2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
36 #include <linux/smp_lock.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/sunrpc/svcauth_gss.h>
41 #include <linux/sunrpc/clnt.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 time_t nfsd4_lease = 90; /* default lease time */
49 time_t nfsd4_grace = 90;
50 static time_t boot_time;
51 static u32 current_ownerid = 1;
52 static u32 current_fileid = 1;
53 static u32 current_delegid = 1;
55 static stateid_t zerostateid; /* bits all 0 */
56 static stateid_t onestateid; /* bits all 1 */
57 static u64 current_sessionid = 1;
59 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
62 /* forward declarations */
63 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
64 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
65 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
66 static void nfs4_set_recdir(char *recdir);
70 /* Currently used for almost all code touching nfsv4 state: */
71 static DEFINE_MUTEX(client_mutex);
74 * Currently used for the del_recall_lru and file hash table. In an
75 * effort to decrease the scope of the client_mutex, this spinlock may
76 * eventually cover more:
78 static DEFINE_SPINLOCK(recall_lock);
80 static struct kmem_cache *stateowner_slab = NULL;
81 static struct kmem_cache *file_slab = NULL;
82 static struct kmem_cache *stateid_slab = NULL;
83 static struct kmem_cache *deleg_slab = NULL;
88 mutex_lock(&client_mutex);
92 nfs4_unlock_state(void)
94 mutex_unlock(&client_mutex);
98 opaque_hashval(const void *ptr, int nbytes)
100 unsigned char *cptr = (unsigned char *) ptr;
110 static struct list_head del_recall_lru;
113 put_nfs4_file(struct nfs4_file *fi)
115 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
116 list_del(&fi->fi_hash);
117 spin_unlock(&recall_lock);
119 kmem_cache_free(file_slab, fi);
124 get_nfs4_file(struct nfs4_file *fi)
126 atomic_inc(&fi->fi_ref);
129 static int num_delegations;
130 unsigned int max_delegations;
133 * Open owner state (share locks)
136 /* hash tables for nfs4_stateowner */
137 #define OWNER_HASH_BITS 8
138 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
139 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
141 #define ownerid_hashval(id) \
142 ((id) & OWNER_HASH_MASK)
143 #define ownerstr_hashval(clientid, ownername) \
144 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
146 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
147 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
149 /* hash table for nfs4_file */
150 #define FILE_HASH_BITS 8
151 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
153 /* hash table for (open)nfs4_stateid */
154 #define STATEID_HASH_BITS 10
155 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
156 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
158 #define file_hashval(x) \
159 hash_ptr(x, FILE_HASH_BITS)
160 #define stateid_hashval(owner_id, file_id) \
161 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
163 static struct list_head file_hashtbl[FILE_HASH_SIZE];
164 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
166 static struct nfs4_delegation *
167 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
169 struct nfs4_delegation *dp;
170 struct nfs4_file *fp = stp->st_file;
171 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
173 dprintk("NFSD alloc_init_deleg\n");
174 if (fp->fi_had_conflict)
176 if (num_delegations > max_delegations)
178 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
182 INIT_LIST_HEAD(&dp->dl_perfile);
183 INIT_LIST_HEAD(&dp->dl_perclnt);
184 INIT_LIST_HEAD(&dp->dl_recall_lru);
189 get_file(stp->st_vfs_file);
190 dp->dl_vfs_file = stp->st_vfs_file;
192 dp->dl_ident = cb->cb_ident;
193 dp->dl_stateid.si_boot = get_seconds();
194 dp->dl_stateid.si_stateownerid = current_delegid++;
195 dp->dl_stateid.si_fileid = 0;
196 dp->dl_stateid.si_generation = 0;
197 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
199 atomic_set(&dp->dl_count, 1);
200 list_add(&dp->dl_perfile, &fp->fi_delegations);
201 list_add(&dp->dl_perclnt, &clp->cl_delegations);
202 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
207 nfs4_put_delegation(struct nfs4_delegation *dp)
209 if (atomic_dec_and_test(&dp->dl_count)) {
210 dprintk("NFSD: freeing dp %p\n",dp);
211 put_nfs4_file(dp->dl_file);
212 kmem_cache_free(deleg_slab, dp);
217 /* Remove the associated file_lock first, then remove the delegation.
218 * lease_modify() is called to remove the FS_LEASE file_lock from
219 * the i_flock list, eventually calling nfsd's lock_manager
220 * fl_release_callback.
223 nfs4_close_delegation(struct nfs4_delegation *dp)
225 struct file *filp = dp->dl_vfs_file;
227 dprintk("NFSD: close_delegation dp %p\n",dp);
228 dp->dl_vfs_file = NULL;
229 /* The following nfsd_close may not actually close the file,
230 * but we want to remove the lease in any case. */
232 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
236 /* Called under the state lock. */
238 unhash_delegation(struct nfs4_delegation *dp)
240 list_del_init(&dp->dl_perfile);
241 list_del_init(&dp->dl_perclnt);
242 spin_lock(&recall_lock);
243 list_del_init(&dp->dl_recall_lru);
244 spin_unlock(&recall_lock);
245 nfs4_close_delegation(dp);
246 nfs4_put_delegation(dp);
253 /* client_lock protects the session hash table */
254 static DEFINE_SPINLOCK(client_lock);
256 /* Hash tables for nfs4_clientid state */
257 #define CLIENT_HASH_BITS 4
258 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
261 #define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
263 #define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
275 * client_lru holds client queue ordered by nfs4_client.cl_time
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
281 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
282 static int reclaim_str_hashtbl_size = 0;
283 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
284 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
285 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
286 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
287 static struct list_head client_lru;
288 static struct list_head close_lru;
290 static void unhash_generic_stateid(struct nfs4_stateid *stp)
292 list_del(&stp->st_hash);
293 list_del(&stp->st_perfile);
294 list_del(&stp->st_perstateowner);
297 static void free_generic_stateid(struct nfs4_stateid *stp)
299 put_nfs4_file(stp->st_file);
300 kmem_cache_free(stateid_slab, stp);
303 static void release_lock_stateid(struct nfs4_stateid *stp)
305 unhash_generic_stateid(stp);
306 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
307 free_generic_stateid(stp);
310 static void unhash_lockowner(struct nfs4_stateowner *sop)
312 struct nfs4_stateid *stp;
314 list_del(&sop->so_idhash);
315 list_del(&sop->so_strhash);
316 list_del(&sop->so_perstateid);
317 while (!list_empty(&sop->so_stateids)) {
318 stp = list_first_entry(&sop->so_stateids,
319 struct nfs4_stateid, st_perstateowner);
320 release_lock_stateid(stp);
324 static void release_lockowner(struct nfs4_stateowner *sop)
326 unhash_lockowner(sop);
327 nfs4_put_stateowner(sop);
331 release_stateid_lockowners(struct nfs4_stateid *open_stp)
333 struct nfs4_stateowner *lock_sop;
335 while (!list_empty(&open_stp->st_lockowners)) {
336 lock_sop = list_entry(open_stp->st_lockowners.next,
337 struct nfs4_stateowner, so_perstateid);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop->so_is_open_owner);
340 release_lockowner(lock_sop);
344 static void release_open_stateid(struct nfs4_stateid *stp)
346 unhash_generic_stateid(stp);
347 release_stateid_lockowners(stp);
348 nfsd_close(stp->st_vfs_file);
349 free_generic_stateid(stp);
352 static void unhash_openowner(struct nfs4_stateowner *sop)
354 struct nfs4_stateid *stp;
356 list_del(&sop->so_idhash);
357 list_del(&sop->so_strhash);
358 list_del(&sop->so_perclient);
359 list_del(&sop->so_perstateid); /* XXX: necessary? */
360 while (!list_empty(&sop->so_stateids)) {
361 stp = list_first_entry(&sop->so_stateids,
362 struct nfs4_stateid, st_perstateowner);
363 release_open_stateid(stp);
367 static void release_openowner(struct nfs4_stateowner *sop)
369 unhash_openowner(sop);
370 list_del(&sop->so_close_lru);
371 nfs4_put_stateowner(sop);
374 #define SESSION_HASH_SIZE 512
375 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
378 hash_sessionid(struct nfs4_sessionid *sessionid)
380 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
382 return sid->sequence % SESSION_HASH_SIZE;
386 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
388 u32 *ptr = (u32 *)(&sessionid->data[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
393 gen_sessionid(struct nfsd4_session *ses)
395 struct nfs4_client *clp = ses->se_client;
396 struct nfsd4_sessionid *sid;
398 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
399 sid->clientid = clp->cl_clientid;
400 sid->sequence = current_sessionid++;
405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
416 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
428 static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
430 int mem, size = fchan->maxresp_cached;
432 if (fchan->maxreqs < 1)
435 if (size < NFSD_MIN_HDR_SEQ_SZ)
436 size = NFSD_MIN_HDR_SEQ_SZ;
437 size -= NFSD_MIN_HDR_SEQ_SZ;
438 if (size > NFSD_SLOT_CACHE_SIZE)
439 size = NFSD_SLOT_CACHE_SIZE;
441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem = fchan->maxreqs * size;
443 if (mem > NFSD_MAX_MEM_PER_SESSION) {
444 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
445 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
446 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
447 mem = fchan->maxreqs * size;
450 spin_lock(&nfsd_drc_lock);
451 /* bound the total session drc memory ussage */
452 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
453 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
454 mem = fchan->maxreqs * size;
456 nfsd_drc_mem_used += mem;
457 spin_unlock(&nfsd_drc_lock);
459 if (fchan->maxreqs == 0)
460 return nfserr_serverfault;
462 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
467 * fchan holds the client values on input, and the server values on output
468 * sv_max_mesg is the maximum payload plus one page for overhead.
470 static int init_forechannel_attrs(struct svc_rqst *rqstp,
471 struct nfsd4_channel_attrs *session_fchan,
472 struct nfsd4_channel_attrs *fchan)
475 __u32 maxcount = nfsd_serv->sv_max_mesg;
477 /* headerpadsz set to zero in encode routine */
479 /* Use the client's max request and max response size if possible */
480 if (fchan->maxreq_sz > maxcount)
481 fchan->maxreq_sz = maxcount;
482 session_fchan->maxreq_sz = fchan->maxreq_sz;
484 if (fchan->maxresp_sz > maxcount)
485 fchan->maxresp_sz = maxcount;
486 session_fchan->maxresp_sz = fchan->maxresp_sz;
488 /* Use the client's maxops if possible */
489 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
490 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
491 session_fchan->maxops = fchan->maxops;
493 /* FIXME: Error means no more DRC pages so the server should
494 * recover pages from existing sessions. For now fail session
497 status = set_forechannel_drc_size(fchan);
499 session_fchan->maxresp_cached = fchan->maxresp_cached;
500 session_fchan->maxreqs = fchan->maxreqs;
502 dprintk("%s status %d\n", __func__, status);
507 free_session_slots(struct nfsd4_session *ses)
511 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
512 kfree(ses->se_slots[i]);
516 * We don't actually need to cache the rpc and session headers, so we
517 * can allocate a little less for each slot:
519 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
521 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
525 alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
526 struct nfsd4_create_session *cses)
528 struct nfsd4_session *new, tmp;
529 struct nfsd4_slot *sp;
530 int idx, slotsize, cachesize, i;
533 memset(&tmp, 0, sizeof(tmp));
535 /* FIXME: For now, we just accept the client back channel attributes. */
536 tmp.se_bchannel = cses->back_channel;
537 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
538 &cses->fore_channel);
542 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
543 + sizeof(struct nfsd4_session) > PAGE_SIZE);
545 status = nfserr_serverfault;
546 /* allocate struct nfsd4_session and slot table pointers in one piece */
547 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
548 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
552 memcpy(new, &tmp, sizeof(*new));
554 /* allocate each struct nfsd4_slot and data cache in one piece */
555 cachesize = slot_bytes(&new->se_fchannel);
556 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
557 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
560 new->se_slots[i] = sp;
563 new->se_client = clp;
565 idx = hash_sessionid(&new->se_sessionid);
566 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
567 NFS4_MAX_SESSIONID_LEN);
569 new->se_flags = cses->flags;
570 kref_init(&new->se_ref);
571 spin_lock(&client_lock);
572 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
573 list_add(&new->se_perclnt, &clp->cl_sessions);
574 spin_unlock(&client_lock);
580 free_session_slots(new);
585 /* caller must hold client_lock */
586 static struct nfsd4_session *
587 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
589 struct nfsd4_session *elem;
592 dump_sessionid(__func__, sessionid);
593 idx = hash_sessionid(sessionid);
594 dprintk("%s: idx is %d\n", __func__, idx);
595 /* Search in the appropriate list */
596 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
597 dump_sessionid("list traversal", &elem->se_sessionid);
598 if (!memcmp(elem->se_sessionid.data, sessionid->data,
599 NFS4_MAX_SESSIONID_LEN)) {
604 dprintk("%s: session not found\n", __func__);
608 /* caller must hold client_lock */
610 unhash_session(struct nfsd4_session *ses)
612 list_del(&ses->se_hash);
613 list_del(&ses->se_perclnt);
617 release_session(struct nfsd4_session *ses)
619 spin_lock(&client_lock);
621 spin_unlock(&client_lock);
622 nfsd4_put_session(ses);
626 free_session(struct kref *kref)
628 struct nfsd4_session *ses;
631 ses = container_of(kref, struct nfsd4_session, se_ref);
632 spin_lock(&nfsd_drc_lock);
633 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
634 nfsd_drc_mem_used -= mem;
635 spin_unlock(&nfsd_drc_lock);
636 free_session_slots(ses);
641 renew_client(struct nfs4_client *clp)
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();
653 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
655 STALE_CLIENTID(clientid_t *clid)
657 if (clid->cl_boot == boot_time)
659 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
660 clid->cl_boot, clid->cl_id, boot_time);
665 * XXX Should we use a slab cache ?
666 * This type of memory management is somewhat inefficient, but we use it
667 * anyway since SETCLIENTID is not a common operation.
669 static struct nfs4_client *alloc_client(struct xdr_netobj name)
671 struct nfs4_client *clp;
673 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
676 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
677 if (clp->cl_name.data == NULL) {
681 memcpy(clp->cl_name.data, name.data, name.len);
682 clp->cl_name.len = name.len;
687 free_client(struct nfs4_client *clp)
689 if (clp->cl_cred.cr_group_info)
690 put_group_info(clp->cl_cred.cr_group_info);
691 kfree(clp->cl_principal);
692 kfree(clp->cl_name.data);
697 expire_client(struct nfs4_client *clp)
699 struct nfs4_stateowner *sop;
700 struct nfs4_delegation *dp;
701 struct list_head reaplist;
703 INIT_LIST_HEAD(&reaplist);
704 spin_lock(&recall_lock);
705 while (!list_empty(&clp->cl_delegations)) {
706 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
707 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
709 list_del_init(&dp->dl_perclnt);
710 list_move(&dp->dl_recall_lru, &reaplist);
712 spin_unlock(&recall_lock);
713 while (!list_empty(&reaplist)) {
714 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
715 list_del_init(&dp->dl_recall_lru);
716 unhash_delegation(dp);
718 list_del(&clp->cl_idhash);
719 list_del(&clp->cl_strhash);
720 list_del(&clp->cl_lru);
721 while (!list_empty(&clp->cl_openowners)) {
722 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
723 release_openowner(sop);
725 while (!list_empty(&clp->cl_sessions)) {
726 struct nfsd4_session *ses;
727 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
729 release_session(ses);
731 nfsd4_set_callback_client(clp, NULL);
732 if (clp->cl_cb_conn.cb_xprt)
733 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
737 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
739 memcpy(target->cl_verifier.data, source->data,
740 sizeof(target->cl_verifier.data));
743 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
745 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
746 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
749 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
751 target->cr_uid = source->cr_uid;
752 target->cr_gid = source->cr_gid;
753 target->cr_group_info = source->cr_group_info;
754 get_group_info(target->cr_group_info);
757 static int same_name(const char *n1, const char *n2)
759 return 0 == memcmp(n1, n2, HEXDIR_LEN);
763 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
765 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
769 same_clid(clientid_t *cl1, clientid_t *cl2)
771 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
774 /* XXX what about NGROUP */
776 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
778 return cr1->cr_uid == cr2->cr_uid;
781 static void gen_clid(struct nfs4_client *clp)
783 static u32 current_clientid = 1;
785 clp->cl_clientid.cl_boot = boot_time;
786 clp->cl_clientid.cl_id = current_clientid++;
789 static void gen_confirm(struct nfs4_client *clp)
794 p = (u32 *)clp->cl_confirm.data;
795 *p++ = get_seconds();
799 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
800 struct svc_rqst *rqstp, nfs4_verifier *verf)
802 struct nfs4_client *clp;
803 struct sockaddr *sa = svc_addr(rqstp);
806 clp = alloc_client(name);
810 princ = svc_gss_principal(rqstp);
812 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
813 if (clp->cl_principal == NULL) {
819 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
820 atomic_set(&clp->cl_cb_set, 0);
821 INIT_LIST_HEAD(&clp->cl_idhash);
822 INIT_LIST_HEAD(&clp->cl_strhash);
823 INIT_LIST_HEAD(&clp->cl_openowners);
824 INIT_LIST_HEAD(&clp->cl_delegations);
825 INIT_LIST_HEAD(&clp->cl_sessions);
826 INIT_LIST_HEAD(&clp->cl_lru);
827 clear_bit(0, &clp->cl_cb_slot_busy);
828 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
829 copy_verf(clp, verf);
830 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
831 clp->cl_flavor = rqstp->rq_flavor;
832 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
838 static int check_name(struct xdr_netobj name)
842 if (name.len > NFS4_OPAQUE_LIMIT) {
843 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
850 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
852 unsigned int idhashval;
854 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
855 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
856 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
857 list_add_tail(&clp->cl_lru, &client_lru);
858 clp->cl_time = get_seconds();
862 move_to_confirmed(struct nfs4_client *clp)
864 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
865 unsigned int strhashval;
867 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
868 list_del_init(&clp->cl_strhash);
869 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
870 strhashval = clientstr_hashval(clp->cl_recdir);
871 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
875 static struct nfs4_client *
876 find_confirmed_client(clientid_t *clid)
878 struct nfs4_client *clp;
879 unsigned int idhashval = clientid_hashval(clid->cl_id);
881 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
882 if (same_clid(&clp->cl_clientid, clid))
888 static struct nfs4_client *
889 find_unconfirmed_client(clientid_t *clid)
891 struct nfs4_client *clp;
892 unsigned int idhashval = clientid_hashval(clid->cl_id);
894 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
895 if (same_clid(&clp->cl_clientid, clid))
902 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
903 * parameter. Matching is based on the fact the at least one of the
904 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
906 * FIXME: we need to unify the clientid namespaces for nfsv4.x
907 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
908 * and SET_CLIENTID{,_CONFIRM}
911 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
913 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
914 return use_exchange_id == has_exchange_flags;
917 static struct nfs4_client *
918 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
919 bool use_exchange_id)
921 struct nfs4_client *clp;
923 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
924 if (same_name(clp->cl_recdir, dname) &&
925 match_clientid_establishment(clp, use_exchange_id))
931 static struct nfs4_client *
932 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
933 bool use_exchange_id)
935 struct nfs4_client *clp;
937 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
938 if (same_name(clp->cl_recdir, dname) &&
939 match_clientid_establishment(clp, use_exchange_id))
946 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
948 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
949 unsigned short expected_family;
951 /* Currently, we only support tcp and tcp6 for the callback channel */
952 if (se->se_callback_netid_len == 3 &&
953 !memcmp(se->se_callback_netid_val, "tcp", 3))
954 expected_family = AF_INET;
955 else if (se->se_callback_netid_len == 4 &&
956 !memcmp(se->se_callback_netid_val, "tcp6", 4))
957 expected_family = AF_INET6;
961 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
962 se->se_callback_addr_len,
963 (struct sockaddr *) &cb->cb_addr,
964 sizeof(cb->cb_addr));
966 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
969 if (cb->cb_addr.ss_family == AF_INET6)
970 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
972 cb->cb_minorversion = 0;
973 cb->cb_prog = se->se_callback_prog;
974 cb->cb_ident = se->se_callback_ident;
977 cb->cb_addr.ss_family = AF_UNSPEC;
979 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
980 "will not receive delegations\n",
981 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
987 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
990 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
992 struct nfsd4_slot *slot = resp->cstate.slot;
995 dprintk("--> %s slot %p\n", __func__, slot);
997 slot->sl_opcnt = resp->opcnt;
998 slot->sl_status = resp->cstate.status;
1000 if (nfsd4_not_cached(resp)) {
1001 slot->sl_datalen = 0;
1004 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1005 base = (char *)resp->cstate.datap -
1006 (char *)resp->xbuf->head[0].iov_base;
1007 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1009 WARN("%s: sessions DRC could not cache compound\n", __func__);
1014 * Encode the replay sequence operation from the slot values.
1015 * If cachethis is FALSE encode the uncached rep error on the next
1016 * operation which sets resp->p and increments resp->opcnt for
1017 * nfs4svc_encode_compoundres.
1021 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1022 struct nfsd4_compoundres *resp)
1024 struct nfsd4_op *op;
1025 struct nfsd4_slot *slot = resp->cstate.slot;
1027 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1028 resp->opcnt, resp->cstate.slot->sl_cachethis);
1030 /* Encode the replayed sequence operation */
1031 op = &args->ops[resp->opcnt - 1];
1032 nfsd4_encode_operation(resp, op);
1034 /* Return nfserr_retry_uncached_rep in next operation. */
1035 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1036 op = &args->ops[resp->opcnt++];
1037 op->status = nfserr_retry_uncached_rep;
1038 nfsd4_encode_operation(resp, op);
1044 * The sequence operation is not cached because we can use the slot and
1048 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1049 struct nfsd4_sequence *seq)
1051 struct nfsd4_slot *slot = resp->cstate.slot;
1054 dprintk("--> %s slot %p\n", __func__, slot);
1056 /* Either returns 0 or nfserr_retry_uncached */
1057 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1058 if (status == nfserr_retry_uncached_rep)
1061 /* The sequence operation has been encoded, cstate->datap set. */
1062 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1064 resp->opcnt = slot->sl_opcnt;
1065 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1066 status = slot->sl_status;
1072 * Set the exchange_id flags returned by the server.
1075 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1077 /* pNFS is not supported */
1078 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1080 /* Referrals are supported, Migration is not. */
1081 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1083 /* set the wire flags to return to client. */
1084 clid->flags = new->cl_exchange_flags;
1088 nfsd4_exchange_id(struct svc_rqst *rqstp,
1089 struct nfsd4_compound_state *cstate,
1090 struct nfsd4_exchange_id *exid)
1092 struct nfs4_client *unconf, *conf, *new;
1094 unsigned int strhashval;
1095 char dname[HEXDIR_LEN];
1096 char addr_str[INET6_ADDRSTRLEN];
1097 nfs4_verifier verf = exid->verifier;
1098 struct sockaddr *sa = svc_addr(rqstp);
1100 rpc_ntop(sa, addr_str, sizeof(addr_str));
1101 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1102 "ip_addr=%s flags %x, spa_how %d\n",
1103 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1104 addr_str, exid->flags, exid->spa_how);
1106 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1107 return nfserr_inval;
1109 /* Currently only support SP4_NONE */
1110 switch (exid->spa_how) {
1114 return nfserr_encr_alg_unsupp;
1116 BUG(); /* checked by xdr code */
1118 return nfserr_serverfault; /* no excuse :-/ */
1121 status = nfs4_make_rec_clidname(dname, &exid->clname);
1126 strhashval = clientstr_hashval(dname);
1131 conf = find_confirmed_client_by_str(dname, strhashval, true);
1133 if (!same_verf(&verf, &conf->cl_verifier)) {
1134 /* 18.35.4 case 8 */
1135 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1136 status = nfserr_not_same;
1139 /* Client reboot: destroy old state */
1140 expire_client(conf);
1143 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1144 /* 18.35.4 case 9 */
1145 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1146 status = nfserr_perm;
1149 expire_client(conf);
1153 * Set bit when the owner id and verifier map to an already
1154 * confirmed client id (18.35.3).
1156 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1159 * Falling into 18.35.4 case 2, possible router replay.
1160 * Leave confirmed record intact and return same result.
1162 copy_verf(conf, &verf);
1167 /* 18.35.4 case 7 */
1168 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1169 status = nfserr_noent;
1173 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1176 * Possible retry or client restart. Per 18.35.4 case 4,
1177 * a new unconfirmed record should be generated regardless
1178 * of whether any properties have changed.
1180 expire_client(unconf);
1185 new = create_client(exid->clname, dname, rqstp, &verf);
1187 status = nfserr_serverfault;
1192 add_to_unconfirmed(new, strhashval);
1194 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1195 exid->clientid.cl_id = new->cl_clientid.cl_id;
1198 nfsd4_set_ex_flags(new, exid);
1200 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1201 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1205 nfs4_unlock_state();
1207 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1212 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1214 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1217 /* The slot is in use, and no response has been sent. */
1219 if (seqid == slot_seqid)
1220 return nfserr_jukebox;
1222 return nfserr_seq_misordered;
1225 if (likely(seqid == slot_seqid + 1))
1228 if (seqid == slot_seqid)
1229 return nfserr_replay_cache;
1231 if (seqid == 1 && (slot_seqid + 1) == 0)
1233 /* Misordered replay or misordered new request */
1234 return nfserr_seq_misordered;
1238 * Cache the create session result into the create session single DRC
1239 * slot cache by saving the xdr structure. sl_seqid has been set.
1240 * Do this for solo or embedded create session operations.
1243 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1244 struct nfsd4_clid_slot *slot, int nfserr)
1246 slot->sl_status = nfserr;
1247 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1251 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1252 struct nfsd4_clid_slot *slot)
1254 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1255 return slot->sl_status;
1259 nfsd4_create_session(struct svc_rqst *rqstp,
1260 struct nfsd4_compound_state *cstate,
1261 struct nfsd4_create_session *cr_ses)
1263 struct sockaddr *sa = svc_addr(rqstp);
1264 struct nfs4_client *conf, *unconf;
1265 struct nfsd4_clid_slot *cs_slot = NULL;
1269 unconf = find_unconfirmed_client(&cr_ses->clientid);
1270 conf = find_confirmed_client(&cr_ses->clientid);
1273 cs_slot = &conf->cl_cs_slot;
1274 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1275 if (status == nfserr_replay_cache) {
1276 dprintk("Got a create_session replay! seqid= %d\n",
1278 /* Return the cached reply status */
1279 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1281 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1282 status = nfserr_seq_misordered;
1283 dprintk("Sequence misordered!\n");
1284 dprintk("Expected seqid= %d but got seqid= %d\n",
1285 cs_slot->sl_seqid, cr_ses->seqid);
1288 cs_slot->sl_seqid++;
1289 } else if (unconf) {
1290 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1291 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1292 status = nfserr_clid_inuse;
1296 cs_slot = &unconf->cl_cs_slot;
1297 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1299 /* an unconfirmed replay returns misordered */
1300 status = nfserr_seq_misordered;
1304 cs_slot->sl_seqid++; /* from 0 to 1 */
1305 move_to_confirmed(unconf);
1307 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1308 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1309 svc_xprt_get(rqstp->rq_xprt);
1311 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1313 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1314 unconf->cl_cb_conn.cb_minorversion =
1315 cstate->minorversion;
1316 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1317 unconf->cl_cb_seq_nr = 1;
1318 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1322 status = nfserr_stale_clientid;
1327 * We do not support RDMA or persistent sessions
1329 cr_ses->flags &= ~SESSION4_PERSIST;
1330 cr_ses->flags &= ~SESSION4_RDMA;
1332 status = alloc_init_session(rqstp, conf, cr_ses);
1336 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1337 NFS4_MAX_SESSIONID_LEN);
1338 cr_ses->seqid = cs_slot->sl_seqid;
1341 /* cache solo and embedded create sessions under the state lock */
1342 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1344 nfs4_unlock_state();
1345 dprintk("%s returns %d\n", __func__, ntohl(status));
1349 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1351 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1352 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1354 return argp->opcnt == resp->opcnt;
1357 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1361 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1365 nfsd4_destroy_session(struct svc_rqst *r,
1366 struct nfsd4_compound_state *cstate,
1367 struct nfsd4_destroy_session *sessionid)
1369 struct nfsd4_session *ses;
1370 u32 status = nfserr_badsession;
1373 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1374 * - Should we return nfserr_back_chan_busy if waiting for
1375 * callbacks on to-be-destroyed session?
1376 * - Do we need to clear any callback info from previous session?
1379 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1380 if (!nfsd4_last_compound_op(r))
1381 return nfserr_not_only_op;
1383 dump_sessionid(__func__, &sessionid->sessionid);
1384 spin_lock(&client_lock);
1385 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1387 spin_unlock(&client_lock);
1391 unhash_session(ses);
1392 spin_unlock(&client_lock);
1394 /* wait for callbacks */
1395 nfsd4_set_callback_client(ses->se_client, NULL);
1396 nfsd4_put_session(ses);
1399 dprintk("%s returns %d\n", __func__, ntohl(status));
1404 nfsd4_sequence(struct svc_rqst *rqstp,
1405 struct nfsd4_compound_state *cstate,
1406 struct nfsd4_sequence *seq)
1408 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1409 struct nfsd4_session *session;
1410 struct nfsd4_slot *slot;
1413 if (resp->opcnt != 1)
1414 return nfserr_sequence_pos;
1416 spin_lock(&client_lock);
1417 status = nfserr_badsession;
1418 session = find_in_sessionid_hashtbl(&seq->sessionid);
1422 status = nfserr_badslot;
1423 if (seq->slotid >= session->se_fchannel.maxreqs)
1426 slot = session->se_slots[seq->slotid];
1427 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1429 /* We do not negotiate the number of slots yet, so set the
1430 * maxslots to the session maxreqs which is used to encode
1431 * sr_highest_slotid and the sr_target_slot id to maxslots */
1432 seq->maxslots = session->se_fchannel.maxreqs;
1434 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1435 if (status == nfserr_replay_cache) {
1436 cstate->slot = slot;
1437 cstate->session = session;
1438 /* Return the cached reply status and set cstate->status
1439 * for nfsd4_proc_compound processing */
1440 status = nfsd4_replay_cache_entry(resp, seq);
1441 cstate->status = nfserr_replay_cache;
1447 /* Success! bump slot seqid */
1448 slot->sl_inuse = true;
1449 slot->sl_seqid = seq->seqid;
1450 slot->sl_cachethis = seq->cachethis;
1452 cstate->slot = slot;
1453 cstate->session = session;
1456 /* Hold a session reference until done processing the compound. */
1457 if (cstate->session)
1458 nfsd4_get_session(cstate->session);
1459 spin_unlock(&client_lock);
1460 /* Renew the clientid on success and on replay */
1461 if (cstate->session) {
1463 renew_client(session->se_client);
1464 nfs4_unlock_state();
1466 dprintk("%s: return %d\n", __func__, ntohl(status));
1471 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1472 struct nfsd4_setclientid *setclid)
1474 struct sockaddr *sa = svc_addr(rqstp);
1475 struct xdr_netobj clname = {
1476 .len = setclid->se_namelen,
1477 .data = setclid->se_name,
1479 nfs4_verifier clverifier = setclid->se_verf;
1480 unsigned int strhashval;
1481 struct nfs4_client *conf, *unconf, *new;
1483 char dname[HEXDIR_LEN];
1485 if (!check_name(clname))
1486 return nfserr_inval;
1488 status = nfs4_make_rec_clidname(dname, &clname);
1493 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1494 * We get here on a DRC miss.
1497 strhashval = clientstr_hashval(dname);
1500 conf = find_confirmed_client_by_str(dname, strhashval, false);
1502 /* RFC 3530 14.2.33 CASE 0: */
1503 status = nfserr_clid_inuse;
1504 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1505 char addr_str[INET6_ADDRSTRLEN];
1506 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1508 dprintk("NFSD: setclientid: string in use by client "
1509 "at %s\n", addr_str);
1514 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1515 * has a description of SETCLIENTID request processing consisting
1516 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1518 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1519 status = nfserr_resource;
1522 * RFC 3530 14.2.33 CASE 4:
1523 * placed first, because it is the normal case
1526 expire_client(unconf);
1527 new = create_client(clname, dname, rqstp, &clverifier);
1531 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1533 * RFC 3530 14.2.33 CASE 1:
1534 * probable callback update
1537 /* Note this is removing unconfirmed {*x***},
1538 * which is stronger than RFC recommended {vxc**}.
1539 * This has the advantage that there is at most
1540 * one {*x***} in either list at any time.
1542 expire_client(unconf);
1544 new = create_client(clname, dname, rqstp, &clverifier);
1547 copy_clid(new, conf);
1548 } else if (!unconf) {
1550 * RFC 3530 14.2.33 CASE 2:
1551 * probable client reboot; state will be removed if
1554 new = create_client(clname, dname, rqstp, &clverifier);
1560 * RFC 3530 14.2.33 CASE 3:
1561 * probable client reboot; state will be removed if
1564 expire_client(unconf);
1565 new = create_client(clname, dname, rqstp, &clverifier);
1570 gen_callback(new, setclid, rpc_get_scope_id(sa));
1571 add_to_unconfirmed(new, strhashval);
1572 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1573 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1574 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1577 nfs4_unlock_state();
1583 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1584 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1585 * bullets, labeled as CASE1 - CASE4 below.
1588 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1589 struct nfsd4_compound_state *cstate,
1590 struct nfsd4_setclientid_confirm *setclientid_confirm)
1592 struct sockaddr *sa = svc_addr(rqstp);
1593 struct nfs4_client *conf, *unconf;
1594 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1595 clientid_t * clid = &setclientid_confirm->sc_clientid;
1598 if (STALE_CLIENTID(clid))
1599 return nfserr_stale_clientid;
1601 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1602 * We get here on a DRC miss.
1607 conf = find_confirmed_client(clid);
1608 unconf = find_unconfirmed_client(clid);
1610 status = nfserr_clid_inuse;
1611 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1613 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1617 * section 14.2.34 of RFC 3530 has a description of
1618 * SETCLIENTID_CONFIRM request processing consisting
1619 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1621 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1623 * RFC 3530 14.2.34 CASE 1:
1626 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1627 status = nfserr_clid_inuse;
1629 atomic_set(&conf->cl_cb_set, 0);
1630 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1631 expire_client(unconf);
1635 } else if (conf && !unconf) {
1637 * RFC 3530 14.2.34 CASE 2:
1638 * probable retransmitted request; play it safe and
1641 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1642 status = nfserr_clid_inuse;
1645 } else if (!conf && unconf
1646 && same_verf(&unconf->cl_confirm, &confirm)) {
1648 * RFC 3530 14.2.34 CASE 3:
1649 * Normal case; new or rebooted client:
1651 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1652 status = nfserr_clid_inuse;
1655 clientstr_hashval(unconf->cl_recdir);
1656 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1659 nfsd4_remove_clid_dir(conf);
1660 expire_client(conf);
1662 move_to_confirmed(unconf);
1664 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1667 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1668 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1671 * RFC 3530 14.2.34 CASE 4:
1672 * Client probably hasn't noticed that we rebooted yet.
1674 status = nfserr_stale_clientid;
1676 /* check that we have hit one of the cases...*/
1677 status = nfserr_clid_inuse;
1680 nfs4_unlock_state();
1684 /* OPEN Share state helper functions */
1685 static inline struct nfs4_file *
1686 alloc_init_file(struct inode *ino)
1688 struct nfs4_file *fp;
1689 unsigned int hashval = file_hashval(ino);
1691 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1693 atomic_set(&fp->fi_ref, 1);
1694 INIT_LIST_HEAD(&fp->fi_hash);
1695 INIT_LIST_HEAD(&fp->fi_stateids);
1696 INIT_LIST_HEAD(&fp->fi_delegations);
1697 spin_lock(&recall_lock);
1698 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1699 spin_unlock(&recall_lock);
1700 fp->fi_inode = igrab(ino);
1701 fp->fi_id = current_fileid++;
1702 fp->fi_had_conflict = false;
1709 nfsd4_free_slab(struct kmem_cache **slab)
1713 kmem_cache_destroy(*slab);
1718 nfsd4_free_slabs(void)
1720 nfsd4_free_slab(&stateowner_slab);
1721 nfsd4_free_slab(&file_slab);
1722 nfsd4_free_slab(&stateid_slab);
1723 nfsd4_free_slab(&deleg_slab);
1727 nfsd4_init_slabs(void)
1729 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1730 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1731 if (stateowner_slab == NULL)
1733 file_slab = kmem_cache_create("nfsd4_files",
1734 sizeof(struct nfs4_file), 0, 0, NULL);
1735 if (file_slab == NULL)
1737 stateid_slab = kmem_cache_create("nfsd4_stateids",
1738 sizeof(struct nfs4_stateid), 0, 0, NULL);
1739 if (stateid_slab == NULL)
1741 deleg_slab = kmem_cache_create("nfsd4_delegations",
1742 sizeof(struct nfs4_delegation), 0, 0, NULL);
1743 if (deleg_slab == NULL)
1748 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1753 nfs4_free_stateowner(struct kref *kref)
1755 struct nfs4_stateowner *sop =
1756 container_of(kref, struct nfs4_stateowner, so_ref);
1757 kfree(sop->so_owner.data);
1758 kmem_cache_free(stateowner_slab, sop);
1761 static inline struct nfs4_stateowner *
1762 alloc_stateowner(struct xdr_netobj *owner)
1764 struct nfs4_stateowner *sop;
1766 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1767 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1768 memcpy(sop->so_owner.data, owner->data, owner->len);
1769 sop->so_owner.len = owner->len;
1770 kref_init(&sop->so_ref);
1773 kmem_cache_free(stateowner_slab, sop);
1778 static struct nfs4_stateowner *
1779 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1780 struct nfs4_stateowner *sop;
1781 struct nfs4_replay *rp;
1782 unsigned int idhashval;
1784 if (!(sop = alloc_stateowner(&open->op_owner)))
1786 idhashval = ownerid_hashval(current_ownerid);
1787 INIT_LIST_HEAD(&sop->so_idhash);
1788 INIT_LIST_HEAD(&sop->so_strhash);
1789 INIT_LIST_HEAD(&sop->so_perclient);
1790 INIT_LIST_HEAD(&sop->so_stateids);
1791 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1792 INIT_LIST_HEAD(&sop->so_close_lru);
1794 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1795 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1796 list_add(&sop->so_perclient, &clp->cl_openowners);
1797 sop->so_is_open_owner = 1;
1798 sop->so_id = current_ownerid++;
1799 sop->so_client = clp;
1800 sop->so_seqid = open->op_seqid;
1801 sop->so_confirmed = 0;
1802 rp = &sop->so_replay;
1803 rp->rp_status = nfserr_serverfault;
1805 rp->rp_buf = rp->rp_ibuf;
1810 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1811 struct nfs4_stateowner *sop = open->op_stateowner;
1812 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1814 INIT_LIST_HEAD(&stp->st_hash);
1815 INIT_LIST_HEAD(&stp->st_perstateowner);
1816 INIT_LIST_HEAD(&stp->st_lockowners);
1817 INIT_LIST_HEAD(&stp->st_perfile);
1818 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1819 list_add(&stp->st_perstateowner, &sop->so_stateids);
1820 list_add(&stp->st_perfile, &fp->fi_stateids);
1821 stp->st_stateowner = sop;
1824 stp->st_stateid.si_boot = get_seconds();
1825 stp->st_stateid.si_stateownerid = sop->so_id;
1826 stp->st_stateid.si_fileid = fp->fi_id;
1827 stp->st_stateid.si_generation = 0;
1828 stp->st_access_bmap = 0;
1829 stp->st_deny_bmap = 0;
1830 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1831 &stp->st_access_bmap);
1832 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1833 stp->st_openstp = NULL;
1837 move_to_close_lru(struct nfs4_stateowner *sop)
1839 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1841 list_move_tail(&sop->so_close_lru, &close_lru);
1842 sop->so_time = get_seconds();
1846 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1849 return (sop->so_owner.len == owner->len) &&
1850 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1851 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1854 static struct nfs4_stateowner *
1855 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1857 struct nfs4_stateowner *so = NULL;
1859 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1860 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1866 /* search file_hashtbl[] for file */
1867 static struct nfs4_file *
1868 find_file(struct inode *ino)
1870 unsigned int hashval = file_hashval(ino);
1871 struct nfs4_file *fp;
1873 spin_lock(&recall_lock);
1874 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1875 if (fp->fi_inode == ino) {
1877 spin_unlock(&recall_lock);
1881 spin_unlock(&recall_lock);
1885 static inline int access_valid(u32 x, u32 minorversion)
1887 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1889 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1891 x &= ~NFS4_SHARE_ACCESS_MASK;
1892 if (minorversion && x) {
1893 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1895 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1897 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1904 static inline int deny_valid(u32 x)
1906 /* Note: unlike access bits, deny bits may be zero. */
1907 return x <= NFS4_SHARE_DENY_BOTH;
1911 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1912 * st_{access,deny}_bmap field of the stateid, in order to track not
1913 * only what share bits are currently in force, but also what
1914 * combinations of share bits previous opens have used. This allows us
1915 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1916 * return an error if the client attempt to downgrade to a combination
1917 * of share bits not explicable by closing some of its previous opens.
1919 * XXX: This enforcement is actually incomplete, since we don't keep
1920 * track of access/deny bit combinations; so, e.g., we allow:
1922 * OPEN allow read, deny write
1923 * OPEN allow both, deny none
1924 * DOWNGRADE allow read, deny none
1926 * which we should reject.
1929 set_access(unsigned int *access, unsigned long bmap) {
1933 for (i = 1; i < 4; i++) {
1934 if (test_bit(i, &bmap))
1940 set_deny(unsigned int *deny, unsigned long bmap) {
1944 for (i = 0; i < 4; i++) {
1945 if (test_bit(i, &bmap))
1951 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1952 unsigned int access, deny;
1954 set_access(&access, stp->st_access_bmap);
1955 set_deny(&deny, stp->st_deny_bmap);
1956 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1962 * Called to check deny when READ with all zero stateid or
1963 * WRITE with all zero or all one stateid
1966 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1968 struct inode *ino = current_fh->fh_dentry->d_inode;
1969 struct nfs4_file *fp;
1970 struct nfs4_stateid *stp;
1973 dprintk("NFSD: nfs4_share_conflict\n");
1975 fp = find_file(ino);
1978 ret = nfserr_locked;
1979 /* Search for conflicting share reservations */
1980 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1981 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1982 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1992 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1994 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1995 drop_file_write_access(filp);
1996 spin_lock(&filp->f_lock);
1997 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1998 spin_unlock(&filp->f_lock);
2003 * Spawn a thread to perform a recall on the delegation represented
2004 * by the lease (file_lock)
2006 * Called from break_lease() with lock_kernel() held.
2007 * Note: we assume break_lease will only call this *once* for any given
2011 void nfsd_break_deleg_cb(struct file_lock *fl)
2013 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2015 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2019 /* We're assuming the state code never drops its reference
2020 * without first removing the lease. Since we're in this lease
2021 * callback (and since the lease code is serialized by the kernel
2022 * lock) we know the server hasn't removed the lease yet, we know
2023 * it's safe to take a reference: */
2024 atomic_inc(&dp->dl_count);
2026 spin_lock(&recall_lock);
2027 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2028 spin_unlock(&recall_lock);
2030 /* only place dl_time is set. protected by lock_kernel*/
2031 dp->dl_time = get_seconds();
2034 * We don't want the locks code to timeout the lease for us;
2035 * we'll remove it ourself if the delegation isn't returned
2038 fl->fl_break_time = 0;
2040 dp->dl_file->fi_had_conflict = true;
2041 nfsd4_cb_recall(dp);
2045 * The file_lock is being reapd.
2047 * Called by locks_free_lock() with lock_kernel() held.
2050 void nfsd_release_deleg_cb(struct file_lock *fl)
2052 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2054 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2056 if (!(fl->fl_flags & FL_LEASE) || !dp)
2058 dp->dl_flock = NULL;
2062 * Set the delegation file_lock back pointer.
2064 * Called from setlease() with lock_kernel() held.
2067 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2069 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2071 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2078 * Called from setlease() with lock_kernel() held
2081 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2083 struct nfs4_delegation *onlistd =
2084 (struct nfs4_delegation *)onlist->fl_owner;
2085 struct nfs4_delegation *tryd =
2086 (struct nfs4_delegation *)try->fl_owner;
2088 if (onlist->fl_lmops != try->fl_lmops)
2091 return onlistd->dl_client == tryd->dl_client;
2096 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2099 return lease_modify(onlist, arg);
2104 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2105 .fl_break = nfsd_break_deleg_cb,
2106 .fl_release_private = nfsd_release_deleg_cb,
2107 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2108 .fl_mylease = nfsd_same_client_deleg_cb,
2109 .fl_change = nfsd_change_deleg_cb,
2114 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2115 struct nfsd4_open *open)
2117 clientid_t *clientid = &open->op_clientid;
2118 struct nfs4_client *clp = NULL;
2119 unsigned int strhashval;
2120 struct nfs4_stateowner *sop = NULL;
2122 if (!check_name(open->op_owner))
2123 return nfserr_inval;
2125 if (STALE_CLIENTID(&open->op_clientid))
2126 return nfserr_stale_clientid;
2128 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2129 sop = find_openstateowner_str(strhashval, open);
2130 open->op_stateowner = sop;
2132 /* Make sure the client's lease hasn't expired. */
2133 clp = find_confirmed_client(clientid);
2135 return nfserr_expired;
2138 /* When sessions are used, skip open sequenceid processing */
2139 if (nfsd4_has_session(cstate))
2141 if (!sop->so_confirmed) {
2142 /* Replace unconfirmed owners without checking for replay. */
2143 clp = sop->so_client;
2144 release_openowner(sop);
2145 open->op_stateowner = NULL;
2148 if (open->op_seqid == sop->so_seqid - 1) {
2149 if (sop->so_replay.rp_buflen)
2150 return nfserr_replay_me;
2151 /* The original OPEN failed so spectacularly
2152 * that we don't even have replay data saved!
2153 * Therefore, we have no choice but to continue
2154 * processing this OPEN; presumably, we'll
2155 * fail again for the same reason.
2157 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2160 if (open->op_seqid != sop->so_seqid)
2161 return nfserr_bad_seqid;
2163 if (open->op_stateowner == NULL) {
2164 sop = alloc_init_open_stateowner(strhashval, clp, open);
2166 return nfserr_resource;
2167 open->op_stateowner = sop;
2169 list_del_init(&sop->so_close_lru);
2170 renew_client(sop->so_client);
2174 static inline __be32
2175 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2177 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2178 return nfserr_openmode;
2183 static struct nfs4_delegation *
2184 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2186 struct nfs4_delegation *dp;
2188 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2189 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2196 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2197 struct nfs4_delegation **dp)
2200 __be32 status = nfserr_bad_stateid;
2202 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2205 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2206 RD_STATE : WR_STATE;
2207 status = nfs4_check_delegmode(*dp, flags);
2211 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2215 open->op_stateowner->so_confirmed = 1;
2220 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2222 struct nfs4_stateid *local;
2223 __be32 status = nfserr_share_denied;
2224 struct nfs4_stateowner *sop = open->op_stateowner;
2226 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2227 /* ignore lock owners */
2228 if (local->st_stateowner->so_is_open_owner == 0)
2230 /* remember if we have seen this open owner */
2231 if (local->st_stateowner == sop)
2233 /* check for conflicting share reservations */
2234 if (!test_share(local, open))
2242 static inline struct nfs4_stateid *
2243 nfs4_alloc_stateid(void)
2245 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2249 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2250 struct nfs4_delegation *dp,
2251 struct svc_fh *cur_fh, int flags)
2253 struct nfs4_stateid *stp;
2255 stp = nfs4_alloc_stateid();
2257 return nfserr_resource;
2260 get_file(dp->dl_vfs_file);
2261 stp->st_vfs_file = dp->dl_vfs_file;
2264 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2267 if (status == nfserr_dropit)
2268 status = nfserr_jukebox;
2269 kmem_cache_free(stateid_slab, stp);
2277 static inline __be32
2278 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2279 struct nfsd4_open *open)
2281 struct iattr iattr = {
2282 .ia_valid = ATTR_SIZE,
2285 if (!open->op_truncate)
2287 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2288 return nfserr_inval;
2289 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2293 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2295 struct file *filp = stp->st_vfs_file;
2296 struct inode *inode = filp->f_path.dentry->d_inode;
2297 unsigned int share_access, new_writer;
2300 set_access(&share_access, stp->st_access_bmap);
2301 new_writer = (~share_access) & open->op_share_access
2302 & NFS4_SHARE_ACCESS_WRITE;
2305 int err = get_write_access(inode);
2307 return nfserrno(err);
2308 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2310 return nfserrno(err);
2311 file_take_write(filp);
2313 status = nfsd4_truncate(rqstp, cur_fh, open);
2316 put_write_access(inode);
2319 /* remember the open */
2320 filp->f_mode |= open->op_share_access;
2321 __set_bit(open->op_share_access, &stp->st_access_bmap);
2322 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2329 nfs4_set_claim_prev(struct nfsd4_open *open)
2331 open->op_stateowner->so_confirmed = 1;
2332 open->op_stateowner->so_client->cl_firststate = 1;
2336 * Attempt to hand out a delegation.
2339 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2341 struct nfs4_delegation *dp;
2342 struct nfs4_stateowner *sop = stp->st_stateowner;
2343 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2344 struct file_lock fl, *flp = &fl;
2345 int status, flag = 0;
2347 flag = NFS4_OPEN_DELEGATE_NONE;
2348 open->op_recall = 0;
2349 switch (open->op_claim_type) {
2350 case NFS4_OPEN_CLAIM_PREVIOUS:
2352 open->op_recall = 1;
2353 flag = open->op_delegate_type;
2354 if (flag == NFS4_OPEN_DELEGATE_NONE)
2357 case NFS4_OPEN_CLAIM_NULL:
2358 /* Let's not give out any delegations till everyone's
2359 * had the chance to reclaim theirs.... */
2360 if (locks_in_grace())
2362 if (!cb_up || !sop->so_confirmed)
2364 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2365 flag = NFS4_OPEN_DELEGATE_WRITE;
2367 flag = NFS4_OPEN_DELEGATE_READ;
2373 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2375 flag = NFS4_OPEN_DELEGATE_NONE;
2378 locks_init_lock(&fl);
2379 fl.fl_lmops = &nfsd_lease_mng_ops;
2380 fl.fl_flags = FL_LEASE;
2381 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2382 fl.fl_end = OFFSET_MAX;
2383 fl.fl_owner = (fl_owner_t)dp;
2384 fl.fl_file = stp->st_vfs_file;
2385 fl.fl_pid = current->tgid;
2387 /* vfs_setlease checks to see if delegation should be handed out.
2388 * the lock_manager callbacks fl_mylease and fl_change are used
2390 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2391 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2392 unhash_delegation(dp);
2393 flag = NFS4_OPEN_DELEGATE_NONE;
2397 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2399 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2400 STATEID_VAL(&dp->dl_stateid));
2402 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2403 && flag == NFS4_OPEN_DELEGATE_NONE
2404 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2405 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2406 open->op_delegate_type = flag;
2410 * called with nfs4_lock_state() held.
2413 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2415 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2416 struct nfs4_file *fp = NULL;
2417 struct inode *ino = current_fh->fh_dentry->d_inode;
2418 struct nfs4_stateid *stp = NULL;
2419 struct nfs4_delegation *dp = NULL;
2422 status = nfserr_inval;
2423 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2424 || !deny_valid(open->op_share_deny))
2427 * Lookup file; if found, lookup stateid and check open request,
2428 * and check for delegations in the process of being recalled.
2429 * If not found, create the nfs4_file struct
2431 fp = find_file(ino);
2433 if ((status = nfs4_check_open(fp, open, &stp)))
2435 status = nfs4_check_deleg(fp, open, &dp);
2439 status = nfserr_bad_stateid;
2440 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2442 status = nfserr_resource;
2443 fp = alloc_init_file(ino);
2449 * OPEN the file, or upgrade an existing OPEN.
2450 * If truncate fails, the OPEN fails.
2453 /* Stateid was found, this is an OPEN upgrade */
2454 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2457 update_stateid(&stp->st_stateid);
2459 /* Stateid was not found, this is a new OPEN */
2461 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2462 flags |= NFSD_MAY_READ;
2463 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2464 flags |= NFSD_MAY_WRITE;
2465 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2468 init_stateid(stp, fp, open);
2469 status = nfsd4_truncate(rqstp, current_fh, open);
2471 release_open_stateid(stp);
2474 if (nfsd4_has_session(&resp->cstate))
2475 update_stateid(&stp->st_stateid);
2477 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2479 if (nfsd4_has_session(&resp->cstate)) {
2480 open->op_stateowner->so_confirmed = 1;
2481 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2485 * Attempt to hand out a delegation. No error return, because the
2486 * OPEN succeeds even if we fail.
2488 nfs4_open_delegation(current_fh, open, stp);
2492 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2493 STATEID_VAL(&stp->st_stateid));
2497 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2498 nfs4_set_claim_prev(open);
2500 * To finish the open response, we just need to set the rflags.
2502 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2503 if (!open->op_stateowner->so_confirmed &&
2504 !nfsd4_has_session(&resp->cstate))
2505 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2511 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2514 struct nfs4_client *clp;
2518 dprintk("process_renew(%08x/%08x): starting\n",
2519 clid->cl_boot, clid->cl_id);
2520 status = nfserr_stale_clientid;
2521 if (STALE_CLIENTID(clid))
2523 clp = find_confirmed_client(clid);
2524 status = nfserr_expired;
2526 /* We assume the client took too long to RENEW. */
2527 dprintk("nfsd4_renew: clientid not found!\n");
2531 status = nfserr_cb_path_down;
2532 if (!list_empty(&clp->cl_delegations)
2533 && !atomic_read(&clp->cl_cb_set))
2537 nfs4_unlock_state();
2541 struct lock_manager nfsd4_manager = {
2545 nfsd4_end_grace(void)
2547 dprintk("NFSD: end of grace period\n");
2548 nfsd4_recdir_purge_old();
2549 locks_end_grace(&nfsd4_manager);
2551 * Now that every NFSv4 client has had the chance to recover and
2552 * to see the (possibly new, possibly shorter) lease time, we
2553 * can safely set the next grace time to the current lease time:
2555 nfsd4_grace = nfsd4_lease;
2559 nfs4_laundromat(void)
2561 struct nfs4_client *clp;
2562 struct nfs4_stateowner *sop;
2563 struct nfs4_delegation *dp;
2564 struct list_head *pos, *next, reaplist;
2565 time_t cutoff = get_seconds() - nfsd4_lease;
2566 time_t t, clientid_val = nfsd4_lease;
2567 time_t u, test_val = nfsd4_lease;
2571 dprintk("NFSD: laundromat service - starting\n");
2572 if (locks_in_grace())
2574 list_for_each_safe(pos, next, &client_lru) {
2575 clp = list_entry(pos, struct nfs4_client, cl_lru);
2576 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2577 t = clp->cl_time - cutoff;
2578 if (clientid_val > t)
2582 dprintk("NFSD: purging unused client (clientid %08x)\n",
2583 clp->cl_clientid.cl_id);
2584 nfsd4_remove_clid_dir(clp);
2587 INIT_LIST_HEAD(&reaplist);
2588 spin_lock(&recall_lock);
2589 list_for_each_safe(pos, next, &del_recall_lru) {
2590 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2591 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2592 u = dp->dl_time - cutoff;
2597 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2599 list_move(&dp->dl_recall_lru, &reaplist);
2601 spin_unlock(&recall_lock);
2602 list_for_each_safe(pos, next, &reaplist) {
2603 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2604 list_del_init(&dp->dl_recall_lru);
2605 unhash_delegation(dp);
2607 test_val = nfsd4_lease;
2608 list_for_each_safe(pos, next, &close_lru) {
2609 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2610 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2611 u = sop->so_time - cutoff;
2616 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2618 release_openowner(sop);
2620 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2621 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2622 nfs4_unlock_state();
2623 return clientid_val;
2626 static struct workqueue_struct *laundry_wq;
2627 static void laundromat_main(struct work_struct *);
2628 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2631 laundromat_main(struct work_struct *not_used)
2635 t = nfs4_laundromat();
2636 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2637 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2640 static struct nfs4_stateowner *
2641 search_close_lru(u32 st_id, int flags)
2643 struct nfs4_stateowner *local = NULL;
2645 if (flags & CLOSE_STATE) {
2646 list_for_each_entry(local, &close_lru, so_close_lru) {
2647 if (local->so_id == st_id)
2655 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2657 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2661 STALE_STATEID(stateid_t *stateid)
2663 if (time_after((unsigned long)boot_time,
2664 (unsigned long)stateid->si_boot)) {
2665 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2666 STATEID_VAL(stateid));
2673 EXPIRED_STATEID(stateid_t *stateid)
2675 if (time_before((unsigned long)boot_time,
2676 ((unsigned long)stateid->si_boot)) &&
2677 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
2678 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2679 STATEID_VAL(stateid));
2686 stateid_error_map(stateid_t *stateid)
2688 if (STALE_STATEID(stateid))
2689 return nfserr_stale_stateid;
2690 if (EXPIRED_STATEID(stateid))
2691 return nfserr_expired;
2693 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2694 STATEID_VAL(stateid));
2695 return nfserr_bad_stateid;
2699 access_permit_read(unsigned long access_bmap)
2701 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2702 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2703 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2707 access_permit_write(unsigned long access_bmap)
2709 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2710 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2714 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2716 __be32 status = nfserr_openmode;
2718 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2720 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2727 static inline __be32
2728 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2730 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2732 else if (locks_in_grace()) {
2733 /* Answer in remaining cases depends on existance of
2734 * conflicting state; so we must wait out the grace period. */
2735 return nfserr_grace;
2736 } else if (flags & WR_STATE)
2737 return nfs4_share_conflict(current_fh,
2738 NFS4_SHARE_DENY_WRITE);
2739 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2740 return nfs4_share_conflict(current_fh,
2741 NFS4_SHARE_DENY_READ);
2745 * Allow READ/WRITE during grace period on recovered state only for files
2746 * that are not able to provide mandatory locking.
2749 grace_disallows_io(struct inode *inode)
2751 return locks_in_grace() && mandatory_lock(inode);
2754 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2757 * When sessions are used the stateid generation number is ignored
2760 if ((flags & HAS_SESSION) && in->si_generation == 0)
2763 /* If the client sends us a stateid from the future, it's buggy: */
2764 if (in->si_generation > ref->si_generation)
2765 return nfserr_bad_stateid;
2767 * The following, however, can happen. For example, if the
2768 * client sends an open and some IO at the same time, the open
2769 * may bump si_generation while the IO is still in flight.
2770 * Thanks to hard links and renames, the client never knows what
2771 * file an open will affect. So it could avoid that situation
2772 * only by serializing all opens and IO from the same open
2773 * owner. To recover from the old_stateid error, the client
2774 * will just have to retry the IO:
2776 if (in->si_generation < ref->si_generation)
2777 return nfserr_old_stateid;
2782 static int is_delegation_stateid(stateid_t *stateid)
2784 return stateid->si_fileid == 0;
2788 * Checks for stateid operations
2791 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2792 stateid_t *stateid, int flags, struct file **filpp)
2794 struct nfs4_stateid *stp = NULL;
2795 struct nfs4_delegation *dp = NULL;
2796 struct svc_fh *current_fh = &cstate->current_fh;
2797 struct inode *ino = current_fh->fh_dentry->d_inode;
2803 if (grace_disallows_io(ino))
2804 return nfserr_grace;
2806 if (nfsd4_has_session(cstate))
2807 flags |= HAS_SESSION;
2809 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2810 return check_special_stateids(current_fh, stateid, flags);
2812 status = nfserr_stale_stateid;
2813 if (STALE_STATEID(stateid))
2816 status = nfserr_bad_stateid;
2817 if (is_delegation_stateid(stateid)) {
2818 dp = find_delegation_stateid(ino, stateid);
2820 status = stateid_error_map(stateid);
2823 status = check_stateid_generation(stateid, &dp->dl_stateid,
2827 status = nfs4_check_delegmode(dp, flags);
2830 renew_client(dp->dl_client);
2832 *filpp = dp->dl_vfs_file;
2833 } else { /* open or lock stateid */
2834 stp = find_stateid(stateid, flags);
2836 status = stateid_error_map(stateid);
2839 if (nfs4_check_fh(current_fh, stp))
2841 if (!stp->st_stateowner->so_confirmed)
2843 status = check_stateid_generation(stateid, &stp->st_stateid,
2847 status = nfs4_check_openmode(stp, flags);
2850 renew_client(stp->st_stateowner->so_client);
2852 *filpp = stp->st_vfs_file;
2862 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2863 RD_STATE : WR_STATE;
2867 * Checks for sequence id mutating operations.
2870 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2871 stateid_t *stateid, int flags,
2872 struct nfs4_stateowner **sopp,
2873 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2875 struct nfs4_stateid *stp;
2876 struct nfs4_stateowner *sop;
2877 struct svc_fh *current_fh = &cstate->current_fh;
2880 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2881 seqid, STATEID_VAL(stateid));
2886 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2887 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2888 return nfserr_bad_stateid;
2891 if (STALE_STATEID(stateid))
2892 return nfserr_stale_stateid;
2894 if (nfsd4_has_session(cstate))
2895 flags |= HAS_SESSION;
2898 * We return BAD_STATEID if filehandle doesn't match stateid,
2899 * the confirmed flag is incorrecly set, or the generation
2900 * number is incorrect.
2902 stp = find_stateid(stateid, flags);
2905 * Also, we should make sure this isn't just the result of
2908 sop = search_close_lru(stateid->si_stateownerid, flags);
2910 return stateid_error_map(stateid);
2916 *sopp = sop = stp->st_stateowner;
2919 clientid_t *lockclid = &lock->v.new.clientid;
2920 struct nfs4_client *clp = sop->so_client;
2924 lkflg = setlkflg(lock->lk_type);
2926 if (lock->lk_is_new) {
2927 if (!sop->so_is_open_owner)
2928 return nfserr_bad_stateid;
2929 if (!(flags & HAS_SESSION) &&
2930 !same_clid(&clp->cl_clientid, lockclid))
2931 return nfserr_bad_stateid;
2932 /* stp is the open stateid */
2933 status = nfs4_check_openmode(stp, lkflg);
2937 /* stp is the lock stateid */
2938 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2944 if (nfs4_check_fh(current_fh, stp)) {
2945 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2946 return nfserr_bad_stateid;
2950 * We now validate the seqid and stateid generation numbers.
2951 * For the moment, we ignore the possibility of
2952 * generation number wraparound.
2954 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
2957 if (sop->so_confirmed && flags & CONFIRM) {
2958 dprintk("NFSD: preprocess_seqid_op: expected"
2959 " unconfirmed stateowner!\n");
2960 return nfserr_bad_stateid;
2962 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2963 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2964 " confirmed yet!\n");
2965 return nfserr_bad_stateid;
2967 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
2970 renew_client(sop->so_client);
2974 if (seqid == sop->so_seqid - 1) {
2975 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2976 /* indicate replay to calling function */
2977 return nfserr_replay_me;
2979 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2980 sop->so_seqid, seqid);
2982 return nfserr_bad_seqid;
2986 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2987 struct nfsd4_open_confirm *oc)
2990 struct nfs4_stateowner *sop;
2991 struct nfs4_stateid *stp;
2993 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2994 (int)cstate->current_fh.fh_dentry->d_name.len,
2995 cstate->current_fh.fh_dentry->d_name.name);
2997 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3003 if ((status = nfs4_preprocess_seqid_op(cstate,
3004 oc->oc_seqid, &oc->oc_req_stateid,
3005 CONFIRM | OPEN_STATE,
3006 &oc->oc_stateowner, &stp, NULL)))
3009 sop = oc->oc_stateowner;
3010 sop->so_confirmed = 1;
3011 update_stateid(&stp->st_stateid);
3012 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3013 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3014 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3016 nfsd4_create_clid_dir(sop->so_client);
3018 if (oc->oc_stateowner) {
3019 nfs4_get_stateowner(oc->oc_stateowner);
3020 cstate->replay_owner = oc->oc_stateowner;
3022 nfs4_unlock_state();
3028 * unset all bits in union bitmap (bmap) that
3029 * do not exist in share (from successful OPEN_DOWNGRADE)
3032 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3035 for (i = 1; i < 4; i++) {
3036 if ((i & access) != i)
3037 __clear_bit(i, bmap);
3042 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3045 for (i = 0; i < 4; i++) {
3046 if ((i & deny) != i)
3047 __clear_bit(i, bmap);
3052 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3053 struct nfsd4_compound_state *cstate,
3054 struct nfsd4_open_downgrade *od)
3057 struct nfs4_stateid *stp;
3058 unsigned int share_access;
3060 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3061 (int)cstate->current_fh.fh_dentry->d_name.len,
3062 cstate->current_fh.fh_dentry->d_name.name);
3064 if (!access_valid(od->od_share_access, cstate->minorversion)
3065 || !deny_valid(od->od_share_deny))
3066 return nfserr_inval;
3069 if ((status = nfs4_preprocess_seqid_op(cstate,
3073 &od->od_stateowner, &stp, NULL)))
3076 status = nfserr_inval;
3077 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3078 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3079 stp->st_access_bmap, od->od_share_access);
3082 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3083 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3084 stp->st_deny_bmap, od->od_share_deny);
3087 set_access(&share_access, stp->st_access_bmap);
3088 nfs4_file_downgrade(stp->st_vfs_file,
3089 share_access & ~od->od_share_access);
3091 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3092 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3094 update_stateid(&stp->st_stateid);
3095 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3098 if (od->od_stateowner) {
3099 nfs4_get_stateowner(od->od_stateowner);
3100 cstate->replay_owner = od->od_stateowner;
3102 nfs4_unlock_state();
3107 * nfs4_unlock_state() called after encode
3110 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3111 struct nfsd4_close *close)
3114 struct nfs4_stateid *stp;
3116 dprintk("NFSD: nfsd4_close on file %.*s\n",
3117 (int)cstate->current_fh.fh_dentry->d_name.len,
3118 cstate->current_fh.fh_dentry->d_name.name);
3121 /* check close_lru for replay */
3122 if ((status = nfs4_preprocess_seqid_op(cstate,
3125 OPEN_STATE | CLOSE_STATE,
3126 &close->cl_stateowner, &stp, NULL)))
3129 update_stateid(&stp->st_stateid);
3130 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3132 /* release_stateid() calls nfsd_close() if needed */
3133 release_open_stateid(stp);
3135 /* place unused nfs4_stateowners on so_close_lru list to be
3136 * released by the laundromat service after the lease period
3137 * to enable us to handle CLOSE replay
3139 if (list_empty(&close->cl_stateowner->so_stateids))
3140 move_to_close_lru(close->cl_stateowner);
3142 if (close->cl_stateowner) {
3143 nfs4_get_stateowner(close->cl_stateowner);
3144 cstate->replay_owner = close->cl_stateowner;
3146 nfs4_unlock_state();
3151 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3152 struct nfsd4_delegreturn *dr)
3154 struct nfs4_delegation *dp;
3155 stateid_t *stateid = &dr->dr_stateid;
3156 struct inode *inode;
3160 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3162 inode = cstate->current_fh.fh_dentry->d_inode;
3164 if (nfsd4_has_session(cstate))
3165 flags |= HAS_SESSION;
3167 status = nfserr_bad_stateid;
3168 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3170 status = nfserr_stale_stateid;
3171 if (STALE_STATEID(stateid))
3173 status = nfserr_bad_stateid;
3174 if (!is_delegation_stateid(stateid))
3176 dp = find_delegation_stateid(inode, stateid);
3178 status = stateid_error_map(stateid);
3181 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3184 renew_client(dp->dl_client);
3186 unhash_delegation(dp);
3188 nfs4_unlock_state();
3195 * Lock owner state (byte-range locks)
3197 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3198 #define LOCK_HASH_BITS 8
3199 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3200 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3203 end_offset(u64 start, u64 len)
3208 return end >= start ? end: NFS4_MAX_UINT64;
3211 /* last octet in a range */
3213 last_byte_offset(u64 start, u64 len)
3219 return end > start ? end - 1: NFS4_MAX_UINT64;
3222 #define lockownerid_hashval(id) \
3223 ((id) & LOCK_HASH_MASK)
3225 static inline unsigned int
3226 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3227 struct xdr_netobj *ownername)
3229 return (file_hashval(inode) + cl_id
3230 + opaque_hashval(ownername->data, ownername->len))
3234 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3235 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3236 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3238 static struct nfs4_stateid *
3239 find_stateid(stateid_t *stid, int flags)
3241 struct nfs4_stateid *local;
3242 u32 st_id = stid->si_stateownerid;
3243 u32 f_id = stid->si_fileid;
3244 unsigned int hashval;
3246 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3247 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3248 hashval = stateid_hashval(st_id, f_id);
3249 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3250 if ((local->st_stateid.si_stateownerid == st_id) &&
3251 (local->st_stateid.si_fileid == f_id))
3256 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3257 hashval = stateid_hashval(st_id, f_id);
3258 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3259 if ((local->st_stateid.si_stateownerid == st_id) &&
3260 (local->st_stateid.si_fileid == f_id))
3267 static struct nfs4_delegation *
3268 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3270 struct nfs4_file *fp;
3271 struct nfs4_delegation *dl;
3273 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3276 fp = find_file(ino);
3279 dl = find_delegation_file(fp, stid);
3285 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3286 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3287 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3288 * locking, this prevents us from being completely protocol-compliant. The
3289 * real solution to this problem is to start using unsigned file offsets in
3290 * the VFS, but this is a very deep change!
3293 nfs4_transform_lock_offset(struct file_lock *lock)
3295 if (lock->fl_start < 0)
3296 lock->fl_start = OFFSET_MAX;
3297 if (lock->fl_end < 0)
3298 lock->fl_end = OFFSET_MAX;
3301 /* Hack!: For now, we're defining this just so we can use a pointer to it
3302 * as a unique cookie to identify our (NFSv4's) posix locks. */
3303 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3307 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3309 struct nfs4_stateowner *sop;
3312 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3313 sop = (struct nfs4_stateowner *) fl->fl_owner;
3314 hval = lockownerid_hashval(sop->so_id);
3315 kref_get(&sop->so_ref);
3317 deny->ld_clientid = sop->so_client->cl_clientid;
3319 deny->ld_sop = NULL;
3320 deny->ld_clientid.cl_boot = 0;
3321 deny->ld_clientid.cl_id = 0;
3323 deny->ld_start = fl->fl_start;
3324 deny->ld_length = NFS4_MAX_UINT64;
3325 if (fl->fl_end != NFS4_MAX_UINT64)
3326 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3327 deny->ld_type = NFS4_READ_LT;
3328 if (fl->fl_type != F_RDLCK)
3329 deny->ld_type = NFS4_WRITE_LT;
3332 static struct nfs4_stateowner *
3333 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3334 struct xdr_netobj *owner)
3336 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3337 struct nfs4_stateowner *op;
3339 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3340 if (same_owner_str(op, owner, clid))
3347 * Alloc a lock owner structure.
3348 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3351 * strhashval = lock_ownerstr_hashval
3354 static struct nfs4_stateowner *
3355 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3356 struct nfs4_stateowner *sop;
3357 struct nfs4_replay *rp;
3358 unsigned int idhashval;
3360 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3362 idhashval = lockownerid_hashval(current_ownerid);
3363 INIT_LIST_HEAD(&sop->so_idhash);
3364 INIT_LIST_HEAD(&sop->so_strhash);
3365 INIT_LIST_HEAD(&sop->so_perclient);
3366 INIT_LIST_HEAD(&sop->so_stateids);
3367 INIT_LIST_HEAD(&sop->so_perstateid);
3368 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3370 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3371 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3372 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3373 sop->so_is_open_owner = 0;
3374 sop->so_id = current_ownerid++;
3375 sop->so_client = clp;
3376 /* It is the openowner seqid that will be incremented in encode in the
3377 * case of new lockowners; so increment the lock seqid manually: */
3378 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3379 sop->so_confirmed = 1;
3380 rp = &sop->so_replay;
3381 rp->rp_status = nfserr_serverfault;
3383 rp->rp_buf = rp->rp_ibuf;
3387 static struct nfs4_stateid *
3388 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3390 struct nfs4_stateid *stp;
3391 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3393 stp = nfs4_alloc_stateid();
3396 INIT_LIST_HEAD(&stp->st_hash);
3397 INIT_LIST_HEAD(&stp->st_perfile);
3398 INIT_LIST_HEAD(&stp->st_perstateowner);
3399 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3400 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3401 list_add(&stp->st_perfile, &fp->fi_stateids);
3402 list_add(&stp->st_perstateowner, &sop->so_stateids);
3403 stp->st_stateowner = sop;
3406 stp->st_stateid.si_boot = get_seconds();
3407 stp->st_stateid.si_stateownerid = sop->so_id;
3408 stp->st_stateid.si_fileid = fp->fi_id;
3409 stp->st_stateid.si_generation = 0;
3410 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3411 stp->st_access_bmap = open_stp->st_access_bmap;
3412 stp->st_deny_bmap = open_stp->st_deny_bmap;
3413 stp->st_openstp = open_stp;
3420 check_lock_length(u64 offset, u64 length)
3422 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3423 LOFF_OVERFLOW(offset, length)));
3430 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3431 struct nfsd4_lock *lock)
3433 struct nfs4_stateowner *open_sop = NULL;
3434 struct nfs4_stateowner *lock_sop = NULL;
3435 struct nfs4_stateid *lock_stp;
3437 struct file_lock file_lock;
3438 struct file_lock conflock;
3440 unsigned int strhashval;
3444 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3445 (long long) lock->lk_offset,
3446 (long long) lock->lk_length);
3448 if (check_lock_length(lock->lk_offset, lock->lk_length))
3449 return nfserr_inval;
3451 if ((status = fh_verify(rqstp, &cstate->current_fh,
3452 S_IFREG, NFSD_MAY_LOCK))) {
3453 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3459 if (lock->lk_is_new) {
3461 * Client indicates that this is a new lockowner.
3462 * Use open owner and open stateid to create lock owner and
3465 struct nfs4_stateid *open_stp = NULL;
3466 struct nfs4_file *fp;
3468 status = nfserr_stale_clientid;
3469 if (!nfsd4_has_session(cstate) &&
3470 STALE_CLIENTID(&lock->lk_new_clientid))
3473 /* validate and update open stateid and open seqid */
3474 status = nfs4_preprocess_seqid_op(cstate,
3475 lock->lk_new_open_seqid,
3476 &lock->lk_new_open_stateid,
3478 &lock->lk_replay_owner, &open_stp,
3482 open_sop = lock->lk_replay_owner;
3483 /* create lockowner and lock stateid */
3484 fp = open_stp->st_file;
3485 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3486 open_sop->so_client->cl_clientid.cl_id,
3487 &lock->v.new.owner);
3488 /* XXX: Do we need to check for duplicate stateowners on
3489 * the same file, or should they just be allowed (and
3490 * create new stateids)? */
3491 status = nfserr_resource;
3492 lock_sop = alloc_init_lock_stateowner(strhashval,
3493 open_sop->so_client, open_stp, lock);
3494 if (lock_sop == NULL)
3496 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3497 if (lock_stp == NULL)
3500 /* lock (lock owner + lock stateid) already exists */
3501 status = nfs4_preprocess_seqid_op(cstate,
3502 lock->lk_old_lock_seqid,
3503 &lock->lk_old_lock_stateid,
3505 &lock->lk_replay_owner, &lock_stp, lock);
3508 lock_sop = lock->lk_replay_owner;
3510 /* lock->lk_replay_owner and lock_stp have been created or found */
3511 filp = lock_stp->st_vfs_file;
3513 status = nfserr_grace;
3514 if (locks_in_grace() && !lock->lk_reclaim)
3516 status = nfserr_no_grace;
3517 if (!locks_in_grace() && lock->lk_reclaim)
3520 locks_init_lock(&file_lock);
3521 switch (lock->lk_type) {
3524 file_lock.fl_type = F_RDLCK;
3528 case NFS4_WRITEW_LT:
3529 file_lock.fl_type = F_WRLCK;
3533 status = nfserr_inval;
3536 file_lock.fl_owner = (fl_owner_t)lock_sop;
3537 file_lock.fl_pid = current->tgid;
3538 file_lock.fl_file = filp;
3539 file_lock.fl_flags = FL_POSIX;
3540 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3542 file_lock.fl_start = lock->lk_offset;
3543 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3544 nfs4_transform_lock_offset(&file_lock);
3547 * Try to lock the file in the VFS.
3548 * Note: locks.c uses the BKL to protect the inode's lock list.
3551 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3553 case 0: /* success! */
3554 update_stateid(&lock_stp->st_stateid);
3555 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3559 case (EAGAIN): /* conflock holds conflicting lock */
3560 status = nfserr_denied;
3561 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3562 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3565 status = nfserr_deadlock;
3568 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3569 status = nfserr_resource;
3573 if (status && lock->lk_is_new && lock_sop)
3574 release_lockowner(lock_sop);
3575 if (lock->lk_replay_owner) {
3576 nfs4_get_stateowner(lock->lk_replay_owner);
3577 cstate->replay_owner = lock->lk_replay_owner;
3579 nfs4_unlock_state();
3584 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3585 * so we do a temporary open here just to get an open file to pass to
3586 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3589 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3594 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3597 err = vfs_test_lock(file, lock);
3606 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3607 struct nfsd4_lockt *lockt)
3609 struct inode *inode;
3610 struct file_lock file_lock;
3614 if (locks_in_grace())
3615 return nfserr_grace;
3617 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3618 return nfserr_inval;
3620 lockt->lt_stateowner = NULL;
3623 status = nfserr_stale_clientid;
3624 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3627 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3628 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3629 if (status == nfserr_symlink)
3630 status = nfserr_inval;
3634 inode = cstate->current_fh.fh_dentry->d_inode;
3635 locks_init_lock(&file_lock);
3636 switch (lockt->lt_type) {
3639 file_lock.fl_type = F_RDLCK;
3642 case NFS4_WRITEW_LT:
3643 file_lock.fl_type = F_WRLCK;
3646 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3647 status = nfserr_inval;
3651 lockt->lt_stateowner = find_lockstateowner_str(inode,
3652 &lockt->lt_clientid, &lockt->lt_owner);
3653 if (lockt->lt_stateowner)
3654 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3655 file_lock.fl_pid = current->tgid;
3656 file_lock.fl_flags = FL_POSIX;
3658 file_lock.fl_start = lockt->lt_offset;
3659 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3661 nfs4_transform_lock_offset(&file_lock);
3664 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3666 status = nfserrno(error);
3669 if (file_lock.fl_type != F_UNLCK) {
3670 status = nfserr_denied;
3671 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3674 nfs4_unlock_state();
3679 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3680 struct nfsd4_locku *locku)
3682 struct nfs4_stateid *stp;
3683 struct file *filp = NULL;
3684 struct file_lock file_lock;
3688 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3689 (long long) locku->lu_offset,
3690 (long long) locku->lu_length);
3692 if (check_lock_length(locku->lu_offset, locku->lu_length))
3693 return nfserr_inval;
3697 if ((status = nfs4_preprocess_seqid_op(cstate,
3701 &locku->lu_stateowner, &stp, NULL)))
3704 filp = stp->st_vfs_file;
3706 locks_init_lock(&file_lock);
3707 file_lock.fl_type = F_UNLCK;
3708 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3709 file_lock.fl_pid = current->tgid;
3710 file_lock.fl_file = filp;
3711 file_lock.fl_flags = FL_POSIX;
3712 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3713 file_lock.fl_start = locku->lu_offset;
3715 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3716 nfs4_transform_lock_offset(&file_lock);
3719 * Try to unlock the file in the VFS.
3721 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3723 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3727 * OK, unlock succeeded; the only thing left to do is update the stateid.
3729 update_stateid(&stp->st_stateid);
3730 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3733 if (locku->lu_stateowner) {
3734 nfs4_get_stateowner(locku->lu_stateowner);
3735 cstate->replay_owner = locku->lu_stateowner;
3737 nfs4_unlock_state();
3741 status = nfserrno(err);
3747 * 1: locks held by lockowner
3748 * 0: no locks held by lockowner
3751 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3753 struct file_lock **flpp;
3754 struct inode *inode = filp->f_path.dentry->d_inode;
3758 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3759 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3770 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3771 struct nfsd4_compound_state *cstate,
3772 struct nfsd4_release_lockowner *rlockowner)
3774 clientid_t *clid = &rlockowner->rl_clientid;
3775 struct nfs4_stateowner *sop;
3776 struct nfs4_stateid *stp;
3777 struct xdr_netobj *owner = &rlockowner->rl_owner;
3778 struct list_head matches;
3782 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3783 clid->cl_boot, clid->cl_id);
3785 /* XXX check for lease expiration */
3787 status = nfserr_stale_clientid;
3788 if (STALE_CLIENTID(clid))
3793 status = nfserr_locks_held;
3794 /* XXX: we're doing a linear search through all the lockowners.
3795 * Yipes! For now we'll just hope clients aren't really using
3796 * release_lockowner much, but eventually we have to fix these
3797 * data structures. */
3798 INIT_LIST_HEAD(&matches);
3799 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3800 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3801 if (!same_owner_str(sop, owner, clid))
3803 list_for_each_entry(stp, &sop->so_stateids,
3805 if (check_for_locks(stp->st_vfs_file, sop))
3807 /* Note: so_perclient unused for lockowners,
3808 * so it's OK to fool with here. */
3809 list_add(&sop->so_perclient, &matches);
3813 /* Clients probably won't expect us to return with some (but not all)
3814 * of the lockowner state released; so don't release any until all
3815 * have been checked. */
3817 while (!list_empty(&matches)) {
3818 sop = list_entry(matches.next, struct nfs4_stateowner,
3820 /* unhash_stateowner deletes so_perclient only
3821 * for openowners. */
3822 list_del(&sop->so_perclient);
3823 release_lockowner(sop);
3826 nfs4_unlock_state();
3830 static inline struct nfs4_client_reclaim *
3833 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3837 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3839 unsigned int strhashval = clientstr_hashval(name);
3840 struct nfs4_client *clp;
3842 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3847 * failure => all reset bets are off, nfserr_no_grace...
3850 nfs4_client_to_reclaim(const char *name)
3852 unsigned int strhashval;
3853 struct nfs4_client_reclaim *crp = NULL;
3855 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3856 crp = alloc_reclaim();
3859 strhashval = clientstr_hashval(name);
3860 INIT_LIST_HEAD(&crp->cr_strhash);
3861 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3862 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3863 reclaim_str_hashtbl_size++;
3868 nfs4_release_reclaim(void)
3870 struct nfs4_client_reclaim *crp = NULL;
3873 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3874 while (!list_empty(&reclaim_str_hashtbl[i])) {
3875 crp = list_entry(reclaim_str_hashtbl[i].next,
3876 struct nfs4_client_reclaim, cr_strhash);
3877 list_del(&crp->cr_strhash);
3879 reclaim_str_hashtbl_size--;
3882 BUG_ON(reclaim_str_hashtbl_size);
3886 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3887 static struct nfs4_client_reclaim *
3888 nfs4_find_reclaim_client(clientid_t *clid)
3890 unsigned int strhashval;
3891 struct nfs4_client *clp;
3892 struct nfs4_client_reclaim *crp = NULL;
3895 /* find clientid in conf_id_hashtbl */
3896 clp = find_confirmed_client(clid);
3900 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3901 clp->cl_name.len, clp->cl_name.data,
3904 /* find clp->cl_name in reclaim_str_hashtbl */
3905 strhashval = clientstr_hashval(clp->cl_recdir);
3906 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3907 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3915 * Called from OPEN. Look for clientid in reclaim list.
3918 nfs4_check_open_reclaim(clientid_t *clid)
3920 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3923 /* initialization to perform at module load time: */
3926 nfs4_state_init(void)
3930 status = nfsd4_init_slabs();
3933 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3934 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3935 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3936 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3937 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3938 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3940 for (i = 0; i < SESSION_HASH_SIZE; i++)
3941 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3942 for (i = 0; i < FILE_HASH_SIZE; i++) {
3943 INIT_LIST_HEAD(&file_hashtbl[i]);
3945 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3946 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3947 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3949 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3950 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3951 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3953 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3954 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3955 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3957 memset(&onestateid, ~0, sizeof(stateid_t));
3958 INIT_LIST_HEAD(&close_lru);
3959 INIT_LIST_HEAD(&client_lru);
3960 INIT_LIST_HEAD(&del_recall_lru);
3961 reclaim_str_hashtbl_size = 0;
3966 nfsd4_load_reboot_recovery_data(void)
3971 nfsd4_init_recdir(user_recovery_dirname);
3972 status = nfsd4_recdir_load();
3973 nfs4_unlock_state();
3975 printk("NFSD: Failure reading reboot recovery data\n");
3979 * Since the lifetime of a delegation isn't limited to that of an open, a
3980 * client may quite reasonably hang on to a delegation as long as it has
3981 * the inode cached. This becomes an obvious problem the first time a
3982 * client's inode cache approaches the size of the server's total memory.
3984 * For now we avoid this problem by imposing a hard limit on the number
3985 * of delegations, which varies according to the server's memory size.
3988 set_max_delegations(void)
3991 * Allow at most 4 delegations per megabyte of RAM. Quick
3992 * estimates suggest that in the worst case (where every delegation
3993 * is for a different inode), a delegation could take about 1.5K,
3994 * giving a worst case usage of about 6% of memory.
3996 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3999 /* initialization to perform when the nfsd service is started: */
4002 __nfs4_state_start(void)
4006 boot_time = get_seconds();
4007 locks_start_grace(&nfsd4_manager);
4008 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4010 ret = set_callback_cred();
4013 laundry_wq = create_singlethread_workqueue("nfsd4");
4014 if (laundry_wq == NULL)
4016 ret = nfsd4_create_callback_queue();
4018 goto out_free_laundry;
4019 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4020 set_max_delegations();
4023 destroy_workqueue(laundry_wq);
4028 nfs4_state_start(void)
4034 nfsd4_load_reboot_recovery_data();
4035 ret = __nfs4_state_start();
4043 __nfs4_state_shutdown(void)
4046 struct nfs4_client *clp = NULL;
4047 struct nfs4_delegation *dp = NULL;
4048 struct list_head *pos, *next, reaplist;
4050 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4051 while (!list_empty(&conf_id_hashtbl[i])) {
4052 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4055 while (!list_empty(&unconf_str_hashtbl[i])) {
4056 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4060 INIT_LIST_HEAD(&reaplist);
4061 spin_lock(&recall_lock);
4062 list_for_each_safe(pos, next, &del_recall_lru) {
4063 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4064 list_move(&dp->dl_recall_lru, &reaplist);
4066 spin_unlock(&recall_lock);
4067 list_for_each_safe(pos, next, &reaplist) {
4068 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4069 list_del_init(&dp->dl_recall_lru);
4070 unhash_delegation(dp);
4073 nfsd4_shutdown_recdir();
4078 nfs4_state_shutdown(void)
4080 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4081 destroy_workqueue(laundry_wq);
4082 locks_end_grace(&nfsd4_manager);
4084 nfs4_release_reclaim();
4085 __nfs4_state_shutdown();
4086 nfsd4_destroy_callback_queue();
4087 nfs4_unlock_state();
4091 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4092 * accessed when nfsd is starting.
4095 nfs4_set_recdir(char *recdir)
4097 strcpy(user_recovery_dirname, recdir);
4101 * Change the NFSv4 recovery directory to recdir.
4104 nfs4_reset_recoverydir(char *recdir)
4109 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4113 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4114 nfs4_set_recdir(recdir);
4122 nfs4_recoverydir(void)
4124 return user_recovery_dirname;