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 client lru list and session hash table */
254 static DEFINE_SPINLOCK(client_lock);
256 /* Hash tables for nfs4_clientid state */
257 #define CLIENT_HASH_BITS 4
258 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
261 #define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
263 #define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
275 * client_lru holds client queue ordered by nfs4_client.cl_time
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
281 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
282 static int reclaim_str_hashtbl_size = 0;
283 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
284 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
285 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
286 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
287 static struct list_head client_lru;
288 static struct list_head close_lru;
290 static void unhash_generic_stateid(struct nfs4_stateid *stp)
292 list_del(&stp->st_hash);
293 list_del(&stp->st_perfile);
294 list_del(&stp->st_perstateowner);
297 static void free_generic_stateid(struct nfs4_stateid *stp)
299 put_nfs4_file(stp->st_file);
300 kmem_cache_free(stateid_slab, stp);
303 static void release_lock_stateid(struct nfs4_stateid *stp)
305 unhash_generic_stateid(stp);
306 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
307 free_generic_stateid(stp);
310 static void unhash_lockowner(struct nfs4_stateowner *sop)
312 struct nfs4_stateid *stp;
314 list_del(&sop->so_idhash);
315 list_del(&sop->so_strhash);
316 list_del(&sop->so_perstateid);
317 while (!list_empty(&sop->so_stateids)) {
318 stp = list_first_entry(&sop->so_stateids,
319 struct nfs4_stateid, st_perstateowner);
320 release_lock_stateid(stp);
324 static void release_lockowner(struct nfs4_stateowner *sop)
326 unhash_lockowner(sop);
327 nfs4_put_stateowner(sop);
331 release_stateid_lockowners(struct nfs4_stateid *open_stp)
333 struct nfs4_stateowner *lock_sop;
335 while (!list_empty(&open_stp->st_lockowners)) {
336 lock_sop = list_entry(open_stp->st_lockowners.next,
337 struct nfs4_stateowner, so_perstateid);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop->so_is_open_owner);
340 release_lockowner(lock_sop);
344 static void release_open_stateid(struct nfs4_stateid *stp)
346 unhash_generic_stateid(stp);
347 release_stateid_lockowners(stp);
348 nfsd_close(stp->st_vfs_file);
349 free_generic_stateid(stp);
352 static void unhash_openowner(struct nfs4_stateowner *sop)
354 struct nfs4_stateid *stp;
356 list_del(&sop->so_idhash);
357 list_del(&sop->so_strhash);
358 list_del(&sop->so_perclient);
359 list_del(&sop->so_perstateid); /* XXX: necessary? */
360 while (!list_empty(&sop->so_stateids)) {
361 stp = list_first_entry(&sop->so_stateids,
362 struct nfs4_stateid, st_perstateowner);
363 release_open_stateid(stp);
367 static void release_openowner(struct nfs4_stateowner *sop)
369 unhash_openowner(sop);
370 list_del(&sop->so_close_lru);
371 nfs4_put_stateowner(sop);
374 #define SESSION_HASH_SIZE 512
375 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
378 hash_sessionid(struct nfs4_sessionid *sessionid)
380 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
382 return sid->sequence % SESSION_HASH_SIZE;
386 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
388 u32 *ptr = (u32 *)(&sessionid->data[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
393 gen_sessionid(struct nfsd4_session *ses)
395 struct nfs4_client *clp = ses->se_client;
396 struct nfsd4_sessionid *sid;
398 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
399 sid->clientid = clp->cl_clientid;
400 sid->sequence = current_sessionid++;
405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
416 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
428 static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
430 int mem, size = fchan->maxresp_cached;
432 if (fchan->maxreqs < 1)
435 if (size < NFSD_MIN_HDR_SEQ_SZ)
436 size = NFSD_MIN_HDR_SEQ_SZ;
437 size -= NFSD_MIN_HDR_SEQ_SZ;
438 if (size > NFSD_SLOT_CACHE_SIZE)
439 size = NFSD_SLOT_CACHE_SIZE;
441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem = fchan->maxreqs * size;
443 if (mem > NFSD_MAX_MEM_PER_SESSION) {
444 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
445 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
446 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
447 mem = fchan->maxreqs * size;
450 spin_lock(&nfsd_drc_lock);
451 /* bound the total session drc memory ussage */
452 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
453 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
454 mem = fchan->maxreqs * size;
456 nfsd_drc_mem_used += mem;
457 spin_unlock(&nfsd_drc_lock);
459 if (fchan->maxreqs == 0)
460 return nfserr_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);
631 /* must be called under the client_lock */
633 renew_client_locked(struct nfs4_client *clp)
635 if (is_client_expired(clp)) {
636 dprintk("%s: client (clientid %08x/%08x) already expired\n",
638 clp->cl_clientid.cl_boot,
639 clp->cl_clientid.cl_id);
644 * Move client to the end to the LRU list.
646 dprintk("renewing client (clientid %08x/%08x)\n",
647 clp->cl_clientid.cl_boot,
648 clp->cl_clientid.cl_id);
649 list_move_tail(&clp->cl_lru, &client_lru);
650 clp->cl_time = get_seconds();
654 renew_client(struct nfs4_client *clp)
656 spin_lock(&client_lock);
657 renew_client_locked(clp);
658 spin_unlock(&client_lock);
661 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
663 STALE_CLIENTID(clientid_t *clid)
665 if (clid->cl_boot == boot_time)
667 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
668 clid->cl_boot, clid->cl_id, boot_time);
673 * XXX Should we use a slab cache ?
674 * This type of memory management is somewhat inefficient, but we use it
675 * anyway since SETCLIENTID is not a common operation.
677 static struct nfs4_client *alloc_client(struct xdr_netobj name)
679 struct nfs4_client *clp;
681 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
684 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
685 if (clp->cl_name.data == NULL) {
689 memcpy(clp->cl_name.data, name.data, name.len);
690 clp->cl_name.len = name.len;
695 free_client(struct nfs4_client *clp)
697 if (clp->cl_cred.cr_group_info)
698 put_group_info(clp->cl_cred.cr_group_info);
699 kfree(clp->cl_principal);
700 kfree(clp->cl_name.data);
704 /* must be called under the client_lock */
706 unhash_client_locked(struct nfs4_client *clp)
708 mark_client_expired(clp);
709 list_del(&clp->cl_lru);
710 while (!list_empty(&clp->cl_sessions)) {
711 struct nfsd4_session *ses;
712 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
715 nfsd4_put_session(ses);
720 expire_client(struct nfs4_client *clp)
722 struct nfs4_stateowner *sop;
723 struct nfs4_delegation *dp;
724 struct list_head reaplist;
726 INIT_LIST_HEAD(&reaplist);
727 spin_lock(&recall_lock);
728 while (!list_empty(&clp->cl_delegations)) {
729 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
730 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
732 list_del_init(&dp->dl_perclnt);
733 list_move(&dp->dl_recall_lru, &reaplist);
735 spin_unlock(&recall_lock);
736 while (!list_empty(&reaplist)) {
737 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
738 list_del_init(&dp->dl_recall_lru);
739 unhash_delegation(dp);
741 while (!list_empty(&clp->cl_openowners)) {
742 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
743 release_openowner(sop);
745 nfsd4_set_callback_client(clp, NULL);
746 if (clp->cl_cb_conn.cb_xprt)
747 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
748 list_del(&clp->cl_idhash);
749 list_del(&clp->cl_strhash);
750 spin_lock(&client_lock);
751 unhash_client_locked(clp);
752 if (atomic_read(&clp->cl_refcount) == 0)
754 spin_unlock(&client_lock);
757 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
759 memcpy(target->cl_verifier.data, source->data,
760 sizeof(target->cl_verifier.data));
763 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
765 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
766 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
769 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
771 target->cr_uid = source->cr_uid;
772 target->cr_gid = source->cr_gid;
773 target->cr_group_info = source->cr_group_info;
774 get_group_info(target->cr_group_info);
777 static int same_name(const char *n1, const char *n2)
779 return 0 == memcmp(n1, n2, HEXDIR_LEN);
783 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
785 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
789 same_clid(clientid_t *cl1, clientid_t *cl2)
791 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
794 /* XXX what about NGROUP */
796 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
798 return cr1->cr_uid == cr2->cr_uid;
801 static void gen_clid(struct nfs4_client *clp)
803 static u32 current_clientid = 1;
805 clp->cl_clientid.cl_boot = boot_time;
806 clp->cl_clientid.cl_id = current_clientid++;
809 static void gen_confirm(struct nfs4_client *clp)
814 p = (u32 *)clp->cl_confirm.data;
815 *p++ = get_seconds();
819 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
820 struct svc_rqst *rqstp, nfs4_verifier *verf)
822 struct nfs4_client *clp;
823 struct sockaddr *sa = svc_addr(rqstp);
826 clp = alloc_client(name);
830 princ = svc_gss_principal(rqstp);
832 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
833 if (clp->cl_principal == NULL) {
839 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
840 atomic_set(&clp->cl_refcount, 0);
841 atomic_set(&clp->cl_cb_set, 0);
842 INIT_LIST_HEAD(&clp->cl_idhash);
843 INIT_LIST_HEAD(&clp->cl_strhash);
844 INIT_LIST_HEAD(&clp->cl_openowners);
845 INIT_LIST_HEAD(&clp->cl_delegations);
846 INIT_LIST_HEAD(&clp->cl_sessions);
847 INIT_LIST_HEAD(&clp->cl_lru);
848 clp->cl_time = get_seconds();
849 clear_bit(0, &clp->cl_cb_slot_busy);
850 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
851 copy_verf(clp, verf);
852 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
853 clp->cl_flavor = rqstp->rq_flavor;
854 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
860 static int check_name(struct xdr_netobj name)
864 if (name.len > NFS4_OPAQUE_LIMIT) {
865 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
872 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
874 unsigned int idhashval;
876 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
877 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
878 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
883 move_to_confirmed(struct nfs4_client *clp)
885 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
886 unsigned int strhashval;
888 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
889 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
890 strhashval = clientstr_hashval(clp->cl_recdir);
891 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
895 static struct nfs4_client *
896 find_confirmed_client(clientid_t *clid)
898 struct nfs4_client *clp;
899 unsigned int idhashval = clientid_hashval(clid->cl_id);
901 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
902 if (same_clid(&clp->cl_clientid, clid))
908 static struct nfs4_client *
909 find_unconfirmed_client(clientid_t *clid)
911 struct nfs4_client *clp;
912 unsigned int idhashval = clientid_hashval(clid->cl_id);
914 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
915 if (same_clid(&clp->cl_clientid, clid))
922 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
923 * parameter. Matching is based on the fact the at least one of the
924 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
926 * FIXME: we need to unify the clientid namespaces for nfsv4.x
927 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
928 * and SET_CLIENTID{,_CONFIRM}
931 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
933 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
934 return use_exchange_id == has_exchange_flags;
937 static struct nfs4_client *
938 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
939 bool use_exchange_id)
941 struct nfs4_client *clp;
943 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
944 if (same_name(clp->cl_recdir, dname) &&
945 match_clientid_establishment(clp, use_exchange_id))
951 static struct nfs4_client *
952 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
953 bool use_exchange_id)
955 struct nfs4_client *clp;
957 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
958 if (same_name(clp->cl_recdir, dname) &&
959 match_clientid_establishment(clp, use_exchange_id))
966 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
968 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
969 unsigned short expected_family;
971 /* Currently, we only support tcp and tcp6 for the callback channel */
972 if (se->se_callback_netid_len == 3 &&
973 !memcmp(se->se_callback_netid_val, "tcp", 3))
974 expected_family = AF_INET;
975 else if (se->se_callback_netid_len == 4 &&
976 !memcmp(se->se_callback_netid_val, "tcp6", 4))
977 expected_family = AF_INET6;
981 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
982 se->se_callback_addr_len,
983 (struct sockaddr *) &cb->cb_addr,
984 sizeof(cb->cb_addr));
986 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
989 if (cb->cb_addr.ss_family == AF_INET6)
990 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
992 cb->cb_minorversion = 0;
993 cb->cb_prog = se->se_callback_prog;
994 cb->cb_ident = se->se_callback_ident;
997 cb->cb_addr.ss_family = AF_UNSPEC;
999 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1000 "will not receive delegations\n",
1001 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1007 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1010 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1012 struct nfsd4_slot *slot = resp->cstate.slot;
1015 dprintk("--> %s slot %p\n", __func__, slot);
1017 slot->sl_opcnt = resp->opcnt;
1018 slot->sl_status = resp->cstate.status;
1020 if (nfsd4_not_cached(resp)) {
1021 slot->sl_datalen = 0;
1024 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1025 base = (char *)resp->cstate.datap -
1026 (char *)resp->xbuf->head[0].iov_base;
1027 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1029 WARN("%s: sessions DRC could not cache compound\n", __func__);
1034 * Encode the replay sequence operation from the slot values.
1035 * If cachethis is FALSE encode the uncached rep error on the next
1036 * operation which sets resp->p and increments resp->opcnt for
1037 * nfs4svc_encode_compoundres.
1041 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1042 struct nfsd4_compoundres *resp)
1044 struct nfsd4_op *op;
1045 struct nfsd4_slot *slot = resp->cstate.slot;
1047 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1048 resp->opcnt, resp->cstate.slot->sl_cachethis);
1050 /* Encode the replayed sequence operation */
1051 op = &args->ops[resp->opcnt - 1];
1052 nfsd4_encode_operation(resp, op);
1054 /* Return nfserr_retry_uncached_rep in next operation. */
1055 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1056 op = &args->ops[resp->opcnt++];
1057 op->status = nfserr_retry_uncached_rep;
1058 nfsd4_encode_operation(resp, op);
1064 * The sequence operation is not cached because we can use the slot and
1068 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1069 struct nfsd4_sequence *seq)
1071 struct nfsd4_slot *slot = resp->cstate.slot;
1074 dprintk("--> %s slot %p\n", __func__, slot);
1076 /* Either returns 0 or nfserr_retry_uncached */
1077 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1078 if (status == nfserr_retry_uncached_rep)
1081 /* The sequence operation has been encoded, cstate->datap set. */
1082 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1084 resp->opcnt = slot->sl_opcnt;
1085 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1086 status = slot->sl_status;
1092 * Set the exchange_id flags returned by the server.
1095 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1097 /* pNFS is not supported */
1098 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1100 /* Referrals are supported, Migration is not. */
1101 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1103 /* set the wire flags to return to client. */
1104 clid->flags = new->cl_exchange_flags;
1108 nfsd4_exchange_id(struct svc_rqst *rqstp,
1109 struct nfsd4_compound_state *cstate,
1110 struct nfsd4_exchange_id *exid)
1112 struct nfs4_client *unconf, *conf, *new;
1114 unsigned int strhashval;
1115 char dname[HEXDIR_LEN];
1116 char addr_str[INET6_ADDRSTRLEN];
1117 nfs4_verifier verf = exid->verifier;
1118 struct sockaddr *sa = svc_addr(rqstp);
1120 rpc_ntop(sa, addr_str, sizeof(addr_str));
1121 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1122 "ip_addr=%s flags %x, spa_how %d\n",
1123 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1124 addr_str, exid->flags, exid->spa_how);
1126 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1127 return nfserr_inval;
1129 /* Currently only support SP4_NONE */
1130 switch (exid->spa_how) {
1134 return nfserr_encr_alg_unsupp;
1136 BUG(); /* checked by xdr code */
1138 return nfserr_serverfault; /* no excuse :-/ */
1141 status = nfs4_make_rec_clidname(dname, &exid->clname);
1146 strhashval = clientstr_hashval(dname);
1151 conf = find_confirmed_client_by_str(dname, strhashval, true);
1153 if (!same_verf(&verf, &conf->cl_verifier)) {
1154 /* 18.35.4 case 8 */
1155 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1156 status = nfserr_not_same;
1159 /* Client reboot: destroy old state */
1160 expire_client(conf);
1163 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1164 /* 18.35.4 case 9 */
1165 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1166 status = nfserr_perm;
1169 expire_client(conf);
1173 * Set bit when the owner id and verifier map to an already
1174 * confirmed client id (18.35.3).
1176 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1179 * Falling into 18.35.4 case 2, possible router replay.
1180 * Leave confirmed record intact and return same result.
1182 copy_verf(conf, &verf);
1187 /* 18.35.4 case 7 */
1188 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1189 status = nfserr_noent;
1193 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1196 * Possible retry or client restart. Per 18.35.4 case 4,
1197 * a new unconfirmed record should be generated regardless
1198 * of whether any properties have changed.
1200 expire_client(unconf);
1205 new = create_client(exid->clname, dname, rqstp, &verf);
1207 status = nfserr_serverfault;
1212 add_to_unconfirmed(new, strhashval);
1214 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1215 exid->clientid.cl_id = new->cl_clientid.cl_id;
1218 nfsd4_set_ex_flags(new, exid);
1220 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1221 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1225 nfs4_unlock_state();
1227 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1232 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1234 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1237 /* The slot is in use, and no response has been sent. */
1239 if (seqid == slot_seqid)
1240 return nfserr_jukebox;
1242 return nfserr_seq_misordered;
1245 if (likely(seqid == slot_seqid + 1))
1248 if (seqid == slot_seqid)
1249 return nfserr_replay_cache;
1251 if (seqid == 1 && (slot_seqid + 1) == 0)
1253 /* Misordered replay or misordered new request */
1254 return nfserr_seq_misordered;
1258 * Cache the create session result into the create session single DRC
1259 * slot cache by saving the xdr structure. sl_seqid has been set.
1260 * Do this for solo or embedded create session operations.
1263 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1264 struct nfsd4_clid_slot *slot, int nfserr)
1266 slot->sl_status = nfserr;
1267 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1271 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1272 struct nfsd4_clid_slot *slot)
1274 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1275 return slot->sl_status;
1279 nfsd4_create_session(struct svc_rqst *rqstp,
1280 struct nfsd4_compound_state *cstate,
1281 struct nfsd4_create_session *cr_ses)
1283 struct sockaddr *sa = svc_addr(rqstp);
1284 struct nfs4_client *conf, *unconf;
1285 struct nfsd4_clid_slot *cs_slot = NULL;
1289 unconf = find_unconfirmed_client(&cr_ses->clientid);
1290 conf = find_confirmed_client(&cr_ses->clientid);
1293 cs_slot = &conf->cl_cs_slot;
1294 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1295 if (status == nfserr_replay_cache) {
1296 dprintk("Got a create_session replay! seqid= %d\n",
1298 /* Return the cached reply status */
1299 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1301 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1302 status = nfserr_seq_misordered;
1303 dprintk("Sequence misordered!\n");
1304 dprintk("Expected seqid= %d but got seqid= %d\n",
1305 cs_slot->sl_seqid, cr_ses->seqid);
1308 cs_slot->sl_seqid++;
1309 } else if (unconf) {
1310 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1311 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1312 status = nfserr_clid_inuse;
1316 cs_slot = &unconf->cl_cs_slot;
1317 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1319 /* an unconfirmed replay returns misordered */
1320 status = nfserr_seq_misordered;
1324 cs_slot->sl_seqid++; /* from 0 to 1 */
1325 move_to_confirmed(unconf);
1327 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1328 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1329 svc_xprt_get(rqstp->rq_xprt);
1331 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1333 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1334 unconf->cl_cb_conn.cb_minorversion =
1335 cstate->minorversion;
1336 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1337 unconf->cl_cb_seq_nr = 1;
1338 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1342 status = nfserr_stale_clientid;
1347 * We do not support RDMA or persistent sessions
1349 cr_ses->flags &= ~SESSION4_PERSIST;
1350 cr_ses->flags &= ~SESSION4_RDMA;
1352 status = alloc_init_session(rqstp, conf, cr_ses);
1356 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1357 NFS4_MAX_SESSIONID_LEN);
1358 cr_ses->seqid = cs_slot->sl_seqid;
1361 /* cache solo and embedded create sessions under the state lock */
1362 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1364 nfs4_unlock_state();
1365 dprintk("%s returns %d\n", __func__, ntohl(status));
1369 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1371 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1372 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1374 return argp->opcnt == resp->opcnt;
1377 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1381 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1385 nfsd4_destroy_session(struct svc_rqst *r,
1386 struct nfsd4_compound_state *cstate,
1387 struct nfsd4_destroy_session *sessionid)
1389 struct nfsd4_session *ses;
1390 u32 status = nfserr_badsession;
1393 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1394 * - Should we return nfserr_back_chan_busy if waiting for
1395 * callbacks on to-be-destroyed session?
1396 * - Do we need to clear any callback info from previous session?
1399 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1400 if (!nfsd4_last_compound_op(r))
1401 return nfserr_not_only_op;
1403 dump_sessionid(__func__, &sessionid->sessionid);
1404 spin_lock(&client_lock);
1405 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1407 spin_unlock(&client_lock);
1411 unhash_session(ses);
1412 spin_unlock(&client_lock);
1414 /* wait for callbacks */
1415 nfsd4_set_callback_client(ses->se_client, NULL);
1416 nfsd4_put_session(ses);
1419 dprintk("%s returns %d\n", __func__, ntohl(status));
1424 nfsd4_sequence(struct svc_rqst *rqstp,
1425 struct nfsd4_compound_state *cstate,
1426 struct nfsd4_sequence *seq)
1428 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1429 struct nfsd4_session *session;
1430 struct nfsd4_slot *slot;
1433 if (resp->opcnt != 1)
1434 return nfserr_sequence_pos;
1436 spin_lock(&client_lock);
1437 status = nfserr_badsession;
1438 session = find_in_sessionid_hashtbl(&seq->sessionid);
1442 status = nfserr_badslot;
1443 if (seq->slotid >= session->se_fchannel.maxreqs)
1446 slot = session->se_slots[seq->slotid];
1447 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1449 /* We do not negotiate the number of slots yet, so set the
1450 * maxslots to the session maxreqs which is used to encode
1451 * sr_highest_slotid and the sr_target_slot id to maxslots */
1452 seq->maxslots = session->se_fchannel.maxreqs;
1454 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1455 if (status == nfserr_replay_cache) {
1456 cstate->slot = slot;
1457 cstate->session = session;
1458 /* Return the cached reply status and set cstate->status
1459 * for nfsd4_proc_compound processing */
1460 status = nfsd4_replay_cache_entry(resp, seq);
1461 cstate->status = nfserr_replay_cache;
1467 /* Success! bump slot seqid */
1468 slot->sl_inuse = true;
1469 slot->sl_seqid = seq->seqid;
1470 slot->sl_cachethis = seq->cachethis;
1472 cstate->slot = slot;
1473 cstate->session = session;
1476 /* Hold a session reference until done processing the compound. */
1477 if (cstate->session) {
1478 nfsd4_get_session(cstate->session);
1479 /* Renew the clientid on success and on replay */
1480 renew_client_locked(session->se_client);
1482 spin_unlock(&client_lock);
1483 dprintk("%s: return %d\n", __func__, ntohl(status));
1488 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1489 struct nfsd4_setclientid *setclid)
1491 struct sockaddr *sa = svc_addr(rqstp);
1492 struct xdr_netobj clname = {
1493 .len = setclid->se_namelen,
1494 .data = setclid->se_name,
1496 nfs4_verifier clverifier = setclid->se_verf;
1497 unsigned int strhashval;
1498 struct nfs4_client *conf, *unconf, *new;
1500 char dname[HEXDIR_LEN];
1502 if (!check_name(clname))
1503 return nfserr_inval;
1505 status = nfs4_make_rec_clidname(dname, &clname);
1510 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1511 * We get here on a DRC miss.
1514 strhashval = clientstr_hashval(dname);
1517 conf = find_confirmed_client_by_str(dname, strhashval, false);
1519 /* RFC 3530 14.2.33 CASE 0: */
1520 status = nfserr_clid_inuse;
1521 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1522 char addr_str[INET6_ADDRSTRLEN];
1523 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1525 dprintk("NFSD: setclientid: string in use by client "
1526 "at %s\n", addr_str);
1531 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1532 * has a description of SETCLIENTID request processing consisting
1533 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1535 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1536 status = nfserr_resource;
1539 * RFC 3530 14.2.33 CASE 4:
1540 * placed first, because it is the normal case
1543 expire_client(unconf);
1544 new = create_client(clname, dname, rqstp, &clverifier);
1548 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1550 * RFC 3530 14.2.33 CASE 1:
1551 * probable callback update
1554 /* Note this is removing unconfirmed {*x***},
1555 * which is stronger than RFC recommended {vxc**}.
1556 * This has the advantage that there is at most
1557 * one {*x***} in either list at any time.
1559 expire_client(unconf);
1561 new = create_client(clname, dname, rqstp, &clverifier);
1564 copy_clid(new, conf);
1565 } else if (!unconf) {
1567 * RFC 3530 14.2.33 CASE 2:
1568 * probable client reboot; state will be removed if
1571 new = create_client(clname, dname, rqstp, &clverifier);
1577 * RFC 3530 14.2.33 CASE 3:
1578 * probable client reboot; state will be removed if
1581 expire_client(unconf);
1582 new = create_client(clname, dname, rqstp, &clverifier);
1587 gen_callback(new, setclid, rpc_get_scope_id(sa));
1588 add_to_unconfirmed(new, strhashval);
1589 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1590 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1591 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1594 nfs4_unlock_state();
1600 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1601 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1602 * bullets, labeled as CASE1 - CASE4 below.
1605 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1606 struct nfsd4_compound_state *cstate,
1607 struct nfsd4_setclientid_confirm *setclientid_confirm)
1609 struct sockaddr *sa = svc_addr(rqstp);
1610 struct nfs4_client *conf, *unconf;
1611 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1612 clientid_t * clid = &setclientid_confirm->sc_clientid;
1615 if (STALE_CLIENTID(clid))
1616 return nfserr_stale_clientid;
1618 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1619 * We get here on a DRC miss.
1624 conf = find_confirmed_client(clid);
1625 unconf = find_unconfirmed_client(clid);
1627 status = nfserr_clid_inuse;
1628 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1630 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1634 * section 14.2.34 of RFC 3530 has a description of
1635 * SETCLIENTID_CONFIRM request processing consisting
1636 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1638 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1640 * RFC 3530 14.2.34 CASE 1:
1643 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1644 status = nfserr_clid_inuse;
1646 atomic_set(&conf->cl_cb_set, 0);
1647 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1648 expire_client(unconf);
1652 } else if (conf && !unconf) {
1654 * RFC 3530 14.2.34 CASE 2:
1655 * probable retransmitted request; play it safe and
1658 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1659 status = nfserr_clid_inuse;
1662 } else if (!conf && unconf
1663 && same_verf(&unconf->cl_confirm, &confirm)) {
1665 * RFC 3530 14.2.34 CASE 3:
1666 * Normal case; new or rebooted client:
1668 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1669 status = nfserr_clid_inuse;
1672 clientstr_hashval(unconf->cl_recdir);
1673 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1676 nfsd4_remove_clid_dir(conf);
1677 expire_client(conf);
1679 move_to_confirmed(unconf);
1681 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1684 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1685 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1688 * RFC 3530 14.2.34 CASE 4:
1689 * Client probably hasn't noticed that we rebooted yet.
1691 status = nfserr_stale_clientid;
1693 /* check that we have hit one of the cases...*/
1694 status = nfserr_clid_inuse;
1697 nfs4_unlock_state();
1701 /* OPEN Share state helper functions */
1702 static inline struct nfs4_file *
1703 alloc_init_file(struct inode *ino)
1705 struct nfs4_file *fp;
1706 unsigned int hashval = file_hashval(ino);
1708 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1710 atomic_set(&fp->fi_ref, 1);
1711 INIT_LIST_HEAD(&fp->fi_hash);
1712 INIT_LIST_HEAD(&fp->fi_stateids);
1713 INIT_LIST_HEAD(&fp->fi_delegations);
1714 spin_lock(&recall_lock);
1715 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1716 spin_unlock(&recall_lock);
1717 fp->fi_inode = igrab(ino);
1718 fp->fi_id = current_fileid++;
1719 fp->fi_had_conflict = false;
1726 nfsd4_free_slab(struct kmem_cache **slab)
1730 kmem_cache_destroy(*slab);
1735 nfsd4_free_slabs(void)
1737 nfsd4_free_slab(&stateowner_slab);
1738 nfsd4_free_slab(&file_slab);
1739 nfsd4_free_slab(&stateid_slab);
1740 nfsd4_free_slab(&deleg_slab);
1744 nfsd4_init_slabs(void)
1746 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1747 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1748 if (stateowner_slab == NULL)
1750 file_slab = kmem_cache_create("nfsd4_files",
1751 sizeof(struct nfs4_file), 0, 0, NULL);
1752 if (file_slab == NULL)
1754 stateid_slab = kmem_cache_create("nfsd4_stateids",
1755 sizeof(struct nfs4_stateid), 0, 0, NULL);
1756 if (stateid_slab == NULL)
1758 deleg_slab = kmem_cache_create("nfsd4_delegations",
1759 sizeof(struct nfs4_delegation), 0, 0, NULL);
1760 if (deleg_slab == NULL)
1765 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1770 nfs4_free_stateowner(struct kref *kref)
1772 struct nfs4_stateowner *sop =
1773 container_of(kref, struct nfs4_stateowner, so_ref);
1774 kfree(sop->so_owner.data);
1775 kmem_cache_free(stateowner_slab, sop);
1778 static inline struct nfs4_stateowner *
1779 alloc_stateowner(struct xdr_netobj *owner)
1781 struct nfs4_stateowner *sop;
1783 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1784 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1785 memcpy(sop->so_owner.data, owner->data, owner->len);
1786 sop->so_owner.len = owner->len;
1787 kref_init(&sop->so_ref);
1790 kmem_cache_free(stateowner_slab, sop);
1795 static struct nfs4_stateowner *
1796 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1797 struct nfs4_stateowner *sop;
1798 struct nfs4_replay *rp;
1799 unsigned int idhashval;
1801 if (!(sop = alloc_stateowner(&open->op_owner)))
1803 idhashval = ownerid_hashval(current_ownerid);
1804 INIT_LIST_HEAD(&sop->so_idhash);
1805 INIT_LIST_HEAD(&sop->so_strhash);
1806 INIT_LIST_HEAD(&sop->so_perclient);
1807 INIT_LIST_HEAD(&sop->so_stateids);
1808 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1809 INIT_LIST_HEAD(&sop->so_close_lru);
1811 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1812 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1813 list_add(&sop->so_perclient, &clp->cl_openowners);
1814 sop->so_is_open_owner = 1;
1815 sop->so_id = current_ownerid++;
1816 sop->so_client = clp;
1817 sop->so_seqid = open->op_seqid;
1818 sop->so_confirmed = 0;
1819 rp = &sop->so_replay;
1820 rp->rp_status = nfserr_serverfault;
1822 rp->rp_buf = rp->rp_ibuf;
1827 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1828 struct nfs4_stateowner *sop = open->op_stateowner;
1829 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1831 INIT_LIST_HEAD(&stp->st_hash);
1832 INIT_LIST_HEAD(&stp->st_perstateowner);
1833 INIT_LIST_HEAD(&stp->st_lockowners);
1834 INIT_LIST_HEAD(&stp->st_perfile);
1835 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1836 list_add(&stp->st_perstateowner, &sop->so_stateids);
1837 list_add(&stp->st_perfile, &fp->fi_stateids);
1838 stp->st_stateowner = sop;
1841 stp->st_stateid.si_boot = get_seconds();
1842 stp->st_stateid.si_stateownerid = sop->so_id;
1843 stp->st_stateid.si_fileid = fp->fi_id;
1844 stp->st_stateid.si_generation = 0;
1845 stp->st_access_bmap = 0;
1846 stp->st_deny_bmap = 0;
1847 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1848 &stp->st_access_bmap);
1849 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1850 stp->st_openstp = NULL;
1854 move_to_close_lru(struct nfs4_stateowner *sop)
1856 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1858 list_move_tail(&sop->so_close_lru, &close_lru);
1859 sop->so_time = get_seconds();
1863 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1866 return (sop->so_owner.len == owner->len) &&
1867 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1868 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1871 static struct nfs4_stateowner *
1872 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1874 struct nfs4_stateowner *so = NULL;
1876 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1877 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1883 /* search file_hashtbl[] for file */
1884 static struct nfs4_file *
1885 find_file(struct inode *ino)
1887 unsigned int hashval = file_hashval(ino);
1888 struct nfs4_file *fp;
1890 spin_lock(&recall_lock);
1891 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1892 if (fp->fi_inode == ino) {
1894 spin_unlock(&recall_lock);
1898 spin_unlock(&recall_lock);
1902 static inline int access_valid(u32 x, u32 minorversion)
1904 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1906 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1908 x &= ~NFS4_SHARE_ACCESS_MASK;
1909 if (minorversion && x) {
1910 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1912 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1914 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1921 static inline int deny_valid(u32 x)
1923 /* Note: unlike access bits, deny bits may be zero. */
1924 return x <= NFS4_SHARE_DENY_BOTH;
1928 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1929 * st_{access,deny}_bmap field of the stateid, in order to track not
1930 * only what share bits are currently in force, but also what
1931 * combinations of share bits previous opens have used. This allows us
1932 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1933 * return an error if the client attempt to downgrade to a combination
1934 * of share bits not explicable by closing some of its previous opens.
1936 * XXX: This enforcement is actually incomplete, since we don't keep
1937 * track of access/deny bit combinations; so, e.g., we allow:
1939 * OPEN allow read, deny write
1940 * OPEN allow both, deny none
1941 * DOWNGRADE allow read, deny none
1943 * which we should reject.
1946 set_access(unsigned int *access, unsigned long bmap) {
1950 for (i = 1; i < 4; i++) {
1951 if (test_bit(i, &bmap))
1957 set_deny(unsigned int *deny, unsigned long bmap) {
1961 for (i = 0; i < 4; i++) {
1962 if (test_bit(i, &bmap))
1968 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1969 unsigned int access, deny;
1971 set_access(&access, stp->st_access_bmap);
1972 set_deny(&deny, stp->st_deny_bmap);
1973 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1979 * Called to check deny when READ with all zero stateid or
1980 * WRITE with all zero or all one stateid
1983 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1985 struct inode *ino = current_fh->fh_dentry->d_inode;
1986 struct nfs4_file *fp;
1987 struct nfs4_stateid *stp;
1990 dprintk("NFSD: nfs4_share_conflict\n");
1992 fp = find_file(ino);
1995 ret = nfserr_locked;
1996 /* Search for conflicting share reservations */
1997 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1998 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1999 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2009 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2011 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
2012 drop_file_write_access(filp);
2013 spin_lock(&filp->f_lock);
2014 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2015 spin_unlock(&filp->f_lock);
2020 * Spawn a thread to perform a recall on the delegation represented
2021 * by the lease (file_lock)
2023 * Called from break_lease() with lock_kernel() held.
2024 * Note: we assume break_lease will only call this *once* for any given
2028 void nfsd_break_deleg_cb(struct file_lock *fl)
2030 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2032 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2036 /* We're assuming the state code never drops its reference
2037 * without first removing the lease. Since we're in this lease
2038 * callback (and since the lease code is serialized by the kernel
2039 * lock) we know the server hasn't removed the lease yet, we know
2040 * it's safe to take a reference: */
2041 atomic_inc(&dp->dl_count);
2043 spin_lock(&recall_lock);
2044 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2045 spin_unlock(&recall_lock);
2047 /* only place dl_time is set. protected by lock_kernel*/
2048 dp->dl_time = get_seconds();
2051 * We don't want the locks code to timeout the lease for us;
2052 * we'll remove it ourself if the delegation isn't returned
2055 fl->fl_break_time = 0;
2057 dp->dl_file->fi_had_conflict = true;
2058 nfsd4_cb_recall(dp);
2062 * The file_lock is being reapd.
2064 * Called by locks_free_lock() with lock_kernel() held.
2067 void nfsd_release_deleg_cb(struct file_lock *fl)
2069 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2071 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2073 if (!(fl->fl_flags & FL_LEASE) || !dp)
2075 dp->dl_flock = NULL;
2079 * Set the delegation file_lock back pointer.
2081 * Called from setlease() with lock_kernel() held.
2084 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2086 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2088 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2095 * Called from setlease() with lock_kernel() held
2098 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2100 struct nfs4_delegation *onlistd =
2101 (struct nfs4_delegation *)onlist->fl_owner;
2102 struct nfs4_delegation *tryd =
2103 (struct nfs4_delegation *)try->fl_owner;
2105 if (onlist->fl_lmops != try->fl_lmops)
2108 return onlistd->dl_client == tryd->dl_client;
2113 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2116 return lease_modify(onlist, arg);
2121 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2122 .fl_break = nfsd_break_deleg_cb,
2123 .fl_release_private = nfsd_release_deleg_cb,
2124 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2125 .fl_mylease = nfsd_same_client_deleg_cb,
2126 .fl_change = nfsd_change_deleg_cb,
2131 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2132 struct nfsd4_open *open)
2134 clientid_t *clientid = &open->op_clientid;
2135 struct nfs4_client *clp = NULL;
2136 unsigned int strhashval;
2137 struct nfs4_stateowner *sop = NULL;
2139 if (!check_name(open->op_owner))
2140 return nfserr_inval;
2142 if (STALE_CLIENTID(&open->op_clientid))
2143 return nfserr_stale_clientid;
2145 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2146 sop = find_openstateowner_str(strhashval, open);
2147 open->op_stateowner = sop;
2149 /* Make sure the client's lease hasn't expired. */
2150 clp = find_confirmed_client(clientid);
2152 return nfserr_expired;
2155 /* When sessions are used, skip open sequenceid processing */
2156 if (nfsd4_has_session(cstate))
2158 if (!sop->so_confirmed) {
2159 /* Replace unconfirmed owners without checking for replay. */
2160 clp = sop->so_client;
2161 release_openowner(sop);
2162 open->op_stateowner = NULL;
2165 if (open->op_seqid == sop->so_seqid - 1) {
2166 if (sop->so_replay.rp_buflen)
2167 return nfserr_replay_me;
2168 /* The original OPEN failed so spectacularly
2169 * that we don't even have replay data saved!
2170 * Therefore, we have no choice but to continue
2171 * processing this OPEN; presumably, we'll
2172 * fail again for the same reason.
2174 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2177 if (open->op_seqid != sop->so_seqid)
2178 return nfserr_bad_seqid;
2180 if (open->op_stateowner == NULL) {
2181 sop = alloc_init_open_stateowner(strhashval, clp, open);
2183 return nfserr_resource;
2184 open->op_stateowner = sop;
2186 list_del_init(&sop->so_close_lru);
2187 renew_client(sop->so_client);
2191 static inline __be32
2192 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2194 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2195 return nfserr_openmode;
2200 static struct nfs4_delegation *
2201 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2203 struct nfs4_delegation *dp;
2205 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2206 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2213 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2214 struct nfs4_delegation **dp)
2217 __be32 status = nfserr_bad_stateid;
2219 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2222 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2223 RD_STATE : WR_STATE;
2224 status = nfs4_check_delegmode(*dp, flags);
2228 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2232 open->op_stateowner->so_confirmed = 1;
2237 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2239 struct nfs4_stateid *local;
2240 __be32 status = nfserr_share_denied;
2241 struct nfs4_stateowner *sop = open->op_stateowner;
2243 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2244 /* ignore lock owners */
2245 if (local->st_stateowner->so_is_open_owner == 0)
2247 /* remember if we have seen this open owner */
2248 if (local->st_stateowner == sop)
2250 /* check for conflicting share reservations */
2251 if (!test_share(local, open))
2259 static inline struct nfs4_stateid *
2260 nfs4_alloc_stateid(void)
2262 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2266 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2267 struct nfs4_delegation *dp,
2268 struct svc_fh *cur_fh, int flags)
2270 struct nfs4_stateid *stp;
2272 stp = nfs4_alloc_stateid();
2274 return nfserr_resource;
2277 get_file(dp->dl_vfs_file);
2278 stp->st_vfs_file = dp->dl_vfs_file;
2281 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2284 if (status == nfserr_dropit)
2285 status = nfserr_jukebox;
2286 kmem_cache_free(stateid_slab, stp);
2294 static inline __be32
2295 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2296 struct nfsd4_open *open)
2298 struct iattr iattr = {
2299 .ia_valid = ATTR_SIZE,
2302 if (!open->op_truncate)
2304 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2305 return nfserr_inval;
2306 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2310 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2312 struct file *filp = stp->st_vfs_file;
2313 struct inode *inode = filp->f_path.dentry->d_inode;
2314 unsigned int share_access, new_writer;
2317 set_access(&share_access, stp->st_access_bmap);
2318 new_writer = (~share_access) & open->op_share_access
2319 & NFS4_SHARE_ACCESS_WRITE;
2322 int err = get_write_access(inode);
2324 return nfserrno(err);
2325 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2327 return nfserrno(err);
2328 file_take_write(filp);
2330 status = nfsd4_truncate(rqstp, cur_fh, open);
2333 put_write_access(inode);
2336 /* remember the open */
2337 filp->f_mode |= open->op_share_access;
2338 __set_bit(open->op_share_access, &stp->st_access_bmap);
2339 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2346 nfs4_set_claim_prev(struct nfsd4_open *open)
2348 open->op_stateowner->so_confirmed = 1;
2349 open->op_stateowner->so_client->cl_firststate = 1;
2353 * Attempt to hand out a delegation.
2356 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2358 struct nfs4_delegation *dp;
2359 struct nfs4_stateowner *sop = stp->st_stateowner;
2360 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2361 struct file_lock fl, *flp = &fl;
2362 int status, flag = 0;
2364 flag = NFS4_OPEN_DELEGATE_NONE;
2365 open->op_recall = 0;
2366 switch (open->op_claim_type) {
2367 case NFS4_OPEN_CLAIM_PREVIOUS:
2369 open->op_recall = 1;
2370 flag = open->op_delegate_type;
2371 if (flag == NFS4_OPEN_DELEGATE_NONE)
2374 case NFS4_OPEN_CLAIM_NULL:
2375 /* Let's not give out any delegations till everyone's
2376 * had the chance to reclaim theirs.... */
2377 if (locks_in_grace())
2379 if (!cb_up || !sop->so_confirmed)
2381 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2382 flag = NFS4_OPEN_DELEGATE_WRITE;
2384 flag = NFS4_OPEN_DELEGATE_READ;
2390 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2392 flag = NFS4_OPEN_DELEGATE_NONE;
2395 locks_init_lock(&fl);
2396 fl.fl_lmops = &nfsd_lease_mng_ops;
2397 fl.fl_flags = FL_LEASE;
2398 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2399 fl.fl_end = OFFSET_MAX;
2400 fl.fl_owner = (fl_owner_t)dp;
2401 fl.fl_file = stp->st_vfs_file;
2402 fl.fl_pid = current->tgid;
2404 /* vfs_setlease checks to see if delegation should be handed out.
2405 * the lock_manager callbacks fl_mylease and fl_change are used
2407 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2408 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2409 unhash_delegation(dp);
2410 flag = NFS4_OPEN_DELEGATE_NONE;
2414 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2416 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2417 STATEID_VAL(&dp->dl_stateid));
2419 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2420 && flag == NFS4_OPEN_DELEGATE_NONE
2421 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2422 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2423 open->op_delegate_type = flag;
2427 * called with nfs4_lock_state() held.
2430 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2432 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2433 struct nfs4_file *fp = NULL;
2434 struct inode *ino = current_fh->fh_dentry->d_inode;
2435 struct nfs4_stateid *stp = NULL;
2436 struct nfs4_delegation *dp = NULL;
2439 status = nfserr_inval;
2440 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2441 || !deny_valid(open->op_share_deny))
2444 * Lookup file; if found, lookup stateid and check open request,
2445 * and check for delegations in the process of being recalled.
2446 * If not found, create the nfs4_file struct
2448 fp = find_file(ino);
2450 if ((status = nfs4_check_open(fp, open, &stp)))
2452 status = nfs4_check_deleg(fp, open, &dp);
2456 status = nfserr_bad_stateid;
2457 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2459 status = nfserr_resource;
2460 fp = alloc_init_file(ino);
2466 * OPEN the file, or upgrade an existing OPEN.
2467 * If truncate fails, the OPEN fails.
2470 /* Stateid was found, this is an OPEN upgrade */
2471 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2474 update_stateid(&stp->st_stateid);
2476 /* Stateid was not found, this is a new OPEN */
2478 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2479 flags |= NFSD_MAY_READ;
2480 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2481 flags |= NFSD_MAY_WRITE;
2482 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2485 init_stateid(stp, fp, open);
2486 status = nfsd4_truncate(rqstp, current_fh, open);
2488 release_open_stateid(stp);
2491 if (nfsd4_has_session(&resp->cstate))
2492 update_stateid(&stp->st_stateid);
2494 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2496 if (nfsd4_has_session(&resp->cstate)) {
2497 open->op_stateowner->so_confirmed = 1;
2498 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2502 * Attempt to hand out a delegation. No error return, because the
2503 * OPEN succeeds even if we fail.
2505 nfs4_open_delegation(current_fh, open, stp);
2509 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2510 STATEID_VAL(&stp->st_stateid));
2514 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2515 nfs4_set_claim_prev(open);
2517 * To finish the open response, we just need to set the rflags.
2519 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2520 if (!open->op_stateowner->so_confirmed &&
2521 !nfsd4_has_session(&resp->cstate))
2522 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2528 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2531 struct nfs4_client *clp;
2535 dprintk("process_renew(%08x/%08x): starting\n",
2536 clid->cl_boot, clid->cl_id);
2537 status = nfserr_stale_clientid;
2538 if (STALE_CLIENTID(clid))
2540 clp = find_confirmed_client(clid);
2541 status = nfserr_expired;
2543 /* We assume the client took too long to RENEW. */
2544 dprintk("nfsd4_renew: clientid not found!\n");
2548 status = nfserr_cb_path_down;
2549 if (!list_empty(&clp->cl_delegations)
2550 && !atomic_read(&clp->cl_cb_set))
2554 nfs4_unlock_state();
2558 struct lock_manager nfsd4_manager = {
2562 nfsd4_end_grace(void)
2564 dprintk("NFSD: end of grace period\n");
2565 nfsd4_recdir_purge_old();
2566 locks_end_grace(&nfsd4_manager);
2568 * Now that every NFSv4 client has had the chance to recover and
2569 * to see the (possibly new, possibly shorter) lease time, we
2570 * can safely set the next grace time to the current lease time:
2572 nfsd4_grace = nfsd4_lease;
2576 nfs4_laundromat(void)
2578 struct nfs4_client *clp;
2579 struct nfs4_stateowner *sop;
2580 struct nfs4_delegation *dp;
2581 struct list_head *pos, *next, reaplist;
2582 time_t cutoff = get_seconds() - nfsd4_lease;
2583 time_t t, clientid_val = nfsd4_lease;
2584 time_t u, test_val = nfsd4_lease;
2588 dprintk("NFSD: laundromat service - starting\n");
2589 if (locks_in_grace())
2591 INIT_LIST_HEAD(&reaplist);
2592 spin_lock(&client_lock);
2593 list_for_each_safe(pos, next, &client_lru) {
2594 clp = list_entry(pos, struct nfs4_client, cl_lru);
2595 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2596 t = clp->cl_time - cutoff;
2597 if (clientid_val > t)
2601 list_move(&clp->cl_lru, &reaplist);
2603 spin_unlock(&client_lock);
2604 list_for_each_safe(pos, next, &reaplist) {
2605 clp = list_entry(pos, struct nfs4_client, cl_lru);
2606 dprintk("NFSD: purging unused client (clientid %08x)\n",
2607 clp->cl_clientid.cl_id);
2608 nfsd4_remove_clid_dir(clp);
2611 spin_lock(&recall_lock);
2612 list_for_each_safe(pos, next, &del_recall_lru) {
2613 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2614 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2615 u = dp->dl_time - cutoff;
2620 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2622 list_move(&dp->dl_recall_lru, &reaplist);
2624 spin_unlock(&recall_lock);
2625 list_for_each_safe(pos, next, &reaplist) {
2626 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2627 list_del_init(&dp->dl_recall_lru);
2628 unhash_delegation(dp);
2630 test_val = nfsd4_lease;
2631 list_for_each_safe(pos, next, &close_lru) {
2632 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2633 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2634 u = sop->so_time - cutoff;
2639 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2641 release_openowner(sop);
2643 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2644 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2645 nfs4_unlock_state();
2646 return clientid_val;
2649 static struct workqueue_struct *laundry_wq;
2650 static void laundromat_main(struct work_struct *);
2651 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2654 laundromat_main(struct work_struct *not_used)
2658 t = nfs4_laundromat();
2659 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2660 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2663 static struct nfs4_stateowner *
2664 search_close_lru(u32 st_id, int flags)
2666 struct nfs4_stateowner *local = NULL;
2668 if (flags & CLOSE_STATE) {
2669 list_for_each_entry(local, &close_lru, so_close_lru) {
2670 if (local->so_id == st_id)
2678 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2680 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2684 STALE_STATEID(stateid_t *stateid)
2686 if (time_after((unsigned long)boot_time,
2687 (unsigned long)stateid->si_boot)) {
2688 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2689 STATEID_VAL(stateid));
2696 EXPIRED_STATEID(stateid_t *stateid)
2698 if (time_before((unsigned long)boot_time,
2699 ((unsigned long)stateid->si_boot)) &&
2700 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
2701 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2702 STATEID_VAL(stateid));
2709 stateid_error_map(stateid_t *stateid)
2711 if (STALE_STATEID(stateid))
2712 return nfserr_stale_stateid;
2713 if (EXPIRED_STATEID(stateid))
2714 return nfserr_expired;
2716 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2717 STATEID_VAL(stateid));
2718 return nfserr_bad_stateid;
2722 access_permit_read(unsigned long access_bmap)
2724 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2725 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2726 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2730 access_permit_write(unsigned long access_bmap)
2732 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2733 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2737 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2739 __be32 status = nfserr_openmode;
2741 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2743 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2750 static inline __be32
2751 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2753 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2755 else if (locks_in_grace()) {
2756 /* Answer in remaining cases depends on existance of
2757 * conflicting state; so we must wait out the grace period. */
2758 return nfserr_grace;
2759 } else if (flags & WR_STATE)
2760 return nfs4_share_conflict(current_fh,
2761 NFS4_SHARE_DENY_WRITE);
2762 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2763 return nfs4_share_conflict(current_fh,
2764 NFS4_SHARE_DENY_READ);
2768 * Allow READ/WRITE during grace period on recovered state only for files
2769 * that are not able to provide mandatory locking.
2772 grace_disallows_io(struct inode *inode)
2774 return locks_in_grace() && mandatory_lock(inode);
2777 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2780 * When sessions are used the stateid generation number is ignored
2783 if ((flags & HAS_SESSION) && in->si_generation == 0)
2786 /* If the client sends us a stateid from the future, it's buggy: */
2787 if (in->si_generation > ref->si_generation)
2788 return nfserr_bad_stateid;
2790 * The following, however, can happen. For example, if the
2791 * client sends an open and some IO at the same time, the open
2792 * may bump si_generation while the IO is still in flight.
2793 * Thanks to hard links and renames, the client never knows what
2794 * file an open will affect. So it could avoid that situation
2795 * only by serializing all opens and IO from the same open
2796 * owner. To recover from the old_stateid error, the client
2797 * will just have to retry the IO:
2799 if (in->si_generation < ref->si_generation)
2800 return nfserr_old_stateid;
2805 static int is_delegation_stateid(stateid_t *stateid)
2807 return stateid->si_fileid == 0;
2811 * Checks for stateid operations
2814 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2815 stateid_t *stateid, int flags, struct file **filpp)
2817 struct nfs4_stateid *stp = NULL;
2818 struct nfs4_delegation *dp = NULL;
2819 struct svc_fh *current_fh = &cstate->current_fh;
2820 struct inode *ino = current_fh->fh_dentry->d_inode;
2826 if (grace_disallows_io(ino))
2827 return nfserr_grace;
2829 if (nfsd4_has_session(cstate))
2830 flags |= HAS_SESSION;
2832 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2833 return check_special_stateids(current_fh, stateid, flags);
2835 status = nfserr_stale_stateid;
2836 if (STALE_STATEID(stateid))
2839 status = nfserr_bad_stateid;
2840 if (is_delegation_stateid(stateid)) {
2841 dp = find_delegation_stateid(ino, stateid);
2843 status = stateid_error_map(stateid);
2846 status = check_stateid_generation(stateid, &dp->dl_stateid,
2850 status = nfs4_check_delegmode(dp, flags);
2853 renew_client(dp->dl_client);
2855 *filpp = dp->dl_vfs_file;
2856 } else { /* open or lock stateid */
2857 stp = find_stateid(stateid, flags);
2859 status = stateid_error_map(stateid);
2862 if (nfs4_check_fh(current_fh, stp))
2864 if (!stp->st_stateowner->so_confirmed)
2866 status = check_stateid_generation(stateid, &stp->st_stateid,
2870 status = nfs4_check_openmode(stp, flags);
2873 renew_client(stp->st_stateowner->so_client);
2875 *filpp = stp->st_vfs_file;
2885 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2886 RD_STATE : WR_STATE;
2890 * Checks for sequence id mutating operations.
2893 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2894 stateid_t *stateid, int flags,
2895 struct nfs4_stateowner **sopp,
2896 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2898 struct nfs4_stateid *stp;
2899 struct nfs4_stateowner *sop;
2900 struct svc_fh *current_fh = &cstate->current_fh;
2903 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2904 seqid, STATEID_VAL(stateid));
2909 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2910 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2911 return nfserr_bad_stateid;
2914 if (STALE_STATEID(stateid))
2915 return nfserr_stale_stateid;
2917 if (nfsd4_has_session(cstate))
2918 flags |= HAS_SESSION;
2921 * We return BAD_STATEID if filehandle doesn't match stateid,
2922 * the confirmed flag is incorrecly set, or the generation
2923 * number is incorrect.
2925 stp = find_stateid(stateid, flags);
2928 * Also, we should make sure this isn't just the result of
2931 sop = search_close_lru(stateid->si_stateownerid, flags);
2933 return stateid_error_map(stateid);
2939 *sopp = sop = stp->st_stateowner;
2942 clientid_t *lockclid = &lock->v.new.clientid;
2943 struct nfs4_client *clp = sop->so_client;
2947 lkflg = setlkflg(lock->lk_type);
2949 if (lock->lk_is_new) {
2950 if (!sop->so_is_open_owner)
2951 return nfserr_bad_stateid;
2952 if (!(flags & HAS_SESSION) &&
2953 !same_clid(&clp->cl_clientid, lockclid))
2954 return nfserr_bad_stateid;
2955 /* stp is the open stateid */
2956 status = nfs4_check_openmode(stp, lkflg);
2960 /* stp is the lock stateid */
2961 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2967 if (nfs4_check_fh(current_fh, stp)) {
2968 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2969 return nfserr_bad_stateid;
2973 * We now validate the seqid and stateid generation numbers.
2974 * For the moment, we ignore the possibility of
2975 * generation number wraparound.
2977 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
2980 if (sop->so_confirmed && flags & CONFIRM) {
2981 dprintk("NFSD: preprocess_seqid_op: expected"
2982 " unconfirmed stateowner!\n");
2983 return nfserr_bad_stateid;
2985 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2986 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2987 " confirmed yet!\n");
2988 return nfserr_bad_stateid;
2990 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
2993 renew_client(sop->so_client);
2997 if (seqid == sop->so_seqid - 1) {
2998 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2999 /* indicate replay to calling function */
3000 return nfserr_replay_me;
3002 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3003 sop->so_seqid, seqid);
3005 return nfserr_bad_seqid;
3009 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3010 struct nfsd4_open_confirm *oc)
3013 struct nfs4_stateowner *sop;
3014 struct nfs4_stateid *stp;
3016 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3017 (int)cstate->current_fh.fh_dentry->d_name.len,
3018 cstate->current_fh.fh_dentry->d_name.name);
3020 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3026 if ((status = nfs4_preprocess_seqid_op(cstate,
3027 oc->oc_seqid, &oc->oc_req_stateid,
3028 CONFIRM | OPEN_STATE,
3029 &oc->oc_stateowner, &stp, NULL)))
3032 sop = oc->oc_stateowner;
3033 sop->so_confirmed = 1;
3034 update_stateid(&stp->st_stateid);
3035 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3036 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3037 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3039 nfsd4_create_clid_dir(sop->so_client);
3041 if (oc->oc_stateowner) {
3042 nfs4_get_stateowner(oc->oc_stateowner);
3043 cstate->replay_owner = oc->oc_stateowner;
3045 nfs4_unlock_state();
3051 * unset all bits in union bitmap (bmap) that
3052 * do not exist in share (from successful OPEN_DOWNGRADE)
3055 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3058 for (i = 1; i < 4; i++) {
3059 if ((i & access) != i)
3060 __clear_bit(i, bmap);
3065 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3068 for (i = 0; i < 4; i++) {
3069 if ((i & deny) != i)
3070 __clear_bit(i, bmap);
3075 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3076 struct nfsd4_compound_state *cstate,
3077 struct nfsd4_open_downgrade *od)
3080 struct nfs4_stateid *stp;
3081 unsigned int share_access;
3083 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3084 (int)cstate->current_fh.fh_dentry->d_name.len,
3085 cstate->current_fh.fh_dentry->d_name.name);
3087 if (!access_valid(od->od_share_access, cstate->minorversion)
3088 || !deny_valid(od->od_share_deny))
3089 return nfserr_inval;
3092 if ((status = nfs4_preprocess_seqid_op(cstate,
3096 &od->od_stateowner, &stp, NULL)))
3099 status = nfserr_inval;
3100 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3101 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3102 stp->st_access_bmap, od->od_share_access);
3105 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3106 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3107 stp->st_deny_bmap, od->od_share_deny);
3110 set_access(&share_access, stp->st_access_bmap);
3111 nfs4_file_downgrade(stp->st_vfs_file,
3112 share_access & ~od->od_share_access);
3114 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3115 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3117 update_stateid(&stp->st_stateid);
3118 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3121 if (od->od_stateowner) {
3122 nfs4_get_stateowner(od->od_stateowner);
3123 cstate->replay_owner = od->od_stateowner;
3125 nfs4_unlock_state();
3130 * nfs4_unlock_state() called after encode
3133 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3134 struct nfsd4_close *close)
3137 struct nfs4_stateid *stp;
3139 dprintk("NFSD: nfsd4_close on file %.*s\n",
3140 (int)cstate->current_fh.fh_dentry->d_name.len,
3141 cstate->current_fh.fh_dentry->d_name.name);
3144 /* check close_lru for replay */
3145 if ((status = nfs4_preprocess_seqid_op(cstate,
3148 OPEN_STATE | CLOSE_STATE,
3149 &close->cl_stateowner, &stp, NULL)))
3152 update_stateid(&stp->st_stateid);
3153 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3155 /* release_stateid() calls nfsd_close() if needed */
3156 release_open_stateid(stp);
3158 /* place unused nfs4_stateowners on so_close_lru list to be
3159 * released by the laundromat service after the lease period
3160 * to enable us to handle CLOSE replay
3162 if (list_empty(&close->cl_stateowner->so_stateids))
3163 move_to_close_lru(close->cl_stateowner);
3165 if (close->cl_stateowner) {
3166 nfs4_get_stateowner(close->cl_stateowner);
3167 cstate->replay_owner = close->cl_stateowner;
3169 nfs4_unlock_state();
3174 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3175 struct nfsd4_delegreturn *dr)
3177 struct nfs4_delegation *dp;
3178 stateid_t *stateid = &dr->dr_stateid;
3179 struct inode *inode;
3183 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3185 inode = cstate->current_fh.fh_dentry->d_inode;
3187 if (nfsd4_has_session(cstate))
3188 flags |= HAS_SESSION;
3190 status = nfserr_bad_stateid;
3191 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3193 status = nfserr_stale_stateid;
3194 if (STALE_STATEID(stateid))
3196 status = nfserr_bad_stateid;
3197 if (!is_delegation_stateid(stateid))
3199 dp = find_delegation_stateid(inode, stateid);
3201 status = stateid_error_map(stateid);
3204 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3207 renew_client(dp->dl_client);
3209 unhash_delegation(dp);
3211 nfs4_unlock_state();
3218 * Lock owner state (byte-range locks)
3220 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3221 #define LOCK_HASH_BITS 8
3222 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3223 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3226 end_offset(u64 start, u64 len)
3231 return end >= start ? end: NFS4_MAX_UINT64;
3234 /* last octet in a range */
3236 last_byte_offset(u64 start, u64 len)
3242 return end > start ? end - 1: NFS4_MAX_UINT64;
3245 #define lockownerid_hashval(id) \
3246 ((id) & LOCK_HASH_MASK)
3248 static inline unsigned int
3249 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3250 struct xdr_netobj *ownername)
3252 return (file_hashval(inode) + cl_id
3253 + opaque_hashval(ownername->data, ownername->len))
3257 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3258 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3259 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3261 static struct nfs4_stateid *
3262 find_stateid(stateid_t *stid, int flags)
3264 struct nfs4_stateid *local;
3265 u32 st_id = stid->si_stateownerid;
3266 u32 f_id = stid->si_fileid;
3267 unsigned int hashval;
3269 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3270 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3271 hashval = stateid_hashval(st_id, f_id);
3272 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3273 if ((local->st_stateid.si_stateownerid == st_id) &&
3274 (local->st_stateid.si_fileid == f_id))
3279 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3280 hashval = stateid_hashval(st_id, f_id);
3281 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3282 if ((local->st_stateid.si_stateownerid == st_id) &&
3283 (local->st_stateid.si_fileid == f_id))
3290 static struct nfs4_delegation *
3291 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3293 struct nfs4_file *fp;
3294 struct nfs4_delegation *dl;
3296 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3299 fp = find_file(ino);
3302 dl = find_delegation_file(fp, stid);
3308 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3309 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3310 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3311 * locking, this prevents us from being completely protocol-compliant. The
3312 * real solution to this problem is to start using unsigned file offsets in
3313 * the VFS, but this is a very deep change!
3316 nfs4_transform_lock_offset(struct file_lock *lock)
3318 if (lock->fl_start < 0)
3319 lock->fl_start = OFFSET_MAX;
3320 if (lock->fl_end < 0)
3321 lock->fl_end = OFFSET_MAX;
3324 /* Hack!: For now, we're defining this just so we can use a pointer to it
3325 * as a unique cookie to identify our (NFSv4's) posix locks. */
3326 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3330 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3332 struct nfs4_stateowner *sop;
3335 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3336 sop = (struct nfs4_stateowner *) fl->fl_owner;
3337 hval = lockownerid_hashval(sop->so_id);
3338 kref_get(&sop->so_ref);
3340 deny->ld_clientid = sop->so_client->cl_clientid;
3342 deny->ld_sop = NULL;
3343 deny->ld_clientid.cl_boot = 0;
3344 deny->ld_clientid.cl_id = 0;
3346 deny->ld_start = fl->fl_start;
3347 deny->ld_length = NFS4_MAX_UINT64;
3348 if (fl->fl_end != NFS4_MAX_UINT64)
3349 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3350 deny->ld_type = NFS4_READ_LT;
3351 if (fl->fl_type != F_RDLCK)
3352 deny->ld_type = NFS4_WRITE_LT;
3355 static struct nfs4_stateowner *
3356 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3357 struct xdr_netobj *owner)
3359 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3360 struct nfs4_stateowner *op;
3362 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3363 if (same_owner_str(op, owner, clid))
3370 * Alloc a lock owner structure.
3371 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3374 * strhashval = lock_ownerstr_hashval
3377 static struct nfs4_stateowner *
3378 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3379 struct nfs4_stateowner *sop;
3380 struct nfs4_replay *rp;
3381 unsigned int idhashval;
3383 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3385 idhashval = lockownerid_hashval(current_ownerid);
3386 INIT_LIST_HEAD(&sop->so_idhash);
3387 INIT_LIST_HEAD(&sop->so_strhash);
3388 INIT_LIST_HEAD(&sop->so_perclient);
3389 INIT_LIST_HEAD(&sop->so_stateids);
3390 INIT_LIST_HEAD(&sop->so_perstateid);
3391 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3393 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3394 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3395 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3396 sop->so_is_open_owner = 0;
3397 sop->so_id = current_ownerid++;
3398 sop->so_client = clp;
3399 /* It is the openowner seqid that will be incremented in encode in the
3400 * case of new lockowners; so increment the lock seqid manually: */
3401 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3402 sop->so_confirmed = 1;
3403 rp = &sop->so_replay;
3404 rp->rp_status = nfserr_serverfault;
3406 rp->rp_buf = rp->rp_ibuf;
3410 static struct nfs4_stateid *
3411 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3413 struct nfs4_stateid *stp;
3414 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3416 stp = nfs4_alloc_stateid();
3419 INIT_LIST_HEAD(&stp->st_hash);
3420 INIT_LIST_HEAD(&stp->st_perfile);
3421 INIT_LIST_HEAD(&stp->st_perstateowner);
3422 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3423 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3424 list_add(&stp->st_perfile, &fp->fi_stateids);
3425 list_add(&stp->st_perstateowner, &sop->so_stateids);
3426 stp->st_stateowner = sop;
3429 stp->st_stateid.si_boot = get_seconds();
3430 stp->st_stateid.si_stateownerid = sop->so_id;
3431 stp->st_stateid.si_fileid = fp->fi_id;
3432 stp->st_stateid.si_generation = 0;
3433 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3434 stp->st_access_bmap = open_stp->st_access_bmap;
3435 stp->st_deny_bmap = open_stp->st_deny_bmap;
3436 stp->st_openstp = open_stp;
3443 check_lock_length(u64 offset, u64 length)
3445 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3446 LOFF_OVERFLOW(offset, length)));
3453 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3454 struct nfsd4_lock *lock)
3456 struct nfs4_stateowner *open_sop = NULL;
3457 struct nfs4_stateowner *lock_sop = NULL;
3458 struct nfs4_stateid *lock_stp;
3460 struct file_lock file_lock;
3461 struct file_lock conflock;
3463 unsigned int strhashval;
3467 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3468 (long long) lock->lk_offset,
3469 (long long) lock->lk_length);
3471 if (check_lock_length(lock->lk_offset, lock->lk_length))
3472 return nfserr_inval;
3474 if ((status = fh_verify(rqstp, &cstate->current_fh,
3475 S_IFREG, NFSD_MAY_LOCK))) {
3476 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3482 if (lock->lk_is_new) {
3484 * Client indicates that this is a new lockowner.
3485 * Use open owner and open stateid to create lock owner and
3488 struct nfs4_stateid *open_stp = NULL;
3489 struct nfs4_file *fp;
3491 status = nfserr_stale_clientid;
3492 if (!nfsd4_has_session(cstate) &&
3493 STALE_CLIENTID(&lock->lk_new_clientid))
3496 /* validate and update open stateid and open seqid */
3497 status = nfs4_preprocess_seqid_op(cstate,
3498 lock->lk_new_open_seqid,
3499 &lock->lk_new_open_stateid,
3501 &lock->lk_replay_owner, &open_stp,
3505 open_sop = lock->lk_replay_owner;
3506 /* create lockowner and lock stateid */
3507 fp = open_stp->st_file;
3508 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3509 open_sop->so_client->cl_clientid.cl_id,
3510 &lock->v.new.owner);
3511 /* XXX: Do we need to check for duplicate stateowners on
3512 * the same file, or should they just be allowed (and
3513 * create new stateids)? */
3514 status = nfserr_resource;
3515 lock_sop = alloc_init_lock_stateowner(strhashval,
3516 open_sop->so_client, open_stp, lock);
3517 if (lock_sop == NULL)
3519 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3520 if (lock_stp == NULL)
3523 /* lock (lock owner + lock stateid) already exists */
3524 status = nfs4_preprocess_seqid_op(cstate,
3525 lock->lk_old_lock_seqid,
3526 &lock->lk_old_lock_stateid,
3528 &lock->lk_replay_owner, &lock_stp, lock);
3531 lock_sop = lock->lk_replay_owner;
3533 /* lock->lk_replay_owner and lock_stp have been created or found */
3534 filp = lock_stp->st_vfs_file;
3536 status = nfserr_grace;
3537 if (locks_in_grace() && !lock->lk_reclaim)
3539 status = nfserr_no_grace;
3540 if (!locks_in_grace() && lock->lk_reclaim)
3543 locks_init_lock(&file_lock);
3544 switch (lock->lk_type) {
3547 file_lock.fl_type = F_RDLCK;
3551 case NFS4_WRITEW_LT:
3552 file_lock.fl_type = F_WRLCK;
3556 status = nfserr_inval;
3559 file_lock.fl_owner = (fl_owner_t)lock_sop;
3560 file_lock.fl_pid = current->tgid;
3561 file_lock.fl_file = filp;
3562 file_lock.fl_flags = FL_POSIX;
3563 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3565 file_lock.fl_start = lock->lk_offset;
3566 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3567 nfs4_transform_lock_offset(&file_lock);
3570 * Try to lock the file in the VFS.
3571 * Note: locks.c uses the BKL to protect the inode's lock list.
3574 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3576 case 0: /* success! */
3577 update_stateid(&lock_stp->st_stateid);
3578 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3582 case (EAGAIN): /* conflock holds conflicting lock */
3583 status = nfserr_denied;
3584 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3585 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3588 status = nfserr_deadlock;
3591 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3592 status = nfserr_resource;
3596 if (status && lock->lk_is_new && lock_sop)
3597 release_lockowner(lock_sop);
3598 if (lock->lk_replay_owner) {
3599 nfs4_get_stateowner(lock->lk_replay_owner);
3600 cstate->replay_owner = lock->lk_replay_owner;
3602 nfs4_unlock_state();
3607 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3608 * so we do a temporary open here just to get an open file to pass to
3609 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3612 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3617 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3620 err = vfs_test_lock(file, lock);
3629 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3630 struct nfsd4_lockt *lockt)
3632 struct inode *inode;
3633 struct file_lock file_lock;
3637 if (locks_in_grace())
3638 return nfserr_grace;
3640 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3641 return nfserr_inval;
3643 lockt->lt_stateowner = NULL;
3646 status = nfserr_stale_clientid;
3647 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3650 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3651 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3652 if (status == nfserr_symlink)
3653 status = nfserr_inval;
3657 inode = cstate->current_fh.fh_dentry->d_inode;
3658 locks_init_lock(&file_lock);
3659 switch (lockt->lt_type) {
3662 file_lock.fl_type = F_RDLCK;
3665 case NFS4_WRITEW_LT:
3666 file_lock.fl_type = F_WRLCK;
3669 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3670 status = nfserr_inval;
3674 lockt->lt_stateowner = find_lockstateowner_str(inode,
3675 &lockt->lt_clientid, &lockt->lt_owner);
3676 if (lockt->lt_stateowner)
3677 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3678 file_lock.fl_pid = current->tgid;
3679 file_lock.fl_flags = FL_POSIX;
3681 file_lock.fl_start = lockt->lt_offset;
3682 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3684 nfs4_transform_lock_offset(&file_lock);
3687 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3689 status = nfserrno(error);
3692 if (file_lock.fl_type != F_UNLCK) {
3693 status = nfserr_denied;
3694 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3697 nfs4_unlock_state();
3702 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3703 struct nfsd4_locku *locku)
3705 struct nfs4_stateid *stp;
3706 struct file *filp = NULL;
3707 struct file_lock file_lock;
3711 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3712 (long long) locku->lu_offset,
3713 (long long) locku->lu_length);
3715 if (check_lock_length(locku->lu_offset, locku->lu_length))
3716 return nfserr_inval;
3720 if ((status = nfs4_preprocess_seqid_op(cstate,
3724 &locku->lu_stateowner, &stp, NULL)))
3727 filp = stp->st_vfs_file;
3729 locks_init_lock(&file_lock);
3730 file_lock.fl_type = F_UNLCK;
3731 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3732 file_lock.fl_pid = current->tgid;
3733 file_lock.fl_file = filp;
3734 file_lock.fl_flags = FL_POSIX;
3735 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3736 file_lock.fl_start = locku->lu_offset;
3738 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3739 nfs4_transform_lock_offset(&file_lock);
3742 * Try to unlock the file in the VFS.
3744 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3746 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3750 * OK, unlock succeeded; the only thing left to do is update the stateid.
3752 update_stateid(&stp->st_stateid);
3753 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3756 if (locku->lu_stateowner) {
3757 nfs4_get_stateowner(locku->lu_stateowner);
3758 cstate->replay_owner = locku->lu_stateowner;
3760 nfs4_unlock_state();
3764 status = nfserrno(err);
3770 * 1: locks held by lockowner
3771 * 0: no locks held by lockowner
3774 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3776 struct file_lock **flpp;
3777 struct inode *inode = filp->f_path.dentry->d_inode;
3781 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3782 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3793 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3794 struct nfsd4_compound_state *cstate,
3795 struct nfsd4_release_lockowner *rlockowner)
3797 clientid_t *clid = &rlockowner->rl_clientid;
3798 struct nfs4_stateowner *sop;
3799 struct nfs4_stateid *stp;
3800 struct xdr_netobj *owner = &rlockowner->rl_owner;
3801 struct list_head matches;
3805 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3806 clid->cl_boot, clid->cl_id);
3808 /* XXX check for lease expiration */
3810 status = nfserr_stale_clientid;
3811 if (STALE_CLIENTID(clid))
3816 status = nfserr_locks_held;
3817 /* XXX: we're doing a linear search through all the lockowners.
3818 * Yipes! For now we'll just hope clients aren't really using
3819 * release_lockowner much, but eventually we have to fix these
3820 * data structures. */
3821 INIT_LIST_HEAD(&matches);
3822 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3823 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3824 if (!same_owner_str(sop, owner, clid))
3826 list_for_each_entry(stp, &sop->so_stateids,
3828 if (check_for_locks(stp->st_vfs_file, sop))
3830 /* Note: so_perclient unused for lockowners,
3831 * so it's OK to fool with here. */
3832 list_add(&sop->so_perclient, &matches);
3836 /* Clients probably won't expect us to return with some (but not all)
3837 * of the lockowner state released; so don't release any until all
3838 * have been checked. */
3840 while (!list_empty(&matches)) {
3841 sop = list_entry(matches.next, struct nfs4_stateowner,
3843 /* unhash_stateowner deletes so_perclient only
3844 * for openowners. */
3845 list_del(&sop->so_perclient);
3846 release_lockowner(sop);
3849 nfs4_unlock_state();
3853 static inline struct nfs4_client_reclaim *
3856 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3860 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3862 unsigned int strhashval = clientstr_hashval(name);
3863 struct nfs4_client *clp;
3865 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3870 * failure => all reset bets are off, nfserr_no_grace...
3873 nfs4_client_to_reclaim(const char *name)
3875 unsigned int strhashval;
3876 struct nfs4_client_reclaim *crp = NULL;
3878 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3879 crp = alloc_reclaim();
3882 strhashval = clientstr_hashval(name);
3883 INIT_LIST_HEAD(&crp->cr_strhash);
3884 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3885 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3886 reclaim_str_hashtbl_size++;
3891 nfs4_release_reclaim(void)
3893 struct nfs4_client_reclaim *crp = NULL;
3896 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3897 while (!list_empty(&reclaim_str_hashtbl[i])) {
3898 crp = list_entry(reclaim_str_hashtbl[i].next,
3899 struct nfs4_client_reclaim, cr_strhash);
3900 list_del(&crp->cr_strhash);
3902 reclaim_str_hashtbl_size--;
3905 BUG_ON(reclaim_str_hashtbl_size);
3909 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3910 static struct nfs4_client_reclaim *
3911 nfs4_find_reclaim_client(clientid_t *clid)
3913 unsigned int strhashval;
3914 struct nfs4_client *clp;
3915 struct nfs4_client_reclaim *crp = NULL;
3918 /* find clientid in conf_id_hashtbl */
3919 clp = find_confirmed_client(clid);
3923 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3924 clp->cl_name.len, clp->cl_name.data,
3927 /* find clp->cl_name in reclaim_str_hashtbl */
3928 strhashval = clientstr_hashval(clp->cl_recdir);
3929 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3930 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3938 * Called from OPEN. Look for clientid in reclaim list.
3941 nfs4_check_open_reclaim(clientid_t *clid)
3943 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3946 /* initialization to perform at module load time: */
3949 nfs4_state_init(void)
3953 status = nfsd4_init_slabs();
3956 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3957 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3958 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3959 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3960 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3961 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3963 for (i = 0; i < SESSION_HASH_SIZE; i++)
3964 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3965 for (i = 0; i < FILE_HASH_SIZE; i++) {
3966 INIT_LIST_HEAD(&file_hashtbl[i]);
3968 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3969 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3970 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3972 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3973 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3974 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3976 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3977 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3978 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3980 memset(&onestateid, ~0, sizeof(stateid_t));
3981 INIT_LIST_HEAD(&close_lru);
3982 INIT_LIST_HEAD(&client_lru);
3983 INIT_LIST_HEAD(&del_recall_lru);
3984 reclaim_str_hashtbl_size = 0;
3989 nfsd4_load_reboot_recovery_data(void)
3994 nfsd4_init_recdir(user_recovery_dirname);
3995 status = nfsd4_recdir_load();
3996 nfs4_unlock_state();
3998 printk("NFSD: Failure reading reboot recovery data\n");
4002 * Since the lifetime of a delegation isn't limited to that of an open, a
4003 * client may quite reasonably hang on to a delegation as long as it has
4004 * the inode cached. This becomes an obvious problem the first time a
4005 * client's inode cache approaches the size of the server's total memory.
4007 * For now we avoid this problem by imposing a hard limit on the number
4008 * of delegations, which varies according to the server's memory size.
4011 set_max_delegations(void)
4014 * Allow at most 4 delegations per megabyte of RAM. Quick
4015 * estimates suggest that in the worst case (where every delegation
4016 * is for a different inode), a delegation could take about 1.5K,
4017 * giving a worst case usage of about 6% of memory.
4019 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4022 /* initialization to perform when the nfsd service is started: */
4025 __nfs4_state_start(void)
4029 boot_time = get_seconds();
4030 locks_start_grace(&nfsd4_manager);
4031 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4033 ret = set_callback_cred();
4036 laundry_wq = create_singlethread_workqueue("nfsd4");
4037 if (laundry_wq == NULL)
4039 ret = nfsd4_create_callback_queue();
4041 goto out_free_laundry;
4042 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4043 set_max_delegations();
4046 destroy_workqueue(laundry_wq);
4051 nfs4_state_start(void)
4057 nfsd4_load_reboot_recovery_data();
4058 ret = __nfs4_state_start();
4066 __nfs4_state_shutdown(void)
4069 struct nfs4_client *clp = NULL;
4070 struct nfs4_delegation *dp = NULL;
4071 struct list_head *pos, *next, reaplist;
4073 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4074 while (!list_empty(&conf_id_hashtbl[i])) {
4075 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4078 while (!list_empty(&unconf_str_hashtbl[i])) {
4079 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4083 INIT_LIST_HEAD(&reaplist);
4084 spin_lock(&recall_lock);
4085 list_for_each_safe(pos, next, &del_recall_lru) {
4086 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4087 list_move(&dp->dl_recall_lru, &reaplist);
4089 spin_unlock(&recall_lock);
4090 list_for_each_safe(pos, next, &reaplist) {
4091 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4092 list_del_init(&dp->dl_recall_lru);
4093 unhash_delegation(dp);
4096 nfsd4_shutdown_recdir();
4101 nfs4_state_shutdown(void)
4103 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4104 destroy_workqueue(laundry_wq);
4105 locks_end_grace(&nfsd4_manager);
4107 nfs4_release_reclaim();
4108 __nfs4_state_shutdown();
4109 nfsd4_destroy_callback_queue();
4110 nfs4_unlock_state();
4114 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4115 * accessed when nfsd is starting.
4118 nfs4_set_recdir(char *recdir)
4120 strcpy(user_recovery_dirname, recdir);
4124 * Change the NFSv4 recovery directory to recdir.
4127 nfs4_reset_recoverydir(char *recdir)
4132 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4136 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4137 nfs4_set_recdir(recdir);
4145 nfs4_recoverydir(void)
4147 return user_recovery_dirname;