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/namei.h>
38 #include <linux/swap.h>
39 #include <linux/sunrpc/svcauth_gss.h>
40 #include <linux/sunrpc/clnt.h>
44 #define NFSDDBG_FACILITY NFSDDBG_PROC
47 time_t nfsd4_lease = 90; /* default lease time */
48 time_t nfsd4_grace = 90;
49 static time_t boot_time;
50 static u32 current_ownerid = 1;
51 static u32 current_fileid = 1;
52 static u32 current_delegid = 1;
54 static stateid_t zerostateid; /* bits all 0 */
55 static stateid_t onestateid; /* bits all 1 */
56 static u64 current_sessionid = 1;
58 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
61 /* forward declarations */
62 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
63 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
64 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
65 static void nfs4_set_recdir(char *recdir);
69 /* Currently used for almost all code touching nfsv4 state: */
70 static DEFINE_MUTEX(client_mutex);
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
77 static DEFINE_SPINLOCK(recall_lock);
79 static struct kmem_cache *stateowner_slab = NULL;
80 static struct kmem_cache *file_slab = NULL;
81 static struct kmem_cache *stateid_slab = NULL;
82 static struct kmem_cache *deleg_slab = NULL;
87 mutex_lock(&client_mutex);
91 nfs4_unlock_state(void)
93 mutex_unlock(&client_mutex);
97 opaque_hashval(const void *ptr, int nbytes)
99 unsigned char *cptr = (unsigned char *) ptr;
109 static struct list_head del_recall_lru;
112 put_nfs4_file(struct nfs4_file *fi)
114 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
115 list_del(&fi->fi_hash);
116 spin_unlock(&recall_lock);
118 kmem_cache_free(file_slab, fi);
123 get_nfs4_file(struct nfs4_file *fi)
125 atomic_inc(&fi->fi_ref);
128 static int num_delegations;
129 unsigned int max_delegations;
132 * Open owner state (share locks)
135 /* hash tables for nfs4_stateowner */
136 #define OWNER_HASH_BITS 8
137 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140 #define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142 #define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
146 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
148 /* hash table for nfs4_file */
149 #define FILE_HASH_BITS 8
150 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
152 /* hash table for (open)nfs4_stateid */
153 #define STATEID_HASH_BITS 10
154 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157 #define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159 #define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162 static struct list_head file_hashtbl[FILE_HASH_SIZE];
163 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
165 static struct nfs4_delegation *
166 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
168 struct nfs4_delegation *dp;
169 struct nfs4_file *fp = stp->st_file;
170 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
172 dprintk("NFSD alloc_init_deleg\n");
173 if (fp->fi_had_conflict)
175 if (num_delegations > max_delegations)
177 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
181 INIT_LIST_HEAD(&dp->dl_perfile);
182 INIT_LIST_HEAD(&dp->dl_perclnt);
183 INIT_LIST_HEAD(&dp->dl_recall_lru);
188 get_file(stp->st_vfs_file);
189 dp->dl_vfs_file = stp->st_vfs_file;
191 dp->dl_ident = cb->cb_ident;
192 dp->dl_stateid.si_boot = get_seconds();
193 dp->dl_stateid.si_stateownerid = current_delegid++;
194 dp->dl_stateid.si_fileid = 0;
195 dp->dl_stateid.si_generation = 0;
196 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
198 atomic_set(&dp->dl_count, 1);
199 list_add(&dp->dl_perfile, &fp->fi_delegations);
200 list_add(&dp->dl_perclnt, &clp->cl_delegations);
201 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
206 nfs4_put_delegation(struct nfs4_delegation *dp)
208 if (atomic_dec_and_test(&dp->dl_count)) {
209 dprintk("NFSD: freeing dp %p\n",dp);
210 put_nfs4_file(dp->dl_file);
211 kmem_cache_free(deleg_slab, dp);
216 /* Remove the associated file_lock first, then remove the delegation.
217 * lease_modify() is called to remove the FS_LEASE file_lock from
218 * the i_flock list, eventually calling nfsd's lock_manager
219 * fl_release_callback.
222 nfs4_close_delegation(struct nfs4_delegation *dp)
224 struct file *filp = dp->dl_vfs_file;
226 dprintk("NFSD: close_delegation dp %p\n",dp);
227 dp->dl_vfs_file = NULL;
228 /* The following nfsd_close may not actually close the file,
229 * but we want to remove the lease in any case. */
231 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
235 /* Called under the state lock. */
237 unhash_delegation(struct nfs4_delegation *dp)
239 list_del_init(&dp->dl_perfile);
240 list_del_init(&dp->dl_perclnt);
241 spin_lock(&recall_lock);
242 list_del_init(&dp->dl_recall_lru);
243 spin_unlock(&recall_lock);
244 nfs4_close_delegation(dp);
245 nfs4_put_delegation(dp);
252 /* Hash tables for nfs4_clientid state */
253 #define CLIENT_HASH_BITS 4
254 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
255 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
257 #define clientid_hashval(id) \
258 ((id) & CLIENT_HASH_MASK)
259 #define clientstr_hashval(name) \
260 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
262 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
263 * used in reboot/reset lease grace period processing
265 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
266 * setclientid_confirmed info.
268 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
271 * client_lru holds client queue ordered by nfs4_client.cl_time
274 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
275 * for last close replay.
277 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
278 static int reclaim_str_hashtbl_size = 0;
279 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
280 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
281 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
282 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
283 static struct list_head client_lru;
284 static struct list_head close_lru;
286 static void unhash_generic_stateid(struct nfs4_stateid *stp)
288 list_del(&stp->st_hash);
289 list_del(&stp->st_perfile);
290 list_del(&stp->st_perstateowner);
293 static void free_generic_stateid(struct nfs4_stateid *stp)
295 put_nfs4_file(stp->st_file);
296 kmem_cache_free(stateid_slab, stp);
299 static void release_lock_stateid(struct nfs4_stateid *stp)
301 unhash_generic_stateid(stp);
302 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
303 free_generic_stateid(stp);
306 static void unhash_lockowner(struct nfs4_stateowner *sop)
308 struct nfs4_stateid *stp;
310 list_del(&sop->so_idhash);
311 list_del(&sop->so_strhash);
312 list_del(&sop->so_perstateid);
313 while (!list_empty(&sop->so_stateids)) {
314 stp = list_first_entry(&sop->so_stateids,
315 struct nfs4_stateid, st_perstateowner);
316 release_lock_stateid(stp);
320 static void release_lockowner(struct nfs4_stateowner *sop)
322 unhash_lockowner(sop);
323 nfs4_put_stateowner(sop);
327 release_stateid_lockowners(struct nfs4_stateid *open_stp)
329 struct nfs4_stateowner *lock_sop;
331 while (!list_empty(&open_stp->st_lockowners)) {
332 lock_sop = list_entry(open_stp->st_lockowners.next,
333 struct nfs4_stateowner, so_perstateid);
334 /* list_del(&open_stp->st_lockowners); */
335 BUG_ON(lock_sop->so_is_open_owner);
336 release_lockowner(lock_sop);
340 static void release_open_stateid(struct nfs4_stateid *stp)
342 unhash_generic_stateid(stp);
343 release_stateid_lockowners(stp);
344 nfsd_close(stp->st_vfs_file);
345 free_generic_stateid(stp);
348 static void unhash_openowner(struct nfs4_stateowner *sop)
350 struct nfs4_stateid *stp;
352 list_del(&sop->so_idhash);
353 list_del(&sop->so_strhash);
354 list_del(&sop->so_perclient);
355 list_del(&sop->so_perstateid); /* XXX: necessary? */
356 while (!list_empty(&sop->so_stateids)) {
357 stp = list_first_entry(&sop->so_stateids,
358 struct nfs4_stateid, st_perstateowner);
359 release_open_stateid(stp);
363 static void release_openowner(struct nfs4_stateowner *sop)
365 unhash_openowner(sop);
366 list_del(&sop->so_close_lru);
367 nfs4_put_stateowner(sop);
370 static DEFINE_SPINLOCK(sessionid_lock);
371 #define SESSION_HASH_SIZE 512
372 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
375 hash_sessionid(struct nfs4_sessionid *sessionid)
377 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
379 return sid->sequence % SESSION_HASH_SIZE;
383 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
385 u32 *ptr = (u32 *)(&sessionid->data[0]);
386 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
390 gen_sessionid(struct nfsd4_session *ses)
392 struct nfs4_client *clp = ses->se_client;
393 struct nfsd4_sessionid *sid;
395 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
396 sid->clientid = clp->cl_clientid;
397 sid->sequence = current_sessionid++;
402 * The protocol defines ca_maxresponssize_cached to include the size of
403 * the rpc header, but all we need to cache is the data starting after
404 * the end of the initial SEQUENCE operation--the rest we regenerate
405 * each time. Therefore we can advertise a ca_maxresponssize_cached
406 * value that is the number of bytes in our cache plus a few additional
407 * bytes. In order to stay on the safe side, and not promise more than
408 * we can cache, those additional bytes must be the minimum possible: 24
409 * bytes of rpc header (xid through accept state, with AUTH_NULL
410 * verifier), 12 for the compound header (with zero-length tag), and 44
411 * for the SEQUENCE op response:
413 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
416 * Give the client the number of ca_maxresponsesize_cached slots it
417 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
418 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
419 * than NFSD_MAX_SLOTS_PER_SESSION.
421 * If we run out of reserved DRC memory we should (up to a point)
422 * re-negotiate active sessions and reduce their slot usage to make
423 * rooom for new connections. For now we just fail the create session.
425 static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
427 int mem, size = fchan->maxresp_cached;
429 if (fchan->maxreqs < 1)
432 if (size < NFSD_MIN_HDR_SEQ_SZ)
433 size = NFSD_MIN_HDR_SEQ_SZ;
434 size -= NFSD_MIN_HDR_SEQ_SZ;
435 if (size > NFSD_SLOT_CACHE_SIZE)
436 size = NFSD_SLOT_CACHE_SIZE;
438 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
439 mem = fchan->maxreqs * size;
440 if (mem > NFSD_MAX_MEM_PER_SESSION) {
441 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
442 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
443 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
444 mem = fchan->maxreqs * size;
447 spin_lock(&nfsd_drc_lock);
448 /* bound the total session drc memory ussage */
449 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
450 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
451 mem = fchan->maxreqs * size;
453 nfsd_drc_mem_used += mem;
454 spin_unlock(&nfsd_drc_lock);
456 if (fchan->maxreqs == 0)
457 return nfserr_serverfault;
459 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
464 * fchan holds the client values on input, and the server values on output
465 * sv_max_mesg is the maximum payload plus one page for overhead.
467 static int init_forechannel_attrs(struct svc_rqst *rqstp,
468 struct nfsd4_channel_attrs *session_fchan,
469 struct nfsd4_channel_attrs *fchan)
472 __u32 maxcount = nfsd_serv->sv_max_mesg;
474 /* headerpadsz set to zero in encode routine */
476 /* Use the client's max request and max response size if possible */
477 if (fchan->maxreq_sz > maxcount)
478 fchan->maxreq_sz = maxcount;
479 session_fchan->maxreq_sz = fchan->maxreq_sz;
481 if (fchan->maxresp_sz > maxcount)
482 fchan->maxresp_sz = maxcount;
483 session_fchan->maxresp_sz = fchan->maxresp_sz;
485 /* Use the client's maxops if possible */
486 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
487 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
488 session_fchan->maxops = fchan->maxops;
490 /* FIXME: Error means no more DRC pages so the server should
491 * recover pages from existing sessions. For now fail session
494 status = set_forechannel_drc_size(fchan);
496 session_fchan->maxresp_cached = fchan->maxresp_cached;
497 session_fchan->maxreqs = fchan->maxreqs;
499 dprintk("%s status %d\n", __func__, status);
504 free_session_slots(struct nfsd4_session *ses)
508 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
509 kfree(ses->se_slots[i]);
513 * We don't actually need to cache the rpc and session headers, so we
514 * can allocate a little less for each slot:
516 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
518 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
522 alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
523 struct nfsd4_create_session *cses)
525 struct nfsd4_session *new, tmp;
526 struct nfsd4_slot *sp;
527 int idx, slotsize, cachesize, i;
530 memset(&tmp, 0, sizeof(tmp));
532 /* FIXME: For now, we just accept the client back channel attributes. */
533 tmp.se_bchannel = cses->back_channel;
534 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
535 &cses->fore_channel);
539 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
540 + sizeof(struct nfsd4_session) > PAGE_SIZE);
542 status = nfserr_serverfault;
543 /* allocate struct nfsd4_session and slot table pointers in one piece */
544 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
545 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
549 memcpy(new, &tmp, sizeof(*new));
551 /* allocate each struct nfsd4_slot and data cache in one piece */
552 cachesize = slot_bytes(&new->se_fchannel);
553 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
554 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
557 new->se_slots[i] = sp;
560 new->se_client = clp;
562 idx = hash_sessionid(&new->se_sessionid);
563 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
564 NFS4_MAX_SESSIONID_LEN);
566 new->se_flags = cses->flags;
567 kref_init(&new->se_ref);
568 spin_lock(&sessionid_lock);
569 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
570 list_add(&new->se_perclnt, &clp->cl_sessions);
571 spin_unlock(&sessionid_lock);
577 free_session_slots(new);
582 /* caller must hold sessionid_lock */
583 static struct nfsd4_session *
584 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
586 struct nfsd4_session *elem;
589 dump_sessionid(__func__, sessionid);
590 idx = hash_sessionid(sessionid);
591 dprintk("%s: idx is %d\n", __func__, idx);
592 /* Search in the appropriate list */
593 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
594 dump_sessionid("list traversal", &elem->se_sessionid);
595 if (!memcmp(elem->se_sessionid.data, sessionid->data,
596 NFS4_MAX_SESSIONID_LEN)) {
601 dprintk("%s: session not found\n", __func__);
605 /* caller must hold sessionid_lock */
607 unhash_session(struct nfsd4_session *ses)
609 list_del(&ses->se_hash);
610 list_del(&ses->se_perclnt);
614 release_session(struct nfsd4_session *ses)
616 spin_lock(&sessionid_lock);
618 spin_unlock(&sessionid_lock);
619 nfsd4_put_session(ses);
623 free_session(struct kref *kref)
625 struct nfsd4_session *ses;
628 ses = container_of(kref, struct nfsd4_session, se_ref);
629 spin_lock(&nfsd_drc_lock);
630 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
631 nfsd_drc_mem_used -= mem;
632 spin_unlock(&nfsd_drc_lock);
633 free_session_slots(ses);
638 renew_client(struct nfs4_client *clp)
641 * Move client to the end to the LRU list.
643 dprintk("renewing client (clientid %08x/%08x)\n",
644 clp->cl_clientid.cl_boot,
645 clp->cl_clientid.cl_id);
646 list_move_tail(&clp->cl_lru, &client_lru);
647 clp->cl_time = get_seconds();
650 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
652 STALE_CLIENTID(clientid_t *clid)
654 if (clid->cl_boot == boot_time)
656 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
657 clid->cl_boot, clid->cl_id, boot_time);
662 * XXX Should we use a slab cache ?
663 * This type of memory management is somewhat inefficient, but we use it
664 * anyway since SETCLIENTID is not a common operation.
666 static struct nfs4_client *alloc_client(struct xdr_netobj name)
668 struct nfs4_client *clp;
670 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
673 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
674 if (clp->cl_name.data == NULL) {
678 memcpy(clp->cl_name.data, name.data, name.len);
679 clp->cl_name.len = name.len;
684 free_client(struct nfs4_client *clp)
686 if (clp->cl_cred.cr_group_info)
687 put_group_info(clp->cl_cred.cr_group_info);
688 kfree(clp->cl_principal);
689 kfree(clp->cl_name.data);
694 expire_client(struct nfs4_client *clp)
696 struct nfs4_stateowner *sop;
697 struct nfs4_delegation *dp;
698 struct list_head reaplist;
700 INIT_LIST_HEAD(&reaplist);
701 spin_lock(&recall_lock);
702 while (!list_empty(&clp->cl_delegations)) {
703 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
704 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
706 list_del_init(&dp->dl_perclnt);
707 list_move(&dp->dl_recall_lru, &reaplist);
709 spin_unlock(&recall_lock);
710 while (!list_empty(&reaplist)) {
711 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
712 list_del_init(&dp->dl_recall_lru);
713 unhash_delegation(dp);
715 list_del(&clp->cl_idhash);
716 list_del(&clp->cl_strhash);
717 list_del(&clp->cl_lru);
718 while (!list_empty(&clp->cl_openowners)) {
719 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
720 release_openowner(sop);
722 while (!list_empty(&clp->cl_sessions)) {
723 struct nfsd4_session *ses;
724 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
726 release_session(ses);
728 nfsd4_set_callback_client(clp, NULL);
730 svc_xprt_put(clp->cl_cb_xprt);
734 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
736 memcpy(target->cl_verifier.data, source->data,
737 sizeof(target->cl_verifier.data));
740 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
742 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
743 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
746 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
748 target->cr_uid = source->cr_uid;
749 target->cr_gid = source->cr_gid;
750 target->cr_group_info = source->cr_group_info;
751 get_group_info(target->cr_group_info);
754 static int same_name(const char *n1, const char *n2)
756 return 0 == memcmp(n1, n2, HEXDIR_LEN);
760 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
762 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
766 same_clid(clientid_t *cl1, clientid_t *cl2)
768 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
771 /* XXX what about NGROUP */
773 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
775 return cr1->cr_uid == cr2->cr_uid;
778 static void gen_clid(struct nfs4_client *clp)
780 static u32 current_clientid = 1;
782 clp->cl_clientid.cl_boot = boot_time;
783 clp->cl_clientid.cl_id = current_clientid++;
786 static void gen_confirm(struct nfs4_client *clp)
791 p = (u32 *)clp->cl_confirm.data;
792 *p++ = get_seconds();
796 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
797 struct svc_rqst *rqstp, nfs4_verifier *verf)
799 struct nfs4_client *clp;
800 struct sockaddr *sa = svc_addr(rqstp);
803 clp = alloc_client(name);
807 princ = svc_gss_principal(rqstp);
809 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
810 if (clp->cl_principal == NULL) {
816 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
817 atomic_set(&clp->cl_cb_conn.cb_set, 0);
818 INIT_LIST_HEAD(&clp->cl_idhash);
819 INIT_LIST_HEAD(&clp->cl_strhash);
820 INIT_LIST_HEAD(&clp->cl_openowners);
821 INIT_LIST_HEAD(&clp->cl_delegations);
822 INIT_LIST_HEAD(&clp->cl_sessions);
823 INIT_LIST_HEAD(&clp->cl_lru);
824 clear_bit(0, &clp->cl_cb_slot_busy);
825 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
826 copy_verf(clp, verf);
827 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
828 clp->cl_flavor = rqstp->rq_flavor;
829 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
835 static int check_name(struct xdr_netobj name)
839 if (name.len > NFS4_OPAQUE_LIMIT) {
840 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
847 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
849 unsigned int idhashval;
851 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
852 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
853 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
854 list_add_tail(&clp->cl_lru, &client_lru);
855 clp->cl_time = get_seconds();
859 move_to_confirmed(struct nfs4_client *clp)
861 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
862 unsigned int strhashval;
864 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
865 list_del_init(&clp->cl_strhash);
866 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
867 strhashval = clientstr_hashval(clp->cl_recdir);
868 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
872 static struct nfs4_client *
873 find_confirmed_client(clientid_t *clid)
875 struct nfs4_client *clp;
876 unsigned int idhashval = clientid_hashval(clid->cl_id);
878 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
879 if (same_clid(&clp->cl_clientid, clid))
885 static struct nfs4_client *
886 find_unconfirmed_client(clientid_t *clid)
888 struct nfs4_client *clp;
889 unsigned int idhashval = clientid_hashval(clid->cl_id);
891 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
892 if (same_clid(&clp->cl_clientid, clid))
899 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
900 * parameter. Matching is based on the fact the at least one of the
901 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
903 * FIXME: we need to unify the clientid namespaces for nfsv4.x
904 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
905 * and SET_CLIENTID{,_CONFIRM}
908 match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
910 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
911 return use_exchange_id == has_exchange_flags;
914 static struct nfs4_client *
915 find_confirmed_client_by_str(const char *dname, unsigned int hashval,
916 bool use_exchange_id)
918 struct nfs4_client *clp;
920 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
921 if (same_name(clp->cl_recdir, dname) &&
922 match_clientid_establishment(clp, use_exchange_id))
928 static struct nfs4_client *
929 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
930 bool use_exchange_id)
932 struct nfs4_client *clp;
934 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
935 if (same_name(clp->cl_recdir, dname) &&
936 match_clientid_establishment(clp, use_exchange_id))
943 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
945 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
946 unsigned short expected_family;
948 /* Currently, we only support tcp and tcp6 for the callback channel */
949 if (se->se_callback_netid_len == 3 &&
950 !memcmp(se->se_callback_netid_val, "tcp", 3))
951 expected_family = AF_INET;
952 else if (se->se_callback_netid_len == 4 &&
953 !memcmp(se->se_callback_netid_val, "tcp6", 4))
954 expected_family = AF_INET6;
958 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
959 se->se_callback_addr_len,
960 (struct sockaddr *) &cb->cb_addr,
961 sizeof(cb->cb_addr));
963 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
966 if (cb->cb_addr.ss_family == AF_INET6)
967 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
969 cb->cb_minorversion = 0;
970 cb->cb_prog = se->se_callback_prog;
971 cb->cb_ident = se->se_callback_ident;
974 cb->cb_addr.ss_family = AF_UNSPEC;
976 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
977 "will not receive delegations\n",
978 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
984 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
987 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
989 struct nfsd4_slot *slot = resp->cstate.slot;
992 dprintk("--> %s slot %p\n", __func__, slot);
994 slot->sl_opcnt = resp->opcnt;
995 slot->sl_status = resp->cstate.status;
997 if (nfsd4_not_cached(resp)) {
998 slot->sl_datalen = 0;
1001 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1002 base = (char *)resp->cstate.datap -
1003 (char *)resp->xbuf->head[0].iov_base;
1004 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1006 WARN("%s: sessions DRC could not cache compound\n", __func__);
1011 * Encode the replay sequence operation from the slot values.
1012 * If cachethis is FALSE encode the uncached rep error on the next
1013 * operation which sets resp->p and increments resp->opcnt for
1014 * nfs4svc_encode_compoundres.
1018 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1019 struct nfsd4_compoundres *resp)
1021 struct nfsd4_op *op;
1022 struct nfsd4_slot *slot = resp->cstate.slot;
1024 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1025 resp->opcnt, resp->cstate.slot->sl_cachethis);
1027 /* Encode the replayed sequence operation */
1028 op = &args->ops[resp->opcnt - 1];
1029 nfsd4_encode_operation(resp, op);
1031 /* Return nfserr_retry_uncached_rep in next operation. */
1032 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1033 op = &args->ops[resp->opcnt++];
1034 op->status = nfserr_retry_uncached_rep;
1035 nfsd4_encode_operation(resp, op);
1041 * The sequence operation is not cached because we can use the slot and
1045 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1046 struct nfsd4_sequence *seq)
1048 struct nfsd4_slot *slot = resp->cstate.slot;
1051 dprintk("--> %s slot %p\n", __func__, slot);
1053 /* Either returns 0 or nfserr_retry_uncached */
1054 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1055 if (status == nfserr_retry_uncached_rep)
1058 /* The sequence operation has been encoded, cstate->datap set. */
1059 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1061 resp->opcnt = slot->sl_opcnt;
1062 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1063 status = slot->sl_status;
1069 * Set the exchange_id flags returned by the server.
1072 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1074 /* pNFS is not supported */
1075 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1077 /* Referrals are supported, Migration is not. */
1078 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1080 /* set the wire flags to return to client. */
1081 clid->flags = new->cl_exchange_flags;
1085 nfsd4_exchange_id(struct svc_rqst *rqstp,
1086 struct nfsd4_compound_state *cstate,
1087 struct nfsd4_exchange_id *exid)
1089 struct nfs4_client *unconf, *conf, *new;
1091 unsigned int strhashval;
1092 char dname[HEXDIR_LEN];
1093 char addr_str[INET6_ADDRSTRLEN];
1094 nfs4_verifier verf = exid->verifier;
1095 struct sockaddr *sa = svc_addr(rqstp);
1097 rpc_ntop(sa, addr_str, sizeof(addr_str));
1098 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1099 "ip_addr=%s flags %x, spa_how %d\n",
1100 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1101 addr_str, exid->flags, exid->spa_how);
1103 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1104 return nfserr_inval;
1106 /* Currently only support SP4_NONE */
1107 switch (exid->spa_how) {
1111 return nfserr_encr_alg_unsupp;
1113 BUG(); /* checked by xdr code */
1115 return nfserr_serverfault; /* no excuse :-/ */
1118 status = nfs4_make_rec_clidname(dname, &exid->clname);
1123 strhashval = clientstr_hashval(dname);
1128 conf = find_confirmed_client_by_str(dname, strhashval, true);
1130 if (!same_verf(&verf, &conf->cl_verifier)) {
1131 /* 18.35.4 case 8 */
1132 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1133 status = nfserr_not_same;
1136 /* Client reboot: destroy old state */
1137 expire_client(conf);
1140 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1141 /* 18.35.4 case 9 */
1142 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1143 status = nfserr_perm;
1146 expire_client(conf);
1150 * Set bit when the owner id and verifier map to an already
1151 * confirmed client id (18.35.3).
1153 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1156 * Falling into 18.35.4 case 2, possible router replay.
1157 * Leave confirmed record intact and return same result.
1159 copy_verf(conf, &verf);
1164 /* 18.35.4 case 7 */
1165 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1166 status = nfserr_noent;
1170 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
1173 * Possible retry or client restart. Per 18.35.4 case 4,
1174 * a new unconfirmed record should be generated regardless
1175 * of whether any properties have changed.
1177 expire_client(unconf);
1182 new = create_client(exid->clname, dname, rqstp, &verf);
1184 status = nfserr_serverfault;
1189 add_to_unconfirmed(new, strhashval);
1191 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1192 exid->clientid.cl_id = new->cl_clientid.cl_id;
1195 nfsd4_set_ex_flags(new, exid);
1197 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1198 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1202 nfs4_unlock_state();
1204 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1209 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1211 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1214 /* The slot is in use, and no response has been sent. */
1216 if (seqid == slot_seqid)
1217 return nfserr_jukebox;
1219 return nfserr_seq_misordered;
1222 if (likely(seqid == slot_seqid + 1))
1225 if (seqid == slot_seqid)
1226 return nfserr_replay_cache;
1228 if (seqid == 1 && (slot_seqid + 1) == 0)
1230 /* Misordered replay or misordered new request */
1231 return nfserr_seq_misordered;
1235 * Cache the create session result into the create session single DRC
1236 * slot cache by saving the xdr structure. sl_seqid has been set.
1237 * Do this for solo or embedded create session operations.
1240 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1241 struct nfsd4_clid_slot *slot, int nfserr)
1243 slot->sl_status = nfserr;
1244 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1248 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1249 struct nfsd4_clid_slot *slot)
1251 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1252 return slot->sl_status;
1256 nfsd4_create_session(struct svc_rqst *rqstp,
1257 struct nfsd4_compound_state *cstate,
1258 struct nfsd4_create_session *cr_ses)
1260 struct sockaddr *sa = svc_addr(rqstp);
1261 struct nfs4_client *conf, *unconf;
1262 struct nfsd4_clid_slot *cs_slot = NULL;
1266 unconf = find_unconfirmed_client(&cr_ses->clientid);
1267 conf = find_confirmed_client(&cr_ses->clientid);
1270 cs_slot = &conf->cl_cs_slot;
1271 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1272 if (status == nfserr_replay_cache) {
1273 dprintk("Got a create_session replay! seqid= %d\n",
1275 /* Return the cached reply status */
1276 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1278 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1279 status = nfserr_seq_misordered;
1280 dprintk("Sequence misordered!\n");
1281 dprintk("Expected seqid= %d but got seqid= %d\n",
1282 cs_slot->sl_seqid, cr_ses->seqid);
1285 cs_slot->sl_seqid++;
1286 } else if (unconf) {
1287 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1288 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1289 status = nfserr_clid_inuse;
1293 cs_slot = &unconf->cl_cs_slot;
1294 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1296 /* an unconfirmed replay returns misordered */
1297 status = nfserr_seq_misordered;
1301 cs_slot->sl_seqid++; /* from 0 to 1 */
1302 move_to_confirmed(unconf);
1304 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1305 unconf->cl_cb_xprt = rqstp->rq_xprt;
1306 svc_xprt_get(unconf->cl_cb_xprt);
1308 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1310 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1311 unconf->cl_cb_conn.cb_minorversion =
1312 cstate->minorversion;
1313 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1314 unconf->cl_cb_seq_nr = 1;
1315 nfsd4_probe_callback(unconf);
1319 status = nfserr_stale_clientid;
1324 * We do not support RDMA or persistent sessions
1326 cr_ses->flags &= ~SESSION4_PERSIST;
1327 cr_ses->flags &= ~SESSION4_RDMA;
1329 status = alloc_init_session(rqstp, conf, cr_ses);
1333 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1334 NFS4_MAX_SESSIONID_LEN);
1335 cr_ses->seqid = cs_slot->sl_seqid;
1338 /* cache solo and embedded create sessions under the state lock */
1339 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1341 nfs4_unlock_state();
1342 dprintk("%s returns %d\n", __func__, ntohl(status));
1347 nfsd4_destroy_session(struct svc_rqst *r,
1348 struct nfsd4_compound_state *cstate,
1349 struct nfsd4_destroy_session *sessionid)
1351 struct nfsd4_session *ses;
1352 u32 status = nfserr_badsession;
1355 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1356 * - Should we return nfserr_back_chan_busy if waiting for
1357 * callbacks on to-be-destroyed session?
1358 * - Do we need to clear any callback info from previous session?
1361 dump_sessionid(__func__, &sessionid->sessionid);
1362 spin_lock(&sessionid_lock);
1363 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1365 spin_unlock(&sessionid_lock);
1369 unhash_session(ses);
1370 spin_unlock(&sessionid_lock);
1372 /* wait for callbacks */
1373 nfsd4_set_callback_client(ses->se_client, NULL);
1374 nfsd4_put_session(ses);
1377 dprintk("%s returns %d\n", __func__, ntohl(status));
1382 nfsd4_sequence(struct svc_rqst *rqstp,
1383 struct nfsd4_compound_state *cstate,
1384 struct nfsd4_sequence *seq)
1386 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1387 struct nfsd4_session *session;
1388 struct nfsd4_slot *slot;
1391 if (resp->opcnt != 1)
1392 return nfserr_sequence_pos;
1394 spin_lock(&sessionid_lock);
1395 status = nfserr_badsession;
1396 session = find_in_sessionid_hashtbl(&seq->sessionid);
1400 status = nfserr_badslot;
1401 if (seq->slotid >= session->se_fchannel.maxreqs)
1404 slot = session->se_slots[seq->slotid];
1405 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1407 /* We do not negotiate the number of slots yet, so set the
1408 * maxslots to the session maxreqs which is used to encode
1409 * sr_highest_slotid and the sr_target_slot id to maxslots */
1410 seq->maxslots = session->se_fchannel.maxreqs;
1412 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1413 if (status == nfserr_replay_cache) {
1414 cstate->slot = slot;
1415 cstate->session = session;
1416 /* Return the cached reply status and set cstate->status
1417 * for nfsd4_proc_compound processing */
1418 status = nfsd4_replay_cache_entry(resp, seq);
1419 cstate->status = nfserr_replay_cache;
1425 /* Success! bump slot seqid */
1426 slot->sl_inuse = true;
1427 slot->sl_seqid = seq->seqid;
1428 slot->sl_cachethis = seq->cachethis;
1430 cstate->slot = slot;
1431 cstate->session = session;
1433 /* Hold a session reference until done processing the compound:
1434 * nfsd4_put_session called only if the cstate slot is set.
1436 nfsd4_get_session(session);
1438 spin_unlock(&sessionid_lock);
1439 /* Renew the clientid on success and on replay */
1440 if (cstate->session) {
1442 renew_client(session->se_client);
1443 nfs4_unlock_state();
1445 dprintk("%s: return %d\n", __func__, ntohl(status));
1450 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1451 struct nfsd4_setclientid *setclid)
1453 struct sockaddr *sa = svc_addr(rqstp);
1454 struct xdr_netobj clname = {
1455 .len = setclid->se_namelen,
1456 .data = setclid->se_name,
1458 nfs4_verifier clverifier = setclid->se_verf;
1459 unsigned int strhashval;
1460 struct nfs4_client *conf, *unconf, *new;
1462 char dname[HEXDIR_LEN];
1464 if (!check_name(clname))
1465 return nfserr_inval;
1467 status = nfs4_make_rec_clidname(dname, &clname);
1472 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1473 * We get here on a DRC miss.
1476 strhashval = clientstr_hashval(dname);
1479 conf = find_confirmed_client_by_str(dname, strhashval, false);
1481 /* RFC 3530 14.2.33 CASE 0: */
1482 status = nfserr_clid_inuse;
1483 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1484 char addr_str[INET6_ADDRSTRLEN];
1485 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1487 dprintk("NFSD: setclientid: string in use by client "
1488 "at %s\n", addr_str);
1493 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1494 * has a description of SETCLIENTID request processing consisting
1495 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1497 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1498 status = nfserr_resource;
1501 * RFC 3530 14.2.33 CASE 4:
1502 * placed first, because it is the normal case
1505 expire_client(unconf);
1506 new = create_client(clname, dname, rqstp, &clverifier);
1510 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1512 * RFC 3530 14.2.33 CASE 1:
1513 * probable callback update
1516 /* Note this is removing unconfirmed {*x***},
1517 * which is stronger than RFC recommended {vxc**}.
1518 * This has the advantage that there is at most
1519 * one {*x***} in either list at any time.
1521 expire_client(unconf);
1523 new = create_client(clname, dname, rqstp, &clverifier);
1526 copy_clid(new, conf);
1527 } else if (!unconf) {
1529 * RFC 3530 14.2.33 CASE 2:
1530 * probable client reboot; state will be removed if
1533 new = create_client(clname, dname, rqstp, &clverifier);
1539 * RFC 3530 14.2.33 CASE 3:
1540 * probable client reboot; state will be removed if
1543 expire_client(unconf);
1544 new = create_client(clname, dname, rqstp, &clverifier);
1549 gen_callback(new, setclid, rpc_get_scope_id(sa));
1550 add_to_unconfirmed(new, strhashval);
1551 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1552 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1553 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1556 nfs4_unlock_state();
1562 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1563 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1564 * bullets, labeled as CASE1 - CASE4 below.
1567 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1568 struct nfsd4_compound_state *cstate,
1569 struct nfsd4_setclientid_confirm *setclientid_confirm)
1571 struct sockaddr *sa = svc_addr(rqstp);
1572 struct nfs4_client *conf, *unconf;
1573 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1574 clientid_t * clid = &setclientid_confirm->sc_clientid;
1577 if (STALE_CLIENTID(clid))
1578 return nfserr_stale_clientid;
1580 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1581 * We get here on a DRC miss.
1586 conf = find_confirmed_client(clid);
1587 unconf = find_unconfirmed_client(clid);
1589 status = nfserr_clid_inuse;
1590 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
1592 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
1596 * section 14.2.34 of RFC 3530 has a description of
1597 * SETCLIENTID_CONFIRM request processing consisting
1598 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1600 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
1602 * RFC 3530 14.2.34 CASE 1:
1605 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1606 status = nfserr_clid_inuse;
1608 /* XXX: We just turn off callbacks until we can handle
1609 * change request correctly. */
1610 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1611 expire_client(unconf);
1615 } else if (conf && !unconf) {
1617 * RFC 3530 14.2.34 CASE 2:
1618 * probable retransmitted request; play it safe and
1621 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1622 status = nfserr_clid_inuse;
1625 } else if (!conf && unconf
1626 && same_verf(&unconf->cl_confirm, &confirm)) {
1628 * RFC 3530 14.2.34 CASE 3:
1629 * Normal case; new or rebooted client:
1631 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1632 status = nfserr_clid_inuse;
1635 clientstr_hashval(unconf->cl_recdir);
1636 conf = find_confirmed_client_by_str(unconf->cl_recdir,
1639 nfsd4_remove_clid_dir(conf);
1640 expire_client(conf);
1642 move_to_confirmed(unconf);
1644 nfsd4_probe_callback(conf);
1647 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1648 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
1651 * RFC 3530 14.2.34 CASE 4:
1652 * Client probably hasn't noticed that we rebooted yet.
1654 status = nfserr_stale_clientid;
1656 /* check that we have hit one of the cases...*/
1657 status = nfserr_clid_inuse;
1660 nfs4_unlock_state();
1664 /* OPEN Share state helper functions */
1665 static inline struct nfs4_file *
1666 alloc_init_file(struct inode *ino)
1668 struct nfs4_file *fp;
1669 unsigned int hashval = file_hashval(ino);
1671 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1673 atomic_set(&fp->fi_ref, 1);
1674 INIT_LIST_HEAD(&fp->fi_hash);
1675 INIT_LIST_HEAD(&fp->fi_stateids);
1676 INIT_LIST_HEAD(&fp->fi_delegations);
1677 spin_lock(&recall_lock);
1678 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1679 spin_unlock(&recall_lock);
1680 fp->fi_inode = igrab(ino);
1681 fp->fi_id = current_fileid++;
1682 fp->fi_had_conflict = false;
1689 nfsd4_free_slab(struct kmem_cache **slab)
1693 kmem_cache_destroy(*slab);
1698 nfsd4_free_slabs(void)
1700 nfsd4_free_slab(&stateowner_slab);
1701 nfsd4_free_slab(&file_slab);
1702 nfsd4_free_slab(&stateid_slab);
1703 nfsd4_free_slab(&deleg_slab);
1707 nfsd4_init_slabs(void)
1709 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1710 sizeof(struct nfs4_stateowner), 0, 0, NULL);
1711 if (stateowner_slab == NULL)
1713 file_slab = kmem_cache_create("nfsd4_files",
1714 sizeof(struct nfs4_file), 0, 0, NULL);
1715 if (file_slab == NULL)
1717 stateid_slab = kmem_cache_create("nfsd4_stateids",
1718 sizeof(struct nfs4_stateid), 0, 0, NULL);
1719 if (stateid_slab == NULL)
1721 deleg_slab = kmem_cache_create("nfsd4_delegations",
1722 sizeof(struct nfs4_delegation), 0, 0, NULL);
1723 if (deleg_slab == NULL)
1728 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1733 nfs4_free_stateowner(struct kref *kref)
1735 struct nfs4_stateowner *sop =
1736 container_of(kref, struct nfs4_stateowner, so_ref);
1737 kfree(sop->so_owner.data);
1738 kmem_cache_free(stateowner_slab, sop);
1741 static inline struct nfs4_stateowner *
1742 alloc_stateowner(struct xdr_netobj *owner)
1744 struct nfs4_stateowner *sop;
1746 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1747 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1748 memcpy(sop->so_owner.data, owner->data, owner->len);
1749 sop->so_owner.len = owner->len;
1750 kref_init(&sop->so_ref);
1753 kmem_cache_free(stateowner_slab, sop);
1758 static struct nfs4_stateowner *
1759 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1760 struct nfs4_stateowner *sop;
1761 struct nfs4_replay *rp;
1762 unsigned int idhashval;
1764 if (!(sop = alloc_stateowner(&open->op_owner)))
1766 idhashval = ownerid_hashval(current_ownerid);
1767 INIT_LIST_HEAD(&sop->so_idhash);
1768 INIT_LIST_HEAD(&sop->so_strhash);
1769 INIT_LIST_HEAD(&sop->so_perclient);
1770 INIT_LIST_HEAD(&sop->so_stateids);
1771 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1772 INIT_LIST_HEAD(&sop->so_close_lru);
1774 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1775 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1776 list_add(&sop->so_perclient, &clp->cl_openowners);
1777 sop->so_is_open_owner = 1;
1778 sop->so_id = current_ownerid++;
1779 sop->so_client = clp;
1780 sop->so_seqid = open->op_seqid;
1781 sop->so_confirmed = 0;
1782 rp = &sop->so_replay;
1783 rp->rp_status = nfserr_serverfault;
1785 rp->rp_buf = rp->rp_ibuf;
1790 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1791 struct nfs4_stateowner *sop = open->op_stateowner;
1792 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1794 INIT_LIST_HEAD(&stp->st_hash);
1795 INIT_LIST_HEAD(&stp->st_perstateowner);
1796 INIT_LIST_HEAD(&stp->st_lockowners);
1797 INIT_LIST_HEAD(&stp->st_perfile);
1798 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1799 list_add(&stp->st_perstateowner, &sop->so_stateids);
1800 list_add(&stp->st_perfile, &fp->fi_stateids);
1801 stp->st_stateowner = sop;
1804 stp->st_stateid.si_boot = get_seconds();
1805 stp->st_stateid.si_stateownerid = sop->so_id;
1806 stp->st_stateid.si_fileid = fp->fi_id;
1807 stp->st_stateid.si_generation = 0;
1808 stp->st_access_bmap = 0;
1809 stp->st_deny_bmap = 0;
1810 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1811 &stp->st_access_bmap);
1812 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1813 stp->st_openstp = NULL;
1817 move_to_close_lru(struct nfs4_stateowner *sop)
1819 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1821 list_move_tail(&sop->so_close_lru, &close_lru);
1822 sop->so_time = get_seconds();
1826 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1829 return (sop->so_owner.len == owner->len) &&
1830 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1831 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1834 static struct nfs4_stateowner *
1835 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1837 struct nfs4_stateowner *so = NULL;
1839 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1840 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1846 /* search file_hashtbl[] for file */
1847 static struct nfs4_file *
1848 find_file(struct inode *ino)
1850 unsigned int hashval = file_hashval(ino);
1851 struct nfs4_file *fp;
1853 spin_lock(&recall_lock);
1854 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1855 if (fp->fi_inode == ino) {
1857 spin_unlock(&recall_lock);
1861 spin_unlock(&recall_lock);
1865 static inline int access_valid(u32 x, u32 minorversion)
1867 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
1869 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1871 x &= ~NFS4_SHARE_ACCESS_MASK;
1872 if (minorversion && x) {
1873 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1875 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1877 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1884 static inline int deny_valid(u32 x)
1886 /* Note: unlike access bits, deny bits may be zero. */
1887 return x <= NFS4_SHARE_DENY_BOTH;
1891 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1892 * st_{access,deny}_bmap field of the stateid, in order to track not
1893 * only what share bits are currently in force, but also what
1894 * combinations of share bits previous opens have used. This allows us
1895 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1896 * return an error if the client attempt to downgrade to a combination
1897 * of share bits not explicable by closing some of its previous opens.
1899 * XXX: This enforcement is actually incomplete, since we don't keep
1900 * track of access/deny bit combinations; so, e.g., we allow:
1902 * OPEN allow read, deny write
1903 * OPEN allow both, deny none
1904 * DOWNGRADE allow read, deny none
1906 * which we should reject.
1909 set_access(unsigned int *access, unsigned long bmap) {
1913 for (i = 1; i < 4; i++) {
1914 if (test_bit(i, &bmap))
1920 set_deny(unsigned int *deny, unsigned long bmap) {
1924 for (i = 0; i < 4; i++) {
1925 if (test_bit(i, &bmap))
1931 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1932 unsigned int access, deny;
1934 set_access(&access, stp->st_access_bmap);
1935 set_deny(&deny, stp->st_deny_bmap);
1936 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1942 * Called to check deny when READ with all zero stateid or
1943 * WRITE with all zero or all one stateid
1946 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1948 struct inode *ino = current_fh->fh_dentry->d_inode;
1949 struct nfs4_file *fp;
1950 struct nfs4_stateid *stp;
1953 dprintk("NFSD: nfs4_share_conflict\n");
1955 fp = find_file(ino);
1958 ret = nfserr_locked;
1959 /* Search for conflicting share reservations */
1960 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1961 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1962 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1972 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1974 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1975 drop_file_write_access(filp);
1976 spin_lock(&filp->f_lock);
1977 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1978 spin_unlock(&filp->f_lock);
1983 * Spawn a thread to perform a recall on the delegation represented
1984 * by the lease (file_lock)
1986 * Called from break_lease() with lock_kernel() held.
1987 * Note: we assume break_lease will only call this *once* for any given
1991 void nfsd_break_deleg_cb(struct file_lock *fl)
1993 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1995 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1999 /* We're assuming the state code never drops its reference
2000 * without first removing the lease. Since we're in this lease
2001 * callback (and since the lease code is serialized by the kernel
2002 * lock) we know the server hasn't removed the lease yet, we know
2003 * it's safe to take a reference: */
2004 atomic_inc(&dp->dl_count);
2006 spin_lock(&recall_lock);
2007 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2008 spin_unlock(&recall_lock);
2010 /* only place dl_time is set. protected by lock_kernel*/
2011 dp->dl_time = get_seconds();
2014 * We don't want the locks code to timeout the lease for us;
2015 * we'll remove it ourself if the delegation isn't returned
2018 fl->fl_break_time = 0;
2020 dp->dl_file->fi_had_conflict = true;
2021 nfsd4_cb_recall(dp);
2025 * The file_lock is being reapd.
2027 * Called by locks_free_lock() with lock_kernel() held.
2030 void nfsd_release_deleg_cb(struct file_lock *fl)
2032 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2034 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2036 if (!(fl->fl_flags & FL_LEASE) || !dp)
2038 dp->dl_flock = NULL;
2042 * Set the delegation file_lock back pointer.
2044 * Called from setlease() with lock_kernel() held.
2047 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2049 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2051 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2058 * Called from setlease() with lock_kernel() held
2061 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2063 struct nfs4_delegation *onlistd =
2064 (struct nfs4_delegation *)onlist->fl_owner;
2065 struct nfs4_delegation *tryd =
2066 (struct nfs4_delegation *)try->fl_owner;
2068 if (onlist->fl_lmops != try->fl_lmops)
2071 return onlistd->dl_client == tryd->dl_client;
2076 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2079 return lease_modify(onlist, arg);
2084 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2085 .fl_break = nfsd_break_deleg_cb,
2086 .fl_release_private = nfsd_release_deleg_cb,
2087 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2088 .fl_mylease = nfsd_same_client_deleg_cb,
2089 .fl_change = nfsd_change_deleg_cb,
2094 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2095 struct nfsd4_open *open)
2097 clientid_t *clientid = &open->op_clientid;
2098 struct nfs4_client *clp = NULL;
2099 unsigned int strhashval;
2100 struct nfs4_stateowner *sop = NULL;
2102 if (!check_name(open->op_owner))
2103 return nfserr_inval;
2105 if (STALE_CLIENTID(&open->op_clientid))
2106 return nfserr_stale_clientid;
2108 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2109 sop = find_openstateowner_str(strhashval, open);
2110 open->op_stateowner = sop;
2112 /* Make sure the client's lease hasn't expired. */
2113 clp = find_confirmed_client(clientid);
2115 return nfserr_expired;
2118 /* When sessions are used, skip open sequenceid processing */
2119 if (nfsd4_has_session(cstate))
2121 if (!sop->so_confirmed) {
2122 /* Replace unconfirmed owners without checking for replay. */
2123 clp = sop->so_client;
2124 release_openowner(sop);
2125 open->op_stateowner = NULL;
2128 if (open->op_seqid == sop->so_seqid - 1) {
2129 if (sop->so_replay.rp_buflen)
2130 return nfserr_replay_me;
2131 /* The original OPEN failed so spectacularly
2132 * that we don't even have replay data saved!
2133 * Therefore, we have no choice but to continue
2134 * processing this OPEN; presumably, we'll
2135 * fail again for the same reason.
2137 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2140 if (open->op_seqid != sop->so_seqid)
2141 return nfserr_bad_seqid;
2143 if (open->op_stateowner == NULL) {
2144 sop = alloc_init_open_stateowner(strhashval, clp, open);
2146 return nfserr_resource;
2147 open->op_stateowner = sop;
2149 list_del_init(&sop->so_close_lru);
2150 renew_client(sop->so_client);
2154 static inline __be32
2155 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2157 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2158 return nfserr_openmode;
2163 static struct nfs4_delegation *
2164 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2166 struct nfs4_delegation *dp;
2168 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
2169 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2176 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2177 struct nfs4_delegation **dp)
2180 __be32 status = nfserr_bad_stateid;
2182 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2185 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2186 RD_STATE : WR_STATE;
2187 status = nfs4_check_delegmode(*dp, flags);
2191 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2195 open->op_stateowner->so_confirmed = 1;
2200 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2202 struct nfs4_stateid *local;
2203 __be32 status = nfserr_share_denied;
2204 struct nfs4_stateowner *sop = open->op_stateowner;
2206 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2207 /* ignore lock owners */
2208 if (local->st_stateowner->so_is_open_owner == 0)
2210 /* remember if we have seen this open owner */
2211 if (local->st_stateowner == sop)
2213 /* check for conflicting share reservations */
2214 if (!test_share(local, open))
2222 static inline struct nfs4_stateid *
2223 nfs4_alloc_stateid(void)
2225 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2229 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
2230 struct nfs4_delegation *dp,
2231 struct svc_fh *cur_fh, int flags)
2233 struct nfs4_stateid *stp;
2235 stp = nfs4_alloc_stateid();
2237 return nfserr_resource;
2240 get_file(dp->dl_vfs_file);
2241 stp->st_vfs_file = dp->dl_vfs_file;
2244 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2247 if (status == nfserr_dropit)
2248 status = nfserr_jukebox;
2249 kmem_cache_free(stateid_slab, stp);
2257 static inline __be32
2258 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2259 struct nfsd4_open *open)
2261 struct iattr iattr = {
2262 .ia_valid = ATTR_SIZE,
2265 if (!open->op_truncate)
2267 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2268 return nfserr_inval;
2269 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2273 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2275 struct file *filp = stp->st_vfs_file;
2276 struct inode *inode = filp->f_path.dentry->d_inode;
2277 unsigned int share_access, new_writer;
2280 set_access(&share_access, stp->st_access_bmap);
2281 new_writer = (~share_access) & open->op_share_access
2282 & NFS4_SHARE_ACCESS_WRITE;
2285 int err = get_write_access(inode);
2287 return nfserrno(err);
2288 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2290 return nfserrno(err);
2291 file_take_write(filp);
2293 status = nfsd4_truncate(rqstp, cur_fh, open);
2296 put_write_access(inode);
2299 /* remember the open */
2300 filp->f_mode |= open->op_share_access;
2301 __set_bit(open->op_share_access, &stp->st_access_bmap);
2302 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2309 nfs4_set_claim_prev(struct nfsd4_open *open)
2311 open->op_stateowner->so_confirmed = 1;
2312 open->op_stateowner->so_client->cl_firststate = 1;
2316 * Attempt to hand out a delegation.
2319 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2321 struct nfs4_delegation *dp;
2322 struct nfs4_stateowner *sop = stp->st_stateowner;
2323 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
2324 struct file_lock fl, *flp = &fl;
2325 int status, flag = 0;
2327 flag = NFS4_OPEN_DELEGATE_NONE;
2328 open->op_recall = 0;
2329 switch (open->op_claim_type) {
2330 case NFS4_OPEN_CLAIM_PREVIOUS:
2331 if (!atomic_read(&cb->cb_set))
2332 open->op_recall = 1;
2333 flag = open->op_delegate_type;
2334 if (flag == NFS4_OPEN_DELEGATE_NONE)
2337 case NFS4_OPEN_CLAIM_NULL:
2338 /* Let's not give out any delegations till everyone's
2339 * had the chance to reclaim theirs.... */
2340 if (locks_in_grace())
2342 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2344 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2345 flag = NFS4_OPEN_DELEGATE_WRITE;
2347 flag = NFS4_OPEN_DELEGATE_READ;
2353 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2355 flag = NFS4_OPEN_DELEGATE_NONE;
2358 locks_init_lock(&fl);
2359 fl.fl_lmops = &nfsd_lease_mng_ops;
2360 fl.fl_flags = FL_LEASE;
2361 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2362 fl.fl_end = OFFSET_MAX;
2363 fl.fl_owner = (fl_owner_t)dp;
2364 fl.fl_file = stp->st_vfs_file;
2365 fl.fl_pid = current->tgid;
2367 /* vfs_setlease checks to see if delegation should be handed out.
2368 * the lock_manager callbacks fl_mylease and fl_change are used
2370 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
2371 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
2372 unhash_delegation(dp);
2373 flag = NFS4_OPEN_DELEGATE_NONE;
2377 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2379 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2380 STATEID_VAL(&dp->dl_stateid));
2382 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2383 && flag == NFS4_OPEN_DELEGATE_NONE
2384 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2385 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2386 open->op_delegate_type = flag;
2390 * called with nfs4_lock_state() held.
2393 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2395 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2396 struct nfs4_file *fp = NULL;
2397 struct inode *ino = current_fh->fh_dentry->d_inode;
2398 struct nfs4_stateid *stp = NULL;
2399 struct nfs4_delegation *dp = NULL;
2402 status = nfserr_inval;
2403 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
2404 || !deny_valid(open->op_share_deny))
2407 * Lookup file; if found, lookup stateid and check open request,
2408 * and check for delegations in the process of being recalled.
2409 * If not found, create the nfs4_file struct
2411 fp = find_file(ino);
2413 if ((status = nfs4_check_open(fp, open, &stp)))
2415 status = nfs4_check_deleg(fp, open, &dp);
2419 status = nfserr_bad_stateid;
2420 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2422 status = nfserr_resource;
2423 fp = alloc_init_file(ino);
2429 * OPEN the file, or upgrade an existing OPEN.
2430 * If truncate fails, the OPEN fails.
2433 /* Stateid was found, this is an OPEN upgrade */
2434 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2437 update_stateid(&stp->st_stateid);
2439 /* Stateid was not found, this is a new OPEN */
2441 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
2442 flags |= NFSD_MAY_READ;
2443 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2444 flags |= NFSD_MAY_WRITE;
2445 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2448 init_stateid(stp, fp, open);
2449 status = nfsd4_truncate(rqstp, current_fh, open);
2451 release_open_stateid(stp);
2454 if (nfsd4_has_session(&resp->cstate))
2455 update_stateid(&stp->st_stateid);
2457 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2459 if (nfsd4_has_session(&resp->cstate)) {
2460 open->op_stateowner->so_confirmed = 1;
2461 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2465 * Attempt to hand out a delegation. No error return, because the
2466 * OPEN succeeds even if we fail.
2468 nfs4_open_delegation(current_fh, open, stp);
2472 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2473 STATEID_VAL(&stp->st_stateid));
2477 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2478 nfs4_set_claim_prev(open);
2480 * To finish the open response, we just need to set the rflags.
2482 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2483 if (!open->op_stateowner->so_confirmed &&
2484 !nfsd4_has_session(&resp->cstate))
2485 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2491 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2494 struct nfs4_client *clp;
2498 dprintk("process_renew(%08x/%08x): starting\n",
2499 clid->cl_boot, clid->cl_id);
2500 status = nfserr_stale_clientid;
2501 if (STALE_CLIENTID(clid))
2503 clp = find_confirmed_client(clid);
2504 status = nfserr_expired;
2506 /* We assume the client took too long to RENEW. */
2507 dprintk("nfsd4_renew: clientid not found!\n");
2511 status = nfserr_cb_path_down;
2512 if (!list_empty(&clp->cl_delegations)
2513 && !atomic_read(&clp->cl_cb_conn.cb_set))
2517 nfs4_unlock_state();
2521 struct lock_manager nfsd4_manager = {
2525 nfsd4_end_grace(void)
2527 dprintk("NFSD: end of grace period\n");
2528 nfsd4_recdir_purge_old();
2529 locks_end_grace(&nfsd4_manager);
2531 * Now that every NFSv4 client has had the chance to recover and
2532 * to see the (possibly new, possibly shorter) lease time, we
2533 * can safely set the next grace time to the current lease time:
2535 nfsd4_grace = nfsd4_lease;
2539 nfs4_laundromat(void)
2541 struct nfs4_client *clp;
2542 struct nfs4_stateowner *sop;
2543 struct nfs4_delegation *dp;
2544 struct list_head *pos, *next, reaplist;
2545 time_t cutoff = get_seconds() - nfsd4_lease;
2546 time_t t, clientid_val = nfsd4_lease;
2547 time_t u, test_val = nfsd4_lease;
2551 dprintk("NFSD: laundromat service - starting\n");
2552 if (locks_in_grace())
2554 list_for_each_safe(pos, next, &client_lru) {
2555 clp = list_entry(pos, struct nfs4_client, cl_lru);
2556 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2557 t = clp->cl_time - cutoff;
2558 if (clientid_val > t)
2562 dprintk("NFSD: purging unused client (clientid %08x)\n",
2563 clp->cl_clientid.cl_id);
2564 nfsd4_remove_clid_dir(clp);
2567 INIT_LIST_HEAD(&reaplist);
2568 spin_lock(&recall_lock);
2569 list_for_each_safe(pos, next, &del_recall_lru) {
2570 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2571 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2572 u = dp->dl_time - cutoff;
2577 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2579 list_move(&dp->dl_recall_lru, &reaplist);
2581 spin_unlock(&recall_lock);
2582 list_for_each_safe(pos, next, &reaplist) {
2583 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2584 list_del_init(&dp->dl_recall_lru);
2585 unhash_delegation(dp);
2587 test_val = nfsd4_lease;
2588 list_for_each_safe(pos, next, &close_lru) {
2589 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2590 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2591 u = sop->so_time - cutoff;
2596 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2598 release_openowner(sop);
2600 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2601 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2602 nfs4_unlock_state();
2603 return clientid_val;
2606 static struct workqueue_struct *laundry_wq;
2607 static void laundromat_main(struct work_struct *);
2608 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2611 laundromat_main(struct work_struct *not_used)
2615 t = nfs4_laundromat();
2616 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
2617 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
2620 static struct nfs4_stateowner *
2621 search_close_lru(u32 st_id, int flags)
2623 struct nfs4_stateowner *local = NULL;
2625 if (flags & CLOSE_STATE) {
2626 list_for_each_entry(local, &close_lru, so_close_lru) {
2627 if (local->so_id == st_id)
2635 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2637 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
2641 STALE_STATEID(stateid_t *stateid)
2643 if (time_after((unsigned long)boot_time,
2644 (unsigned long)stateid->si_boot)) {
2645 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2646 STATEID_VAL(stateid));
2653 EXPIRED_STATEID(stateid_t *stateid)
2655 if (time_before((unsigned long)boot_time,
2656 ((unsigned long)stateid->si_boot)) &&
2657 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
2658 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2659 STATEID_VAL(stateid));
2666 stateid_error_map(stateid_t *stateid)
2668 if (STALE_STATEID(stateid))
2669 return nfserr_stale_stateid;
2670 if (EXPIRED_STATEID(stateid))
2671 return nfserr_expired;
2673 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2674 STATEID_VAL(stateid));
2675 return nfserr_bad_stateid;
2679 access_permit_read(unsigned long access_bmap)
2681 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2682 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2683 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2687 access_permit_write(unsigned long access_bmap)
2689 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2690 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2694 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2696 __be32 status = nfserr_openmode;
2698 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2700 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2707 static inline __be32
2708 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2710 if (ONE_STATEID(stateid) && (flags & RD_STATE))
2712 else if (locks_in_grace()) {
2713 /* Answer in remaining cases depends on existance of
2714 * conflicting state; so we must wait out the grace period. */
2715 return nfserr_grace;
2716 } else if (flags & WR_STATE)
2717 return nfs4_share_conflict(current_fh,
2718 NFS4_SHARE_DENY_WRITE);
2719 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2720 return nfs4_share_conflict(current_fh,
2721 NFS4_SHARE_DENY_READ);
2725 * Allow READ/WRITE during grace period on recovered state only for files
2726 * that are not able to provide mandatory locking.
2729 grace_disallows_io(struct inode *inode)
2731 return locks_in_grace() && mandatory_lock(inode);
2734 static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
2737 * When sessions are used the stateid generation number is ignored
2740 if ((flags & HAS_SESSION) && in->si_generation == 0)
2743 /* If the client sends us a stateid from the future, it's buggy: */
2744 if (in->si_generation > ref->si_generation)
2745 return nfserr_bad_stateid;
2747 * The following, however, can happen. For example, if the
2748 * client sends an open and some IO at the same time, the open
2749 * may bump si_generation while the IO is still in flight.
2750 * Thanks to hard links and renames, the client never knows what
2751 * file an open will affect. So it could avoid that situation
2752 * only by serializing all opens and IO from the same open
2753 * owner. To recover from the old_stateid error, the client
2754 * will just have to retry the IO:
2756 if (in->si_generation < ref->si_generation)
2757 return nfserr_old_stateid;
2762 static int is_delegation_stateid(stateid_t *stateid)
2764 return stateid->si_fileid == 0;
2768 * Checks for stateid operations
2771 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2772 stateid_t *stateid, int flags, struct file **filpp)
2774 struct nfs4_stateid *stp = NULL;
2775 struct nfs4_delegation *dp = NULL;
2776 struct svc_fh *current_fh = &cstate->current_fh;
2777 struct inode *ino = current_fh->fh_dentry->d_inode;
2783 if (grace_disallows_io(ino))
2784 return nfserr_grace;
2786 if (nfsd4_has_session(cstate))
2787 flags |= HAS_SESSION;
2789 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2790 return check_special_stateids(current_fh, stateid, flags);
2792 status = nfserr_stale_stateid;
2793 if (STALE_STATEID(stateid))
2796 status = nfserr_bad_stateid;
2797 if (is_delegation_stateid(stateid)) {
2798 dp = find_delegation_stateid(ino, stateid);
2800 status = stateid_error_map(stateid);
2803 status = check_stateid_generation(stateid, &dp->dl_stateid,
2807 status = nfs4_check_delegmode(dp, flags);
2810 renew_client(dp->dl_client);
2812 *filpp = dp->dl_vfs_file;
2813 } else { /* open or lock stateid */
2814 stp = find_stateid(stateid, flags);
2816 status = stateid_error_map(stateid);
2819 if (nfs4_check_fh(current_fh, stp))
2821 if (!stp->st_stateowner->so_confirmed)
2823 status = check_stateid_generation(stateid, &stp->st_stateid,
2827 status = nfs4_check_openmode(stp, flags);
2830 renew_client(stp->st_stateowner->so_client);
2832 *filpp = stp->st_vfs_file;
2842 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2843 RD_STATE : WR_STATE;
2847 * Checks for sequence id mutating operations.
2850 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2851 stateid_t *stateid, int flags,
2852 struct nfs4_stateowner **sopp,
2853 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2855 struct nfs4_stateid *stp;
2856 struct nfs4_stateowner *sop;
2857 struct svc_fh *current_fh = &cstate->current_fh;
2860 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2861 seqid, STATEID_VAL(stateid));
2866 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2867 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2868 return nfserr_bad_stateid;
2871 if (STALE_STATEID(stateid))
2872 return nfserr_stale_stateid;
2874 if (nfsd4_has_session(cstate))
2875 flags |= HAS_SESSION;
2878 * We return BAD_STATEID if filehandle doesn't match stateid,
2879 * the confirmed flag is incorrecly set, or the generation
2880 * number is incorrect.
2882 stp = find_stateid(stateid, flags);
2885 * Also, we should make sure this isn't just the result of
2888 sop = search_close_lru(stateid->si_stateownerid, flags);
2890 return stateid_error_map(stateid);
2896 *sopp = sop = stp->st_stateowner;
2899 clientid_t *lockclid = &lock->v.new.clientid;
2900 struct nfs4_client *clp = sop->so_client;
2904 lkflg = setlkflg(lock->lk_type);
2906 if (lock->lk_is_new) {
2907 if (!sop->so_is_open_owner)
2908 return nfserr_bad_stateid;
2909 if (!(flags & HAS_SESSION) &&
2910 !same_clid(&clp->cl_clientid, lockclid))
2911 return nfserr_bad_stateid;
2912 /* stp is the open stateid */
2913 status = nfs4_check_openmode(stp, lkflg);
2917 /* stp is the lock stateid */
2918 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2924 if (nfs4_check_fh(current_fh, stp)) {
2925 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2926 return nfserr_bad_stateid;
2930 * We now validate the seqid and stateid generation numbers.
2931 * For the moment, we ignore the possibility of
2932 * generation number wraparound.
2934 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
2937 if (sop->so_confirmed && flags & CONFIRM) {
2938 dprintk("NFSD: preprocess_seqid_op: expected"
2939 " unconfirmed stateowner!\n");
2940 return nfserr_bad_stateid;
2942 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2943 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2944 " confirmed yet!\n");
2945 return nfserr_bad_stateid;
2947 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
2950 renew_client(sop->so_client);
2954 if (seqid == sop->so_seqid - 1) {
2955 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2956 /* indicate replay to calling function */
2957 return nfserr_replay_me;
2959 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2960 sop->so_seqid, seqid);
2962 return nfserr_bad_seqid;
2966 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2967 struct nfsd4_open_confirm *oc)
2970 struct nfs4_stateowner *sop;
2971 struct nfs4_stateid *stp;
2973 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2974 (int)cstate->current_fh.fh_dentry->d_name.len,
2975 cstate->current_fh.fh_dentry->d_name.name);
2977 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2983 if ((status = nfs4_preprocess_seqid_op(cstate,
2984 oc->oc_seqid, &oc->oc_req_stateid,
2985 CONFIRM | OPEN_STATE,
2986 &oc->oc_stateowner, &stp, NULL)))
2989 sop = oc->oc_stateowner;
2990 sop->so_confirmed = 1;
2991 update_stateid(&stp->st_stateid);
2992 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2993 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
2994 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
2996 nfsd4_create_clid_dir(sop->so_client);
2998 if (oc->oc_stateowner) {
2999 nfs4_get_stateowner(oc->oc_stateowner);
3000 cstate->replay_owner = oc->oc_stateowner;
3002 nfs4_unlock_state();
3008 * unset all bits in union bitmap (bmap) that
3009 * do not exist in share (from successful OPEN_DOWNGRADE)
3012 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3015 for (i = 1; i < 4; i++) {
3016 if ((i & access) != i)
3017 __clear_bit(i, bmap);
3022 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3025 for (i = 0; i < 4; i++) {
3026 if ((i & deny) != i)
3027 __clear_bit(i, bmap);
3032 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3033 struct nfsd4_compound_state *cstate,
3034 struct nfsd4_open_downgrade *od)
3037 struct nfs4_stateid *stp;
3038 unsigned int share_access;
3040 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3041 (int)cstate->current_fh.fh_dentry->d_name.len,
3042 cstate->current_fh.fh_dentry->d_name.name);
3044 if (!access_valid(od->od_share_access, cstate->minorversion)
3045 || !deny_valid(od->od_share_deny))
3046 return nfserr_inval;
3049 if ((status = nfs4_preprocess_seqid_op(cstate,
3053 &od->od_stateowner, &stp, NULL)))
3056 status = nfserr_inval;
3057 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3058 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3059 stp->st_access_bmap, od->od_share_access);
3062 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3063 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3064 stp->st_deny_bmap, od->od_share_deny);
3067 set_access(&share_access, stp->st_access_bmap);
3068 nfs4_file_downgrade(stp->st_vfs_file,
3069 share_access & ~od->od_share_access);
3071 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3072 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3074 update_stateid(&stp->st_stateid);
3075 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3078 if (od->od_stateowner) {
3079 nfs4_get_stateowner(od->od_stateowner);
3080 cstate->replay_owner = od->od_stateowner;
3082 nfs4_unlock_state();
3087 * nfs4_unlock_state() called after encode
3090 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3091 struct nfsd4_close *close)
3094 struct nfs4_stateid *stp;
3096 dprintk("NFSD: nfsd4_close on file %.*s\n",
3097 (int)cstate->current_fh.fh_dentry->d_name.len,
3098 cstate->current_fh.fh_dentry->d_name.name);
3101 /* check close_lru for replay */
3102 if ((status = nfs4_preprocess_seqid_op(cstate,
3105 OPEN_STATE | CLOSE_STATE,
3106 &close->cl_stateowner, &stp, NULL)))
3109 update_stateid(&stp->st_stateid);
3110 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3112 /* release_stateid() calls nfsd_close() if needed */
3113 release_open_stateid(stp);
3115 /* place unused nfs4_stateowners on so_close_lru list to be
3116 * released by the laundromat service after the lease period
3117 * to enable us to handle CLOSE replay
3119 if (list_empty(&close->cl_stateowner->so_stateids))
3120 move_to_close_lru(close->cl_stateowner);
3122 if (close->cl_stateowner) {
3123 nfs4_get_stateowner(close->cl_stateowner);
3124 cstate->replay_owner = close->cl_stateowner;
3126 nfs4_unlock_state();
3131 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3132 struct nfsd4_delegreturn *dr)
3134 struct nfs4_delegation *dp;
3135 stateid_t *stateid = &dr->dr_stateid;
3136 struct inode *inode;
3140 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3142 inode = cstate->current_fh.fh_dentry->d_inode;
3144 if (nfsd4_has_session(cstate))
3145 flags |= HAS_SESSION;
3147 status = nfserr_bad_stateid;
3148 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3150 status = nfserr_stale_stateid;
3151 if (STALE_STATEID(stateid))
3153 status = nfserr_bad_stateid;
3154 if (!is_delegation_stateid(stateid))
3156 dp = find_delegation_stateid(inode, stateid);
3158 status = stateid_error_map(stateid);
3161 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
3164 renew_client(dp->dl_client);
3166 unhash_delegation(dp);
3168 nfs4_unlock_state();
3175 * Lock owner state (byte-range locks)
3177 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3178 #define LOCK_HASH_BITS 8
3179 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3180 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3183 end_offset(u64 start, u64 len)
3188 return end >= start ? end: NFS4_MAX_UINT64;
3191 /* last octet in a range */
3193 last_byte_offset(u64 start, u64 len)
3199 return end > start ? end - 1: NFS4_MAX_UINT64;
3202 #define lockownerid_hashval(id) \
3203 ((id) & LOCK_HASH_MASK)
3205 static inline unsigned int
3206 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3207 struct xdr_netobj *ownername)
3209 return (file_hashval(inode) + cl_id
3210 + opaque_hashval(ownername->data, ownername->len))
3214 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3215 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3216 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3218 static struct nfs4_stateid *
3219 find_stateid(stateid_t *stid, int flags)
3221 struct nfs4_stateid *local;
3222 u32 st_id = stid->si_stateownerid;
3223 u32 f_id = stid->si_fileid;
3224 unsigned int hashval;
3226 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
3227 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
3228 hashval = stateid_hashval(st_id, f_id);
3229 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3230 if ((local->st_stateid.si_stateownerid == st_id) &&
3231 (local->st_stateid.si_fileid == f_id))
3236 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
3237 hashval = stateid_hashval(st_id, f_id);
3238 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3239 if ((local->st_stateid.si_stateownerid == st_id) &&
3240 (local->st_stateid.si_fileid == f_id))
3247 static struct nfs4_delegation *
3248 find_delegation_stateid(struct inode *ino, stateid_t *stid)
3250 struct nfs4_file *fp;
3251 struct nfs4_delegation *dl;
3253 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3256 fp = find_file(ino);
3259 dl = find_delegation_file(fp, stid);
3265 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3266 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3267 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3268 * locking, this prevents us from being completely protocol-compliant. The
3269 * real solution to this problem is to start using unsigned file offsets in
3270 * the VFS, but this is a very deep change!
3273 nfs4_transform_lock_offset(struct file_lock *lock)
3275 if (lock->fl_start < 0)
3276 lock->fl_start = OFFSET_MAX;
3277 if (lock->fl_end < 0)
3278 lock->fl_end = OFFSET_MAX;
3281 /* Hack!: For now, we're defining this just so we can use a pointer to it
3282 * as a unique cookie to identify our (NFSv4's) posix locks. */
3283 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3287 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3289 struct nfs4_stateowner *sop;
3292 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3293 sop = (struct nfs4_stateowner *) fl->fl_owner;
3294 hval = lockownerid_hashval(sop->so_id);
3295 kref_get(&sop->so_ref);
3297 deny->ld_clientid = sop->so_client->cl_clientid;
3299 deny->ld_sop = NULL;
3300 deny->ld_clientid.cl_boot = 0;
3301 deny->ld_clientid.cl_id = 0;
3303 deny->ld_start = fl->fl_start;
3304 deny->ld_length = NFS4_MAX_UINT64;
3305 if (fl->fl_end != NFS4_MAX_UINT64)
3306 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3307 deny->ld_type = NFS4_READ_LT;
3308 if (fl->fl_type != F_RDLCK)
3309 deny->ld_type = NFS4_WRITE_LT;
3312 static struct nfs4_stateowner *
3313 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3314 struct xdr_netobj *owner)
3316 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3317 struct nfs4_stateowner *op;
3319 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3320 if (same_owner_str(op, owner, clid))
3327 * Alloc a lock owner structure.
3328 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3331 * strhashval = lock_ownerstr_hashval
3334 static struct nfs4_stateowner *
3335 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3336 struct nfs4_stateowner *sop;
3337 struct nfs4_replay *rp;
3338 unsigned int idhashval;
3340 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3342 idhashval = lockownerid_hashval(current_ownerid);
3343 INIT_LIST_HEAD(&sop->so_idhash);
3344 INIT_LIST_HEAD(&sop->so_strhash);
3345 INIT_LIST_HEAD(&sop->so_perclient);
3346 INIT_LIST_HEAD(&sop->so_stateids);
3347 INIT_LIST_HEAD(&sop->so_perstateid);
3348 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3350 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3351 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3352 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
3353 sop->so_is_open_owner = 0;
3354 sop->so_id = current_ownerid++;
3355 sop->so_client = clp;
3356 /* It is the openowner seqid that will be incremented in encode in the
3357 * case of new lockowners; so increment the lock seqid manually: */
3358 sop->so_seqid = lock->lk_new_lock_seqid + 1;
3359 sop->so_confirmed = 1;
3360 rp = &sop->so_replay;
3361 rp->rp_status = nfserr_serverfault;
3363 rp->rp_buf = rp->rp_ibuf;
3367 static struct nfs4_stateid *
3368 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3370 struct nfs4_stateid *stp;
3371 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3373 stp = nfs4_alloc_stateid();
3376 INIT_LIST_HEAD(&stp->st_hash);
3377 INIT_LIST_HEAD(&stp->st_perfile);
3378 INIT_LIST_HEAD(&stp->st_perstateowner);
3379 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
3380 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
3381 list_add(&stp->st_perfile, &fp->fi_stateids);
3382 list_add(&stp->st_perstateowner, &sop->so_stateids);
3383 stp->st_stateowner = sop;
3386 stp->st_stateid.si_boot = get_seconds();
3387 stp->st_stateid.si_stateownerid = sop->so_id;
3388 stp->st_stateid.si_fileid = fp->fi_id;
3389 stp->st_stateid.si_generation = 0;
3390 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3391 stp->st_access_bmap = open_stp->st_access_bmap;
3392 stp->st_deny_bmap = open_stp->st_deny_bmap;
3393 stp->st_openstp = open_stp;
3400 check_lock_length(u64 offset, u64 length)
3402 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3403 LOFF_OVERFLOW(offset, length)));
3410 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3411 struct nfsd4_lock *lock)
3413 struct nfs4_stateowner *open_sop = NULL;
3414 struct nfs4_stateowner *lock_sop = NULL;
3415 struct nfs4_stateid *lock_stp;
3417 struct file_lock file_lock;
3418 struct file_lock conflock;
3420 unsigned int strhashval;
3424 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3425 (long long) lock->lk_offset,
3426 (long long) lock->lk_length);
3428 if (check_lock_length(lock->lk_offset, lock->lk_length))
3429 return nfserr_inval;
3431 if ((status = fh_verify(rqstp, &cstate->current_fh,
3432 S_IFREG, NFSD_MAY_LOCK))) {
3433 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3439 if (lock->lk_is_new) {
3441 * Client indicates that this is a new lockowner.
3442 * Use open owner and open stateid to create lock owner and
3445 struct nfs4_stateid *open_stp = NULL;
3446 struct nfs4_file *fp;
3448 status = nfserr_stale_clientid;
3449 if (!nfsd4_has_session(cstate) &&
3450 STALE_CLIENTID(&lock->lk_new_clientid))
3453 /* validate and update open stateid and open seqid */
3454 status = nfs4_preprocess_seqid_op(cstate,
3455 lock->lk_new_open_seqid,
3456 &lock->lk_new_open_stateid,
3458 &lock->lk_replay_owner, &open_stp,
3462 open_sop = lock->lk_replay_owner;
3463 /* create lockowner and lock stateid */
3464 fp = open_stp->st_file;
3465 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3466 open_sop->so_client->cl_clientid.cl_id,
3467 &lock->v.new.owner);
3468 /* XXX: Do we need to check for duplicate stateowners on
3469 * the same file, or should they just be allowed (and
3470 * create new stateids)? */
3471 status = nfserr_resource;
3472 lock_sop = alloc_init_lock_stateowner(strhashval,
3473 open_sop->so_client, open_stp, lock);
3474 if (lock_sop == NULL)
3476 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3477 if (lock_stp == NULL)
3480 /* lock (lock owner + lock stateid) already exists */
3481 status = nfs4_preprocess_seqid_op(cstate,
3482 lock->lk_old_lock_seqid,
3483 &lock->lk_old_lock_stateid,
3485 &lock->lk_replay_owner, &lock_stp, lock);
3488 lock_sop = lock->lk_replay_owner;
3490 /* lock->lk_replay_owner and lock_stp have been created or found */
3491 filp = lock_stp->st_vfs_file;
3493 status = nfserr_grace;
3494 if (locks_in_grace() && !lock->lk_reclaim)
3496 status = nfserr_no_grace;
3497 if (!locks_in_grace() && lock->lk_reclaim)
3500 locks_init_lock(&file_lock);
3501 switch (lock->lk_type) {
3504 file_lock.fl_type = F_RDLCK;
3508 case NFS4_WRITEW_LT:
3509 file_lock.fl_type = F_WRLCK;
3513 status = nfserr_inval;
3516 file_lock.fl_owner = (fl_owner_t)lock_sop;
3517 file_lock.fl_pid = current->tgid;
3518 file_lock.fl_file = filp;
3519 file_lock.fl_flags = FL_POSIX;
3520 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3522 file_lock.fl_start = lock->lk_offset;
3523 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
3524 nfs4_transform_lock_offset(&file_lock);
3527 * Try to lock the file in the VFS.
3528 * Note: locks.c uses the BKL to protect the inode's lock list.
3531 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
3533 case 0: /* success! */
3534 update_stateid(&lock_stp->st_stateid);
3535 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3539 case (EAGAIN): /* conflock holds conflicting lock */
3540 status = nfserr_denied;
3541 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3542 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3545 status = nfserr_deadlock;
3548 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
3549 status = nfserr_resource;
3553 if (status && lock->lk_is_new && lock_sop)
3554 release_lockowner(lock_sop);
3555 if (lock->lk_replay_owner) {
3556 nfs4_get_stateowner(lock->lk_replay_owner);
3557 cstate->replay_owner = lock->lk_replay_owner;
3559 nfs4_unlock_state();
3564 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3565 * so we do a temporary open here just to get an open file to pass to
3566 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3569 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3574 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3577 err = vfs_test_lock(file, lock);
3586 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3587 struct nfsd4_lockt *lockt)
3589 struct inode *inode;
3590 struct file_lock file_lock;
3594 if (locks_in_grace())
3595 return nfserr_grace;
3597 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3598 return nfserr_inval;
3600 lockt->lt_stateowner = NULL;
3603 status = nfserr_stale_clientid;
3604 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
3607 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
3608 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
3609 if (status == nfserr_symlink)
3610 status = nfserr_inval;
3614 inode = cstate->current_fh.fh_dentry->d_inode;
3615 locks_init_lock(&file_lock);
3616 switch (lockt->lt_type) {
3619 file_lock.fl_type = F_RDLCK;
3622 case NFS4_WRITEW_LT:
3623 file_lock.fl_type = F_WRLCK;
3626 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
3627 status = nfserr_inval;
3631 lockt->lt_stateowner = find_lockstateowner_str(inode,
3632 &lockt->lt_clientid, &lockt->lt_owner);
3633 if (lockt->lt_stateowner)
3634 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3635 file_lock.fl_pid = current->tgid;
3636 file_lock.fl_flags = FL_POSIX;
3638 file_lock.fl_start = lockt->lt_offset;
3639 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
3641 nfs4_transform_lock_offset(&file_lock);
3644 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
3646 status = nfserrno(error);
3649 if (file_lock.fl_type != F_UNLCK) {
3650 status = nfserr_denied;
3651 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
3654 nfs4_unlock_state();
3659 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3660 struct nfsd4_locku *locku)
3662 struct nfs4_stateid *stp;
3663 struct file *filp = NULL;
3664 struct file_lock file_lock;
3668 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3669 (long long) locku->lu_offset,
3670 (long long) locku->lu_length);
3672 if (check_lock_length(locku->lu_offset, locku->lu_length))
3673 return nfserr_inval;
3677 if ((status = nfs4_preprocess_seqid_op(cstate,
3681 &locku->lu_stateowner, &stp, NULL)))
3684 filp = stp->st_vfs_file;
3686 locks_init_lock(&file_lock);
3687 file_lock.fl_type = F_UNLCK;
3688 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3689 file_lock.fl_pid = current->tgid;
3690 file_lock.fl_file = filp;
3691 file_lock.fl_flags = FL_POSIX;
3692 file_lock.fl_lmops = &nfsd_posix_mng_ops;
3693 file_lock.fl_start = locku->lu_offset;
3695 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
3696 nfs4_transform_lock_offset(&file_lock);
3699 * Try to unlock the file in the VFS.
3701 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
3703 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
3707 * OK, unlock succeeded; the only thing left to do is update the stateid.
3709 update_stateid(&stp->st_stateid);
3710 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3713 if (locku->lu_stateowner) {
3714 nfs4_get_stateowner(locku->lu_stateowner);
3715 cstate->replay_owner = locku->lu_stateowner;
3717 nfs4_unlock_state();
3721 status = nfserrno(err);
3727 * 1: locks held by lockowner
3728 * 0: no locks held by lockowner
3731 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3733 struct file_lock **flpp;
3734 struct inode *inode = filp->f_path.dentry->d_inode;
3738 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3739 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
3750 nfsd4_release_lockowner(struct svc_rqst *rqstp,
3751 struct nfsd4_compound_state *cstate,
3752 struct nfsd4_release_lockowner *rlockowner)
3754 clientid_t *clid = &rlockowner->rl_clientid;
3755 struct nfs4_stateowner *sop;
3756 struct nfs4_stateid *stp;
3757 struct xdr_netobj *owner = &rlockowner->rl_owner;
3758 struct list_head matches;
3762 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3763 clid->cl_boot, clid->cl_id);
3765 /* XXX check for lease expiration */
3767 status = nfserr_stale_clientid;
3768 if (STALE_CLIENTID(clid))
3773 status = nfserr_locks_held;
3774 /* XXX: we're doing a linear search through all the lockowners.
3775 * Yipes! For now we'll just hope clients aren't really using
3776 * release_lockowner much, but eventually we have to fix these
3777 * data structures. */
3778 INIT_LIST_HEAD(&matches);
3779 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3780 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3781 if (!same_owner_str(sop, owner, clid))
3783 list_for_each_entry(stp, &sop->so_stateids,
3785 if (check_for_locks(stp->st_vfs_file, sop))
3787 /* Note: so_perclient unused for lockowners,
3788 * so it's OK to fool with here. */
3789 list_add(&sop->so_perclient, &matches);
3793 /* Clients probably won't expect us to return with some (but not all)
3794 * of the lockowner state released; so don't release any until all
3795 * have been checked. */
3797 while (!list_empty(&matches)) {
3798 sop = list_entry(matches.next, struct nfs4_stateowner,
3800 /* unhash_stateowner deletes so_perclient only
3801 * for openowners. */
3802 list_del(&sop->so_perclient);
3803 release_lockowner(sop);
3806 nfs4_unlock_state();
3810 static inline struct nfs4_client_reclaim *
3813 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3817 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
3819 unsigned int strhashval = clientstr_hashval(name);
3820 struct nfs4_client *clp;
3822 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
3827 * failure => all reset bets are off, nfserr_no_grace...
3830 nfs4_client_to_reclaim(const char *name)
3832 unsigned int strhashval;
3833 struct nfs4_client_reclaim *crp = NULL;
3835 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3836 crp = alloc_reclaim();
3839 strhashval = clientstr_hashval(name);
3840 INIT_LIST_HEAD(&crp->cr_strhash);
3841 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3842 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3843 reclaim_str_hashtbl_size++;
3848 nfs4_release_reclaim(void)
3850 struct nfs4_client_reclaim *crp = NULL;
3853 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3854 while (!list_empty(&reclaim_str_hashtbl[i])) {
3855 crp = list_entry(reclaim_str_hashtbl[i].next,
3856 struct nfs4_client_reclaim, cr_strhash);
3857 list_del(&crp->cr_strhash);
3859 reclaim_str_hashtbl_size--;
3862 BUG_ON(reclaim_str_hashtbl_size);
3866 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3867 static struct nfs4_client_reclaim *
3868 nfs4_find_reclaim_client(clientid_t *clid)
3870 unsigned int strhashval;
3871 struct nfs4_client *clp;
3872 struct nfs4_client_reclaim *crp = NULL;
3875 /* find clientid in conf_id_hashtbl */
3876 clp = find_confirmed_client(clid);
3880 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3881 clp->cl_name.len, clp->cl_name.data,
3884 /* find clp->cl_name in reclaim_str_hashtbl */
3885 strhashval = clientstr_hashval(clp->cl_recdir);
3886 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3887 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3895 * Called from OPEN. Look for clientid in reclaim list.
3898 nfs4_check_open_reclaim(clientid_t *clid)
3900 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3903 /* initialization to perform at module load time: */
3906 nfs4_state_init(void)
3910 status = nfsd4_init_slabs();
3913 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3914 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3915 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3916 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3917 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3918 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3920 for (i = 0; i < SESSION_HASH_SIZE; i++)
3921 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
3922 for (i = 0; i < FILE_HASH_SIZE; i++) {
3923 INIT_LIST_HEAD(&file_hashtbl[i]);
3925 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3926 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3927 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3929 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3930 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3931 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3933 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3934 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3935 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3937 memset(&onestateid, ~0, sizeof(stateid_t));
3938 INIT_LIST_HEAD(&close_lru);
3939 INIT_LIST_HEAD(&client_lru);
3940 INIT_LIST_HEAD(&del_recall_lru);
3941 reclaim_str_hashtbl_size = 0;
3946 nfsd4_load_reboot_recovery_data(void)
3951 nfsd4_init_recdir(user_recovery_dirname);
3952 status = nfsd4_recdir_load();
3953 nfs4_unlock_state();
3955 printk("NFSD: Failure reading reboot recovery data\n");
3959 * Since the lifetime of a delegation isn't limited to that of an open, a
3960 * client may quite reasonably hang on to a delegation as long as it has
3961 * the inode cached. This becomes an obvious problem the first time a
3962 * client's inode cache approaches the size of the server's total memory.
3964 * For now we avoid this problem by imposing a hard limit on the number
3965 * of delegations, which varies according to the server's memory size.
3968 set_max_delegations(void)
3971 * Allow at most 4 delegations per megabyte of RAM. Quick
3972 * estimates suggest that in the worst case (where every delegation
3973 * is for a different inode), a delegation could take about 1.5K,
3974 * giving a worst case usage of about 6% of memory.
3976 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3979 /* initialization to perform when the nfsd service is started: */
3982 __nfs4_state_start(void)
3986 boot_time = get_seconds();
3987 locks_start_grace(&nfsd4_manager);
3988 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3990 ret = set_callback_cred();
3993 laundry_wq = create_singlethread_workqueue("nfsd4");
3994 if (laundry_wq == NULL)
3996 ret = nfsd4_create_callback_queue();
3998 goto out_free_laundry;
3999 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4000 set_max_delegations();
4003 destroy_workqueue(laundry_wq);
4008 nfs4_state_start(void)
4014 nfsd4_load_reboot_recovery_data();
4015 ret = __nfs4_state_start();
4023 __nfs4_state_shutdown(void)
4026 struct nfs4_client *clp = NULL;
4027 struct nfs4_delegation *dp = NULL;
4028 struct list_head *pos, *next, reaplist;
4030 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4031 while (!list_empty(&conf_id_hashtbl[i])) {
4032 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4035 while (!list_empty(&unconf_str_hashtbl[i])) {
4036 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4040 INIT_LIST_HEAD(&reaplist);
4041 spin_lock(&recall_lock);
4042 list_for_each_safe(pos, next, &del_recall_lru) {
4043 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4044 list_move(&dp->dl_recall_lru, &reaplist);
4046 spin_unlock(&recall_lock);
4047 list_for_each_safe(pos, next, &reaplist) {
4048 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4049 list_del_init(&dp->dl_recall_lru);
4050 unhash_delegation(dp);
4053 nfsd4_shutdown_recdir();
4058 nfs4_state_shutdown(void)
4060 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4061 destroy_workqueue(laundry_wq);
4062 locks_end_grace(&nfsd4_manager);
4064 nfs4_release_reclaim();
4065 __nfs4_state_shutdown();
4066 nfsd4_destroy_callback_queue();
4067 nfs4_unlock_state();
4071 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4072 * accessed when nfsd is starting.
4075 nfs4_set_recdir(char *recdir)
4077 strcpy(user_recovery_dirname, recdir);
4081 * Change the NFSv4 recovery directory to recdir.
4084 nfs4_reset_recoverydir(char *recdir)
4089 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
4093 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
4094 nfs4_set_recdir(recdir);
4102 nfs4_recoverydir(void)
4104 return user_recovery_dirname;