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 /* Hash tables for nfs4_clientid state */
254 #define CLIENT_HASH_BITS 4
255 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
256 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
258 #define clientid_hashval(id) \
259 ((id) & CLIENT_HASH_MASK)
260 #define clientstr_hashval(name) \
261 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
263 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
264 * used in reboot/reset lease grace period processing
266 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
267 * setclientid_confirmed info.
269 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
272 * client_lru holds client queue ordered by nfs4_client.cl_time
275 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
276 * for last close replay.
278 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
279 static int reclaim_str_hashtbl_size = 0;
280 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
281 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
282 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
283 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
284 static struct list_head client_lru;
285 static struct list_head close_lru;
287 static void unhash_generic_stateid(struct nfs4_stateid *stp)
289 list_del(&stp->st_hash);
290 list_del(&stp->st_perfile);
291 list_del(&stp->st_perstateowner);
294 static void free_generic_stateid(struct nfs4_stateid *stp)
296 put_nfs4_file(stp->st_file);
297 kmem_cache_free(stateid_slab, stp);
300 static void release_lock_stateid(struct nfs4_stateid *stp)
302 unhash_generic_stateid(stp);
303 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
304 free_generic_stateid(stp);
307 static void unhash_lockowner(struct nfs4_stateowner *sop)
309 struct nfs4_stateid *stp;
311 list_del(&sop->so_idhash);
312 list_del(&sop->so_strhash);
313 list_del(&sop->so_perstateid);
314 while (!list_empty(&sop->so_stateids)) {
315 stp = list_first_entry(&sop->so_stateids,
316 struct nfs4_stateid, st_perstateowner);
317 release_lock_stateid(stp);
321 static void release_lockowner(struct nfs4_stateowner *sop)
323 unhash_lockowner(sop);
324 nfs4_put_stateowner(sop);
328 release_stateid_lockowners(struct nfs4_stateid *open_stp)
330 struct nfs4_stateowner *lock_sop;
332 while (!list_empty(&open_stp->st_lockowners)) {
333 lock_sop = list_entry(open_stp->st_lockowners.next,
334 struct nfs4_stateowner, so_perstateid);
335 /* list_del(&open_stp->st_lockowners); */
336 BUG_ON(lock_sop->so_is_open_owner);
337 release_lockowner(lock_sop);
341 static void release_open_stateid(struct nfs4_stateid *stp)
343 unhash_generic_stateid(stp);
344 release_stateid_lockowners(stp);
345 nfsd_close(stp->st_vfs_file);
346 free_generic_stateid(stp);
349 static void unhash_openowner(struct nfs4_stateowner *sop)
351 struct nfs4_stateid *stp;
353 list_del(&sop->so_idhash);
354 list_del(&sop->so_strhash);
355 list_del(&sop->so_perclient);
356 list_del(&sop->so_perstateid); /* XXX: necessary? */
357 while (!list_empty(&sop->so_stateids)) {
358 stp = list_first_entry(&sop->so_stateids,
359 struct nfs4_stateid, st_perstateowner);
360 release_open_stateid(stp);
364 static void release_openowner(struct nfs4_stateowner *sop)
366 unhash_openowner(sop);
367 list_del(&sop->so_close_lru);
368 nfs4_put_stateowner(sop);
371 static DEFINE_SPINLOCK(sessionid_lock);
372 #define SESSION_HASH_SIZE 512
373 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
376 hash_sessionid(struct nfs4_sessionid *sessionid)
378 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
380 return sid->sequence % SESSION_HASH_SIZE;
384 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
386 u32 *ptr = (u32 *)(&sessionid->data[0]);
387 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
391 gen_sessionid(struct nfsd4_session *ses)
393 struct nfs4_client *clp = ses->se_client;
394 struct nfsd4_sessionid *sid;
396 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
397 sid->clientid = clp->cl_clientid;
398 sid->sequence = current_sessionid++;
403 * The protocol defines ca_maxresponssize_cached to include the size of
404 * the rpc header, but all we need to cache is the data starting after
405 * the end of the initial SEQUENCE operation--the rest we regenerate
406 * each time. Therefore we can advertise a ca_maxresponssize_cached
407 * value that is the number of bytes in our cache plus a few additional
408 * bytes. In order to stay on the safe side, and not promise more than
409 * we can cache, those additional bytes must be the minimum possible: 24
410 * bytes of rpc header (xid through accept state, with AUTH_NULL
411 * verifier), 12 for the compound header (with zero-length tag), and 44
412 * for the SEQUENCE op response:
414 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
417 * Give the client the number of ca_maxresponsesize_cached slots it
418 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
419 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
420 * than NFSD_MAX_SLOTS_PER_SESSION.
422 * If we run out of reserved DRC memory we should (up to a point)
423 * re-negotiate active sessions and reduce their slot usage to make
424 * rooom for new connections. For now we just fail the create session.
426 static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
428 int mem, size = fchan->maxresp_cached;
430 if (fchan->maxreqs < 1)
433 if (size < NFSD_MIN_HDR_SEQ_SZ)
434 size = NFSD_MIN_HDR_SEQ_SZ;
435 size -= NFSD_MIN_HDR_SEQ_SZ;
436 if (size > NFSD_SLOT_CACHE_SIZE)
437 size = NFSD_SLOT_CACHE_SIZE;
439 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
440 mem = fchan->maxreqs * size;
441 if (mem > NFSD_MAX_MEM_PER_SESSION) {
442 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
443 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
444 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
445 mem = fchan->maxreqs * size;
448 spin_lock(&nfsd_drc_lock);
449 /* bound the total session drc memory ussage */
450 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
451 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
452 mem = fchan->maxreqs * size;
454 nfsd_drc_mem_used += mem;
455 spin_unlock(&nfsd_drc_lock);
457 if (fchan->maxreqs == 0)
458 return nfserr_serverfault;
460 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
465 * fchan holds the client values on input, and the server values on output
466 * sv_max_mesg is the maximum payload plus one page for overhead.
468 static int init_forechannel_attrs(struct svc_rqst *rqstp,
469 struct nfsd4_channel_attrs *session_fchan,
470 struct nfsd4_channel_attrs *fchan)
473 __u32 maxcount = nfsd_serv->sv_max_mesg;
475 /* headerpadsz set to zero in encode routine */
477 /* Use the client's max request and max response size if possible */
478 if (fchan->maxreq_sz > maxcount)
479 fchan->maxreq_sz = maxcount;
480 session_fchan->maxreq_sz = fchan->maxreq_sz;
482 if (fchan->maxresp_sz > maxcount)
483 fchan->maxresp_sz = maxcount;
484 session_fchan->maxresp_sz = fchan->maxresp_sz;
486 /* Use the client's maxops if possible */
487 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
488 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
489 session_fchan->maxops = fchan->maxops;
491 /* FIXME: Error means no more DRC pages so the server should
492 * recover pages from existing sessions. For now fail session
495 status = set_forechannel_drc_size(fchan);
497 session_fchan->maxresp_cached = fchan->maxresp_cached;
498 session_fchan->maxreqs = fchan->maxreqs;
500 dprintk("%s status %d\n", __func__, status);
505 free_session_slots(struct nfsd4_session *ses)
509 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
510 kfree(ses->se_slots[i]);
514 * We don't actually need to cache the rpc and session headers, so we
515 * can allocate a little less for each slot:
517 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
519 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
523 alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
524 struct nfsd4_create_session *cses)
526 struct nfsd4_session *new, tmp;
527 struct nfsd4_slot *sp;
528 int idx, slotsize, cachesize, i;
531 memset(&tmp, 0, sizeof(tmp));
533 /* FIXME: For now, we just accept the client back channel attributes. */
534 tmp.se_bchannel = cses->back_channel;
535 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
536 &cses->fore_channel);
540 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
541 + sizeof(struct nfsd4_session) > PAGE_SIZE);
543 status = nfserr_serverfault;
544 /* allocate struct nfsd4_session and slot table pointers in one piece */
545 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
546 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
550 memcpy(new, &tmp, sizeof(*new));
552 /* allocate each struct nfsd4_slot and data cache in one piece */
553 cachesize = slot_bytes(&new->se_fchannel);
554 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
555 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
558 new->se_slots[i] = sp;
561 new->se_client = clp;
563 idx = hash_sessionid(&new->se_sessionid);
564 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
565 NFS4_MAX_SESSIONID_LEN);
567 new->se_flags = cses->flags;
568 kref_init(&new->se_ref);
569 spin_lock(&sessionid_lock);
570 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
571 list_add(&new->se_perclnt, &clp->cl_sessions);
572 spin_unlock(&sessionid_lock);
578 free_session_slots(new);
583 /* caller must hold sessionid_lock */
584 static struct nfsd4_session *
585 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
587 struct nfsd4_session *elem;
590 dump_sessionid(__func__, sessionid);
591 idx = hash_sessionid(sessionid);
592 dprintk("%s: idx is %d\n", __func__, idx);
593 /* Search in the appropriate list */
594 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
595 dump_sessionid("list traversal", &elem->se_sessionid);
596 if (!memcmp(elem->se_sessionid.data, sessionid->data,
597 NFS4_MAX_SESSIONID_LEN)) {
602 dprintk("%s: session not found\n", __func__);
606 /* caller must hold sessionid_lock */
608 unhash_session(struct nfsd4_session *ses)
610 list_del(&ses->se_hash);
611 list_del(&ses->se_perclnt);
615 release_session(struct nfsd4_session *ses)
617 spin_lock(&sessionid_lock);
619 spin_unlock(&sessionid_lock);
620 nfsd4_put_session(ses);
624 free_session(struct kref *kref)
626 struct nfsd4_session *ses;
629 ses = container_of(kref, struct nfsd4_session, se_ref);
630 spin_lock(&nfsd_drc_lock);
631 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
632 nfsd_drc_mem_used -= mem;
633 spin_unlock(&nfsd_drc_lock);
634 free_session_slots(ses);
639 renew_client(struct nfs4_client *clp)
642 * Move client to the end to the LRU list.
644 dprintk("renewing client (clientid %08x/%08x)\n",
645 clp->cl_clientid.cl_boot,
646 clp->cl_clientid.cl_id);
647 list_move_tail(&clp->cl_lru, &client_lru);
648 clp->cl_time = get_seconds();
651 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
653 STALE_CLIENTID(clientid_t *clid)
655 if (clid->cl_boot == boot_time)
657 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
658 clid->cl_boot, clid->cl_id, boot_time);
663 * XXX Should we use a slab cache ?
664 * This type of memory management is somewhat inefficient, but we use it
665 * anyway since SETCLIENTID is not a common operation.
667 static struct nfs4_client *alloc_client(struct xdr_netobj name)
669 struct nfs4_client *clp;
671 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
674 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
675 if (clp->cl_name.data == NULL) {
679 memcpy(clp->cl_name.data, name.data, name.len);
680 clp->cl_name.len = name.len;
685 free_client(struct nfs4_client *clp)
687 if (clp->cl_cred.cr_group_info)
688 put_group_info(clp->cl_cred.cr_group_info);
689 kfree(clp->cl_principal);
690 kfree(clp->cl_name.data);
695 expire_client(struct nfs4_client *clp)
697 struct nfs4_stateowner *sop;
698 struct nfs4_delegation *dp;
699 struct list_head reaplist;
701 INIT_LIST_HEAD(&reaplist);
702 spin_lock(&recall_lock);
703 while (!list_empty(&clp->cl_delegations)) {
704 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
705 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
707 list_del_init(&dp->dl_perclnt);
708 list_move(&dp->dl_recall_lru, &reaplist);
710 spin_unlock(&recall_lock);
711 while (!list_empty(&reaplist)) {
712 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
713 list_del_init(&dp->dl_recall_lru);
714 unhash_delegation(dp);
716 list_del(&clp->cl_idhash);
717 list_del(&clp->cl_strhash);
718 list_del(&clp->cl_lru);
719 while (!list_empty(&clp->cl_openowners)) {
720 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
721 release_openowner(sop);
723 while (!list_empty(&clp->cl_sessions)) {
724 struct nfsd4_session *ses;
725 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
727 release_session(ses);
729 nfsd4_set_callback_client(clp, NULL);
730 if (clp->cl_cb_conn.cb_xprt)
731 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
735 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
737 memcpy(target->cl_verifier.data, source->data,
738 sizeof(target->cl_verifier.data));
741 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
743 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
744 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
747 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
749 target->cr_uid = source->cr_uid;
750 target->cr_gid = source->cr_gid;
751 target->cr_group_info = source->cr_group_info;
752 get_group_info(target->cr_group_info);
755 static int same_name(const char *n1, const char *n2)
757 return 0 == memcmp(n1, n2, HEXDIR_LEN);
761 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
763 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
767 same_clid(clientid_t *cl1, clientid_t *cl2)
769 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
772 /* XXX what about NGROUP */
774 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
776 return cr1->cr_uid == cr2->cr_uid;
779 static void gen_clid(struct nfs4_client *clp)
781 static u32 current_clientid = 1;
783 clp->cl_clientid.cl_boot = boot_time;
784 clp->cl_clientid.cl_id = current_clientid++;
787 static void gen_confirm(struct nfs4_client *clp)
792 p = (u32 *)clp->cl_confirm.data;
793 *p++ = get_seconds();
797 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
798 struct svc_rqst *rqstp, nfs4_verifier *verf)
800 struct nfs4_client *clp;
801 struct sockaddr *sa = svc_addr(rqstp);
804 clp = alloc_client(name);
808 princ = svc_gss_principal(rqstp);
810 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
811 if (clp->cl_principal == NULL) {
817 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
818 atomic_set(&clp->cl_cb_set, 0);
819 INIT_LIST_HEAD(&clp->cl_idhash);
820 INIT_LIST_HEAD(&clp->cl_strhash);
821 INIT_LIST_HEAD(&clp->cl_openowners);
822 INIT_LIST_HEAD(&clp->cl_delegations);
823 INIT_LIST_HEAD(&clp->cl_sessions);
824 INIT_LIST_HEAD(&clp->cl_lru);
825 clear_bit(0, &clp->cl_cb_slot_busy);
826 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
827 copy_verf(clp, verf);
828 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
829 clp->cl_flavor = rqstp->rq_flavor;
830 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
836 static int check_name(struct xdr_netobj name)
840 if (name.len > NFS4_OPAQUE_LIMIT) {
841 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
848 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
850 unsigned int idhashval;
852 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
853 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
854 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
855 list_add_tail(&clp->cl_lru, &client_lru);
856 clp->cl_time = get_seconds();
860 move_to_confirmed(struct nfs4_client *clp)
862 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
863 unsigned int strhashval;
865 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
866 list_del_init(&clp->cl_strhash);
867 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
868 strhashval = clientstr_hashval(clp->cl_recdir);
869 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
873 static struct nfs4_client *
874 find_confirmed_client(clientid_t *clid)
876 struct nfs4_client *clp;
877 unsigned int idhashval = clientid_hashval(clid->cl_id);
879 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
880 if (same_clid(&clp->cl_clientid, clid))
886 static struct nfs4_client *
887 find_unconfirmed_client(clientid_t *clid)
889 struct nfs4_client *clp;
890 unsigned int idhashval = clientid_hashval(clid->cl_id);
892 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
893 if (same_clid(&clp->cl_clientid, clid))
900 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
901 * parameter. Matching is based on the fact the at least one of the
902 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
904 * FIXME: we need to unify the clientid namespaces for nfsv4.x
905 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
906 * and SET_CLIENTID{,_CONFIRM}
909 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
911 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
912 return use_exchange_id == has_exchange_flags;
915 static struct nfs4_client *
916 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
917 bool use_exchange_id)
919 struct nfs4_client *clp;
921 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
922 if (same_name(clp->cl_recdir, dname) &&
923 match_clientid_establishment(clp, use_exchange_id))
929 static struct nfs4_client *
930 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
931 bool use_exchange_id)
933 struct nfs4_client *clp;
935 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
936 if (same_name(clp->cl_recdir, dname) &&
937 match_clientid_establishment(clp, use_exchange_id))
944 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
946 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
947 unsigned short expected_family;
949 /* Currently, we only support tcp and tcp6 for the callback channel */
950 if (se->se_callback_netid_len == 3 &&
951 !memcmp(se->se_callback_netid_val, "tcp", 3))
952 expected_family = AF_INET;
953 else if (se->se_callback_netid_len == 4 &&
954 !memcmp(se->se_callback_netid_val, "tcp6", 4))
955 expected_family = AF_INET6;
959 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
960 se->se_callback_addr_len,
961 (struct sockaddr *) &cb->cb_addr,
962 sizeof(cb->cb_addr));
964 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
967 if (cb->cb_addr.ss_family == AF_INET6)
968 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
970 cb->cb_minorversion = 0;
971 cb->cb_prog = se->se_callback_prog;
972 cb->cb_ident = se->se_callback_ident;
975 cb->cb_addr.ss_family = AF_UNSPEC;
977 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
978 "will not receive delegations\n",
979 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
985 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
988 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
990 struct nfsd4_slot *slot = resp->cstate.slot;
993 dprintk("--> %s slot %p\n", __func__, slot);
995 slot->sl_opcnt = resp->opcnt;
996 slot->sl_status = resp->cstate.status;
998 if (nfsd4_not_cached(resp)) {
999 slot->sl_datalen = 0;
1002 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1003 base = (char *)resp->cstate.datap -
1004 (char *)resp->xbuf->head[0].iov_base;
1005 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1007 WARN("%s: sessions DRC could not cache compound\n", __func__);
1012 * Encode the replay sequence operation from the slot values.
1013 * If cachethis is FALSE encode the uncached rep error on the next
1014 * operation which sets resp->p and increments resp->opcnt for
1015 * nfs4svc_encode_compoundres.
1019 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1020 struct nfsd4_compoundres *resp)
1022 struct nfsd4_op *op;
1023 struct nfsd4_slot *slot = resp->cstate.slot;
1025 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1026 resp->opcnt, resp->cstate.slot->sl_cachethis);
1028 /* Encode the replayed sequence operation */
1029 op = &args->ops[resp->opcnt - 1];
1030 nfsd4_encode_operation(resp, op);
1032 /* Return nfserr_retry_uncached_rep in next operation. */
1033 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1034 op = &args->ops[resp->opcnt++];
1035 op->status = nfserr_retry_uncached_rep;
1036 nfsd4_encode_operation(resp, op);
1042 * The sequence operation is not cached because we can use the slot and
1046 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1047 struct nfsd4_sequence *seq)
1049 struct nfsd4_slot *slot = resp->cstate.slot;
1052 dprintk("--> %s slot %p\n", __func__, slot);
1054 /* Either returns 0 or nfserr_retry_uncached */
1055 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1056 if (status == nfserr_retry_uncached_rep)
1059 /* The sequence operation has been encoded, cstate->datap set. */
1060 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1062 resp->opcnt = slot->sl_opcnt;
1063 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1064 status = slot->sl_status;
1070 * Set the exchange_id flags returned by the server.
1073 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1075 /* pNFS is not supported */
1076 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1078 /* Referrals are supported, Migration is not. */
1079 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1081 /* set the wire flags to return to client. */
1082 clid->flags = new->cl_exchange_flags;
1086 nfsd4_exchange_id(struct svc_rqst *rqstp,
1087 struct nfsd4_compound_state *cstate,
1088 struct nfsd4_exchange_id *exid)
1090 struct nfs4_client *unconf, *conf, *new;
1092 unsigned int strhashval;
1093 char dname[HEXDIR_LEN];
1094 char addr_str[INET6_ADDRSTRLEN];
1095 nfs4_verifier verf = exid->verifier;
1096 struct sockaddr *sa = svc_addr(rqstp);
1098 rpc_ntop(sa, addr_str, sizeof(addr_str));
1099 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1100 "ip_addr=%s flags %x, spa_how %d\n",
1101 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1102 addr_str, exid->flags, exid->spa_how);
1104 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1105 return nfserr_inval;
1107 /* Currently only support SP4_NONE */
1108 switch (exid->spa_how) {
1112 return nfserr_encr_alg_unsupp;
1114 BUG(); /* checked by xdr code */
1116 return nfserr_serverfault; /* no excuse :-/ */
1119 status = nfs4_make_rec_clidname(dname, &exid->clname);
1124 strhashval = clientstr_hashval(dname);
1129 conf = find_confirmed_client_by_str(dname, strhashval, true);
1131 if (!same_verf(&verf, &conf->cl_verifier)) {
1132 /* 18.35.4 case 8 */
1133 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1134 status = nfserr_not_same;
1137 /* Client reboot: destroy old state */
1138 expire_client(conf);
1141 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1142 /* 18.35.4 case 9 */
1143 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1144 status = nfserr_perm;
1147 expire_client(conf);
1151 * Set bit when the owner id and verifier map to an already
1152 * confirmed client id (18.35.3).
1154 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1157 * Falling into 18.35.4 case 2, possible router replay.
1158 * Leave confirmed record intact and return same result.
1160 copy_verf(conf, &verf);
1165 /* 18.35.4 case 7 */
1166 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1167 status = nfserr_noent;
1171 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1174 * Possible retry or client restart. Per 18.35.4 case 4,
1175 * a new unconfirmed record should be generated regardless
1176 * of whether any properties have changed.
1178 expire_client(unconf);
1183 new = create_client(exid->clname, dname, rqstp, &verf);
1185 status = nfserr_serverfault;
1190 add_to_unconfirmed(new, strhashval);
1192 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1193 exid->clientid.cl_id = new->cl_clientid.cl_id;
1196 nfsd4_set_ex_flags(new, exid);
1198 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1199 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1203 nfs4_unlock_state();
1205 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1210 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1212 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1215 /* The slot is in use, and no response has been sent. */
1217 if (seqid == slot_seqid)
1218 return nfserr_jukebox;
1220 return nfserr_seq_misordered;
1223 if (likely(seqid == slot_seqid + 1))
1226 if (seqid == slot_seqid)
1227 return nfserr_replay_cache;
1229 if (seqid == 1 && (slot_seqid + 1) == 0)
1231 /* Misordered replay or misordered new request */
1232 return nfserr_seq_misordered;
1236 * Cache the create session result into the create session single DRC
1237 * slot cache by saving the xdr structure. sl_seqid has been set.
1238 * Do this for solo or embedded create session operations.
1241 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1242 struct nfsd4_clid_slot *slot, int nfserr)
1244 slot->sl_status = nfserr;
1245 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1249 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1250 struct nfsd4_clid_slot *slot)
1252 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1253 return slot->sl_status;
1257 nfsd4_create_session(struct svc_rqst *rqstp,
1258 struct nfsd4_compound_state *cstate,
1259 struct nfsd4_create_session *cr_ses)
1261 struct sockaddr *sa = svc_addr(rqstp);
1262 struct nfs4_client *conf, *unconf;
1263 struct nfsd4_clid_slot *cs_slot = NULL;
1267 unconf = find_unconfirmed_client(&cr_ses->clientid);
1268 conf = find_confirmed_client(&cr_ses->clientid);
1271 cs_slot = &conf->cl_cs_slot;
1272 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1273 if (status == nfserr_replay_cache) {
1274 dprintk("Got a create_session replay! seqid= %d\n",
1276 /* Return the cached reply status */
1277 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1279 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1280 status = nfserr_seq_misordered;
1281 dprintk("Sequence misordered!\n");
1282 dprintk("Expected seqid= %d but got seqid= %d\n",
1283 cs_slot->sl_seqid, cr_ses->seqid);
1286 cs_slot->sl_seqid++;
1287 } else if (unconf) {
1288 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1289 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1290 status = nfserr_clid_inuse;
1294 cs_slot = &unconf->cl_cs_slot;
1295 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1297 /* an unconfirmed replay returns misordered */
1298 status = nfserr_seq_misordered;
1302 cs_slot->sl_seqid++; /* from 0 to 1 */
1303 move_to_confirmed(unconf);
1305 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1306 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1307 svc_xprt_get(rqstp->rq_xprt);
1309 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1311 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1312 unconf->cl_cb_conn.cb_minorversion =
1313 cstate->minorversion;
1314 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1315 unconf->cl_cb_seq_nr = 1;
1316 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1320 status = nfserr_stale_clientid;
1325 * We do not support RDMA or persistent sessions
1327 cr_ses->flags &= ~SESSION4_PERSIST;
1328 cr_ses->flags &= ~SESSION4_RDMA;
1330 status = alloc_init_session(rqstp, conf, cr_ses);
1334 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1335 NFS4_MAX_SESSIONID_LEN);
1336 cr_ses->seqid = cs_slot->sl_seqid;
1339 /* cache solo and embedded create sessions under the state lock */
1340 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1342 nfs4_unlock_state();
1343 dprintk("%s returns %d\n", __func__, ntohl(status));
1347 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1349 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1350 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1352 return argp->opcnt == resp->opcnt;
1355 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1359 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1363 nfsd4_destroy_session(struct svc_rqst *r,
1364 struct nfsd4_compound_state *cstate,
1365 struct nfsd4_destroy_session *sessionid)
1367 struct nfsd4_session *ses;
1368 u32 status = nfserr_badsession;
1371 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1372 * - Should we return nfserr_back_chan_busy if waiting for
1373 * callbacks on to-be-destroyed session?
1374 * - Do we need to clear any callback info from previous session?
1377 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1378 if (!nfsd4_last_compound_op(r))
1379 return nfserr_not_only_op;
1381 dump_sessionid(__func__, &sessionid->sessionid);
1382 spin_lock(&sessionid_lock);
1383 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1385 spin_unlock(&sessionid_lock);
1389 unhash_session(ses);
1390 spin_unlock(&sessionid_lock);
1392 /* wait for callbacks */
1393 nfsd4_set_callback_client(ses->se_client, NULL);
1394 nfsd4_put_session(ses);
1397 dprintk("%s returns %d\n", __func__, ntohl(status));
1402 nfsd4_sequence(struct svc_rqst *rqstp,
1403 struct nfsd4_compound_state *cstate,
1404 struct nfsd4_sequence *seq)
1406 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1407 struct nfsd4_session *session;
1408 struct nfsd4_slot *slot;
1411 if (resp->opcnt != 1)
1412 return nfserr_sequence_pos;
1414 spin_lock(&sessionid_lock);
1415 status = nfserr_badsession;
1416 session = find_in_sessionid_hashtbl(&seq->sessionid);
1420 status = nfserr_badslot;
1421 if (seq->slotid >= session->se_fchannel.maxreqs)
1424 slot = session->se_slots[seq->slotid];
1425 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1427 /* We do not negotiate the number of slots yet, so set the
1428 * maxslots to the session maxreqs which is used to encode
1429 * sr_highest_slotid and the sr_target_slot id to maxslots */
1430 seq->maxslots = session->se_fchannel.maxreqs;
1432 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1433 if (status == nfserr_replay_cache) {
1434 cstate->slot = slot;
1435 cstate->session = session;
1436 /* Return the cached reply status and set cstate->status
1437 * for nfsd4_proc_compound processing */
1438 status = nfsd4_replay_cache_entry(resp, seq);
1439 cstate->status = nfserr_replay_cache;
1445 /* Success! bump slot seqid */
1446 slot->sl_inuse = true;
1447 slot->sl_seqid = seq->seqid;
1448 slot->sl_cachethis = seq->cachethis;
1450 cstate->slot = slot;
1451 cstate->session = session;
1454 /* Hold a session reference until done processing the compound. */
1455 if (cstate->session)
1456 nfsd4_get_session(cstate->session);
1457 spin_unlock(&sessionid_lock);
1458 /* Renew the clientid on success and on replay */
1459 if (cstate->session) {
1461 renew_client(session->se_client);
1462 nfs4_unlock_state();
1464 dprintk("%s: return %d\n", __func__, ntohl(status));
1469 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1470 struct nfsd4_setclientid *setclid)
1472 struct sockaddr *sa = svc_addr(rqstp);
1473 struct xdr_netobj clname = {
1474 .len = setclid->se_namelen,
1475 .data = setclid->se_name,
1477 nfs4_verifier clverifier = setclid->se_verf;
1478 unsigned int strhashval;
1479 struct nfs4_client *conf, *unconf, *new;
1481 char dname[HEXDIR_LEN];
1483 if (!check_name(clname))
1484 return nfserr_inval;
1486 status = nfs4_make_rec_clidname(dname, &clname);
1491 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1492 * We get here on a DRC miss.
1495 strhashval = clientstr_hashval(dname);
1498 conf = find_confirmed_client_by_str(dname, strhashval, false);
1500 /* RFC 3530 14.2.33 CASE 0: */
1501 status = nfserr_clid_inuse;
1502 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1503 char addr_str[INET6_ADDRSTRLEN];
1504 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1506 dprintk("NFSD: setclientid: string in use by client "
1507 "at %s\n", addr_str);
1512 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1513 * has a description of SETCLIENTID request processing consisting
1514 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1516 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1517 status = nfserr_resource;
1520 * RFC 3530 14.2.33 CASE 4:
1521 * placed first, because it is the normal case
1524 expire_client(unconf);
1525 new = create_client(clname, dname, rqstp, &clverifier);
1529 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1531 * RFC 3530 14.2.33 CASE 1:
1532 * probable callback update
1535 /* Note this is removing unconfirmed {*x***},
1536 * which is stronger than RFC recommended {vxc**}.
1537 * This has the advantage that there is at most
1538 * one {*x***} in either list at any time.
1540 expire_client(unconf);
1542 new = create_client(clname, dname, rqstp, &clverifier);
1545 copy_clid(new, conf);
1546 } else if (!unconf) {
1548 * RFC 3530 14.2.33 CASE 2:
1549 * probable client reboot; state will be removed if
1552 new = create_client(clname, dname, rqstp, &clverifier);
1558 * RFC 3530 14.2.33 CASE 3:
1559 * probable client reboot; state will be removed if
1562 expire_client(unconf);
1563 new = create_client(clname, dname, rqstp, &clverifier);
1568 gen_callback(new, setclid, rpc_get_scope_id(sa));
1569 add_to_unconfirmed(new, strhashval);
1570 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1571 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1572 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1575 nfs4_unlock_state();
1581 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1582 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1583 * bullets, labeled as CASE1 - CASE4 below.
1586 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1587 struct nfsd4_compound_state *cstate,
1588 struct nfsd4_setclientid_confirm *setclientid_confirm)
1590 struct sockaddr *sa = svc_addr(rqstp);
1591 struct nfs4_client *conf, *unconf;
1592 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1593 clientid_t * clid = &setclientid_confirm->sc_clientid;
1596 if (STALE_CLIENTID(clid))
1597 return nfserr_stale_clientid;
1599 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1600 * We get here on a DRC miss.
1605 conf = find_confirmed_client(clid);
1606 unconf = find_unconfirmed_client(clid);
1608 status = nfserr_clid_inuse;
1609 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1611 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1615 * section 14.2.34 of RFC 3530 has a description of
1616 * SETCLIENTID_CONFIRM request processing consisting
1617 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1619 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1621 * RFC 3530 14.2.34 CASE 1:
1624 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1625 status = nfserr_clid_inuse;
1627 atomic_set(&conf->cl_cb_set, 0);
1628 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1629 expire_client(unconf);
1633 } else if (conf && !unconf) {
1635 * RFC 3530 14.2.34 CASE 2:
1636 * probable retransmitted request; play it safe and
1639 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1640 status = nfserr_clid_inuse;
1643 } else if (!conf && unconf
1644 && same_verf(&unconf->cl_confirm, &confirm)) {
1646 * RFC 3530 14.2.34 CASE 3:
1647 * Normal case; new or rebooted client:
1649 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1650 status = nfserr_clid_inuse;
1653 clientstr_hashval(unconf->cl_recdir);
1654 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1657 nfsd4_remove_clid_dir(conf);
1658 expire_client(conf);
1660 move_to_confirmed(unconf);
1662 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1665 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1666 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1669 * RFC 3530 14.2.34 CASE 4:
1670 * Client probably hasn't noticed that we rebooted yet.
1672 status = nfserr_stale_clientid;
1674 /* check that we have hit one of the cases...*/
1675 status = nfserr_clid_inuse;
1678 nfs4_unlock_state();
1682 /* OPEN Share state helper functions */
1683 static inline struct nfs4_file *
1684 alloc_init_file(struct inode *ino)
1686 struct nfs4_file *fp;
1687 unsigned int hashval = file_hashval(ino);
1689 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1691 atomic_set(&fp->fi_ref, 1);
1692 INIT_LIST_HEAD(&fp->fi_hash);
1693 INIT_LIST_HEAD(&fp->fi_stateids);
1694 INIT_LIST_HEAD(&fp->fi_delegations);
1695 spin_lock(&recall_lock);
1696 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1697 spin_unlock(&recall_lock);
1698 fp->fi_inode = igrab(ino);
1699 fp->fi_id = current_fileid++;
1700 fp->fi_had_conflict = false;
1707 nfsd4_free_slab(struct kmem_cache **slab)
1711 kmem_cache_destroy(*slab);
1716 nfsd4_free_slabs(void)
1718 nfsd4_free_slab(&stateowner_slab);
1719 nfsd4_free_slab(&file_slab);
1720 nfsd4_free_slab(&stateid_slab);
1721 nfsd4_free_slab(&deleg_slab);
1725 nfsd4_init_slabs(void)
1727 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1728 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1729 if (stateowner_slab == NULL)
1731 file_slab = kmem_cache_create("nfsd4_files",
1732 sizeof(struct nfs4_file), 0, 0, NULL);
1733 if (file_slab == NULL)
1735 stateid_slab = kmem_cache_create("nfsd4_stateids",
1736 sizeof(struct nfs4_stateid), 0, 0, NULL);
1737 if (stateid_slab == NULL)
1739 deleg_slab = kmem_cache_create("nfsd4_delegations",
1740 sizeof(struct nfs4_delegation), 0, 0, NULL);
1741 if (deleg_slab == NULL)
1746 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1751 nfs4_free_stateowner(struct kref *kref)
1753 struct nfs4_stateowner *sop =
1754 container_of(kref, struct nfs4_stateowner, so_ref);
1755 kfree(sop->so_owner.data);
1756 kmem_cache_free(stateowner_slab, sop);
1759 static inline struct nfs4_stateowner *
1760 alloc_stateowner(struct xdr_netobj *owner)
1762 struct nfs4_stateowner *sop;
1764 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1765 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1766 memcpy(sop->so_owner.data, owner->data, owner->len);
1767 sop->so_owner.len = owner->len;
1768 kref_init(&sop->so_ref);
1771 kmem_cache_free(stateowner_slab, sop);
1776 static struct nfs4_stateowner *
1777 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1778 struct nfs4_stateowner *sop;
1779 struct nfs4_replay *rp;
1780 unsigned int idhashval;
1782 if (!(sop = alloc_stateowner(&open->op_owner)))
1784 idhashval = ownerid_hashval(current_ownerid);
1785 INIT_LIST_HEAD(&sop->so_idhash);
1786 INIT_LIST_HEAD(&sop->so_strhash);
1787 INIT_LIST_HEAD(&sop->so_perclient);
1788 INIT_LIST_HEAD(&sop->so_stateids);
1789 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1790 INIT_LIST_HEAD(&sop->so_close_lru);
1792 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1793 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1794 list_add(&sop->so_perclient, &clp->cl_openowners);
1795 sop->so_is_open_owner = 1;
1796 sop->so_id = current_ownerid++;
1797 sop->so_client = clp;
1798 sop->so_seqid = open->op_seqid;
1799 sop->so_confirmed = 0;
1800 rp = &sop->so_replay;
1801 rp->rp_status = nfserr_serverfault;
1803 rp->rp_buf = rp->rp_ibuf;
1808 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1809 struct nfs4_stateowner *sop = open->op_stateowner;
1810 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1812 INIT_LIST_HEAD(&stp->st_hash);
1813 INIT_LIST_HEAD(&stp->st_perstateowner);
1814 INIT_LIST_HEAD(&stp->st_lockowners);
1815 INIT_LIST_HEAD(&stp->st_perfile);
1816 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1817 list_add(&stp->st_perstateowner, &sop->so_stateids);
1818 list_add(&stp->st_perfile, &fp->fi_stateids);
1819 stp->st_stateowner = sop;
1822 stp->st_stateid.si_boot = get_seconds();
1823 stp->st_stateid.si_stateownerid = sop->so_id;
1824 stp->st_stateid.si_fileid = fp->fi_id;
1825 stp->st_stateid.si_generation = 0;
1826 stp->st_access_bmap = 0;
1827 stp->st_deny_bmap = 0;
1828 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1829 &stp->st_access_bmap);
1830 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1831 stp->st_openstp = NULL;
1835 move_to_close_lru(struct nfs4_stateowner *sop)
1837 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1839 list_move_tail(&sop->so_close_lru, &close_lru);
1840 sop->so_time = get_seconds();
1844 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1847 return (sop->so_owner.len == owner->len) &&
1848 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1849 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1852 static struct nfs4_stateowner *
1853 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1855 struct nfs4_stateowner *so = NULL;
1857 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1858 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1864 /* search file_hashtbl[] for file */
1865 static struct nfs4_file *
1866 find_file(struct inode *ino)
1868 unsigned int hashval = file_hashval(ino);
1869 struct nfs4_file *fp;
1871 spin_lock(&recall_lock);
1872 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1873 if (fp->fi_inode == ino) {
1875 spin_unlock(&recall_lock);
1879 spin_unlock(&recall_lock);
1883 static inline int access_valid(u32 x, u32 minorversion)
1885 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1887 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1889 x &= ~NFS4_SHARE_ACCESS_MASK;
1890 if (minorversion && x) {
1891 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1893 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1895 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1902 static inline int deny_valid(u32 x)
1904 /* Note: unlike access bits, deny bits may be zero. */
1905 return x <= NFS4_SHARE_DENY_BOTH;
1909 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1910 * st_{access,deny}_bmap field of the stateid, in order to track not
1911 * only what share bits are currently in force, but also what
1912 * combinations of share bits previous opens have used. This allows us
1913 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1914 * return an error if the client attempt to downgrade to a combination
1915 * of share bits not explicable by closing some of its previous opens.
1917 * XXX: This enforcement is actually incomplete, since we don't keep
1918 * track of access/deny bit combinations; so, e.g., we allow:
1920 * OPEN allow read, deny write
1921 * OPEN allow both, deny none
1922 * DOWNGRADE allow read, deny none
1924 * which we should reject.
1927 set_access(unsigned int *access, unsigned long bmap) {
1931 for (i = 1; i < 4; i++) {
1932 if (test_bit(i, &bmap))
1938 set_deny(unsigned int *deny, unsigned long bmap) {
1942 for (i = 0; i < 4; i++) {
1943 if (test_bit(i, &bmap))
1949 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1950 unsigned int access, deny;
1952 set_access(&access, stp->st_access_bmap);
1953 set_deny(&deny, stp->st_deny_bmap);
1954 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1960 * Called to check deny when READ with all zero stateid or
1961 * WRITE with all zero or all one stateid
1964 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1966 struct inode *ino = current_fh->fh_dentry->d_inode;
1967 struct nfs4_file *fp;
1968 struct nfs4_stateid *stp;
1971 dprintk("NFSD: nfs4_share_conflict\n");
1973 fp = find_file(ino);
1976 ret = nfserr_locked;
1977 /* Search for conflicting share reservations */
1978 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1979 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1980 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1990 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1992 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1993 drop_file_write_access(filp);
1994 spin_lock(&filp->f_lock);
1995 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1996 spin_unlock(&filp->f_lock);
2001 * Spawn a thread to perform a recall on the delegation represented
2002 * by the lease (file_lock)
2004 * Called from break_lease() with lock_kernel() held.
2005 * Note: we assume break_lease will only call this *once* for any given
2009 void nfsd_break_deleg_cb(struct file_lock *fl)
2011 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2013 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2017 /* We're assuming the state code never drops its reference
2018 * without first removing the lease. Since we're in this lease
2019 * callback (and since the lease code is serialized by the kernel
2020 * lock) we know the server hasn't removed the lease yet, we know
2021 * it's safe to take a reference: */
2022 atomic_inc(&dp->dl_count);
2024 spin_lock(&recall_lock);
2025 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2026 spin_unlock(&recall_lock);
2028 /* only place dl_time is set. protected by lock_kernel*/
2029 dp->dl_time = get_seconds();
2032 * We don't want the locks code to timeout the lease for us;
2033 * we'll remove it ourself if the delegation isn't returned
2036 fl->fl_break_time = 0;
2038 dp->dl_file->fi_had_conflict = true;
2039 nfsd4_cb_recall(dp);
2043 * The file_lock is being reapd.
2045 * Called by locks_free_lock() with lock_kernel() held.
2048 void nfsd_release_deleg_cb(struct file_lock *fl)
2050 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2052 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2054 if (!(fl->fl_flags & FL_LEASE) || !dp)
2056 dp->dl_flock = NULL;
2060 * Set the delegation file_lock back pointer.
2062 * Called from setlease() with lock_kernel() held.
2065 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2067 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2069 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2076 * Called from setlease() with lock_kernel() held
2079 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2081 struct nfs4_delegation *onlistd =
2082 (struct nfs4_delegation *)onlist->fl_owner;
2083 struct nfs4_delegation *tryd =
2084 (struct nfs4_delegation *)try->fl_owner;
2086 if (onlist->fl_lmops != try->fl_lmops)
2089 return onlistd->dl_client == tryd->dl_client;
2094 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2097 return lease_modify(onlist, arg);
2102 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2103 .fl_break = nfsd_break_deleg_cb,
2104 .fl_release_private = nfsd_release_deleg_cb,
2105 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2106 .fl_mylease = nfsd_same_client_deleg_cb,
2107 .fl_change = nfsd_change_deleg_cb,
2112 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2113 struct nfsd4_open *open)
2115 clientid_t *clientid = &open->op_clientid;
2116 struct nfs4_client *clp = NULL;
2117 unsigned int strhashval;
2118 struct nfs4_stateowner *sop = NULL;
2120 if (!check_name(open->op_owner))
2121 return nfserr_inval;
2123 if (STALE_CLIENTID(&open->op_clientid))
2124 return nfserr_stale_clientid;
2126 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2127 sop = find_openstateowner_str(strhashval, open);
2128 open->op_stateowner = sop;
2130 /* Make sure the client's lease hasn't expired. */
2131 clp = find_confirmed_client(clientid);
2133 return nfserr_expired;
2136 /* When sessions are used, skip open sequenceid processing */
2137 if (nfsd4_has_session(cstate))
2139 if (!sop->so_confirmed) {
2140 /* Replace unconfirmed owners without checking for replay. */
2141 clp = sop->so_client;
2142 release_openowner(sop);
2143 open->op_stateowner = NULL;
2146 if (open->op_seqid == sop->so_seqid - 1) {
2147 if (sop->so_replay.rp_buflen)
2148 return nfserr_replay_me;
2149 /* The original OPEN failed so spectacularly
2150 * that we don't even have replay data saved!
2151 * Therefore, we have no choice but to continue
2152 * processing this OPEN; presumably, we'll
2153 * fail again for the same reason.
2155 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2158 if (open->op_seqid != sop->so_seqid)
2159 return nfserr_bad_seqid;
2161 if (open->op_stateowner == NULL) {
2162 sop = alloc_init_open_stateowner(strhashval, clp, open);
2164 return nfserr_resource;
2165 open->op_stateowner = sop;
2167 list_del_init(&sop->so_close_lru);
2168 renew_client(sop->so_client);
2172 static inline __be32
2173 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2175 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2176 return nfserr_openmode;
2181 static struct nfs4_delegation *
2182 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2184 struct nfs4_delegation *dp;
2186 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2187 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2194 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2195 struct nfs4_delegation **dp)
2198 __be32 status = nfserr_bad_stateid;
2200 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2203 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2204 RD_STATE : WR_STATE;
2205 status = nfs4_check_delegmode(*dp, flags);
2209 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2213 open->op_stateowner->so_confirmed = 1;
2218 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2220 struct nfs4_stateid *local;
2221 __be32 status = nfserr_share_denied;
2222 struct nfs4_stateowner *sop = open->op_stateowner;
2224 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2225 /* ignore lock owners */
2226 if (local->st_stateowner->so_is_open_owner == 0)
2228 /* remember if we have seen this open owner */
2229 if (local->st_stateowner == sop)
2231 /* check for conflicting share reservations */
2232 if (!test_share(local, open))
2240 static inline struct nfs4_stateid *
2241 nfs4_alloc_stateid(void)
2243 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2247 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2248 struct nfs4_delegation *dp,
2249 struct svc_fh *cur_fh, int flags)
2251 struct nfs4_stateid *stp;
2253 stp = nfs4_alloc_stateid();
2255 return nfserr_resource;
2258 get_file(dp->dl_vfs_file);
2259 stp->st_vfs_file = dp->dl_vfs_file;
2262 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2265 if (status == nfserr_dropit)
2266 status = nfserr_jukebox;
2267 kmem_cache_free(stateid_slab, stp);
2275 static inline __be32
2276 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2277 struct nfsd4_open *open)
2279 struct iattr iattr = {
2280 .ia_valid = ATTR_SIZE,
2283 if (!open->op_truncate)
2285 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2286 return nfserr_inval;
2287 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2291 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2293 struct file *filp = stp->st_vfs_file;
2294 struct inode *inode = filp->f_path.dentry->d_inode;
2295 unsigned int share_access, new_writer;
2298 set_access(&share_access, stp->st_access_bmap);
2299 new_writer = (~share_access) & open->op_share_access
2300 & NFS4_SHARE_ACCESS_WRITE;
2303 int err = get_write_access(inode);
2305 return nfserrno(err);
2306 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2308 return nfserrno(err);
2309 file_take_write(filp);
2311 status = nfsd4_truncate(rqstp, cur_fh, open);
2314 put_write_access(inode);
2317 /* remember the open */
2318 filp->f_mode |= open->op_share_access;
2319 __set_bit(open->op_share_access, &stp->st_access_bmap);
2320 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2327 nfs4_set_claim_prev(struct nfsd4_open *open)
2329 open->op_stateowner->so_confirmed = 1;
2330 open->op_stateowner->so_client->cl_firststate = 1;
2334 * Attempt to hand out a delegation.
2337 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2339 struct nfs4_delegation *dp;
2340 struct nfs4_stateowner *sop = stp->st_stateowner;
2341 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
2342 struct file_lock fl, *flp = &fl;
2343 int status, flag = 0;
2345 flag = NFS4_OPEN_DELEGATE_NONE;
2346 open->op_recall = 0;
2347 switch (open->op_claim_type) {
2348 case NFS4_OPEN_CLAIM_PREVIOUS:
2350 open->op_recall = 1;
2351 flag = open->op_delegate_type;
2352 if (flag == NFS4_OPEN_DELEGATE_NONE)
2355 case NFS4_OPEN_CLAIM_NULL:
2356 /* Let's not give out any delegations till everyone's
2357 * had the chance to reclaim theirs.... */
2358 if (locks_in_grace())
2360 if (!cb_up || !sop->so_confirmed)
2362 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2363 flag = NFS4_OPEN_DELEGATE_WRITE;
2365 flag = NFS4_OPEN_DELEGATE_READ;
2371 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2373 flag = NFS4_OPEN_DELEGATE_NONE;
2376 locks_init_lock(&fl);
2377 fl.fl_lmops = &nfsd_lease_mng_ops;
2378 fl.fl_flags = FL_LEASE;
2379 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2380 fl.fl_end = OFFSET_MAX;
2381 fl.fl_owner = (fl_owner_t)dp;
2382 fl.fl_file = stp->st_vfs_file;
2383 fl.fl_pid = current->tgid;
2385 /* vfs_setlease checks to see if delegation should be handed out.
2386 * the lock_manager callbacks fl_mylease and fl_change are used
2388 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2389 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2390 unhash_delegation(dp);
2391 flag = NFS4_OPEN_DELEGATE_NONE;
2395 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2397 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2398 STATEID_VAL(&dp->dl_stateid));
2400 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2401 && flag == NFS4_OPEN_DELEGATE_NONE
2402 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2403 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2404 open->op_delegate_type = flag;
2408 * called with nfs4_lock_state() held.
2411 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2413 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2414 struct nfs4_file *fp = NULL;
2415 struct inode *ino = current_fh->fh_dentry->d_inode;
2416 struct nfs4_stateid *stp = NULL;
2417 struct nfs4_delegation *dp = NULL;
2420 status = nfserr_inval;
2421 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2422 || !deny_valid(open->op_share_deny))
2425 * Lookup file; if found, lookup stateid and check open request,
2426 * and check for delegations in the process of being recalled.
2427 * If not found, create the nfs4_file struct
2429 fp = find_file(ino);
2431 if ((status = nfs4_check_open(fp, open, &stp)))
2433 status = nfs4_check_deleg(fp, open, &dp);
2437 status = nfserr_bad_stateid;
2438 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2440 status = nfserr_resource;
2441 fp = alloc_init_file(ino);
2447 * OPEN the file, or upgrade an existing OPEN.
2448 * If truncate fails, the OPEN fails.
2451 /* Stateid was found, this is an OPEN upgrade */
2452 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2455 update_stateid(&stp->st_stateid);
2457 /* Stateid was not found, this is a new OPEN */
2459 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2460 flags |= NFSD_MAY_READ;
2461 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2462 flags |= NFSD_MAY_WRITE;
2463 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2466 init_stateid(stp, fp, open);
2467 status = nfsd4_truncate(rqstp, current_fh, open);
2469 release_open_stateid(stp);
2472 if (nfsd4_has_session(&resp->cstate))
2473 update_stateid(&stp->st_stateid);
2475 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2477 if (nfsd4_has_session(&resp->cstate)) {
2478 open->op_stateowner->so_confirmed = 1;
2479 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2483 * Attempt to hand out a delegation. No error return, because the
2484 * OPEN succeeds even if we fail.
2486 nfs4_open_delegation(current_fh, open, stp);
2490 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2491 STATEID_VAL(&stp->st_stateid));
2495 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2496 nfs4_set_claim_prev(open);
2498 * To finish the open response, we just need to set the rflags.
2500 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2501 if (!open->op_stateowner->so_confirmed &&
2502 !nfsd4_has_session(&resp->cstate))
2503 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2509 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2512 struct nfs4_client *clp;
2516 dprintk("process_renew(%08x/%08x): starting\n",
2517 clid->cl_boot, clid->cl_id);
2518 status = nfserr_stale_clientid;
2519 if (STALE_CLIENTID(clid))
2521 clp = find_confirmed_client(clid);
2522 status = nfserr_expired;
2524 /* We assume the client took too long to RENEW. */
2525 dprintk("nfsd4_renew: clientid not found!\n");
2529 status = nfserr_cb_path_down;
2530 if (!list_empty(&clp->cl_delegations)
2531 && !atomic_read(&clp->cl_cb_set))
2535 nfs4_unlock_state();
2539 struct lock_manager nfsd4_manager = {
2543 nfsd4_end_grace(void)
2545 dprintk("NFSD: end of grace period\n");
2546 nfsd4_recdir_purge_old();
2547 locks_end_grace(&nfsd4_manager);
2549 * Now that every NFSv4 client has had the chance to recover and
2550 * to see the (possibly new, possibly shorter) lease time, we
2551 * can safely set the next grace time to the current lease time:
2553 nfsd4_grace = nfsd4_lease;
2557 nfs4_laundromat(void)
2559 struct nfs4_client *clp;
2560 struct nfs4_stateowner *sop;
2561 struct nfs4_delegation *dp;
2562 struct list_head *pos, *next, reaplist;
2563 time_t cutoff = get_seconds() - nfsd4_lease;
2564 time_t t, clientid_val = nfsd4_lease;
2565 time_t u, test_val = nfsd4_lease;
2569 dprintk("NFSD: laundromat service - starting\n");
2570 if (locks_in_grace())
2572 list_for_each_safe(pos, next, &client_lru) {
2573 clp = list_entry(pos, struct nfs4_client, cl_lru);
2574 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2575 t = clp->cl_time - cutoff;
2576 if (clientid_val > t)
2580 dprintk("NFSD: purging unused client (clientid %08x)\n",
2581 clp->cl_clientid.cl_id);
2582 nfsd4_remove_clid_dir(clp);
2585 INIT_LIST_HEAD(&reaplist);
2586 spin_lock(&recall_lock);
2587 list_for_each_safe(pos, next, &del_recall_lru) {
2588 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2589 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2590 u = dp->dl_time - cutoff;
2595 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2597 list_move(&dp->dl_recall_lru, &reaplist);
2599 spin_unlock(&recall_lock);
2600 list_for_each_safe(pos, next, &reaplist) {
2601 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2602 list_del_init(&dp->dl_recall_lru);
2603 unhash_delegation(dp);
2605 test_val = nfsd4_lease;
2606 list_for_each_safe(pos, next, &close_lru) {
2607 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2608 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2609 u = sop->so_time - cutoff;
2614 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2616 release_openowner(sop);
2618 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2619 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2620 nfs4_unlock_state();
2621 return clientid_val;
2624 static struct workqueue_struct *laundry_wq;
2625 static void laundromat_main(struct work_struct *);
2626 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2629 laundromat_main(struct work_struct *not_used)
2633 t = nfs4_laundromat();
2634 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2635 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2638 static struct nfs4_stateowner *
2639 search_close_lru(u32 st_id, int flags)
2641 struct nfs4_stateowner *local = NULL;
2643 if (flags & CLOSE_STATE) {
2644 list_for_each_entry(local, &close_lru, so_close_lru) {
2645 if (local->so_id == st_id)
2653 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2655 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2659 STALE_STATEID(stateid_t *stateid)
2661 if (time_after((unsigned long)boot_time,
2662 (unsigned long)stateid->si_boot)) {
2663 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2664 STATEID_VAL(stateid));
2671 EXPIRED_STATEID(stateid_t *stateid)
2673 if (time_before((unsigned long)boot_time,
2674 ((unsigned long)stateid->si_boot)) &&
2675 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
2676 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2677 STATEID_VAL(stateid));
2684 stateid_error_map(stateid_t *stateid)
2686 if (STALE_STATEID(stateid))
2687 return nfserr_stale_stateid;
2688 if (EXPIRED_STATEID(stateid))
2689 return nfserr_expired;
2691 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2692 STATEID_VAL(stateid));
2693 return nfserr_bad_stateid;
2697 access_permit_read(unsigned long access_bmap)
2699 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2700 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2701 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2705 access_permit_write(unsigned long access_bmap)
2707 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2708 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2712 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2714 __be32 status = nfserr_openmode;
2716 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2718 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2725 static inline __be32
2726 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2728 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2730 else if (locks_in_grace()) {
2731 /* Answer in remaining cases depends on existance of
2732 * conflicting state; so we must wait out the grace period. */
2733 return nfserr_grace;
2734 } else if (flags & WR_STATE)
2735 return nfs4_share_conflict(current_fh,
2736 NFS4_SHARE_DENY_WRITE);
2737 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2738 return nfs4_share_conflict(current_fh,
2739 NFS4_SHARE_DENY_READ);
2743 * Allow READ/WRITE during grace period on recovered state only for files
2744 * that are not able to provide mandatory locking.
2747 grace_disallows_io(struct inode *inode)
2749 return locks_in_grace() && mandatory_lock(inode);
2752 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2755 * When sessions are used the stateid generation number is ignored
2758 if ((flags & HAS_SESSION) && in->si_generation == 0)
2761 /* If the client sends us a stateid from the future, it's buggy: */
2762 if (in->si_generation > ref->si_generation)
2763 return nfserr_bad_stateid;
2765 * The following, however, can happen. For example, if the
2766 * client sends an open and some IO at the same time, the open
2767 * may bump si_generation while the IO is still in flight.
2768 * Thanks to hard links and renames, the client never knows what
2769 * file an open will affect. So it could avoid that situation
2770 * only by serializing all opens and IO from the same open
2771 * owner. To recover from the old_stateid error, the client
2772 * will just have to retry the IO:
2774 if (in->si_generation < ref->si_generation)
2775 return nfserr_old_stateid;
2780 static int is_delegation_stateid(stateid_t *stateid)
2782 return stateid->si_fileid == 0;
2786 * Checks for stateid operations
2789 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2790 stateid_t *stateid, int flags, struct file **filpp)
2792 struct nfs4_stateid *stp = NULL;
2793 struct nfs4_delegation *dp = NULL;
2794 struct svc_fh *current_fh = &cstate->current_fh;
2795 struct inode *ino = current_fh->fh_dentry->d_inode;
2801 if (grace_disallows_io(ino))
2802 return nfserr_grace;
2804 if (nfsd4_has_session(cstate))
2805 flags |= HAS_SESSION;
2807 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2808 return check_special_stateids(current_fh, stateid, flags);
2810 status = nfserr_stale_stateid;
2811 if (STALE_STATEID(stateid))
2814 status = nfserr_bad_stateid;
2815 if (is_delegation_stateid(stateid)) {
2816 dp = find_delegation_stateid(ino, stateid);
2818 status = stateid_error_map(stateid);
2821 status = check_stateid_generation(stateid, &dp->dl_stateid,
2825 status = nfs4_check_delegmode(dp, flags);
2828 renew_client(dp->dl_client);
2830 *filpp = dp->dl_vfs_file;
2831 } else { /* open or lock stateid */
2832 stp = find_stateid(stateid, flags);
2834 status = stateid_error_map(stateid);
2837 if (nfs4_check_fh(current_fh, stp))
2839 if (!stp->st_stateowner->so_confirmed)
2841 status = check_stateid_generation(stateid, &stp->st_stateid,
2845 status = nfs4_check_openmode(stp, flags);
2848 renew_client(stp->st_stateowner->so_client);
2850 *filpp = stp->st_vfs_file;
2860 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2861 RD_STATE : WR_STATE;
2865 * Checks for sequence id mutating operations.
2868 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2869 stateid_t *stateid, int flags,
2870 struct nfs4_stateowner **sopp,
2871 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2873 struct nfs4_stateid *stp;
2874 struct nfs4_stateowner *sop;
2875 struct svc_fh *current_fh = &cstate->current_fh;
2878 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2879 seqid, STATEID_VAL(stateid));
2884 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2885 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2886 return nfserr_bad_stateid;
2889 if (STALE_STATEID(stateid))
2890 return nfserr_stale_stateid;
2892 if (nfsd4_has_session(cstate))
2893 flags |= HAS_SESSION;
2896 * We return BAD_STATEID if filehandle doesn't match stateid,
2897 * the confirmed flag is incorrecly set, or the generation
2898 * number is incorrect.
2900 stp = find_stateid(stateid, flags);
2903 * Also, we should make sure this isn't just the result of
2906 sop = search_close_lru(stateid->si_stateownerid, flags);
2908 return stateid_error_map(stateid);
2914 *sopp = sop = stp->st_stateowner;
2917 clientid_t *lockclid = &lock->v.new.clientid;
2918 struct nfs4_client *clp = sop->so_client;
2922 lkflg = setlkflg(lock->lk_type);
2924 if (lock->lk_is_new) {
2925 if (!sop->so_is_open_owner)
2926 return nfserr_bad_stateid;
2927 if (!(flags & HAS_SESSION) &&
2928 !same_clid(&clp->cl_clientid, lockclid))
2929 return nfserr_bad_stateid;
2930 /* stp is the open stateid */
2931 status = nfs4_check_openmode(stp, lkflg);
2935 /* stp is the lock stateid */
2936 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2942 if (nfs4_check_fh(current_fh, stp)) {
2943 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2944 return nfserr_bad_stateid;
2948 * We now validate the seqid and stateid generation numbers.
2949 * For the moment, we ignore the possibility of
2950 * generation number wraparound.
2952 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
2955 if (sop->so_confirmed && flags & CONFIRM) {
2956 dprintk("NFSD: preprocess_seqid_op: expected"
2957 " unconfirmed stateowner!\n");
2958 return nfserr_bad_stateid;
2960 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2961 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2962 " confirmed yet!\n");
2963 return nfserr_bad_stateid;
2965 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
2968 renew_client(sop->so_client);
2972 if (seqid == sop->so_seqid - 1) {
2973 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2974 /* indicate replay to calling function */
2975 return nfserr_replay_me;
2977 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2978 sop->so_seqid, seqid);
2980 return nfserr_bad_seqid;
2984 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2985 struct nfsd4_open_confirm *oc)
2988 struct nfs4_stateowner *sop;
2989 struct nfs4_stateid *stp;
2991 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2992 (int)cstate->current_fh.fh_dentry->d_name.len,
2993 cstate->current_fh.fh_dentry->d_name.name);
2995 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3001 if ((status = nfs4_preprocess_seqid_op(cstate,
3002 oc->oc_seqid, &oc->oc_req_stateid,
3003 CONFIRM | OPEN_STATE,
3004 &oc->oc_stateowner, &stp, NULL)))
3007 sop = oc->oc_stateowner;
3008 sop->so_confirmed = 1;
3009 update_stateid(&stp->st_stateid);
3010 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3011 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3012 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
3014 nfsd4_create_clid_dir(sop->so_client);
3016 if (oc->oc_stateowner) {
3017 nfs4_get_stateowner(oc->oc_stateowner);
3018 cstate->replay_owner = oc->oc_stateowner;
3020 nfs4_unlock_state();
3026 * unset all bits in union bitmap (bmap) that
3027 * do not exist in share (from successful OPEN_DOWNGRADE)
3030 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3033 for (i = 1; i < 4; i++) {
3034 if ((i & access) != i)
3035 __clear_bit(i, bmap);
3040 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3043 for (i = 0; i < 4; i++) {
3044 if ((i & deny) != i)
3045 __clear_bit(i, bmap);
3050 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3051 struct nfsd4_compound_state *cstate,
3052 struct nfsd4_open_downgrade *od)
3055 struct nfs4_stateid *stp;
3056 unsigned int share_access;
3058 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3059 (int)cstate->current_fh.fh_dentry->d_name.len,
3060 cstate->current_fh.fh_dentry->d_name.name);
3062 if (!access_valid(od->od_share_access, cstate->minorversion)
3063 || !deny_valid(od->od_share_deny))
3064 return nfserr_inval;
3067 if ((status = nfs4_preprocess_seqid_op(cstate,
3071 &od->od_stateowner, &stp, NULL)))
3074 status = nfserr_inval;
3075 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3076 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3077 stp->st_access_bmap, od->od_share_access);
3080 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3081 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3082 stp->st_deny_bmap, od->od_share_deny);
3085 set_access(&share_access, stp->st_access_bmap);
3086 nfs4_file_downgrade(stp->st_vfs_file,
3087 share_access & ~od->od_share_access);
3089 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3090 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3092 update_stateid(&stp->st_stateid);
3093 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3096 if (od->od_stateowner) {
3097 nfs4_get_stateowner(od->od_stateowner);
3098 cstate->replay_owner = od->od_stateowner;
3100 nfs4_unlock_state();
3105 * nfs4_unlock_state() called after encode
3108 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3109 struct nfsd4_close *close)
3112 struct nfs4_stateid *stp;
3114 dprintk("NFSD: nfsd4_close on file %.*s\n",
3115 (int)cstate->current_fh.fh_dentry->d_name.len,
3116 cstate->current_fh.fh_dentry->d_name.name);
3119 /* check close_lru for replay */
3120 if ((status = nfs4_preprocess_seqid_op(cstate,
3123 OPEN_STATE | CLOSE_STATE,
3124 &close->cl_stateowner, &stp, NULL)))
3127 update_stateid(&stp->st_stateid);
3128 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3130 /* release_stateid() calls nfsd_close() if needed */
3131 release_open_stateid(stp);
3133 /* place unused nfs4_stateowners on so_close_lru list to be
3134 * released by the laundromat service after the lease period
3135 * to enable us to handle CLOSE replay
3137 if (list_empty(&close->cl_stateowner->so_stateids))
3138 move_to_close_lru(close->cl_stateowner);
3140 if (close->cl_stateowner) {
3141 nfs4_get_stateowner(close->cl_stateowner);
3142 cstate->replay_owner = close->cl_stateowner;
3144 nfs4_unlock_state();
3149 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3150 struct nfsd4_delegreturn *dr)
3152 struct nfs4_delegation *dp;
3153 stateid_t *stateid = &dr->dr_stateid;
3154 struct inode *inode;
3158 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3160 inode = cstate->current_fh.fh_dentry->d_inode;
3162 if (nfsd4_has_session(cstate))
3163 flags |= HAS_SESSION;
3165 status = nfserr_bad_stateid;
3166 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3168 status = nfserr_stale_stateid;
3169 if (STALE_STATEID(stateid))
3171 status = nfserr_bad_stateid;
3172 if (!is_delegation_stateid(stateid))
3174 dp = find_delegation_stateid(inode, stateid);
3176 status = stateid_error_map(stateid);
3179 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3182 renew_client(dp->dl_client);
3184 unhash_delegation(dp);
3186 nfs4_unlock_state();
3193 * Lock owner state (byte-range locks)
3195 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3196 #define LOCK_HASH_BITS 8
3197 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3198 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3201 end_offset(u64 start, u64 len)
3206 return end >= start ? end: NFS4_MAX_UINT64;
3209 /* last octet in a range */
3211 last_byte_offset(u64 start, u64 len)
3217 return end > start ? end - 1: NFS4_MAX_UINT64;
3220 #define lockownerid_hashval(id) \
3221 ((id) & LOCK_HASH_MASK)
3223 static inline unsigned int
3224 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3225 struct xdr_netobj *ownername)
3227 return (file_hashval(inode) + cl_id
3228 + opaque_hashval(ownername->data, ownername->len))
3232 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3233 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3234 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3236 static struct nfs4_stateid *
3237 find_stateid(stateid_t *stid, int flags)
3239 struct nfs4_stateid *local;
3240 u32 st_id = stid->si_stateownerid;
3241 u32 f_id = stid->si_fileid;
3242 unsigned int hashval;
3244 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3245 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3246 hashval = stateid_hashval(st_id, f_id);
3247 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3248 if ((local->st_stateid.si_stateownerid == st_id) &&
3249 (local->st_stateid.si_fileid == f_id))
3254 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3255 hashval = stateid_hashval(st_id, f_id);
3256 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3257 if ((local->st_stateid.si_stateownerid == st_id) &&
3258 (local->st_stateid.si_fileid == f_id))
3265 static struct nfs4_delegation *
3266 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3268 struct nfs4_file *fp;
3269 struct nfs4_delegation *dl;
3271 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3274 fp = find_file(ino);
3277 dl = find_delegation_file(fp, stid);
3283 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3284 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3285 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3286 * locking, this prevents us from being completely protocol-compliant. The
3287 * real solution to this problem is to start using unsigned file offsets in
3288 * the VFS, but this is a very deep change!
3291 nfs4_transform_lock_offset(struct file_lock *lock)
3293 if (lock->fl_start < 0)
3294 lock->fl_start = OFFSET_MAX;
3295 if (lock->fl_end < 0)
3296 lock->fl_end = OFFSET_MAX;
3299 /* Hack!: For now, we're defining this just so we can use a pointer to it
3300 * as a unique cookie to identify our (NFSv4's) posix locks. */
3301 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3305 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3307 struct nfs4_stateowner *sop;
3310 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3311 sop = (struct nfs4_stateowner *) fl->fl_owner;
3312 hval = lockownerid_hashval(sop->so_id);
3313 kref_get(&sop->so_ref);
3315 deny->ld_clientid = sop->so_client->cl_clientid;
3317 deny->ld_sop = NULL;
3318 deny->ld_clientid.cl_boot = 0;
3319 deny->ld_clientid.cl_id = 0;
3321 deny->ld_start = fl->fl_start;
3322 deny->ld_length = NFS4_MAX_UINT64;
3323 if (fl->fl_end != NFS4_MAX_UINT64)
3324 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3325 deny->ld_type = NFS4_READ_LT;
3326 if (fl->fl_type != F_RDLCK)
3327 deny->ld_type = NFS4_WRITE_LT;
3330 static struct nfs4_stateowner *
3331 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3332 struct xdr_netobj *owner)
3334 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3335 struct nfs4_stateowner *op;
3337 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3338 if (same_owner_str(op, owner, clid))
3345 * Alloc a lock owner structure.
3346 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3349 * strhashval = lock_ownerstr_hashval
3352 static struct nfs4_stateowner *
3353 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3354 struct nfs4_stateowner *sop;
3355 struct nfs4_replay *rp;
3356 unsigned int idhashval;
3358 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3360 idhashval = lockownerid_hashval(current_ownerid);
3361 INIT_LIST_HEAD(&sop->so_idhash);
3362 INIT_LIST_HEAD(&sop->so_strhash);
3363 INIT_LIST_HEAD(&sop->so_perclient);
3364 INIT_LIST_HEAD(&sop->so_stateids);
3365 INIT_LIST_HEAD(&sop->so_perstateid);
3366 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3368 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3369 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3370 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3371 sop->so_is_open_owner = 0;
3372 sop->so_id = current_ownerid++;
3373 sop->so_client = clp;
3374 /* It is the openowner seqid that will be incremented in encode in the
3375 * case of new lockowners; so increment the lock seqid manually: */
3376 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3377 sop->so_confirmed = 1;
3378 rp = &sop->so_replay;
3379 rp->rp_status = nfserr_serverfault;
3381 rp->rp_buf = rp->rp_ibuf;
3385 static struct nfs4_stateid *
3386 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3388 struct nfs4_stateid *stp;
3389 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3391 stp = nfs4_alloc_stateid();
3394 INIT_LIST_HEAD(&stp->st_hash);
3395 INIT_LIST_HEAD(&stp->st_perfile);
3396 INIT_LIST_HEAD(&stp->st_perstateowner);
3397 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3398 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3399 list_add(&stp->st_perfile, &fp->fi_stateids);
3400 list_add(&stp->st_perstateowner, &sop->so_stateids);
3401 stp->st_stateowner = sop;
3404 stp->st_stateid.si_boot = get_seconds();
3405 stp->st_stateid.si_stateownerid = sop->so_id;
3406 stp->st_stateid.si_fileid = fp->fi_id;
3407 stp->st_stateid.si_generation = 0;
3408 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3409 stp->st_access_bmap = open_stp->st_access_bmap;
3410 stp->st_deny_bmap = open_stp->st_deny_bmap;
3411 stp->st_openstp = open_stp;
3418 check_lock_length(u64 offset, u64 length)
3420 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3421 LOFF_OVERFLOW(offset, length)));
3428 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3429 struct nfsd4_lock *lock)
3431 struct nfs4_stateowner *open_sop = NULL;
3432 struct nfs4_stateowner *lock_sop = NULL;
3433 struct nfs4_stateid *lock_stp;
3435 struct file_lock file_lock;
3436 struct file_lock conflock;
3438 unsigned int strhashval;
3442 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3443 (long long) lock->lk_offset,
3444 (long long) lock->lk_length);
3446 if (check_lock_length(lock->lk_offset, lock->lk_length))
3447 return nfserr_inval;
3449 if ((status = fh_verify(rqstp, &cstate->current_fh,
3450 S_IFREG, NFSD_MAY_LOCK))) {
3451 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3457 if (lock->lk_is_new) {
3459 * Client indicates that this is a new lockowner.
3460 * Use open owner and open stateid to create lock owner and
3463 struct nfs4_stateid *open_stp = NULL;
3464 struct nfs4_file *fp;
3466 status = nfserr_stale_clientid;
3467 if (!nfsd4_has_session(cstate) &&
3468 STALE_CLIENTID(&lock->lk_new_clientid))
3471 /* validate and update open stateid and open seqid */
3472 status = nfs4_preprocess_seqid_op(cstate,
3473 lock->lk_new_open_seqid,
3474 &lock->lk_new_open_stateid,
3476 &lock->lk_replay_owner, &open_stp,
3480 open_sop = lock->lk_replay_owner;
3481 /* create lockowner and lock stateid */
3482 fp = open_stp->st_file;
3483 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3484 open_sop->so_client->cl_clientid.cl_id,
3485 &lock->v.new.owner);
3486 /* XXX: Do we need to check for duplicate stateowners on
3487 * the same file, or should they just be allowed (and
3488 * create new stateids)? */
3489 status = nfserr_resource;
3490 lock_sop = alloc_init_lock_stateowner(strhashval,
3491 open_sop->so_client, open_stp, lock);
3492 if (lock_sop == NULL)
3494 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3495 if (lock_stp == NULL)
3498 /* lock (lock owner + lock stateid) already exists */
3499 status = nfs4_preprocess_seqid_op(cstate,
3500 lock->lk_old_lock_seqid,
3501 &lock->lk_old_lock_stateid,
3503 &lock->lk_replay_owner, &lock_stp, lock);
3506 lock_sop = lock->lk_replay_owner;
3508 /* lock->lk_replay_owner and lock_stp have been created or found */
3509 filp = lock_stp->st_vfs_file;
3511 status = nfserr_grace;
3512 if (locks_in_grace() && !lock->lk_reclaim)
3514 status = nfserr_no_grace;
3515 if (!locks_in_grace() && lock->lk_reclaim)
3518 locks_init_lock(&file_lock);
3519 switch (lock->lk_type) {
3522 file_lock.fl_type = F_RDLCK;
3526 case NFS4_WRITEW_LT:
3527 file_lock.fl_type = F_WRLCK;
3531 status = nfserr_inval;
3534 file_lock.fl_owner = (fl_owner_t)lock_sop;
3535 file_lock.fl_pid = current->tgid;
3536 file_lock.fl_file = filp;
3537 file_lock.fl_flags = FL_POSIX;
3538 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3540 file_lock.fl_start = lock->lk_offset;
3541 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3542 nfs4_transform_lock_offset(&file_lock);
3545 * Try to lock the file in the VFS.
3546 * Note: locks.c uses the BKL to protect the inode's lock list.
3549 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3551 case 0: /* success! */
3552 update_stateid(&lock_stp->st_stateid);
3553 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3557 case (EAGAIN): /* conflock holds conflicting lock */
3558 status = nfserr_denied;
3559 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3560 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3563 status = nfserr_deadlock;
3566 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3567 status = nfserr_resource;
3571 if (status && lock->lk_is_new && lock_sop)
3572 release_lockowner(lock_sop);
3573 if (lock->lk_replay_owner) {
3574 nfs4_get_stateowner(lock->lk_replay_owner);
3575 cstate->replay_owner = lock->lk_replay_owner;
3577 nfs4_unlock_state();
3582 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3583 * so we do a temporary open here just to get an open file to pass to
3584 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3587 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3592 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3595 err = vfs_test_lock(file, lock);
3604 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3605 struct nfsd4_lockt *lockt)
3607 struct inode *inode;
3608 struct file_lock file_lock;
3612 if (locks_in_grace())
3613 return nfserr_grace;
3615 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3616 return nfserr_inval;
3618 lockt->lt_stateowner = NULL;
3621 status = nfserr_stale_clientid;
3622 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3625 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3626 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3627 if (status == nfserr_symlink)
3628 status = nfserr_inval;
3632 inode = cstate->current_fh.fh_dentry->d_inode;
3633 locks_init_lock(&file_lock);
3634 switch (lockt->lt_type) {
3637 file_lock.fl_type = F_RDLCK;
3640 case NFS4_WRITEW_LT:
3641 file_lock.fl_type = F_WRLCK;
3644 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3645 status = nfserr_inval;
3649 lockt->lt_stateowner = find_lockstateowner_str(inode,
3650 &lockt->lt_clientid, &lockt->lt_owner);
3651 if (lockt->lt_stateowner)
3652 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3653 file_lock.fl_pid = current->tgid;
3654 file_lock.fl_flags = FL_POSIX;
3656 file_lock.fl_start = lockt->lt_offset;
3657 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3659 nfs4_transform_lock_offset(&file_lock);
3662 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3664 status = nfserrno(error);
3667 if (file_lock.fl_type != F_UNLCK) {
3668 status = nfserr_denied;
3669 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3672 nfs4_unlock_state();
3677 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3678 struct nfsd4_locku *locku)
3680 struct nfs4_stateid *stp;
3681 struct file *filp = NULL;
3682 struct file_lock file_lock;
3686 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3687 (long long) locku->lu_offset,
3688 (long long) locku->lu_length);
3690 if (check_lock_length(locku->lu_offset, locku->lu_length))
3691 return nfserr_inval;
3695 if ((status = nfs4_preprocess_seqid_op(cstate,
3699 &locku->lu_stateowner, &stp, NULL)))
3702 filp = stp->st_vfs_file;
3704 locks_init_lock(&file_lock);
3705 file_lock.fl_type = F_UNLCK;
3706 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3707 file_lock.fl_pid = current->tgid;
3708 file_lock.fl_file = filp;
3709 file_lock.fl_flags = FL_POSIX;
3710 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3711 file_lock.fl_start = locku->lu_offset;
3713 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3714 nfs4_transform_lock_offset(&file_lock);
3717 * Try to unlock the file in the VFS.
3719 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3721 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3725 * OK, unlock succeeded; the only thing left to do is update the stateid.
3727 update_stateid(&stp->st_stateid);
3728 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3731 if (locku->lu_stateowner) {
3732 nfs4_get_stateowner(locku->lu_stateowner);
3733 cstate->replay_owner = locku->lu_stateowner;
3735 nfs4_unlock_state();
3739 status = nfserrno(err);
3745 * 1: locks held by lockowner
3746 * 0: no locks held by lockowner
3749 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3751 struct file_lock **flpp;
3752 struct inode *inode = filp->f_path.dentry->d_inode;
3756 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3757 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3768 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3769 struct nfsd4_compound_state *cstate,
3770 struct nfsd4_release_lockowner *rlockowner)
3772 clientid_t *clid = &rlockowner->rl_clientid;
3773 struct nfs4_stateowner *sop;
3774 struct nfs4_stateid *stp;
3775 struct xdr_netobj *owner = &rlockowner->rl_owner;
3776 struct list_head matches;
3780 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3781 clid->cl_boot, clid->cl_id);
3783 /* XXX check for lease expiration */
3785 status = nfserr_stale_clientid;
3786 if (STALE_CLIENTID(clid))
3791 status = nfserr_locks_held;
3792 /* XXX: we're doing a linear search through all the lockowners.
3793 * Yipes! For now we'll just hope clients aren't really using
3794 * release_lockowner much, but eventually we have to fix these
3795 * data structures. */
3796 INIT_LIST_HEAD(&matches);
3797 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3798 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3799 if (!same_owner_str(sop, owner, clid))
3801 list_for_each_entry(stp, &sop->so_stateids,
3803 if (check_for_locks(stp->st_vfs_file, sop))
3805 /* Note: so_perclient unused for lockowners,
3806 * so it's OK to fool with here. */
3807 list_add(&sop->so_perclient, &matches);
3811 /* Clients probably won't expect us to return with some (but not all)
3812 * of the lockowner state released; so don't release any until all
3813 * have been checked. */
3815 while (!list_empty(&matches)) {
3816 sop = list_entry(matches.next, struct nfs4_stateowner,
3818 /* unhash_stateowner deletes so_perclient only
3819 * for openowners. */
3820 list_del(&sop->so_perclient);
3821 release_lockowner(sop);
3824 nfs4_unlock_state();
3828 static inline struct nfs4_client_reclaim *
3831 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3835 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3837 unsigned int strhashval = clientstr_hashval(name);
3838 struct nfs4_client *clp;
3840 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3845 * failure => all reset bets are off, nfserr_no_grace...
3848 nfs4_client_to_reclaim(const char *name)
3850 unsigned int strhashval;
3851 struct nfs4_client_reclaim *crp = NULL;
3853 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3854 crp = alloc_reclaim();
3857 strhashval = clientstr_hashval(name);
3858 INIT_LIST_HEAD(&crp->cr_strhash);
3859 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3860 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3861 reclaim_str_hashtbl_size++;
3866 nfs4_release_reclaim(void)
3868 struct nfs4_client_reclaim *crp = NULL;
3871 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3872 while (!list_empty(&reclaim_str_hashtbl[i])) {
3873 crp = list_entry(reclaim_str_hashtbl[i].next,
3874 struct nfs4_client_reclaim, cr_strhash);
3875 list_del(&crp->cr_strhash);
3877 reclaim_str_hashtbl_size--;
3880 BUG_ON(reclaim_str_hashtbl_size);
3884 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3885 static struct nfs4_client_reclaim *
3886 nfs4_find_reclaim_client(clientid_t *clid)
3888 unsigned int strhashval;
3889 struct nfs4_client *clp;
3890 struct nfs4_client_reclaim *crp = NULL;
3893 /* find clientid in conf_id_hashtbl */
3894 clp = find_confirmed_client(clid);
3898 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3899 clp->cl_name.len, clp->cl_name.data,
3902 /* find clp->cl_name in reclaim_str_hashtbl */
3903 strhashval = clientstr_hashval(clp->cl_recdir);
3904 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3905 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3913 * Called from OPEN. Look for clientid in reclaim list.
3916 nfs4_check_open_reclaim(clientid_t *clid)
3918 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3921 /* initialization to perform at module load time: */
3924 nfs4_state_init(void)
3928 status = nfsd4_init_slabs();
3931 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3932 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3933 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3934 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3935 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3936 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3938 for (i = 0; i < SESSION_HASH_SIZE; i++)
3939 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3940 for (i = 0; i < FILE_HASH_SIZE; i++) {
3941 INIT_LIST_HEAD(&file_hashtbl[i]);
3943 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3944 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3945 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3947 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3948 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3949 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3951 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3952 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3953 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3955 memset(&onestateid, ~0, sizeof(stateid_t));
3956 INIT_LIST_HEAD(&close_lru);
3957 INIT_LIST_HEAD(&client_lru);
3958 INIT_LIST_HEAD(&del_recall_lru);
3959 reclaim_str_hashtbl_size = 0;
3964 nfsd4_load_reboot_recovery_data(void)
3969 nfsd4_init_recdir(user_recovery_dirname);
3970 status = nfsd4_recdir_load();
3971 nfs4_unlock_state();
3973 printk("NFSD: Failure reading reboot recovery data\n");
3977 * Since the lifetime of a delegation isn't limited to that of an open, a
3978 * client may quite reasonably hang on to a delegation as long as it has
3979 * the inode cached. This becomes an obvious problem the first time a
3980 * client's inode cache approaches the size of the server's total memory.
3982 * For now we avoid this problem by imposing a hard limit on the number
3983 * of delegations, which varies according to the server's memory size.
3986 set_max_delegations(void)
3989 * Allow at most 4 delegations per megabyte of RAM. Quick
3990 * estimates suggest that in the worst case (where every delegation
3991 * is for a different inode), a delegation could take about 1.5K,
3992 * giving a worst case usage of about 6% of memory.
3994 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3997 /* initialization to perform when the nfsd service is started: */
4000 __nfs4_state_start(void)
4004 boot_time = get_seconds();
4005 locks_start_grace(&nfsd4_manager);
4006 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4008 ret = set_callback_cred();
4011 laundry_wq = create_singlethread_workqueue("nfsd4");
4012 if (laundry_wq == NULL)
4014 ret = nfsd4_create_callback_queue();
4016 goto out_free_laundry;
4017 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4018 set_max_delegations();
4021 destroy_workqueue(laundry_wq);
4026 nfs4_state_start(void)
4032 nfsd4_load_reboot_recovery_data();
4033 ret = __nfs4_state_start();
4041 __nfs4_state_shutdown(void)
4044 struct nfs4_client *clp = NULL;
4045 struct nfs4_delegation *dp = NULL;
4046 struct list_head *pos, *next, reaplist;
4048 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4049 while (!list_empty(&conf_id_hashtbl[i])) {
4050 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4053 while (!list_empty(&unconf_str_hashtbl[i])) {
4054 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4058 INIT_LIST_HEAD(&reaplist);
4059 spin_lock(&recall_lock);
4060 list_for_each_safe(pos, next, &del_recall_lru) {
4061 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4062 list_move(&dp->dl_recall_lru, &reaplist);
4064 spin_unlock(&recall_lock);
4065 list_for_each_safe(pos, next, &reaplist) {
4066 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4067 list_del_init(&dp->dl_recall_lru);
4068 unhash_delegation(dp);
4071 nfsd4_shutdown_recdir();
4076 nfs4_state_shutdown(void)
4078 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4079 destroy_workqueue(laundry_wq);
4080 locks_end_grace(&nfsd4_manager);
4082 nfs4_release_reclaim();
4083 __nfs4_state_shutdown();
4084 nfsd4_destroy_callback_queue();
4085 nfs4_unlock_state();
4089 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4090 * accessed when nfsd is starting.
4093 nfs4_set_recdir(char *recdir)
4095 strcpy(user_recovery_dirname, recdir);
4099 * Change the NFSv4 recovery directory to recdir.
4102 nfs4_reset_recoverydir(char *recdir)
4107 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4111 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4112 nfs4_set_recdir(recdir);
4120 nfs4_recoverydir(void)
4122 return user_recovery_dirname;