserr_seq_misordered;
1276 dprintk("Sequence misordered!\n");
1277 dprintk("Expected seqid= %d but got seqid= %d\n",
1278 cs_slot->sl_seqid, cr_ses->seqid);
1281 cs_slot->sl_seqid++;
1282 } else if (unconf) {
1283 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1284 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1285 status = nfserr_clid_inuse;
1289 cs_slot = &unconf->cl_cs_slot;
1290 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1292 /* an unconfirmed replay returns misordered */
1293 status = nfserr_seq_misordered;
1297 cs_slot->sl_seqid++; /* from 0 to 1 */
1298 move_to_confirmed(unconf);
1300 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1301 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1302 svc_xprt_get(rqstp->rq_xprt);
1304 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1306 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1307 unconf->cl_cb_conn.cb_minorversion =
1308 cstate->minorversion;
1309 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1310 unconf->cl_cb_seq_nr = 1;
1311 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1315 status = nfserr_stale_clientid;
1320 * We do not support RDMA or persistent sessions
1322 cr_ses->flags &= ~SESSION4_PERSIST;
1323 cr_ses->flags &= ~SESSION4_RDMA;
1325 status = alloc_init_session(rqstp, conf, cr_ses);
1329 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1330 NFS4_MAX_SESSIONID_LEN);
1331 cr_ses->seqid = cs_slot->sl_seqid;
1334 /* cache solo and embedded create sessions under the state lock */
1335 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1337 nfs4_unlock_state();
1338 dprintk("%s returns %d\n", __func__, ntohl(status));
1342 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1344 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1345 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1347 return argp->opcnt == resp->opcnt;
1350 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1354 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1358 nfsd4_destroy_session(struct svc_rqst *r,
1359 struct nfsd4_compound_state *cstate,
1360 struct nfsd4_destroy_session *sessionid)
1362 struct nfsd4_session *ses;
1363 u32 status = nfserr_badsession;
1366 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1367 * - Should we return nfserr_back_chan_busy if waiting for
1368 * callbacks on to-be-destroyed session?
1369 * - Do we need to clear any callback info from previous session?
1372 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1373 if (!nfsd4_last_compound_op(r))
1374 return nfserr_not_only_op;
1376 dump_sessionid(__func__, &sessionid->sessionid);
1377 spin_lock(&client_lock);
1378 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1380 spin_unlock(&client_lock);
1384 unhash_session(ses);
1385 spin_unlock(&client_lock);
1387 /* wait for callbacks */
1388 nfsd4_set_callback_client(ses->se_client, NULL);
1389 nfsd4_put_session(ses);
1392 dprintk("%s returns %d\n", __func__, ntohl(status));
1397 nfsd4_sequence(struct svc_rqst *rqstp,
1398 struct nfsd4_compound_state *cstate,
1399 struct nfsd4_sequence *seq)
1401 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1402 struct nfsd4_session *session;
1403 struct nfsd4_slot *slot;
1406 if (resp->opcnt != 1)
1407 return nfserr_sequence_pos;
1409 spin_lock(&client_lock);
1410 status = nfserr_badsession;
1411 session = find_in_sessionid_hashtbl(&seq->sessionid);
1415 status = nfserr_badslot;
1416 if (seq->slotid >= session->se_fchannel.maxreqs)
1419 slot = session->se_slots[seq->slotid];
1420 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1422 /* We do not negotiate the number of slots yet, so set the
1423 * maxslots to the session maxreqs which is used to encode
1424 * sr_highest_slotid and the sr_target_slot id to maxslots */
1425 seq->maxslots = session->se_fchannel.maxreqs;
1427 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1428 if (status == nfserr_replay_cache) {
1429 cstate->slot = slot;
1430 cstate->session = session;
1431 /* Return the cached reply status and set cstate->status
1432 * for nfsd4_proc_compound processing */
1433 status = nfsd4_replay_cache_entry(resp, seq);
1434 cstate->status = nfserr_replay_cache;
1440 /* Success! bump slot seqid */
1441 slot->sl_inuse = true;
1442 slot->sl_seqid = seq->seqid;
1443 slot->sl_cachethis = seq->cachethis;
1445 cstate->slot = slot;
1446 cstate->session = session;
1449 /* Hold a session reference until done processing the compound. */
1450 if (cstate->session)
1451 nfsd4_get_session(cstate->session);
1452 spin_unlock(&client_lock);
1453 /* Renew the clientid on success and on replay */
1454 if (cstate->session) {
1456 renew_client(session->se_client);
1457 nfs4_unlock_state();
1459 dprintk("%s: return %d\n", __func__, ntohl(status));
1464 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1465 struct nfsd4_setclientid *setclid)
1467 struct sockaddr *sa = svc_addr(rqstp);
1468 struct xdr_netobj clname = {
1469 .len = setclid->se_namelen,
1470 .data = setclid->se_name,
1472 nfs4_verifier clverifier = setclid->se_verf;
1473 unsigned int strhashval;
1474 struct nfs4_client *conf, *unconf, *new;
1476 char dname[HEXDIR_LEN];
1478 if (!check_name(clname))
1479 return nfserr_inval;
1481 status = nfs4_make_rec_clidname(dname, &clname);
1486 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1487 * We get here on a DRC miss.
1490 strhashval = clientstr_hashval(dname);
1493 conf = find_confirmed_client_by_str(dname, strhashval, false);
1495 /* RFC 3530 14.2.33 CASE 0: */
1496 status = nfserr_clid_inuse;
1497 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1498 char addr_str[INET6_ADDRSTRLEN];
1499 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1501 dprintk("NFSD: setclientid: string in use by client "
1502 "at %s\n", addr_str);
1507 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1508 * has a description of SETCLIENTID request processing consisting
1509 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1511 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1512 status = nfserr_resource;
1515 * RFC 3530 14.2.33 CASE 4:
1516 * placed first, because it is the normal case
1519 expire_client(unconf);
1520 new = create_client(clname, dname, rqstp, &clverifier);
1524 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1526 * RFC 3530 14.2.33 CASE 1:
1527 * probable callback update
1530 /* Note this is removing unconfirmed {*x***},
1531 * which is stronger than RFC recommended {vxc**}.
1532 * This has the advantage that there is at most
1533 * one {*x***} in either list at any time.
1535 expire_client(unconf);
1537 new = create_client(clname, dname, rqstp, &clverifier);
1540 copy_clid(new, conf);
1541 } else if (!unconf) {
1543 * RFC 3530 14.2.33 CASE 2:
1544 * probable client reboot; state will be removed if
1547 new = create_client(clname, dname, rqstp, &clverifier);
1553 * RFC 3530 14.2.33 CASE 3:
1554 * probable client reboot; state will be removed if
1557 expire_client(unconf);
1558 new = create_client(clname, dname, rqstp, &clverifier);
1563 gen_callback(new, setclid, rpc_get_scope_id(sa));
1564 add_to_unconfirmed(new, strhashval);
1565 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1566 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1567 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1570 nfs4_unlock_state();
1576 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1577 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1578 * bullets, labeled as CASE1 - CASE4 below.
1581 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1582 struct nfsd4_compound_state *cstate,
1583 struct nfsd4_setclientid_confirm *setclientid_confirm)
1585 struct sockaddr *sa = svc_addr(rqstp);
1586 struct nfs4_client *conf, *unconf;
1587 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1588 clientid_t * clid = &setclientid_confirm->sc_clientid;
1591 if (STALE_CLIENTID(clid))
1592 return nfserr_stale_clientid;
1594 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1595 * We get here on a DRC miss.
1600 conf = find_confirmed_client(clid);
1601 unconf = find_unconfirmed_client(clid);
1603 status = nfserr_clid_inuse;
1604 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1606 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1610 * section 14.2.34 of RFC 3530 has a description of
1611 * SETCLIENTID_CONFIRM request processing consisting
1612 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1614 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1616 * RFC 3530 14.2.34 CASE 1:
1619 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1620 status = nfserr_clid_inuse;
1622 atomic_set(&conf->cl_cb_set, 0);
1623 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1624 expire_client(unconf);
1628 } else if (conf && !unconf) {
1630 * RFC 3530 14.2.34 CASE 2:
1631 * probable retransmitted request; play it safe and
1634 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1635 status = nfserr_clid_inuse;
1638 } else if (!conf && unconf
1639 && same_verf(&unconf->cl_confirm, &confirm)) {
1641 * RFC 3530 14.2.34 CASE 3:
1642 * Normal case; new or rebooted client:
1644 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1645 status = nfserr_clid_inuse;
1648 clientstr_hashval(unconf->cl_recdir);
1649 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1652 nfsd4_remove_clid_dir(conf);
1653 expire_client(conf);
1655 move_to_confirmed(unconf);
1657 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1660 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1661 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1664 * RFC 3530 14.2.34 CASE 4:
1665 * Client probably hasn't noticed that we rebooted yet.
1667 status = nfserr_stale_clientid;
1669 /* check that we have hit one of the cases...*/
1670 status = nfserr_clid_inuse;
1673 nfs4_unlock_state();
1677 /* OPEN Share state helper functions */
1678 static inline struct nfs4_file *
1679 alloc_init_file(struct inode *ino)
1681 struct nfs4_file *fp;
1682 unsigned int hashval = file_hashval(ino);
1684 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1686 atomic_set(&fp->fi_ref, 1);
1687 INIT_LIST_HEAD(&fp->fi_hash);
1688 INIT_LIST_HEAD(&fp->fi_stateids);
1689 INIT_LIST_HEAD(&fp->fi_delegations);
1690 spin_lock(&recall_lock);
1691 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1692 spin_unlock(&recall_lock);
1693 fp->fi_inode = igrab(ino);
1694 fp->fi_id = current_fileid++;
1695 fp->fi_had_conflict = false;
1702 nfsd4_free_slab(struct kmem_cache **slab)
1706 kmem_cache_destroy(*slab);
1711 nfsd4_free_slabs(void)
1713 nfsd4_free_slab(&stateowner_slab);
1714 nfsd4_free_slab(&file_slab);
1715 nfsd4_free_slab(&stateid_slab);
1716 nfsd4_free_slab(&deleg_slab);
1720 nfsd4_init_slabs(void)
1722 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1723 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1724 if (stateowner_slab == NULL)
1726 file_slab = kmem_cache_create("nfsd4_files",
1727 sizeof(struct nfs4_file), 0, 0, NULL);
1728 if (file_slab == NULL)
1730 stateid_slab = kmem_cache_create("nfsd4_stateids",
1731 sizeof(struct nfs4_stateid), 0, 0, NULL);
1732 if (stateid_slab == NULL)
1734 deleg_slab = kmem_cache_create("nfsd4_delegations",
1735 sizeof(struct nfs4_delegation), 0, 0, NULL);
1736 if (deleg_slab == NULL)
1741 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1746 nfs4_free_stateowner(struct kref *kref)
1748 struct nfs4_stateowner *sop =
1749 container_of(kref, struct nfs4_stateowner, so_ref);
1750 kfree(sop->so_owner.data);
1751 kmem_cache_free(stateowner_slab, sop);
1754 static inline struct nfs4_stateowner *
1755 alloc_stateowner(struct xdr_netobj *owner)
1757 struct nfs4_stateowner *sop;
1759 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1760 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1761 memcpy(sop->so_owner.data, owner->data, owner->len);
1762 sop->so_owner.len = owner->len;
1763 kref_init(&sop->so_ref);
1766 kmem_cache_free(stateowner_slab, sop);
1771 static struct nfs4_stateowner *
1772 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1773 struct nfs4_stateowner *sop;
1774 struct nfs4_replay *rp;
1775 unsigned int idhashval;
1777 if (!(sop = alloc_stateowner(&open->op_owner)))
1779 idhashval = ownerid_hashval(current_ownerid);
1780 INIT_LIST_HEAD(&sop->so_idhash);
1781 INIT_LIST_HEAD(&sop->so_strhash);
1782 INIT_LIST_HEAD(&sop->so_perclient);
1783 INIT_LIST_HEAD(&sop->so_stateids);
1784 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1785 INIT_LIST_HEAD(&sop->so_close_lru);
1787 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1788 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1789 list_add(&sop->so_perclient, &clp->cl_openowners);
1790 sop->so_is_open_owner = 1;
1791 sop->so_id = current_ownerid++;
1792 sop->so_client = clp;
1793 sop->so_seqid = open->op_seqid;
1794 sop->so_confirmed = 0;
1795 rp = &sop->so_replay;
1796 rp->rp_status = nfserr_serverfault;
1798 rp->rp_buf = rp->rp_ibuf;
1803 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1804 struct nfs4_stateowner *sop = open->op_stateowner;
1805 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1807 INIT_LIST_HEAD(&stp->st_hash);
1808 INIT_LIST_HEAD(&stp->st_perstateowner);
1809 INIT_LIST_HEAD(&stp->st_lockowners);
1810 INIT_LIST_HEAD(&stp->st_perfile);
1811 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1812 list_add(&stp->st_perstateowner, &sop->so_stateids);
1813 list_add(&stp->st_perfile, &fp->fi_stateids);
1814 stp->st_stateowner = sop;
1817 stp->st_stateid.si_boot = get_seconds();
1818 stp->st_stateid.si_stateownerid = sop->so_id;
1819 stp->st_stateid.si_fileid = fp->fi_id;
1820 stp->st_stateid.si_generation = 0;
1821 stp->st_access_bmap = 0;
1822 stp->st_deny_bmap = 0;
1823 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1824 &stp->st_access_bmap);
1825 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1826 stp->st_openstp = NULL;
1830 move_to_close_lru(struct nfs4_stateowner *sop)
1832 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1834 list_move_tail(&sop->so_close_lru, &close_lru);
1835 sop->so_time = get_seconds();
1839 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1842 return (sop->so_owner.len == owner->len) &&
1843 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1844 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1847 static struct nfs4_stateowner *
1848 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1850 struct nfs4_stateowner *so = NULL;
1852 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1853 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1859 /* search file_hashtbl[] for file */
1860 static struct nfs4_file *
1861 find_file(struct inode *ino)
1863 unsigned int hashval = file_hashval(ino);
1864 struct nfs4_file *fp;
1866 spin_lock(&recall_lock);
1867 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1868 if (fp->fi_inode == ino) {
1870 spin_unlock(&recall_lock);
1874 spin_unlock(&recall_lock);
1878 static inline int access_valid(u32 x, u32 minorversion)
1880 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1882 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1884 x &= ~NFS4_SHARE_ACCESS_MASK;
1885 if (minorversion && x) {
1886 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1888 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1890 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1897 static inline int deny_valid(u32 x)
1899 /* Note: unlike access bits, deny bits may be zero. */
1900 return x <= NFS4_SHARE_DENY_BOTH;
1904 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1905 * st_{access,deny}_bmap field of the stateid, in order to track not
1906 * only what share bits are currently in force, but also what
1907 * combinations of share bits previous opens have used. This allows us
1908 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1909 * return an error if the client attempt to downgrade to a combination
1910 * of share bits not explicable by closing some of its previous opens.
1912 * XXX: This enforcement is actually incomplete, since we don't keep
1913 * track of access/deny bit combinations; so, e.g., we allow:
1915 * OPEN allow read, deny write
1916 * OPEN allow both, deny none
1917 * DOWNGRADE allow read, deny none
1919 * which we should reject.
1922 set_access(unsigned int *access, unsigned long bmap) {
1926 for (i = 1; i < 4; i++) {
1927 if (test_bit(i, &bmap))
1933 set_deny(unsigned int *deny, unsigned long bmap) {
1937 for (i = 0; i < 4; i++) {
1938 if (test_bit(i, &bmap))
1944 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1945 unsigned int access, deny;
1947 set_access(&access, stp->st_access_bmap);
1948 set_deny(&deny, stp->st_deny_bmap);
1949 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1955 * Called to check deny when READ with all zero stateid or
1956 * WRITE with all zero or all one stateid
1959 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1961 struct inode *ino = current_fh->fh_dentry->d_inode;
1962 struct nfs4_file *fp;
1963 struct nfs4_stateid *stp;
1966 dprintk("NFSD: nfs4_share_conflict\n");
1968 fp = find_file(ino);
1971 ret = nfserr_locked;
1972 /* Search for conflicting share reservations */
1973 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1974 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1975 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1985 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1987 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1988 drop_file_write_access(filp);
1989 spin_lock(&filp->f_lock);
1990 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1991 spin_unlock(&filp->f_lock);
1996 * Spawn a thread to perform a recall on the delegation represented
1997 * by the lease (file_lock)
1999 * Called from break_lease() with lock_kernel() held.
2000 * Note: we assume break_lease will only call this *once* for any given
2004 void nfsd_break_deleg_cb(struct file_lock *fl)
2006 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2008 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2012 /* We're assuming the state code never drops its reference
2013 * without first removing the lease. Since we're in this lease
2014 * callback (and since the lease code is serialized by the kernel
2015 * lock) we know the server hasn't removed the lease yet, we know
2016 * it's safe to take a reference: */
2017 atomic_inc(&dp->dl_count);
2019 spin_lock(&recall_lock);
2020 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2021 spin_unlock(&recall_lock);
2023 /* only place dl_time is set. protected by lock_kernel*/
2024 dp->dl_time = get_seconds();
2027 * We don't want the locks code to timeout the lease for us;
2028 * we'll remove it ourself if the delegation isn't returned
2031 fl->fl_break_time = 0;
2033 dp->dl_file->fi_had_conflict = true;
2034 nfsd4_cb_recall(dp);
2038 * The file_lock is being reapd.
2040 * Called by locks_free_lock() with lock_kernel() held.
2043 void nfsd_release_deleg_cb(struct file_lock *fl)
2045 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2047 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2049 if (!(fl->fl_flags & FL_LEASE) || !dp)
2051 dp->dl_flock = NULL;
2055 * Set the delegation file_lock back pointer.
2057 * Called from setlease() with lock_kernel() held.
2060 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2062 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2064 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2071 * Called from setlease() with lock_kernel() held
2074 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2076 struct nfs4_delegation *onlistd =
2077 (struct nfs4_delegation *)onlist->fl_owner;
2078 struct nfs4_delegation *tryd =
2079 (struct nfs4_delegation *)try->fl_owner;
2081 if (onlist->fl_lmops != try->fl_lmops)
2084 return onlistd->dl_client == tryd->dl_client;
2089 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2092 return lease_modify(onlist, arg);
2097 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2098 .fl_break = nfsd_break_deleg_cb,
2099 .fl_release_private = nfsd_release_deleg_cb,
2100 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2101 .fl_mylease = nfsd_same_client_deleg_cb,
2102 .fl_change = nfsd_change_deleg_cb,
2107 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2108 struct nfsd4_open *open)
2110 clientid_t *clientid = &open->op_clientid;
2111 struct nfs4_client *clp = NULL;
2112 unsigned int strhashval;
2113 struct nfs4_stateowner *sop = NULL;
2115 if (!check_name(open->op_owner))
2116 return nfserr_inval;
2118 if (STALE_CLIENTID(&open->op_clientid))
2119 return nfserr_stale_clientid;
2121 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2122 sop = find_openstateowner_str(strhashval, open);
2123 open->op_stateowner = sop;
2125 /* Make sure the client's lease hasn't expired. */
2126 clp = find_confirmed_client(clientid);
2128 return nfserr_expired;
2131 /* When sessions are used, skip open sequenceid processing */
2132 if (nfsd4_has_session(cstate))
2134 if (!sop->so_confirmed) {
2135 /* Replace unconfirmed owners without checking for replay. */
2136 clp = sop->so_client;
2137 release_openowner(sop);
2138 open->op_stateowner = NULL;
2141 if (open->op_seqid == sop->so_seqid - 1) {
2142 if (sop->so_replay.rp_buflen)
2143 return nfserr_replay_me;
2144 /* The original OPEN failed so spectacularly
2145 * that we don't even have replay data saved!
2146 * Therefore, we have no choice but to continue
2147 * processing this OPEN; presumably, we'll
2148 * fail again for the same reason.
2150 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2153 if (open->op_seqid != sop->so_seqid)
2154 return nfserr_bad_seqid;
2156 if (open->op_stateowner == NULL) {
2157 sop = alloc_init_open_stateowner(strhashval, clp, open);
2159 return nfserr_resource;
2160 open->op_stateowner = sop;
2162 list_del_init(&sop->so_close_lru);
2163 renew_client(sop->so_client);
2167 static inline __be32
2168 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2170 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2171 return nfserr_openmode;
2176 static struct nfs4_delegation *
2177 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2179 struct nfs4_delegation *dp;
2181 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2182 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2189 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2190 struct nfs4_delegation **dp)
2193 __be32 status = nfserr_bad_stateid;
2195 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2198 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2199 RD_STATE : WR_STATE;
2200 status = nfs4_check_delegmode(*dp, flags);
2204 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2208 open->op_stateowner->so_confirmed = 1;
2213 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2215 struct nfs4_stateid *local;
2216 __be32 status = nfserr_share_denied;
2217 struct nfs4_stateowner *sop = open->op_stateowner;
2219 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2220 /* ignore lock owners */
2221 if (local->st_stateowner->so_is_open_owner == 0)
2223 /* remember if we have seen this open owner */
2224 if (local->st_stateowner == sop)
2226 /* check for conflicting share reservations */
2227 if (!test_share(local, open))
2235 static inline struct nfs4_stateid *
2236 nfs4_alloc_stateid(void)
2238 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2242 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2243 struct nfs4_delegation *dp,
2244 struct svc_fh *cur_fh, int flags)
2246 struct nfs4_stateid *stp;
2248 stp = nfs4_alloc_stateid();
2250 return nfserr_resource;
2253 get_file(dp->dl_vfs_file);
2254 stp->st_vfs_file = dp->dl_vfs_file;
2257 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2260 if (status == nfserr_dropit)
2261 status = nfserr_jukebox;
2262 kmem_cache_free(stateid_slab, stp);
2270 static inline __be32
2271 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2272 struct nfsd4_open *open)
2274 struct iattr iattr = {
2275 .ia_valid = ATTR_SIZE,
2278 if (!open->op_truncate)
2280 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2281 return nfserr_inval;
2282 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2286 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2288 struct file *filp = stp->st_vfs_file;
2289 struct inode *inode = filp->f_path.dentry->d_inode;
2290 unsigned int share_access, new_writer;
2293 set_access(&share_access, stp->st_access_bmap);
2294 new_writer = (~share_access) & open->op_share_access
2295 & NFS4_SHARE_ACCESS_WRITE;
2298 int err = get_write_access(inode);
2300 return nfserrno(err);
2301 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2303 return nfserrno(err);
2304 file_take_write(filp);
2306 status = nfsd4_truncate(rqstp, cur_fh, open);
2309 put_write_access(inode);
2312 /* remember the open */
2313 filp->f_mode |= open->op_share_access;
2314 __set_bit(open->op_share_access, &stp->st_access_bmap);
2315 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2322 nfs4_set_claim_prev(struct nfsd4_open *open)
2324 open->op_stateowner->so_confirmed = 1;
2325 open->op_stateowner->so_client->cl_firststate = 1;
2329 * Attempt to hand out a delegation.
2332 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2334 struct nfs4_delegation *dp;
2335 struct nfs4_stateowner *sop = stp->st_stateowner;
2336 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2337 struct file_lock fl, *flp = &fl;
2338 int status, flag = 0;
2340 flag = NFS4_OPEN_DELEGATE_NONE;
2341 open->op_recall = 0;
2342 switch (open->op_claim_type) {
2343 case NFS4_OPEN_CLAIM_PREVIOUS:
2345 open->op_recall = 1;
2346 flag = open->op_delegate_type;
2347 if (flag == NFS4_OPEN_DELEGATE_NONE)
2350 case NFS4_OPEN_CLAIM_NULL:
2351 /* Let's not give out any delegations till everyone's
2352 * had the chance to reclaim theirs.... */
2353 if (locks_in_grace())
2355 if (!cb_up || !sop->so_confirmed)
2357 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2358 flag = NFS4_OPEN_DELEGATE_WRITE;
2360 flag = NFS4_OPEN_DELEGATE_READ;
2366 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2368 flag = NFS4_OPEN_DELEGATE_NONE;
2371 locks_init_lock(&fl);
2372 fl.fl_lmops = &nfsd_lease_mng_ops;
2373 fl.fl_flags = FL_LEASE;
2374 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2375 fl.fl_end = OFFSET_MAX;
2376 fl.fl_owner = (fl_owner_t)dp;
2377 fl.fl_file = stp->st_vfs_file;
2378 fl.fl_pid = current->tgid;
2380 /* vfs_setlease checks to see if delegation should be handed out.
2381 * the lock_manager callbacks fl_mylease and fl_change are used
2383 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2384 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2385 unhash_delegation(dp);
2386 flag = NFS4_OPEN_DELEGATE_NONE;
2390 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2392 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2393 STATEID_VAL(&dp->dl_stateid));
2395 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2396 && flag == NFS4_OPEN_DELEGATE_NONE
2397 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2398 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2399 open->op_delegate_type = flag;
2403 * called with nfs4_lock_state() held.
2406 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2408 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2409 struct nfs4_file *fp = NULL;
2410 struct inode *ino = current_fh->fh_dentry->d_inode;
2411 struct nfs4_stateid *stp = NULL;
2412 struct nfs4_delegation *dp = NULL;
2415 status = nfserr_inval;
2416 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2417 || !deny_valid(open->op_share_deny))
2420 * Lookup file; if found, lookup stateid and check open request,
2421 * and check for delegations in the process of being recalled.
2422 * If not found, create the nfs4_file struct
2424 fp = find_file(ino);
2426 if ((status = nfs4_check_open(fp, open, &stp)))
2428 status = nfs4_check_deleg(fp, open, &dp);
2432 status = nfserr_bad_stateid;
2433 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2435 status = nfserr_resource;
2436 fp = alloc_init_file(ino);
2442 * OPEN the file, or upgrade an existing OPEN.
2443 * If truncate fails, the OPEN fails.
2446 /* Stateid was found, this is an OPEN upgrade */
2447 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2450 update_stateid(&stp->st_stateid);
2452 /* Stateid was not found, this is a new OPEN */
2454 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2455 flags |= NFSD_MAY_READ;
2456 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2457 flags |= NFSD_MAY_WRITE;
2458 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2461 init_stateid(stp, fp, open);
2462 status = nfsd4_truncate(rqstp, current_fh, open);
2464 release_open_stateid(stp);
2467 if (nfsd4_has_session(&resp->cstate))
2468 update_stateid(&stp->st_stateid);
2470 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2472 if (nfsd4_has_session(&resp->cstate)) {
2473 open->op_stateowner->so_confirmed = 1;
2474 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2478 * Attempt to hand out a delegation. No error return, because the
2479 * OPEN succeeds even if we fail.
2481 nfs4_open_delegation(current_fh, open, stp);
2485 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2486 STATEID_VAL(&stp->st_stateid));
2490 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2491 nfs4_set_claim_prev(open);
2493 * To finish the open response, we just need to set the rflags.
2495 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2496 if (!open->op_stateowner->so_confirmed &&
2497 !nfsd4_has_session(&resp->cstate))
2498 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2504 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2507 struct nfs4_client *clp;
2511 dprintk("process_renew(%08x/%08x): starting\n",
2512 clid->cl_boot, clid->cl_id);
2513 status = nfserr_stale_clientid;
2514 if (STALE_CLIENTID(clid))
2516 clp = find_confirmed_client(clid);
2517 status = nfserr_expired;
2519 /* We assume the client took too long to RENEW. */
2520 dprintk("nfsd4_renew: clientid not found!\n");
2524 status = nfserr_cb_path_down;
2525 if (!list_empty(&clp->cl_delegations)
2526 && !atomic_read(&clp->cl_cb_set))
2530 nfs4_unlock_state();
2534 struct lock_manager nfsd4_manager = {
2538 nfsd4_end_grace(void)
2540 dprintk("NFSD: end of grace period\n");
2541 nfsd4_recdir_purge_old();
2542 locks_end_grace(&nfsd4_manager);
2544 * Now that every NFSv4 client has had the chance to recover and
2545 * to see the (possibly new, possibly shorter) lease time, we
2546 * can safely set the next grace time to the current lease time:
2548 nfsd4_grace = nfsd4_lease;
2552 nfs4_laundromat(void)
2554 struct nfs4_client *clp;
2555 struct nfs4_stateowner *sop;
2556 struct nfs4_delegation *dp;
2557 struct list_head *pos, *next, reaplist;
2558 time_t cutoff = get_seconds() - nfsd4_lease;
2559 time_t t, clientid_val = nfsd4_lease;
2560 time_t u, test_val = nfsd4_lease;
2564 dprintk("NFSD: laundromat service - starting\n");
2565 if (locks_in_grace())
2567 list_for_each_safe(pos, next, &client_lru) {
2568 clp = list_entry(pos, struct nfs4_client, cl_lru);
2569 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2570 t = clp->cl_time - cutoff;
2571 if (clientid_val > t)
2575 dprintk("NFSD: purging unused client (clientid %08x)\n",
2576 clp->cl_clientid.cl_id);
2577 nfsd4_remove_clid_dir(clp);
2580 INIT_LIST_HEAD(&reaplist);
2581 spin_lock(&recall_lock);
2582 list_for_each_safe(pos, next, &del_recall_lru) {
2583 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2584 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2585 u = dp->dl_time - cutoff;
2590 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2592 list_move(&dp->dl_recall_lru, &reaplist);
2594 spin_unlock(&recall_lock);
2595 list_for_each_safe(pos, next, &reaplist) {
2596 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2597 list_del_init(&dp->dl_recall_lru);
2598 unhash_delegation(dp);
2600 test_val = nfsd4_lease;
2601 list_for_each_safe(pos, next, &close_lru) {
2602 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2603 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2604 u = sop->so_time - cutoff;
2609 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2611 release_openowner(sop);
2613 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2614 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2615 nfs4_unlock_state();
2616 return clientid_val;
2619 static struct workqueue_struct *laundry_wq;
2620 static void laundromat_main(struct work_struct *);
2621 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2624 laundromat_main(struct work_struct *not_used)
2628 t = nfs4_laundromat();
2629 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2630 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2633 static struct nfs4_stateowner *
2634 search_close_lru(u32 st_id, int flags)
2636 struct nfs4_stateowner *local = NULL;
2638 if (flags & CLOSE_STATE) {
2639 list_for_each_entry(local, &close_lru, so_close_lru) {
2640 if (local->so_id == st_id)
2648 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2650 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2654 STALE_STATEID(stateid_t *stateid)
2656 if (time_after((unsigned long)boot_time,
2657 (unsigned long)stateid->si_boot)) {
2658 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2659 STATEID_VAL(stateid));
2666 EXPIRED_STATEID(stateid_t *stateid)
2668 if (time_before((unsigned long)boot_time,
2669 ((unsigned long)stateid->si_boot)) &&
2670 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
2671 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2672 STATEID_VAL(stateid));
2679 stateid_error_map(stateid_t *stateid)
2681 if (STALE_STATEID(stateid))
2682 return nfserr_stale_stateid;
2683 if (EXPIRED_STATEID(stateid))
2684 return nfserr_expired;
2686 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2687 STATEID_VAL(stateid));
2688 return nfserr_bad_stateid;
2692 access_permit_read(unsigned long access_bmap)
2694 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2695 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2696 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2700 access_permit_write(unsigned long access_bmap)
2702 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2703 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2707 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2709 __be32 status = nfserr_openmode;
2711 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2713 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2720 static inline __be32
2721 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2723 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2725 else if (locks_in_grace()) {
2726 /* Answer in remaining cases depends on existance of
2727 * conflicting state; so we must wait out the grace period. */
2728 return nfserr_grace;
2729 } else if (flags & WR_STATE)
2730 return nfs4_share_conflict(current_fh,
2731 NFS4_SHARE_DENY_WRITE);
2732 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2733 return nfs4_share_conflict(current_fh,
2734 NFS4_SHARE_DENY_READ);
2738 * Allow READ/WRITE during grace period on recovered state only for files
2739 * that are not able to provide mandatory locking.
2742 grace_disallows_io(struct inode *inode)
2744 return locks_in_grace() && mandatory_lock(inode);
2747 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2750 * When sessions are used the stateid generation number is ignored
2753 if ((flags & HAS_SESSION) && in->si_generation == 0)
2756 /* If the client sends us a stateid from the future, it's buggy: */
2757 if (in->si_generation > ref->si_generation)
2758 return nfserr_bad_stateid;
2760 * The following, however, can happen. For example, if the
2761 * client sends an open and some IO at the same time, the open
2762 * may bump si_generation while the IO is still in flight.
2763 * Thanks to hard links and renames, the client never knows what
2764 * file an open will affect. So it could avoid that situation
2765 * only by serializing all opens and IO from the same open
2766 * owner. To recover from the old_stateid error, the client
2767 * will just have to retry the IO:
2769 if (in->si_generation < ref->si_generation)
2770 return nfserr_old_stateid;
2775 static int is_delegation_stateid(stateid_t *stateid)
2777 return stateid->si_fileid == 0;
2781 * Checks for stateid operations
2784 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2785 stateid_t *stateid, int flags, struct file **filpp)
2787 struct nfs4_stateid *stp = NULL;
2788 struct nfs4_delegation *dp = NULL;
2789 struct svc_fh *current_fh = &cstate->current_fh;
2790 struct inode *ino = current_fh->fh_dentry->d_inode;
2796 if (grace_disallows_io(ino))
2797 return nfserr_grace;
2799 if (nfsd4_has_session(cstate))
2800 flags |= HAS_SESSION;
2802 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2803 return check_special_stateids(current_fh, stateid, flags);
2805 status = nfserr_stale_stateid;
2806 if (STALE_STATEID(stateid))
2809 status = nfserr_bad_stateid;
2810 if (is_delegation_stateid(stateid)) {
2811 dp = find_delegation_stateid(ino, stateid);
2813 status = stateid_error_map(stateid);
2816 status = check_stateid_generation(stateid, &dp->dl_stateid,
2820 status = nfs4_check_delegmode(dp, flags);
2823 renew_client(dp->dl_client);
2825 *filpp = dp->dl_vfs_file;
2826 } else { /* open or lock stateid */
2827 stp = find_stateid(stateid, flags);
2829 status = stateid_error_map(stateid);
2832 if (nfs4_check_fh(current_fh, stp))
2834 if (!stp->st_stateowner->so_confirmed)
2836 status = check_stateid_generation(stateid, &stp->st_stateid,
2840 status = nfs4_check_openmode(stp, flags);
2843 renew_client(stp->st_stateowner->so_client);
2845 *filpp = stp->st_vfs_file;
2855 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2856 RD_STATE : WR_STATE;
2860 * Checks for sequence id mutating operations.
2863 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2864 stateid_t *stateid, int flags,
2865 struct nfs4_stateowner **sopp,
2866 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2868 struct nfs4_stateid *stp;
2869 struct nfs4_stateowner *sop;
2870 struct svc_fh *current_fh = &cstate->current_fh;
2873 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2874 seqid, STATEID_VAL(stateid));
2879 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2880 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2881 return nfserr_bad_stateid;
2884 if (STALE_STATEID(stateid))
2885 return nfserr_stale_stateid;
2887 if (nfsd4_has_session(cstate))
2888 flags |= HAS_SESSION;
2891 * We return BAD_STATEID if filehandle doesn't match stateid,
2892 * the confirmed flag is incorrecly set, or the generation
2893 * number is incorrect.
2895 stp = find_stateid(stateid, flags);
2898 * Also, we should make sure this isn't just the result of
2901 sop = search_close_lru(stateid->si_stateownerid, flags);
2903 return stateid_error_map(stateid);
2909 *sopp = sop = stp->st_stateowner;
2912 clientid_t *lockclid = &lock->v.new.clientid;
2913 struct nfs4_client *clp = sop->so_client;
2917 lkflg = setlkflg(lock->lk_type);
2919 if (lock->lk_is_new) {
2920 if (!sop->so_is_open_owner)
2921 return nfserr_bad_stateid;
2922 if (!(flags & HAS_SESSION) &&
2923 !same_clid(&clp->cl_clientid, lockclid))
2924 return nfserr_bad_stateid;
2925 /* stp is the open stateid */
2926 status = nfs4_check_openmode(stp, lkflg);
2930 /* stp is the lock stateid */
2931 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2937 if (nfs4_check_fh(current_fh, stp)) {
2938 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2939 return nfserr_bad_stateid;
2943 * We now validate the seqid and stateid generation numbers.
2944 * For the moment, we ignore the possibility of
2945 * generation number wraparound.
2947 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
2950 if (sop->so_confirmed && flags & CONFIRM) {
2951 dprintk("NFSD: preprocess_seqid_op: expected"
2952 " unconfirmed stateowner!\n");
2953 return nfserr_bad_stateid;
2955 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2956 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2957 " confirmed yet!\n");
2958 return nfserr_bad_stateid;
2960 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
2963 renew_client(sop->so_client);
2967 if (seqid == sop->so_seqid - 1) {
2968 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2969 /* indicate replay to calling function */
2970 return nfserr_replay_me;
2972 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2973 sop->so_seqid, seqid);
2975 return nfserr_bad_seqid;
2979 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2980 struct nfsd4_open_confirm *oc)
2983 struct nfs4_stateowner *sop;
2984 struct nfs4_stateid *stp;
2986 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2987 (int)cstate->current_fh.fh_dentry->d_name.len,
2988 cstate->current_fh.fh_dentry->d_name.name);
2990 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2996 if ((status = nfs4_preprocess_seqid_op(cstate,
2997 oc->oc_seqid, &oc->oc_req_stateid,
2998 CONFIRM | OPEN_STATE,
2999 &oc->oc_stateowner, &stp, NULL)))
3002 sop = oc->oc_stateowner;
3003 sop->so_confirmed = 1;
3004 update_stateid(&stp->st_stateid);
3005 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3006 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3007 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3009 nfsd4_create_clid_dir(sop->so_client);
3011 if (oc->oc_stateowner) {
3012 nfs4_get_stateowner(oc->oc_stateowner);
3013 cstate->replay_owner = oc->oc_stateowner;
3015 nfs4_unlock_state();
3021 * unset all bits in union bitmap (bmap) that
3022 * do not exist in share (from successful OPEN_DOWNGRADE)
3025 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3028 for (i = 1; i < 4; i++) {
3029 if ((i & access) != i)
3030 __clear_bit(i, bmap);
3035 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3038 for (i = 0; i < 4; i++) {
3039 if ((i & deny) != i)
3040 __clear_bit(i, bmap);
3045 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3046 struct nfsd4_compound_state *cstate,
3047 struct nfsd4_open_downgrade *od)
3050 struct nfs4_stateid *stp;
3051 unsigned int share_access;
3053 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3054 (int)cstate->current_fh.fh_dentry->d_name.len,
3055 cstate->current_fh.fh_dentry->d_name.name);
3057 if (!access_valid(od->od_share_access, cstate->minorversion)
3058 || !deny_valid(od->od_share_deny))
3059 return nfserr_inval;
3062 if ((status = nfs4_preprocess_seqid_op(cstate,
3066 &od->od_stateowner, &stp, NULL)))
3069 status = nfserr_inval;
3070 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3071 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3072 stp->st_access_bmap, od->od_share_access);
3075 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3076 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3077 stp->st_deny_bmap, od->od_share_deny);
3080 set_access(&share_access, stp->st_access_bmap);
3081 nfs4_file_downgrade(stp->st_vfs_file,
3082 share_access & ~od->od_share_access);
3084 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3085 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3087 update_stateid(&stp->st_stateid);
3088 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3091 if (od->od_stateowner) {
3092 nfs4_get_stateowner(od->od_stateowner);
3093 cstate->replay_owner = od->od_stateowner;
3095 nfs4_unlock_state();
3100 * nfs4_unlock_state() called after encode
3103 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3104 struct nfsd4_close *close)
3107 struct nfs4_stateid *stp;
3109 dprintk("NFSD: nfsd4_close on file %.*s\n",
3110 (int)cstate->current_fh.fh_dentry->d_name.len,
3111 cstate->current_fh.fh_dentry->d_name.name);
3114 /* check close_lru for replay */
3115 if ((status = nfs4_preprocess_seqid_op(cstate,
3118 OPEN_STATE | CLOSE_STATE,
3119 &close->cl_stateowner, &stp, NULL)))
3122 update_stateid(&stp->st_stateid);
3123 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3125 /* release_stateid() calls nfsd_close() if needed */
3126 release_open_stateid(stp);
3128 /* place unused nfs4_stateowners on so_close_lru list to be
3129 * released by the laundromat service after the lease period
3130 * to enable us to handle CLOSE replay
3132 if (list_empty(&close->cl_stateowner->so_stateids))
3133 move_to_close_lru(close->cl_stateowner);
3135 if (close->cl_stateowner) {
3136 nfs4_get_stateowner(close->cl_stateowner);
3137 cstate->replay_owner = close->cl_stateowner;
3139 nfs4_unlock_state();
3144 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3145 struct nfsd4_delegreturn *dr)
3147 struct nfs4_delegation *dp;
3148 stateid_t *stateid = &dr->dr_stateid;
3149 struct inode *inode;
3153 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3155 inode = cstate->current_fh.fh_dentry->d_inode;
3157 if (nfsd4_has_session(cstate))
3158 flags |= HAS_SESSION;
3160 status = nfserr_bad_stateid;
3161 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3163 status = nfserr_stale_stateid;
3164 if (STALE_STATEID(stateid))
3166 status = nfserr_bad_stateid;
3167 if (!is_delegation_stateid(stateid))
3169 dp = find_delegation_stateid(inode, stateid);
3171 status = stateid_error_map(stateid);
3174 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3177 renew_client(dp->dl_client);
3179 unhash_delegation(dp);
3181 nfs4_unlock_state();
3188 * Lock owner state (byte-range locks)
3190 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3191 #define LOCK_HASH_BITS 8
3192 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3193 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3196 end_offset(u64 start, u64 len)
3201 return end >= start ? end: NFS4_MAX_UINT64;
3204 /* last octet in a range */
3206 last_byte_offset(u64 start, u64 len)
3212 return end > start ? end - 1: NFS4_MAX_UINT64;
3215 #define lockownerid_hashval(id) \
3216 ((id) & LOCK_HASH_MASK)
3218 static inline unsigned int
3219 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3220 struct xdr_netobj *ownername)
3222 return (file_hashval(inode) + cl_id
3223 + opaque_hashval(ownername->data, ownername->len))
3227 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3228 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3229 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3231 static struct nfs4_stateid *
3232 find_stateid(stateid_t *stid, int flags)
3234 struct nfs4_stateid *local;
3235 u32 st_id = stid->si_stateownerid;
3236 u32 f_id = stid->si_fileid;
3237 unsigned int hashval;
3239 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3240 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3241 hashval = stateid_hashval(st_id, f_id);
3242 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3243 if ((local->st_stateid.si_stateownerid == st_id) &&
3244 (local->st_stateid.si_fileid == f_id))
3249 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3250 hashval = stateid_hashval(st_id, f_id);
3251 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3252 if ((local->st_stateid.si_stateownerid == st_id) &&
3253 (local->st_stateid.si_fileid == f_id))
3260 static struct nfs4_delegation *
3261 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3263 struct nfs4_file *fp;
3264 struct nfs4_delegation *dl;
3266 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3269 fp = find_file(ino);
3272 dl = find_delegation_file(fp, stid);
3278 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3279 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3280 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3281 * locking, this prevents us from being completely protocol-compliant. The
3282 * real solution to this problem is to start using unsigned file offsets in
3283 * the VFS, but this is a very deep change!
3286 nfs4_transform_lock_offset(struct file_lock *lock)
3288 if (lock->fl_start < 0)
3289 lock->fl_start = OFFSET_MAX;
3290 if (lock->fl_end < 0)
3291 lock->fl_end = OFFSET_MAX;
3294 /* Hack!: For now, we're defining this just so we can use a pointer to it
3295 * as a unique cookie to identify our (NFSv4's) posix locks. */
3296 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3300 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3302 struct nfs4_stateowner *sop;
3305 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3306 sop = (struct nfs4_stateowner *) fl->fl_owner;
3307 hval = lockownerid_hashval(sop->so_id);
3308 kref_get(&sop->so_ref);
3310 deny->ld_clientid = sop->so_client->cl_clientid;
3312 deny->ld_sop = NULL;
3313 deny->ld_clientid.cl_boot = 0;
3314 deny->ld_clientid.cl_id = 0;
3316 deny->ld_start = fl->fl_start;
3317 deny->ld_length = NFS4_MAX_UINT64;
3318 if (fl->fl_end != NFS4_MAX_UINT64)
3319 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3320 deny->ld_type = NFS4_READ_LT;
3321 if (fl->fl_type != F_RDLCK)
3322 deny->ld_type = NFS4_WRITE_LT;
3325 static struct nfs4_stateowner *
3326 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3327 struct xdr_netobj *owner)
3329 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3330 struct nfs4_stateowner *op;
3332 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3333 if (same_owner_str(op, owner, clid))
3340 * Alloc a lock owner structure.
3341 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3344 * strhashval = lock_ownerstr_hashval
3347 static struct nfs4_stateowner *
3348 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3349 struct nfs4_stateowner *sop;
3350 struct nfs4_replay *rp;
3351 unsigned int idhashval;
3353 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3355 idhashval = lockownerid_hashval(current_ownerid);
3356 INIT_LIST_HEAD(&sop->so_idhash);
3357 INIT_LIST_HEAD(&sop->so_strhash);
3358 INIT_LIST_HEAD(&sop->so_perclient);
3359 INIT_LIST_HEAD(&sop->so_stateids);
3360 INIT_LIST_HEAD(&sop->so_perstateid);
3361 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3363 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3364 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3365 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3366 sop->so_is_open_owner = 0;
3367 sop->so_id = current_ownerid++;
3368 sop->so_client = clp;
3369 /* It is the openowner seqid that will be incremented in encode in the
3370 * case of new lockowners; so increment the lock seqid manually: */
3371 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3372 sop->so_confirmed = 1;
3373 rp = &sop->so_replay;
3374 rp->rp_status = nfserr_serverfault;
3376 rp->rp_buf = rp->rp_ibuf;
3380 static struct nfs4_stateid *
3381 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3383 struct nfs4_stateid *stp;
3384 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3386 stp = nfs4_alloc_stateid();
3389 INIT_LIST_HEAD(&stp->st_hash);
3390 INIT_LIST_HEAD(&stp->st_perfile);
3391 INIT_LIST_HEAD(&stp->st_perstateowner);
3392 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3393 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3394 list_add(&stp->st_perfile, &fp->fi_stateids);
3395 list_add(&stp->st_perstateowner, &sop->so_stateids);
3396 stp->st_stateowner = sop;
3399 stp->st_stateid.si_boot = get_seconds();
3400 stp->st_stateid.si_stateownerid = sop->so_id;
3401 stp->st_stateid.si_fileid = fp->fi_id;
3402 stp->st_stateid.si_generation = 0;
3403 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3404 stp->st_access_bmap = open_stp->st_access_bmap;
3405 stp->st_deny_bmap = open_stp->st_deny_bmap;
3406 stp->st_openstp = open_stp;
3413 check_lock_length(u64 offset, u64 length)
3415 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3416 LOFF_OVERFLOW(offset, length)));
3423 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3424 struct nfsd4_lock *lock)
3426 struct nfs4_stateowner *open_sop = NULL;
3427 struct nfs4_stateowner *lock_sop = NULL;
3428 struct nfs4_stateid *lock_stp;
3430 struct file_lock file_lock;
3431 struct file_lock conflock;
3433 unsigned int strhashval;
3437 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3438 (long long) lock->lk_offset,
3439 (long long) lock->lk_length);
3441 if (check_lock_length(lock->lk_offset, lock->lk_length))
3442 return nfserr_inval;
3444 if ((status = fh_verify(rqstp, &cstate->current_fh,
3445 S_IFREG, NFSD_MAY_LOCK))) {
3446 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3452 if (lock->lk_is_new) {
3454 * Client indicates that this is a new lockowner.
3455 * Use open owner and open stateid to create lock owner and
3458 struct nfs4_stateid *open_stp = NULL;
3459 struct nfs4_file *fp;
3461 status = nfserr_stale_clientid;
3462 if (!nfsd4_has_session(cstate) &&
3463 STALE_CLIENTID(&lock->lk_new_clientid))
3466 /* validate and update open stateid and open seqid */
3467 status = nfs4_preprocess_seqid_op(cstate,
3468 lock->lk_new_open_seqid,
3469 &lock->lk_new_open_stateid,
3471 &lock->lk_replay_owner, &open_stp,
3475 open_sop = lock->lk_replay_owner;
3476 /* create lockowner and lock stateid */
3477 fp = open_stp->st_file;
3478 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3479 open_sop->so_client->cl_clientid.cl_id,
3480 &lock->v.new.owner);
3481 /* XXX: Do we need to check for duplicate stateowners on
3482 * the same file, or should they just be allowed (and
3483 * create new stateids)? */
3484 status = nfserr_resource;
3485 lock_sop = alloc_init_lock_stateowner(strhashval,
3486 open_sop->so_client, open_stp, lock);
3487 if (lock_sop == NULL)
3489 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3490 if (lock_stp == NULL)
3493 /* lock (lock owner + lock stateid) already exists */
3494 status = nfs4_preprocess_seqid_op(cstate,
3495 lock->lk_old_lock_seqid,
3496 &lock->lk_old_lock_stateid,
3498 &lock->lk_replay_owner, &lock_stp, lock);
3501 lock_sop = lock->lk_replay_owner;
3503 /* lock->lk_replay_owner and lock_stp have been created or found */
3504 filp = lock_stp->st_vfs_file;
3506 status = nfserr_grace;
3507 if (locks_in_grace() && !lock->lk_reclaim)
3509 status = nfserr_no_grace;
3510 if (!locks_in_grace() && lock->lk_reclaim)
3513 locks_init_lock(&file_lock);
3514 switch (lock->lk_type) {
3517 file_lock.fl_type = F_RDLCK;
3521 case NFS4_WRITEW_LT:
3522 file_lock.fl_type = F_WRLCK;
3526 status = nfserr_inval;
3529 file_lock.fl_owner = (fl_owner_t)lock_sop;
3530 file_lock.fl_pid = current->tgid;
3531 file_lock.fl_file = filp;
3532 file_lock.fl_flags = FL_POSIX;
3533 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3535 file_lock.fl_start = lock->lk_offset;
3536 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3537 nfs4_transform_lock_offset(&file_lock);
3540 * Try to lock the file in the VFS.
3541 * Note: locks.c uses the BKL to protect the inode's lock list.
3544 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3546 case 0: /* success! */
3547 update_stateid(&lock_stp->st_stateid);
3548 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3552 case (EAGAIN): /* conflock holds conflicting lock */
3553 status = nfserr_denied;
3554 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3555 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3558 status = nfserr_deadlock;
3561 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3562 status = nfserr_resource;
3566 if (status && lock->lk_is_new && lock_sop)
3567 release_lockowner(lock_sop);
3568 if (lock->lk_replay_owner) {
3569 nfs4_get_stateowner(lock->lk_replay_owner);
3570 cstate->replay_owner = lock->lk_replay_owner;
3572 nfs4_unlock_state();
3577 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3578 * so we do a temporary open here just to get an open file to pass to
3579 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3582 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3587 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3590 err = vfs_test_lock(file, lock);
3599 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3600 struct nfsd4_lockt *lockt)
3602 struct inode *inode;
3603 struct file_lock file_lock;
3607 if (locks_in_grace())
3608 return nfserr_grace;
3610 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3611 return nfserr_inval;
3613 lockt->lt_stateowner = NULL;
3616 status = nfserr_stale_clientid;
3617 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3620 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3621 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3622 if (status == nfserr_symlink)
3623 status = nfserr_inval;
3627 inode = cstate->current_fh.fh_dentry->d_inode;
3628 locks_init_lock(&file_lock);
3629 switch (lockt->lt_type) {
3632 file_lock.fl_type = F_RDLCK;
3635 case NFS4_WRITEW_LT:
3636 file_lock.fl_type = F_WRLCK;
3639 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3640 status = nfserr_inval;
3644 lockt->lt_stateowner = find_lockstateowner_str(inode,
3645 &lockt->lt_clientid, &lockt->lt_owner);
3646 if (lockt->lt_stateowner)
3647 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3648 file_lock.fl_pid = current->tgid;
3649 file_lock.fl_flags = FL_POSIX;
3651 file_lock.fl_start = lockt->lt_offset;
3652 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3654 nfs4_transform_lock_offset(&file_lock);
3657 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3659 status = nfserrno(error);
3662 if (file_lock.fl_type != F_UNLCK) {
3663 status = nfserr_denied;
3664 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3667 nfs4_unlock_state();
3672 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3673 struct nfsd4_locku *locku)
3675 struct nfs4_stateid *stp;
3676 struct file *filp = NULL;
3677 struct file_lock file_lock;
3681 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3682 (long long) locku->lu_offset,
3683 (long long) locku->lu_length);
3685 if (check_lock_length(locku->lu_offset, locku->lu_length))
3686 return nfserr_inval;
3690 if ((status = nfs4_preprocess_seqid_op(cstate,
3694 &locku->lu_stateowner, &stp, NULL)))
3697 filp = stp->st_vfs_file;
3699 locks_init_lock(&file_lock);
3700 file_lock.fl_type = F_UNLCK;
3701 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3702 file_lock.fl_pid = current->tgid;
3703 file_lock.fl_file = filp;
3704 file_lock.fl_flags = FL_POSIX;
3705 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3706 file_lock.fl_start = locku->lu_offset;
3708 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3709 nfs4_transform_lock_offset(&file_lock);
3712 * Try to unlock the file in the VFS.
3714 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3716 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3720 * OK, unlock succeeded; the only thing left to do is update the stateid.
3722 update_stateid(&stp->st_stateid);
3723 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3726 if (locku->lu_stateowner) {
3727 nfs4_get_stateowner(locku->lu_stateowner);
3728 cstate->replay_owner = locku->lu_stateowner;
3730 nfs4_unlock_state();
3734 status = nfserrno(err);
3740 * 1: locks held by lockowner
3741 * 0: no locks held by lockowner
3744 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3746 struct file_lock **flpp;
3747 struct inode *inode = filp->f_path.dentry->d_inode;
3751 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3752 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3763 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3764 struct nfsd4_compound_state *cstate,
3765 struct nfsd4_release_lockowner *rlockowner)
3767 clientid_t *clid = &rlockowner->rl_clientid;
3768 struct nfs4_stateowner *sop;
3769 struct nfs4_stateid *stp;
3770 struct xdr_netobj *owner = &rlockowner->rl_owner;
3771 struct list_head matches;
3775 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3776 clid->cl_boot, clid->cl_id);
3778 /* XXX check for lease expiration */
3780 status = nfserr_stale_clientid;
3781 if (STALE_CLIENTID(clid))
3786 status = nfserr_locks_held;
3787 /* XXX: we're doing a linear search through all the lockowners.
3788 * Yipes! For now we'll just hope clients aren't really using
3789 * release_lockowner much, but eventually we have to fix these
3790 * data structures. */
3791 INIT_LIST_HEAD(&matches);
3792 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3793 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3794 if (!same_owner_str(sop, owner, clid))
3796 list_for_each_entry(stp, &sop->so_stateids,
3798 if (check_for_locks(stp->st_vfs_file, sop))
3800 /* Note: so_perclient unused for lockowners,
3801 * so it's OK to fool with here. */
3802 list_add(&sop->so_perclient, &matches);
3806 /* Clients probably won't expect us to return with some (but not all)
3807 * of the lockowner state released; so don't release any until all
3808 * have been checked. */
3810 while (!list_empty(&matches)) {
3811 sop = list_entry(matches.next, struct nfs4_stateowner,
3813 /* unhash_stateowner deletes so_perclient only
3814 * for openowners. */
3815 list_del(&sop->so_perclient);
3816 release_lockowner(sop);
3819 nfs4_unlock_state();
3823 static inline struct nfs4_client_reclaim *
3826 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3830 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3832 unsigned int strhashval = clientstr_hashval(name);
3833 struct nfs4_client *clp;
3835 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3840 * failure => all reset bets are off, nfserr_no_grace...
3843 nfs4_client_to_reclaim(const char *name)
3845 unsigned int strhashval;
3846 struct nfs4_client_reclaim *crp = NULL;
3848 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3849 crp = alloc_reclaim();
3852 strhashval = clientstr_hashval(name);
3853 INIT_LIST_HEAD(&crp->cr_strhash);
3854 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3855 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3856 reclaim_str_hashtbl_size++;
3861 nfs4_release_reclaim(void)
3863 struct nfs4_client_reclaim *crp = NULL;
3866 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3867 while (!list_empty(&reclaim_str_hashtbl[i])) {
3868 crp = list_entry(reclaim_str_hashtbl[i].next,
3869 struct nfs4_client_reclaim, cr_strhash);
3870 list_del(&crp->cr_strhash);
3872 reclaim_str_hashtbl_size--;
3875 BUG_ON(reclaim_str_hashtbl_size);
3879 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3880 static struct nfs4_client_reclaim *
3881 nfs4_find_reclaim_client(clientid_t *clid)
3883 unsigned int strhashval;
3884 struct nfs4_client *clp;
3885 struct nfs4_client_reclaim *crp = NULL;
3888 /* find clientid in conf_id_hashtbl */
3889 clp = find_confirmed_client(clid);
3893 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3894 clp->cl_name.len, clp->cl_name.data,
3897 /* find clp->cl_name in reclaim_str_hashtbl */
3898 strhashval = clientstr_hashval(clp->cl_recdir);
3899 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3900 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3908 * Called from OPEN. Look for clientid in reclaim list.
3911 nfs4_check_open_reclaim(clientid_t *clid)
3913 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3916 /* initialization to perform at module load time: */
3919 nfs4_state_init(void)
3923 status = nfsd4_init_slabs();
3926 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3927 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3928 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3929 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3930 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3931 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3933 for (i = 0; i < SESSION_HASH_SIZE; i++)
3934 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3935 for (i = 0; i < FILE_HASH_SIZE; i++) {
3936 INIT_LIST_HEAD(&file_hashtbl[i]);
3938 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3939 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3940 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3942 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3943 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3944 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3946 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3947 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3948 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3950 memset(&onestateid, ~0, sizeof(stateid_t));
3951 INIT_LIST_HEAD(&close_lru);
3952 INIT_LIST_HEAD(&client_lru);
3953 INIT_LIST_HEAD(&del_recall_lru);
3954 reclaim_str_hashtbl_size = 0;
3959 nfsd4_load_reboot_recovery_data(void)
3964 nfsd4_init_recdir(user_recovery_dirname);
3965 status = nfsd4_recdir_load();
3966 nfs4_unlock_state();
3968 printk("NFSD: Failure reading reboot recovery data\n");
3972 * Since the lifetime of a delegation isn't limited to that of an open, a
3973 * client may quite reasonably hang on to a delegation as long as it has
3974 * the inode cached. This becomes an obvious problem the first time a
3975 * client's inode cache approaches the size of the server's total memory.
3977 * For now we avoid this problem by imposing a hard limit on the number
3978 * of delegations, which varies according to the server's memory size.
3981 set_max_delegations(void)
3984 * Allow at most 4 delegations per megabyte of RAM. Quick
3985 * estimates suggest that in the worst case (where every delegation
3986 * is for a different inode), a delegation could take about 1.5K,
3987 * giving a worst case usage of about 6% of memory.
3989 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3992 /* initialization to perform when the nfsd service is started: */
3995 __nfs4_state_start(void)
3999 boot_time = get_seconds();
4000 locks_start_grace(&nfsd4_manager);
4001 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4003 ret = set_callback_cred();
4006 laundry_wq = create_singlethread_workqueue("nfsd4");
4007 if (laundry_wq == NULL)
4009 ret = nfsd4_create_callback_queue();
4011 goto out_free_laundry;
4012 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4013 set_max_delegations();
4016 destroy_workqueue(laundry_wq);
4021 nfs4_state_start(void)
4027 nfsd4_load_reboot_recovery_data();
4028 ret = __nfs4_state_start();
4036 __nfs4_state_shutdown(void)
4039 struct nfs4_client *clp = NULL;
4040 struct nfs4_delegation *dp = NULL;
4041 struct list_head *pos, *next, reaplist;
4043 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4044 while (!list_empty(&conf_id_hashtbl[i])) {
4045 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4048 while (!list_empty(&unconf_str_hashtbl[i])) {
4049 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4053 INIT_LIST_HEAD(&reaplist);
4054 spin_lock(&recall_lock);
4055 list_for_each_safe(pos, next, &del_recall_lru) {
4056 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4057 list_move(&dp->dl_recall_lru, &reaplist);
4059 spin_unlock(&recall_lock);
4060 list_for_each_safe(pos, next, &reaplist) {
4061 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4062 list_del_init(&dp->dl_recall_lru);
4063 unhash_delegation(dp);
4066 nfsd4_shutdown_recdir();
4071 nfs4_state_shutdown(void)
4073 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4074 destroy_workqueue(laundry_wq);
4075 locks_end_grace(&nfsd4_manager);
4077 nfs4_release_reclaim();
4078 __nfs4_state_shutdown();
4079 nfsd4_destroy_callback_queue();
4080 nfs4_unlock_state();
4084 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4085 * accessed when nfsd is starting.
4088 nfs4_set_recdir(char *recdir)
4090 strcpy(user_recovery_dirname, recdir);
4094 * Change the NFSv4 recovery directory to recdir.
4097 nfs4_reset_recoverydir(char *recdir)
4102 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4106 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4107 nfs4_set_recdir(recdir);
4115 nfs4_recoverydir(void)
4117 return user_recovery_dirname;