2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 #include <linux/sunrpc/svcauth_gss.h>
59 #define NFSDDBG_FACILITY NFSDDBG_PROC
62 static time_t lease_time = 90; /* default lease time */
63 static time_t user_lease_time = 90;
64 static time_t boot_time;
65 static u32 current_ownerid = 1;
66 static u32 current_fileid = 1;
67 static u32 current_delegid = 1;
69 static stateid_t zerostateid; /* bits all 0 */
70 static stateid_t onestateid; /* bits all 1 */
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
77 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
78 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
79 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80 static void nfs4_set_recdir(char *recdir);
85 * protects clientid_hashtbl[], clientstr_hashtbl[],
86 * unconfstr_hashtbl[], uncofid_hashtbl[].
88 static DEFINE_MUTEX(client_mutex);
90 static struct kmem_cache *stateowner_slab = NULL;
91 static struct kmem_cache *file_slab = NULL;
92 static struct kmem_cache *stateid_slab = NULL;
93 static struct kmem_cache *deleg_slab = NULL;
98 mutex_lock(&client_mutex);
102 nfs4_unlock_state(void)
104 mutex_unlock(&client_mutex);
108 opaque_hashval(const void *ptr, int nbytes)
110 unsigned char *cptr = (unsigned char *) ptr;
120 /* forward declarations */
121 static void release_stateowner(struct nfs4_stateowner *sop);
127 /* recall_lock protects the del_recall_lru */
128 static DEFINE_SPINLOCK(recall_lock);
129 static struct list_head del_recall_lru;
132 free_nfs4_file(struct kref *kref)
134 struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
135 list_del(&fp->fi_hash);
137 kmem_cache_free(file_slab, fp);
141 put_nfs4_file(struct nfs4_file *fi)
143 kref_put(&fi->fi_ref, free_nfs4_file);
147 get_nfs4_file(struct nfs4_file *fi)
149 kref_get(&fi->fi_ref);
152 static int num_delegations;
153 unsigned int max_delegations;
156 * Open owner state (share locks)
159 /* hash tables for nfs4_stateowner */
160 #define OWNER_HASH_BITS 8
161 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
162 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
164 #define ownerid_hashval(id) \
165 ((id) & OWNER_HASH_MASK)
166 #define ownerstr_hashval(clientid, ownername) \
167 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
169 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
170 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
172 /* hash table for nfs4_file */
173 #define FILE_HASH_BITS 8
174 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
175 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
176 /* hash table for (open)nfs4_stateid */
177 #define STATEID_HASH_BITS 10
178 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
179 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
181 #define file_hashval(x) \
182 hash_ptr(x, FILE_HASH_BITS)
183 #define stateid_hashval(owner_id, file_id) \
184 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
186 static struct list_head file_hashtbl[FILE_HASH_SIZE];
187 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
189 static struct nfs4_delegation *
190 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
192 struct nfs4_delegation *dp;
193 struct nfs4_file *fp = stp->st_file;
194 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
196 dprintk("NFSD alloc_init_deleg\n");
197 if (fp->fi_had_conflict)
199 if (num_delegations > max_delegations)
201 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
205 INIT_LIST_HEAD(&dp->dl_perfile);
206 INIT_LIST_HEAD(&dp->dl_perclnt);
207 INIT_LIST_HEAD(&dp->dl_recall_lru);
212 get_file(stp->st_vfs_file);
213 dp->dl_vfs_file = stp->st_vfs_file;
215 dp->dl_recall.cbr_dp = NULL;
216 dp->dl_recall.cbr_ident = cb->cb_ident;
217 dp->dl_recall.cbr_trunc = 0;
218 dp->dl_stateid.si_boot = boot_time;
219 dp->dl_stateid.si_stateownerid = current_delegid++;
220 dp->dl_stateid.si_fileid = 0;
221 dp->dl_stateid.si_generation = 0;
222 dp->dl_fhlen = current_fh->fh_handle.fh_size;
223 memcpy(dp->dl_fhval, ¤t_fh->fh_handle.fh_base,
224 current_fh->fh_handle.fh_size);
226 atomic_set(&dp->dl_count, 1);
227 list_add(&dp->dl_perfile, &fp->fi_delegations);
228 list_add(&dp->dl_perclnt, &clp->cl_delegations);
233 nfs4_put_delegation(struct nfs4_delegation *dp)
235 if (atomic_dec_and_test(&dp->dl_count)) {
236 dprintk("NFSD: freeing dp %p\n",dp);
237 put_nfs4_file(dp->dl_file);
238 kmem_cache_free(deleg_slab, dp);
243 /* Remove the associated file_lock first, then remove the delegation.
244 * lease_modify() is called to remove the FS_LEASE file_lock from
245 * the i_flock list, eventually calling nfsd's lock_manager
246 * fl_release_callback.
249 nfs4_close_delegation(struct nfs4_delegation *dp)
251 struct file *filp = dp->dl_vfs_file;
253 dprintk("NFSD: close_delegation dp %p\n",dp);
254 dp->dl_vfs_file = NULL;
255 /* The following nfsd_close may not actually close the file,
256 * but we want to remove the lease in any case. */
258 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
262 /* Called under the state lock. */
264 unhash_delegation(struct nfs4_delegation *dp)
266 list_del_init(&dp->dl_perfile);
267 list_del_init(&dp->dl_perclnt);
268 spin_lock(&recall_lock);
269 list_del_init(&dp->dl_recall_lru);
270 spin_unlock(&recall_lock);
271 nfs4_close_delegation(dp);
272 nfs4_put_delegation(dp);
279 /* Hash tables for nfs4_clientid state */
280 #define CLIENT_HASH_BITS 4
281 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
282 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
284 #define clientid_hashval(id) \
285 ((id) & CLIENT_HASH_MASK)
286 #define clientstr_hashval(name) \
287 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
289 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
290 * used in reboot/reset lease grace period processing
292 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
293 * setclientid_confirmed info.
295 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
298 * client_lru holds client queue ordered by nfs4_client.cl_time
301 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
302 * for last close replay.
304 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
305 static int reclaim_str_hashtbl_size = 0;
306 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
307 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
308 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
309 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
310 static struct list_head client_lru;
311 static struct list_head close_lru;
313 static void unhash_generic_stateid(struct nfs4_stateid *stp)
315 list_del(&stp->st_hash);
316 list_del(&stp->st_perfile);
317 list_del(&stp->st_perstateowner);
320 static void free_generic_stateid(struct nfs4_stateid *stp)
322 put_nfs4_file(stp->st_file);
323 kmem_cache_free(stateid_slab, stp);
326 static void release_lock_stateid(struct nfs4_stateid *stp)
328 unhash_generic_stateid(stp);
329 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
330 free_generic_stateid(stp);
333 static void release_open_stateid(struct nfs4_stateid *stp)
335 unhash_generic_stateid(stp);
336 release_stateid_lockowners(stp);
337 nfsd_close(stp->st_vfs_file);
338 free_generic_stateid(stp);
342 renew_client(struct nfs4_client *clp)
345 * Move client to the end to the LRU list.
347 dprintk("renewing client (clientid %08x/%08x)\n",
348 clp->cl_clientid.cl_boot,
349 clp->cl_clientid.cl_id);
350 list_move_tail(&clp->cl_lru, &client_lru);
351 clp->cl_time = get_seconds();
354 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
356 STALE_CLIENTID(clientid_t *clid)
358 if (clid->cl_boot == boot_time)
360 dprintk("NFSD stale clientid (%08x/%08x)\n",
361 clid->cl_boot, clid->cl_id);
366 * XXX Should we use a slab cache ?
367 * This type of memory management is somewhat inefficient, but we use it
368 * anyway since SETCLIENTID is not a common operation.
370 static struct nfs4_client *alloc_client(struct xdr_netobj name)
372 struct nfs4_client *clp;
374 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
377 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
378 if (clp->cl_name.data == NULL) {
382 memcpy(clp->cl_name.data, name.data, name.len);
383 clp->cl_name.len = name.len;
388 shutdown_callback_client(struct nfs4_client *clp)
390 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
394 * Callback threads take a reference on the client, so there
395 * should be no outstanding callbacks at this point.
397 clp->cl_callback.cb_client = NULL;
398 rpc_shutdown_client(clnt);
403 free_client(struct nfs4_client *clp)
405 shutdown_callback_client(clp);
406 if (clp->cl_cred.cr_group_info)
407 put_group_info(clp->cl_cred.cr_group_info);
408 kfree(clp->cl_principal);
409 kfree(clp->cl_name.data);
414 put_nfs4_client(struct nfs4_client *clp)
416 if (atomic_dec_and_test(&clp->cl_count))
421 expire_client(struct nfs4_client *clp)
423 struct nfs4_stateowner *sop;
424 struct nfs4_delegation *dp;
425 struct list_head reaplist;
427 dprintk("NFSD: expire_client cl_count %d\n",
428 atomic_read(&clp->cl_count));
430 INIT_LIST_HEAD(&reaplist);
431 spin_lock(&recall_lock);
432 while (!list_empty(&clp->cl_delegations)) {
433 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
434 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
436 list_del_init(&dp->dl_perclnt);
437 list_move(&dp->dl_recall_lru, &reaplist);
439 spin_unlock(&recall_lock);
440 while (!list_empty(&reaplist)) {
441 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
442 list_del_init(&dp->dl_recall_lru);
443 unhash_delegation(dp);
445 list_del(&clp->cl_idhash);
446 list_del(&clp->cl_strhash);
447 list_del(&clp->cl_lru);
448 while (!list_empty(&clp->cl_openowners)) {
449 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
450 release_stateowner(sop);
452 put_nfs4_client(clp);
455 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
457 struct nfs4_client *clp;
459 clp = alloc_client(name);
462 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
463 atomic_set(&clp->cl_count, 1);
464 atomic_set(&clp->cl_callback.cb_set, 0);
465 INIT_LIST_HEAD(&clp->cl_idhash);
466 INIT_LIST_HEAD(&clp->cl_strhash);
467 INIT_LIST_HEAD(&clp->cl_openowners);
468 INIT_LIST_HEAD(&clp->cl_delegations);
469 INIT_LIST_HEAD(&clp->cl_lru);
473 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
475 memcpy(target->cl_verifier.data, source->data,
476 sizeof(target->cl_verifier.data));
479 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
481 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
482 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
485 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
487 target->cr_uid = source->cr_uid;
488 target->cr_gid = source->cr_gid;
489 target->cr_group_info = source->cr_group_info;
490 get_group_info(target->cr_group_info);
493 static int same_name(const char *n1, const char *n2)
495 return 0 == memcmp(n1, n2, HEXDIR_LEN);
499 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
501 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
505 same_clid(clientid_t *cl1, clientid_t *cl2)
507 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
510 /* XXX what about NGROUP */
512 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
514 return cr1->cr_uid == cr2->cr_uid;
517 static void gen_clid(struct nfs4_client *clp)
519 static u32 current_clientid = 1;
521 clp->cl_clientid.cl_boot = boot_time;
522 clp->cl_clientid.cl_id = current_clientid++;
525 static void gen_confirm(struct nfs4_client *clp)
530 p = (u32 *)clp->cl_confirm.data;
531 *p++ = get_seconds();
535 static int check_name(struct xdr_netobj name)
539 if (name.len > NFS4_OPAQUE_LIMIT) {
540 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
547 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
549 unsigned int idhashval;
551 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
552 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
553 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
554 list_add_tail(&clp->cl_lru, &client_lru);
555 clp->cl_time = get_seconds();
559 move_to_confirmed(struct nfs4_client *clp)
561 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
562 unsigned int strhashval;
564 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
565 list_del_init(&clp->cl_strhash);
566 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
567 strhashval = clientstr_hashval(clp->cl_recdir);
568 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
572 static struct nfs4_client *
573 find_confirmed_client(clientid_t *clid)
575 struct nfs4_client *clp;
576 unsigned int idhashval = clientid_hashval(clid->cl_id);
578 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
579 if (same_clid(&clp->cl_clientid, clid))
585 static struct nfs4_client *
586 find_unconfirmed_client(clientid_t *clid)
588 struct nfs4_client *clp;
589 unsigned int idhashval = clientid_hashval(clid->cl_id);
591 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
592 if (same_clid(&clp->cl_clientid, clid))
598 static struct nfs4_client *
599 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
601 struct nfs4_client *clp;
603 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
604 if (same_name(clp->cl_recdir, dname))
610 static struct nfs4_client *
611 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
613 struct nfs4_client *clp;
615 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
616 if (same_name(clp->cl_recdir, dname))
622 /* a helper function for parse_callback */
624 parse_octet(unsigned int *lenp, char **addrp)
626 unsigned int len = *lenp;
638 if ((c < '0') || (c > '9')) {
644 n = (n * 10) + (c - '0');
655 /* parse and set the setclientid ipv4 callback address */
657 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
662 u32 addrlen = addr_len;
663 char *addr = addr_val;
668 for(i = 4; i > 0 ; i--) {
669 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
672 cbaddr |= (temp << shift);
680 for(i = 2; i > 0 ; i--) {
681 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
684 cbport |= (temp << shift);
693 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
695 struct nfs4_callback *cb = &clp->cl_callback;
697 /* Currently, we only support tcp for the callback channel */
698 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
701 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
702 &cb->cb_addr, &cb->cb_port)))
704 cb->cb_prog = se->se_callback_prog;
705 cb->cb_ident = se->se_callback_ident;
708 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
709 "will not receive delegations\n",
710 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
716 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
717 struct nfsd4_setclientid *setclid)
719 struct sockaddr_in *sin = svc_addr_in(rqstp);
720 struct xdr_netobj clname = {
721 .len = setclid->se_namelen,
722 .data = setclid->se_name,
724 nfs4_verifier clverifier = setclid->se_verf;
725 unsigned int strhashval;
726 struct nfs4_client *conf, *unconf, *new;
729 char dname[HEXDIR_LEN];
731 if (!check_name(clname))
734 status = nfs4_make_rec_clidname(dname, &clname);
739 * XXX The Duplicate Request Cache (DRC) has been checked (??)
740 * We get here on a DRC miss.
743 strhashval = clientstr_hashval(dname);
746 conf = find_confirmed_client_by_str(dname, strhashval);
748 /* RFC 3530 14.2.33 CASE 0: */
749 status = nfserr_clid_inuse;
750 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)
751 || conf->cl_addr != sin->sin_addr.s_addr) {
752 dprintk("NFSD: setclientid: string in use by clientat %pI4\n",
758 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
759 * has a description of SETCLIENTID request processing consisting
760 * of 5 bullet points, labeled as CASE0 - CASE4 below.
762 unconf = find_unconfirmed_client_by_str(dname, strhashval);
763 status = nfserr_resource;
766 * RFC 3530 14.2.33 CASE 4:
767 * placed first, because it is the normal case
770 expire_client(unconf);
771 new = create_client(clname, dname);
775 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
777 * RFC 3530 14.2.33 CASE 1:
778 * probable callback update
781 /* Note this is removing unconfirmed {*x***},
782 * which is stronger than RFC recommended {vxc**}.
783 * This has the advantage that there is at most
784 * one {*x***} in either list at any time.
786 expire_client(unconf);
788 new = create_client(clname, dname);
791 copy_clid(new, conf);
792 } else if (!unconf) {
794 * RFC 3530 14.2.33 CASE 2:
795 * probable client reboot; state will be removed if
798 new = create_client(clname, dname);
804 * RFC 3530 14.2.33 CASE 3:
805 * probable client reboot; state will be removed if
808 expire_client(unconf);
809 new = create_client(clname, dname);
814 copy_verf(new, &clverifier);
815 new->cl_addr = sin->sin_addr.s_addr;
816 new->cl_flavor = rqstp->rq_flavor;
817 princ = svc_gss_principal(rqstp);
819 new->cl_principal = kstrdup(princ, GFP_KERNEL);
820 if (new->cl_principal == NULL) {
825 copy_cred(&new->cl_cred, &rqstp->rq_cred);
827 gen_callback(new, setclid);
828 add_to_unconfirmed(new, strhashval);
829 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
830 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
831 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
840 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
841 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
842 * bullets, labeled as CASE1 - CASE4 below.
845 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
846 struct nfsd4_compound_state *cstate,
847 struct nfsd4_setclientid_confirm *setclientid_confirm)
849 struct sockaddr_in *sin = svc_addr_in(rqstp);
850 struct nfs4_client *conf, *unconf;
851 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
852 clientid_t * clid = &setclientid_confirm->sc_clientid;
855 if (STALE_CLIENTID(clid))
856 return nfserr_stale_clientid;
858 * XXX The Duplicate Request Cache (DRC) has been checked (??)
859 * We get here on a DRC miss.
864 conf = find_confirmed_client(clid);
865 unconf = find_unconfirmed_client(clid);
867 status = nfserr_clid_inuse;
868 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
870 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
874 * section 14.2.34 of RFC 3530 has a description of
875 * SETCLIENTID_CONFIRM request processing consisting
876 * of 4 bullet points, labeled as CASE1 - CASE4 below.
878 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
880 * RFC 3530 14.2.34 CASE 1:
883 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
884 status = nfserr_clid_inuse;
886 /* XXX: We just turn off callbacks until we can handle
887 * change request correctly. */
888 atomic_set(&conf->cl_callback.cb_set, 0);
890 nfsd4_remove_clid_dir(unconf);
891 expire_client(unconf);
895 } else if (conf && !unconf) {
897 * RFC 3530 14.2.34 CASE 2:
898 * probable retransmitted request; play it safe and
901 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
902 status = nfserr_clid_inuse;
905 } else if (!conf && unconf
906 && same_verf(&unconf->cl_confirm, &confirm)) {
908 * RFC 3530 14.2.34 CASE 3:
909 * Normal case; new or rebooted client:
911 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
912 status = nfserr_clid_inuse;
915 clientstr_hashval(unconf->cl_recdir);
916 conf = find_confirmed_client_by_str(unconf->cl_recdir,
919 nfsd4_remove_clid_dir(conf);
922 move_to_confirmed(unconf);
924 nfsd4_probe_callback(conf);
927 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
928 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
931 * RFC 3530 14.2.34 CASE 4:
932 * Client probably hasn't noticed that we rebooted yet.
934 status = nfserr_stale_clientid;
936 /* check that we have hit one of the cases...*/
937 status = nfserr_clid_inuse;
944 /* OPEN Share state helper functions */
945 static inline struct nfs4_file *
946 alloc_init_file(struct inode *ino)
948 struct nfs4_file *fp;
949 unsigned int hashval = file_hashval(ino);
951 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
953 kref_init(&fp->fi_ref);
954 INIT_LIST_HEAD(&fp->fi_hash);
955 INIT_LIST_HEAD(&fp->fi_stateids);
956 INIT_LIST_HEAD(&fp->fi_delegations);
957 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
958 fp->fi_inode = igrab(ino);
959 fp->fi_id = current_fileid++;
960 fp->fi_had_conflict = false;
967 nfsd4_free_slab(struct kmem_cache **slab)
971 kmem_cache_destroy(*slab);
976 nfsd4_free_slabs(void)
978 nfsd4_free_slab(&stateowner_slab);
979 nfsd4_free_slab(&file_slab);
980 nfsd4_free_slab(&stateid_slab);
981 nfsd4_free_slab(&deleg_slab);
985 nfsd4_init_slabs(void)
987 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
988 sizeof(struct nfs4_stateowner), 0, 0, NULL);
989 if (stateowner_slab == NULL)
991 file_slab = kmem_cache_create("nfsd4_files",
992 sizeof(struct nfs4_file), 0, 0, NULL);
993 if (file_slab == NULL)
995 stateid_slab = kmem_cache_create("nfsd4_stateids",
996 sizeof(struct nfs4_stateid), 0, 0, NULL);
997 if (stateid_slab == NULL)
999 deleg_slab = kmem_cache_create("nfsd4_delegations",
1000 sizeof(struct nfs4_delegation), 0, 0, NULL);
1001 if (deleg_slab == NULL)
1006 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1011 nfs4_free_stateowner(struct kref *kref)
1013 struct nfs4_stateowner *sop =
1014 container_of(kref, struct nfs4_stateowner, so_ref);
1015 kfree(sop->so_owner.data);
1016 kmem_cache_free(stateowner_slab, sop);
1019 static inline struct nfs4_stateowner *
1020 alloc_stateowner(struct xdr_netobj *owner)
1022 struct nfs4_stateowner *sop;
1024 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1025 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1026 memcpy(sop->so_owner.data, owner->data, owner->len);
1027 sop->so_owner.len = owner->len;
1028 kref_init(&sop->so_ref);
1031 kmem_cache_free(stateowner_slab, sop);
1036 static struct nfs4_stateowner *
1037 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1038 struct nfs4_stateowner *sop;
1039 struct nfs4_replay *rp;
1040 unsigned int idhashval;
1042 if (!(sop = alloc_stateowner(&open->op_owner)))
1044 idhashval = ownerid_hashval(current_ownerid);
1045 INIT_LIST_HEAD(&sop->so_idhash);
1046 INIT_LIST_HEAD(&sop->so_strhash);
1047 INIT_LIST_HEAD(&sop->so_perclient);
1048 INIT_LIST_HEAD(&sop->so_stateids);
1049 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1050 INIT_LIST_HEAD(&sop->so_close_lru);
1052 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1053 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1054 list_add(&sop->so_perclient, &clp->cl_openowners);
1055 sop->so_is_open_owner = 1;
1056 sop->so_id = current_ownerid++;
1057 sop->so_client = clp;
1058 sop->so_seqid = open->op_seqid;
1059 sop->so_confirmed = 0;
1060 rp = &sop->so_replay;
1061 rp->rp_status = nfserr_serverfault;
1063 rp->rp_buf = rp->rp_ibuf;
1068 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1070 struct nfs4_stateowner *lock_sop;
1072 while (!list_empty(&open_stp->st_lockowners)) {
1073 lock_sop = list_entry(open_stp->st_lockowners.next,
1074 struct nfs4_stateowner, so_perstateid);
1075 /* list_del(&open_stp->st_lockowners); */
1076 BUG_ON(lock_sop->so_is_open_owner);
1077 release_stateowner(lock_sop);
1082 unhash_stateowner(struct nfs4_stateowner *sop)
1084 struct nfs4_stateid *stp;
1086 list_del(&sop->so_idhash);
1087 list_del(&sop->so_strhash);
1088 if (sop->so_is_open_owner)
1089 list_del(&sop->so_perclient);
1090 list_del(&sop->so_perstateid);
1091 while (!list_empty(&sop->so_stateids)) {
1092 stp = list_entry(sop->so_stateids.next,
1093 struct nfs4_stateid, st_perstateowner);
1094 if (sop->so_is_open_owner)
1095 release_open_stateid(stp);
1097 release_lock_stateid(stp);
1102 release_stateowner(struct nfs4_stateowner *sop)
1104 unhash_stateowner(sop);
1105 list_del(&sop->so_close_lru);
1106 nfs4_put_stateowner(sop);
1110 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1111 struct nfs4_stateowner *sop = open->op_stateowner;
1112 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1114 INIT_LIST_HEAD(&stp->st_hash);
1115 INIT_LIST_HEAD(&stp->st_perstateowner);
1116 INIT_LIST_HEAD(&stp->st_lockowners);
1117 INIT_LIST_HEAD(&stp->st_perfile);
1118 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1119 list_add(&stp->st_perstateowner, &sop->so_stateids);
1120 list_add(&stp->st_perfile, &fp->fi_stateids);
1121 stp->st_stateowner = sop;
1124 stp->st_stateid.si_boot = boot_time;
1125 stp->st_stateid.si_stateownerid = sop->so_id;
1126 stp->st_stateid.si_fileid = fp->fi_id;
1127 stp->st_stateid.si_generation = 0;
1128 stp->st_access_bmap = 0;
1129 stp->st_deny_bmap = 0;
1130 __set_bit(open->op_share_access, &stp->st_access_bmap);
1131 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1132 stp->st_openstp = NULL;
1136 move_to_close_lru(struct nfs4_stateowner *sop)
1138 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1140 list_move_tail(&sop->so_close_lru, &close_lru);
1141 sop->so_time = get_seconds();
1145 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1148 return (sop->so_owner.len == owner->len) &&
1149 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1150 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1153 static struct nfs4_stateowner *
1154 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1156 struct nfs4_stateowner *so = NULL;
1158 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1159 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1165 /* search file_hashtbl[] for file */
1166 static struct nfs4_file *
1167 find_file(struct inode *ino)
1169 unsigned int hashval = file_hashval(ino);
1170 struct nfs4_file *fp;
1172 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1173 if (fp->fi_inode == ino) {
1181 static inline int access_valid(u32 x)
1183 if (x < NFS4_SHARE_ACCESS_READ)
1185 if (x > NFS4_SHARE_ACCESS_BOTH)
1190 static inline int deny_valid(u32 x)
1192 /* Note: unlike access bits, deny bits may be zero. */
1193 return x <= NFS4_SHARE_DENY_BOTH;
1197 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1198 * st_{access,deny}_bmap field of the stateid, in order to track not
1199 * only what share bits are currently in force, but also what
1200 * combinations of share bits previous opens have used. This allows us
1201 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1202 * return an error if the client attempt to downgrade to a combination
1203 * of share bits not explicable by closing some of its previous opens.
1205 * XXX: This enforcement is actually incomplete, since we don't keep
1206 * track of access/deny bit combinations; so, e.g., we allow:
1208 * OPEN allow read, deny write
1209 * OPEN allow both, deny none
1210 * DOWNGRADE allow read, deny none
1212 * which we should reject.
1215 set_access(unsigned int *access, unsigned long bmap) {
1219 for (i = 1; i < 4; i++) {
1220 if (test_bit(i, &bmap))
1226 set_deny(unsigned int *deny, unsigned long bmap) {
1230 for (i = 0; i < 4; i++) {
1231 if (test_bit(i, &bmap))
1237 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1238 unsigned int access, deny;
1240 set_access(&access, stp->st_access_bmap);
1241 set_deny(&deny, stp->st_deny_bmap);
1242 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1248 * Called to check deny when READ with all zero stateid or
1249 * WRITE with all zero or all one stateid
1252 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1254 struct inode *ino = current_fh->fh_dentry->d_inode;
1255 struct nfs4_file *fp;
1256 struct nfs4_stateid *stp;
1259 dprintk("NFSD: nfs4_share_conflict\n");
1261 fp = find_file(ino);
1264 ret = nfserr_locked;
1265 /* Search for conflicting share reservations */
1266 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1267 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1268 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1278 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1280 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1281 drop_file_write_access(filp);
1282 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1287 * Recall a delegation
1290 do_recall(void *__dp)
1292 struct nfs4_delegation *dp = __dp;
1294 dp->dl_file->fi_had_conflict = true;
1295 nfsd4_cb_recall(dp);
1300 * Spawn a thread to perform a recall on the delegation represented
1301 * by the lease (file_lock)
1303 * Called from break_lease() with lock_kernel() held.
1304 * Note: we assume break_lease will only call this *once* for any given
1308 void nfsd_break_deleg_cb(struct file_lock *fl)
1310 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1311 struct task_struct *t;
1313 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1317 /* We're assuming the state code never drops its reference
1318 * without first removing the lease. Since we're in this lease
1319 * callback (and since the lease code is serialized by the kernel
1320 * lock) we know the server hasn't removed the lease yet, we know
1321 * it's safe to take a reference: */
1322 atomic_inc(&dp->dl_count);
1323 atomic_inc(&dp->dl_client->cl_count);
1325 spin_lock(&recall_lock);
1326 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1327 spin_unlock(&recall_lock);
1329 /* only place dl_time is set. protected by lock_kernel*/
1330 dp->dl_time = get_seconds();
1333 * We don't want the locks code to timeout the lease for us;
1334 * we'll remove it ourself if the delegation isn't returned
1337 fl->fl_break_time = 0;
1339 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1341 struct nfs4_client *clp = dp->dl_client;
1343 printk(KERN_INFO "NFSD: Callback thread failed for "
1344 "for client (clientid %08x/%08x)\n",
1345 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1346 put_nfs4_client(dp->dl_client);
1347 nfs4_put_delegation(dp);
1352 * The file_lock is being reapd.
1354 * Called by locks_free_lock() with lock_kernel() held.
1357 void nfsd_release_deleg_cb(struct file_lock *fl)
1359 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1361 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1363 if (!(fl->fl_flags & FL_LEASE) || !dp)
1365 dp->dl_flock = NULL;
1369 * Set the delegation file_lock back pointer.
1371 * Called from setlease() with lock_kernel() held.
1374 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1376 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1378 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1385 * Called from setlease() with lock_kernel() held
1388 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1390 struct nfs4_delegation *onlistd =
1391 (struct nfs4_delegation *)onlist->fl_owner;
1392 struct nfs4_delegation *tryd =
1393 (struct nfs4_delegation *)try->fl_owner;
1395 if (onlist->fl_lmops != try->fl_lmops)
1398 return onlistd->dl_client == tryd->dl_client;
1403 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1406 return lease_modify(onlist, arg);
1411 static struct lock_manager_operations nfsd_lease_mng_ops = {
1412 .fl_break = nfsd_break_deleg_cb,
1413 .fl_release_private = nfsd_release_deleg_cb,
1414 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1415 .fl_mylease = nfsd_same_client_deleg_cb,
1416 .fl_change = nfsd_change_deleg_cb,
1421 nfsd4_process_open1(struct nfsd4_open *open)
1423 clientid_t *clientid = &open->op_clientid;
1424 struct nfs4_client *clp = NULL;
1425 unsigned int strhashval;
1426 struct nfs4_stateowner *sop = NULL;
1428 if (!check_name(open->op_owner))
1429 return nfserr_inval;
1431 if (STALE_CLIENTID(&open->op_clientid))
1432 return nfserr_stale_clientid;
1434 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1435 sop = find_openstateowner_str(strhashval, open);
1436 open->op_stateowner = sop;
1438 /* Make sure the client's lease hasn't expired. */
1439 clp = find_confirmed_client(clientid);
1441 return nfserr_expired;
1444 if (!sop->so_confirmed) {
1445 /* Replace unconfirmed owners without checking for replay. */
1446 clp = sop->so_client;
1447 release_stateowner(sop);
1448 open->op_stateowner = NULL;
1451 if (open->op_seqid == sop->so_seqid - 1) {
1452 if (sop->so_replay.rp_buflen)
1453 return nfserr_replay_me;
1454 /* The original OPEN failed so spectacularly
1455 * that we don't even have replay data saved!
1456 * Therefore, we have no choice but to continue
1457 * processing this OPEN; presumably, we'll
1458 * fail again for the same reason.
1460 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1463 if (open->op_seqid != sop->so_seqid)
1464 return nfserr_bad_seqid;
1466 if (open->op_stateowner == NULL) {
1467 sop = alloc_init_open_stateowner(strhashval, clp, open);
1469 return nfserr_resource;
1470 open->op_stateowner = sop;
1472 list_del_init(&sop->so_close_lru);
1473 renew_client(sop->so_client);
1477 static inline __be32
1478 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1480 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1481 return nfserr_openmode;
1486 static struct nfs4_delegation *
1487 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1489 struct nfs4_delegation *dp;
1491 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1492 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1499 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1500 struct nfs4_delegation **dp)
1503 __be32 status = nfserr_bad_stateid;
1505 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1508 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1509 RD_STATE : WR_STATE;
1510 status = nfs4_check_delegmode(*dp, flags);
1514 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1518 open->op_stateowner->so_confirmed = 1;
1523 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1525 struct nfs4_stateid *local;
1526 __be32 status = nfserr_share_denied;
1527 struct nfs4_stateowner *sop = open->op_stateowner;
1529 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1530 /* ignore lock owners */
1531 if (local->st_stateowner->so_is_open_owner == 0)
1533 /* remember if we have seen this open owner */
1534 if (local->st_stateowner == sop)
1536 /* check for conflicting share reservations */
1537 if (!test_share(local, open))
1545 static inline struct nfs4_stateid *
1546 nfs4_alloc_stateid(void)
1548 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1552 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1553 struct nfs4_delegation *dp,
1554 struct svc_fh *cur_fh, int flags)
1556 struct nfs4_stateid *stp;
1558 stp = nfs4_alloc_stateid();
1560 return nfserr_resource;
1563 get_file(dp->dl_vfs_file);
1564 stp->st_vfs_file = dp->dl_vfs_file;
1567 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1570 if (status == nfserr_dropit)
1571 status = nfserr_jukebox;
1572 kmem_cache_free(stateid_slab, stp);
1580 static inline __be32
1581 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1582 struct nfsd4_open *open)
1584 struct iattr iattr = {
1585 .ia_valid = ATTR_SIZE,
1588 if (!open->op_truncate)
1590 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1591 return nfserr_inval;
1592 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1596 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1598 struct file *filp = stp->st_vfs_file;
1599 struct inode *inode = filp->f_path.dentry->d_inode;
1600 unsigned int share_access, new_writer;
1603 set_access(&share_access, stp->st_access_bmap);
1604 new_writer = (~share_access) & open->op_share_access
1605 & NFS4_SHARE_ACCESS_WRITE;
1608 int err = get_write_access(inode);
1610 return nfserrno(err);
1611 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1613 return nfserrno(err);
1614 file_take_write(filp);
1616 status = nfsd4_truncate(rqstp, cur_fh, open);
1619 put_write_access(inode);
1622 /* remember the open */
1623 filp->f_mode |= open->op_share_access;
1624 __set_bit(open->op_share_access, &stp->st_access_bmap);
1625 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1632 nfs4_set_claim_prev(struct nfsd4_open *open)
1634 open->op_stateowner->so_confirmed = 1;
1635 open->op_stateowner->so_client->cl_firststate = 1;
1639 * Attempt to hand out a delegation.
1642 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1644 struct nfs4_delegation *dp;
1645 struct nfs4_stateowner *sop = stp->st_stateowner;
1646 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1647 struct file_lock fl, *flp = &fl;
1648 int status, flag = 0;
1650 flag = NFS4_OPEN_DELEGATE_NONE;
1651 open->op_recall = 0;
1652 switch (open->op_claim_type) {
1653 case NFS4_OPEN_CLAIM_PREVIOUS:
1654 if (!atomic_read(&cb->cb_set))
1655 open->op_recall = 1;
1656 flag = open->op_delegate_type;
1657 if (flag == NFS4_OPEN_DELEGATE_NONE)
1660 case NFS4_OPEN_CLAIM_NULL:
1661 /* Let's not give out any delegations till everyone's
1662 * had the chance to reclaim theirs.... */
1663 if (locks_in_grace())
1665 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1667 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1668 flag = NFS4_OPEN_DELEGATE_WRITE;
1670 flag = NFS4_OPEN_DELEGATE_READ;
1676 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1678 flag = NFS4_OPEN_DELEGATE_NONE;
1681 locks_init_lock(&fl);
1682 fl.fl_lmops = &nfsd_lease_mng_ops;
1683 fl.fl_flags = FL_LEASE;
1684 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1685 fl.fl_end = OFFSET_MAX;
1686 fl.fl_owner = (fl_owner_t)dp;
1687 fl.fl_file = stp->st_vfs_file;
1688 fl.fl_pid = current->tgid;
1690 /* vfs_setlease checks to see if delegation should be handed out.
1691 * the lock_manager callbacks fl_mylease and fl_change are used
1693 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1694 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1695 unhash_delegation(dp);
1696 flag = NFS4_OPEN_DELEGATE_NONE;
1700 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1702 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1703 dp->dl_stateid.si_boot,
1704 dp->dl_stateid.si_stateownerid,
1705 dp->dl_stateid.si_fileid,
1706 dp->dl_stateid.si_generation);
1708 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1709 && flag == NFS4_OPEN_DELEGATE_NONE
1710 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1711 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1712 open->op_delegate_type = flag;
1716 * called with nfs4_lock_state() held.
1719 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1721 struct nfs4_file *fp = NULL;
1722 struct inode *ino = current_fh->fh_dentry->d_inode;
1723 struct nfs4_stateid *stp = NULL;
1724 struct nfs4_delegation *dp = NULL;
1727 status = nfserr_inval;
1728 if (!access_valid(open->op_share_access)
1729 || !deny_valid(open->op_share_deny))
1732 * Lookup file; if found, lookup stateid and check open request,
1733 * and check for delegations in the process of being recalled.
1734 * If not found, create the nfs4_file struct
1736 fp = find_file(ino);
1738 if ((status = nfs4_check_open(fp, open, &stp)))
1740 status = nfs4_check_deleg(fp, open, &dp);
1744 status = nfserr_bad_stateid;
1745 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1747 status = nfserr_resource;
1748 fp = alloc_init_file(ino);
1754 * OPEN the file, or upgrade an existing OPEN.
1755 * If truncate fails, the OPEN fails.
1758 /* Stateid was found, this is an OPEN upgrade */
1759 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1762 update_stateid(&stp->st_stateid);
1764 /* Stateid was not found, this is a new OPEN */
1766 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1767 flags |= NFSD_MAY_READ;
1768 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1769 flags |= NFSD_MAY_WRITE;
1770 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1773 init_stateid(stp, fp, open);
1774 status = nfsd4_truncate(rqstp, current_fh, open);
1776 release_open_stateid(stp);
1780 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1783 * Attempt to hand out a delegation. No error return, because the
1784 * OPEN succeeds even if we fail.
1786 nfs4_open_delegation(current_fh, open, stp);
1790 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1791 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1792 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1796 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1797 nfs4_set_claim_prev(open);
1799 * To finish the open response, we just need to set the rflags.
1801 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1802 if (!open->op_stateowner->so_confirmed)
1803 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1809 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1812 struct nfs4_client *clp;
1816 dprintk("process_renew(%08x/%08x): starting\n",
1817 clid->cl_boot, clid->cl_id);
1818 status = nfserr_stale_clientid;
1819 if (STALE_CLIENTID(clid))
1821 clp = find_confirmed_client(clid);
1822 status = nfserr_expired;
1824 /* We assume the client took too long to RENEW. */
1825 dprintk("nfsd4_renew: clientid not found!\n");
1829 status = nfserr_cb_path_down;
1830 if (!list_empty(&clp->cl_delegations)
1831 && !atomic_read(&clp->cl_callback.cb_set))
1835 nfs4_unlock_state();
1839 struct lock_manager nfsd4_manager = {
1843 nfsd4_end_grace(void)
1845 dprintk("NFSD: end of grace period\n");
1846 nfsd4_recdir_purge_old();
1847 locks_end_grace(&nfsd4_manager);
1851 nfs4_laundromat(void)
1853 struct nfs4_client *clp;
1854 struct nfs4_stateowner *sop;
1855 struct nfs4_delegation *dp;
1856 struct list_head *pos, *next, reaplist;
1857 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1858 time_t t, clientid_val = NFSD_LEASE_TIME;
1859 time_t u, test_val = NFSD_LEASE_TIME;
1863 dprintk("NFSD: laundromat service - starting\n");
1864 if (locks_in_grace())
1866 list_for_each_safe(pos, next, &client_lru) {
1867 clp = list_entry(pos, struct nfs4_client, cl_lru);
1868 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1869 t = clp->cl_time - cutoff;
1870 if (clientid_val > t)
1874 dprintk("NFSD: purging unused client (clientid %08x)\n",
1875 clp->cl_clientid.cl_id);
1876 nfsd4_remove_clid_dir(clp);
1879 INIT_LIST_HEAD(&reaplist);
1880 spin_lock(&recall_lock);
1881 list_for_each_safe(pos, next, &del_recall_lru) {
1882 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1883 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1884 u = dp->dl_time - cutoff;
1889 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1891 list_move(&dp->dl_recall_lru, &reaplist);
1893 spin_unlock(&recall_lock);
1894 list_for_each_safe(pos, next, &reaplist) {
1895 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1896 list_del_init(&dp->dl_recall_lru);
1897 unhash_delegation(dp);
1899 test_val = NFSD_LEASE_TIME;
1900 list_for_each_safe(pos, next, &close_lru) {
1901 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1902 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1903 u = sop->so_time - cutoff;
1908 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1910 release_stateowner(sop);
1912 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1913 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1914 nfs4_unlock_state();
1915 return clientid_val;
1918 static struct workqueue_struct *laundry_wq;
1919 static void laundromat_main(struct work_struct *);
1920 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1923 laundromat_main(struct work_struct *not_used)
1927 t = nfs4_laundromat();
1928 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1929 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1932 static struct nfs4_stateowner *
1933 search_close_lru(u32 st_id, int flags)
1935 struct nfs4_stateowner *local = NULL;
1937 if (flags & CLOSE_STATE) {
1938 list_for_each_entry(local, &close_lru, so_close_lru) {
1939 if (local->so_id == st_id)
1947 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1949 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1953 STALE_STATEID(stateid_t *stateid)
1955 if (stateid->si_boot == boot_time)
1957 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1958 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1959 stateid->si_generation);
1964 access_permit_read(unsigned long access_bmap)
1966 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1967 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1968 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1972 access_permit_write(unsigned long access_bmap)
1974 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1975 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1979 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1981 __be32 status = nfserr_openmode;
1983 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
1985 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
1992 static inline __be32
1993 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
1995 /* Trying to call delegreturn with a special stateid? Yuch: */
1996 if (!(flags & (RD_STATE | WR_STATE)))
1997 return nfserr_bad_stateid;
1998 else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2000 else if (locks_in_grace()) {
2001 /* Answer in remaining cases depends on existance of
2002 * conflicting state; so we must wait out the grace period. */
2003 return nfserr_grace;
2004 } else if (flags & WR_STATE)
2005 return nfs4_share_conflict(current_fh,
2006 NFS4_SHARE_DENY_WRITE);
2007 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2008 return nfs4_share_conflict(current_fh,
2009 NFS4_SHARE_DENY_READ);
2013 * Allow READ/WRITE during grace period on recovered state only for files
2014 * that are not able to provide mandatory locking.
2017 io_during_grace_disallowed(struct inode *inode, int flags)
2019 return locks_in_grace() && (flags & (RD_STATE | WR_STATE))
2020 && mandatory_lock(inode);
2023 static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2025 /* If the client sends us a stateid from the future, it's buggy: */
2026 if (in->si_generation > ref->si_generation)
2027 return nfserr_bad_stateid;
2029 * The following, however, can happen. For example, if the
2030 * client sends an open and some IO at the same time, the open
2031 * may bump si_generation while the IO is still in flight.
2032 * Thanks to hard links and renames, the client never knows what
2033 * file an open will affect. So it could avoid that situation
2034 * only by serializing all opens and IO from the same open
2035 * owner. To recover from the old_stateid error, the client
2036 * will just have to retry the IO:
2038 if (in->si_generation < ref->si_generation)
2039 return nfserr_old_stateid;
2044 * Checks for stateid operations
2047 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2049 struct nfs4_stateid *stp = NULL;
2050 struct nfs4_delegation *dp = NULL;
2052 struct inode *ino = current_fh->fh_dentry->d_inode;
2055 dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2056 stateid->si_boot, stateid->si_stateownerid,
2057 stateid->si_fileid, stateid->si_generation);
2061 if (io_during_grace_disallowed(ino, flags))
2062 return nfserr_grace;
2064 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2065 return check_special_stateids(current_fh, stateid, flags);
2068 status = nfserr_stale_stateid;
2069 if (STALE_STATEID(stateid))
2073 status = nfserr_bad_stateid;
2074 if (!stateid->si_fileid) { /* delegation stateid */
2075 if(!(dp = find_delegation_stateid(ino, stateid))) {
2076 dprintk("NFSD: delegation stateid not found\n");
2079 stidp = &dp->dl_stateid;
2080 } else { /* open or lock stateid */
2081 if (!(stp = find_stateid(stateid, flags))) {
2082 dprintk("NFSD: open or lock stateid not found\n");
2085 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2087 if (!stp->st_stateowner->so_confirmed)
2089 stidp = &stp->st_stateid;
2091 status = check_stateid_generation(stateid, stidp);
2095 if ((status = nfs4_check_openmode(stp,flags)))
2097 renew_client(stp->st_stateowner->so_client);
2099 *filpp = stp->st_vfs_file;
2101 if ((status = nfs4_check_delegmode(dp, flags)))
2103 renew_client(dp->dl_client);
2104 if (flags & DELEG_RET)
2105 unhash_delegation(dp);
2107 *filpp = dp->dl_vfs_file;
2117 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2118 RD_STATE : WR_STATE;
2122 * Checks for sequence id mutating operations.
2125 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2127 struct nfs4_stateid *stp;
2128 struct nfs4_stateowner *sop;
2131 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2132 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2133 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2134 stateid->si_generation);
2139 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2140 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2141 return nfserr_bad_stateid;
2144 if (STALE_STATEID(stateid))
2145 return nfserr_stale_stateid;
2147 * We return BAD_STATEID if filehandle doesn't match stateid,
2148 * the confirmed flag is incorrecly set, or the generation
2149 * number is incorrect.
2151 stp = find_stateid(stateid, flags);
2154 * Also, we should make sure this isn't just the result of
2157 sop = search_close_lru(stateid->si_stateownerid, flags);
2159 return nfserr_bad_stateid;
2165 *sopp = sop = stp->st_stateowner;
2168 clientid_t *lockclid = &lock->v.new.clientid;
2169 struct nfs4_client *clp = sop->so_client;
2173 lkflg = setlkflg(lock->lk_type);
2175 if (lock->lk_is_new) {
2176 if (!sop->so_is_open_owner)
2177 return nfserr_bad_stateid;
2178 if (!same_clid(&clp->cl_clientid, lockclid))
2179 return nfserr_bad_stateid;
2180 /* stp is the open stateid */
2181 status = nfs4_check_openmode(stp, lkflg);
2185 /* stp is the lock stateid */
2186 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2192 if (nfs4_check_fh(current_fh, stp)) {
2193 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2194 return nfserr_bad_stateid;
2198 * We now validate the seqid and stateid generation numbers.
2199 * For the moment, we ignore the possibility of
2200 * generation number wraparound.
2202 if (seqid != sop->so_seqid)
2205 if (sop->so_confirmed && flags & CONFIRM) {
2206 dprintk("NFSD: preprocess_seqid_op: expected"
2207 " unconfirmed stateowner!\n");
2208 return nfserr_bad_stateid;
2210 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2211 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2212 " confirmed yet!\n");
2213 return nfserr_bad_stateid;
2215 status = check_stateid_generation(stateid, &stp->st_stateid);
2218 renew_client(sop->so_client);
2222 if (seqid == sop->so_seqid - 1) {
2223 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2224 /* indicate replay to calling function */
2225 return nfserr_replay_me;
2227 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2228 sop->so_seqid, seqid);
2230 return nfserr_bad_seqid;
2234 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2235 struct nfsd4_open_confirm *oc)
2238 struct nfs4_stateowner *sop;
2239 struct nfs4_stateid *stp;
2241 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2242 (int)cstate->current_fh.fh_dentry->d_name.len,
2243 cstate->current_fh.fh_dentry->d_name.name);
2245 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2251 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2252 oc->oc_seqid, &oc->oc_req_stateid,
2253 CONFIRM | OPEN_STATE,
2254 &oc->oc_stateowner, &stp, NULL)))
2257 sop = oc->oc_stateowner;
2258 sop->so_confirmed = 1;
2259 update_stateid(&stp->st_stateid);
2260 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2261 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2262 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2263 stp->st_stateid.si_boot,
2264 stp->st_stateid.si_stateownerid,
2265 stp->st_stateid.si_fileid,
2266 stp->st_stateid.si_generation);
2268 nfsd4_create_clid_dir(sop->so_client);
2270 if (oc->oc_stateowner) {
2271 nfs4_get_stateowner(oc->oc_stateowner);
2272 cstate->replay_owner = oc->oc_stateowner;
2274 nfs4_unlock_state();
2280 * unset all bits in union bitmap (bmap) that
2281 * do not exist in share (from successful OPEN_DOWNGRADE)
2284 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2287 for (i = 1; i < 4; i++) {
2288 if ((i & access) != i)
2289 __clear_bit(i, bmap);
2294 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2297 for (i = 0; i < 4; i++) {
2298 if ((i & deny) != i)
2299 __clear_bit(i, bmap);
2304 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2305 struct nfsd4_compound_state *cstate,
2306 struct nfsd4_open_downgrade *od)
2309 struct nfs4_stateid *stp;
2310 unsigned int share_access;
2312 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2313 (int)cstate->current_fh.fh_dentry->d_name.len,
2314 cstate->current_fh.fh_dentry->d_name.name);
2316 if (!access_valid(od->od_share_access)
2317 || !deny_valid(od->od_share_deny))
2318 return nfserr_inval;
2321 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2325 &od->od_stateowner, &stp, NULL)))
2328 status = nfserr_inval;
2329 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2330 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2331 stp->st_access_bmap, od->od_share_access);
2334 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2335 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2336 stp->st_deny_bmap, od->od_share_deny);
2339 set_access(&share_access, stp->st_access_bmap);
2340 nfs4_file_downgrade(stp->st_vfs_file,
2341 share_access & ~od->od_share_access);
2343 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2344 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2346 update_stateid(&stp->st_stateid);
2347 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2350 if (od->od_stateowner) {
2351 nfs4_get_stateowner(od->od_stateowner);
2352 cstate->replay_owner = od->od_stateowner;
2354 nfs4_unlock_state();
2359 * nfs4_unlock_state() called after encode
2362 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2363 struct nfsd4_close *close)
2366 struct nfs4_stateid *stp;
2368 dprintk("NFSD: nfsd4_close on file %.*s\n",
2369 (int)cstate->current_fh.fh_dentry->d_name.len,
2370 cstate->current_fh.fh_dentry->d_name.name);
2373 /* check close_lru for replay */
2374 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2377 OPEN_STATE | CLOSE_STATE,
2378 &close->cl_stateowner, &stp, NULL)))
2381 update_stateid(&stp->st_stateid);
2382 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2384 /* release_stateid() calls nfsd_close() if needed */
2385 release_open_stateid(stp);
2387 /* place unused nfs4_stateowners on so_close_lru list to be
2388 * released by the laundromat service after the lease period
2389 * to enable us to handle CLOSE replay
2391 if (list_empty(&close->cl_stateowner->so_stateids))
2392 move_to_close_lru(close->cl_stateowner);
2394 if (close->cl_stateowner) {
2395 nfs4_get_stateowner(close->cl_stateowner);
2396 cstate->replay_owner = close->cl_stateowner;
2398 nfs4_unlock_state();
2403 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2404 struct nfsd4_delegreturn *dr)
2408 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2412 status = nfs4_preprocess_stateid_op(&cstate->current_fh,
2413 &dr->dr_stateid, DELEG_RET, NULL);
2414 nfs4_unlock_state();
2421 * Lock owner state (byte-range locks)
2423 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2424 #define LOCK_HASH_BITS 8
2425 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2426 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2429 end_offset(u64 start, u64 len)
2434 return end >= start ? end: NFS4_MAX_UINT64;
2437 /* last octet in a range */
2439 last_byte_offset(u64 start, u64 len)
2445 return end > start ? end - 1: NFS4_MAX_UINT64;
2448 #define lockownerid_hashval(id) \
2449 ((id) & LOCK_HASH_MASK)
2451 static inline unsigned int
2452 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2453 struct xdr_netobj *ownername)
2455 return (file_hashval(inode) + cl_id
2456 + opaque_hashval(ownername->data, ownername->len))
2460 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2461 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2462 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2464 static struct nfs4_stateid *
2465 find_stateid(stateid_t *stid, int flags)
2467 struct nfs4_stateid *local;
2468 u32 st_id = stid->si_stateownerid;
2469 u32 f_id = stid->si_fileid;
2470 unsigned int hashval;
2472 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2473 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
2474 hashval = stateid_hashval(st_id, f_id);
2475 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2476 if ((local->st_stateid.si_stateownerid == st_id) &&
2477 (local->st_stateid.si_fileid == f_id))
2482 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
2483 hashval = stateid_hashval(st_id, f_id);
2484 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2485 if ((local->st_stateid.si_stateownerid == st_id) &&
2486 (local->st_stateid.si_fileid == f_id))
2493 static struct nfs4_delegation *
2494 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2496 struct nfs4_file *fp;
2497 struct nfs4_delegation *dl;
2499 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2500 stid->si_boot, stid->si_stateownerid,
2501 stid->si_fileid, stid->si_generation);
2503 fp = find_file(ino);
2506 dl = find_delegation_file(fp, stid);
2512 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2513 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2514 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2515 * locking, this prevents us from being completely protocol-compliant. The
2516 * real solution to this problem is to start using unsigned file offsets in
2517 * the VFS, but this is a very deep change!
2520 nfs4_transform_lock_offset(struct file_lock *lock)
2522 if (lock->fl_start < 0)
2523 lock->fl_start = OFFSET_MAX;
2524 if (lock->fl_end < 0)
2525 lock->fl_end = OFFSET_MAX;
2528 /* Hack!: For now, we're defining this just so we can use a pointer to it
2529 * as a unique cookie to identify our (NFSv4's) posix locks. */
2530 static struct lock_manager_operations nfsd_posix_mng_ops = {
2534 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2536 struct nfs4_stateowner *sop;
2539 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2540 sop = (struct nfs4_stateowner *) fl->fl_owner;
2541 hval = lockownerid_hashval(sop->so_id);
2542 kref_get(&sop->so_ref);
2544 deny->ld_clientid = sop->so_client->cl_clientid;
2546 deny->ld_sop = NULL;
2547 deny->ld_clientid.cl_boot = 0;
2548 deny->ld_clientid.cl_id = 0;
2550 deny->ld_start = fl->fl_start;
2551 deny->ld_length = NFS4_MAX_UINT64;
2552 if (fl->fl_end != NFS4_MAX_UINT64)
2553 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2554 deny->ld_type = NFS4_READ_LT;
2555 if (fl->fl_type != F_RDLCK)
2556 deny->ld_type = NFS4_WRITE_LT;
2559 static struct nfs4_stateowner *
2560 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2561 struct xdr_netobj *owner)
2563 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2564 struct nfs4_stateowner *op;
2566 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2567 if (same_owner_str(op, owner, clid))
2574 * Alloc a lock owner structure.
2575 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2578 * strhashval = lock_ownerstr_hashval
2581 static struct nfs4_stateowner *
2582 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2583 struct nfs4_stateowner *sop;
2584 struct nfs4_replay *rp;
2585 unsigned int idhashval;
2587 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2589 idhashval = lockownerid_hashval(current_ownerid);
2590 INIT_LIST_HEAD(&sop->so_idhash);
2591 INIT_LIST_HEAD(&sop->so_strhash);
2592 INIT_LIST_HEAD(&sop->so_perclient);
2593 INIT_LIST_HEAD(&sop->so_stateids);
2594 INIT_LIST_HEAD(&sop->so_perstateid);
2595 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2597 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2598 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2599 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2600 sop->so_is_open_owner = 0;
2601 sop->so_id = current_ownerid++;
2602 sop->so_client = clp;
2603 /* It is the openowner seqid that will be incremented in encode in the
2604 * case of new lockowners; so increment the lock seqid manually: */
2605 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2606 sop->so_confirmed = 1;
2607 rp = &sop->so_replay;
2608 rp->rp_status = nfserr_serverfault;
2610 rp->rp_buf = rp->rp_ibuf;
2614 static struct nfs4_stateid *
2615 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2617 struct nfs4_stateid *stp;
2618 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2620 stp = nfs4_alloc_stateid();
2623 INIT_LIST_HEAD(&stp->st_hash);
2624 INIT_LIST_HEAD(&stp->st_perfile);
2625 INIT_LIST_HEAD(&stp->st_perstateowner);
2626 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2627 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2628 list_add(&stp->st_perfile, &fp->fi_stateids);
2629 list_add(&stp->st_perstateowner, &sop->so_stateids);
2630 stp->st_stateowner = sop;
2633 stp->st_stateid.si_boot = boot_time;
2634 stp->st_stateid.si_stateownerid = sop->so_id;
2635 stp->st_stateid.si_fileid = fp->fi_id;
2636 stp->st_stateid.si_generation = 0;
2637 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2638 stp->st_access_bmap = open_stp->st_access_bmap;
2639 stp->st_deny_bmap = open_stp->st_deny_bmap;
2640 stp->st_openstp = open_stp;
2647 check_lock_length(u64 offset, u64 length)
2649 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
2650 LOFF_OVERFLOW(offset, length)));
2657 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2658 struct nfsd4_lock *lock)
2660 struct nfs4_stateowner *open_sop = NULL;
2661 struct nfs4_stateowner *lock_sop = NULL;
2662 struct nfs4_stateid *lock_stp;
2664 struct file_lock file_lock;
2665 struct file_lock conflock;
2667 unsigned int strhashval;
2671 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2672 (long long) lock->lk_offset,
2673 (long long) lock->lk_length);
2675 if (check_lock_length(lock->lk_offset, lock->lk_length))
2676 return nfserr_inval;
2678 if ((status = fh_verify(rqstp, &cstate->current_fh,
2679 S_IFREG, NFSD_MAY_LOCK))) {
2680 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2686 if (lock->lk_is_new) {
2688 * Client indicates that this is a new lockowner.
2689 * Use open owner and open stateid to create lock owner and
2692 struct nfs4_stateid *open_stp = NULL;
2693 struct nfs4_file *fp;
2695 status = nfserr_stale_clientid;
2696 if (STALE_CLIENTID(&lock->lk_new_clientid))
2699 /* validate and update open stateid and open seqid */
2700 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2701 lock->lk_new_open_seqid,
2702 &lock->lk_new_open_stateid,
2704 &lock->lk_replay_owner, &open_stp,
2708 open_sop = lock->lk_replay_owner;
2709 /* create lockowner and lock stateid */
2710 fp = open_stp->st_file;
2711 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2712 open_sop->so_client->cl_clientid.cl_id,
2713 &lock->v.new.owner);
2714 /* XXX: Do we need to check for duplicate stateowners on
2715 * the same file, or should they just be allowed (and
2716 * create new stateids)? */
2717 status = nfserr_resource;
2718 lock_sop = alloc_init_lock_stateowner(strhashval,
2719 open_sop->so_client, open_stp, lock);
2720 if (lock_sop == NULL)
2722 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2723 if (lock_stp == NULL)
2726 /* lock (lock owner + lock stateid) already exists */
2727 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2728 lock->lk_old_lock_seqid,
2729 &lock->lk_old_lock_stateid,
2731 &lock->lk_replay_owner, &lock_stp, lock);
2734 lock_sop = lock->lk_replay_owner;
2736 /* lock->lk_replay_owner and lock_stp have been created or found */
2737 filp = lock_stp->st_vfs_file;
2739 status = nfserr_grace;
2740 if (locks_in_grace() && !lock->lk_reclaim)
2742 status = nfserr_no_grace;
2743 if (!locks_in_grace() && lock->lk_reclaim)
2746 locks_init_lock(&file_lock);
2747 switch (lock->lk_type) {
2750 file_lock.fl_type = F_RDLCK;
2754 case NFS4_WRITEW_LT:
2755 file_lock.fl_type = F_WRLCK;
2759 status = nfserr_inval;
2762 file_lock.fl_owner = (fl_owner_t)lock_sop;
2763 file_lock.fl_pid = current->tgid;
2764 file_lock.fl_file = filp;
2765 file_lock.fl_flags = FL_POSIX;
2766 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2768 file_lock.fl_start = lock->lk_offset;
2769 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
2770 nfs4_transform_lock_offset(&file_lock);
2773 * Try to lock the file in the VFS.
2774 * Note: locks.c uses the BKL to protect the inode's lock list.
2777 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
2779 case 0: /* success! */
2780 update_stateid(&lock_stp->st_stateid);
2781 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2785 case (EAGAIN): /* conflock holds conflicting lock */
2786 status = nfserr_denied;
2787 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2788 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2791 status = nfserr_deadlock;
2794 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
2795 status = nfserr_resource;
2799 if (status && lock->lk_is_new && lock_sop)
2800 release_stateowner(lock_sop);
2801 if (lock->lk_replay_owner) {
2802 nfs4_get_stateowner(lock->lk_replay_owner);
2803 cstate->replay_owner = lock->lk_replay_owner;
2805 nfs4_unlock_state();
2810 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
2811 * so we do a temporary open here just to get an open file to pass to
2812 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
2815 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
2820 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
2823 err = vfs_test_lock(file, lock);
2832 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2833 struct nfsd4_lockt *lockt)
2835 struct inode *inode;
2836 struct file_lock file_lock;
2840 if (locks_in_grace())
2841 return nfserr_grace;
2843 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2844 return nfserr_inval;
2846 lockt->lt_stateowner = NULL;
2849 status = nfserr_stale_clientid;
2850 if (STALE_CLIENTID(&lockt->lt_clientid))
2853 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2854 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2855 if (status == nfserr_symlink)
2856 status = nfserr_inval;
2860 inode = cstate->current_fh.fh_dentry->d_inode;
2861 locks_init_lock(&file_lock);
2862 switch (lockt->lt_type) {
2865 file_lock.fl_type = F_RDLCK;
2868 case NFS4_WRITEW_LT:
2869 file_lock.fl_type = F_WRLCK;
2872 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2873 status = nfserr_inval;
2877 lockt->lt_stateowner = find_lockstateowner_str(inode,
2878 &lockt->lt_clientid, &lockt->lt_owner);
2879 if (lockt->lt_stateowner)
2880 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2881 file_lock.fl_pid = current->tgid;
2882 file_lock.fl_flags = FL_POSIX;
2884 file_lock.fl_start = lockt->lt_offset;
2885 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
2887 nfs4_transform_lock_offset(&file_lock);
2890 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
2892 status = nfserrno(error);
2895 if (file_lock.fl_type != F_UNLCK) {
2896 status = nfserr_denied;
2897 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
2900 nfs4_unlock_state();
2905 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2906 struct nfsd4_locku *locku)
2908 struct nfs4_stateid *stp;
2909 struct file *filp = NULL;
2910 struct file_lock file_lock;
2914 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2915 (long long) locku->lu_offset,
2916 (long long) locku->lu_length);
2918 if (check_lock_length(locku->lu_offset, locku->lu_length))
2919 return nfserr_inval;
2923 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2927 &locku->lu_stateowner, &stp, NULL)))
2930 filp = stp->st_vfs_file;
2932 locks_init_lock(&file_lock);
2933 file_lock.fl_type = F_UNLCK;
2934 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2935 file_lock.fl_pid = current->tgid;
2936 file_lock.fl_file = filp;
2937 file_lock.fl_flags = FL_POSIX;
2938 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2939 file_lock.fl_start = locku->lu_offset;
2941 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
2942 nfs4_transform_lock_offset(&file_lock);
2945 * Try to unlock the file in the VFS.
2947 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
2949 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2953 * OK, unlock succeeded; the only thing left to do is update the stateid.
2955 update_stateid(&stp->st_stateid);
2956 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2959 if (locku->lu_stateowner) {
2960 nfs4_get_stateowner(locku->lu_stateowner);
2961 cstate->replay_owner = locku->lu_stateowner;
2963 nfs4_unlock_state();
2967 status = nfserrno(err);
2973 * 1: locks held by lockowner
2974 * 0: no locks held by lockowner
2977 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2979 struct file_lock **flpp;
2980 struct inode *inode = filp->f_path.dentry->d_inode;
2984 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2985 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2996 nfsd4_release_lockowner(struct svc_rqst *rqstp,
2997 struct nfsd4_compound_state *cstate,
2998 struct nfsd4_release_lockowner *rlockowner)
3000 clientid_t *clid = &rlockowner->rl_clientid;
3001 struct nfs4_stateowner *sop;
3002 struct nfs4_stateid *stp;
3003 struct xdr_netobj *owner = &rlockowner->rl_owner;
3004 struct list_head matches;
3008 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3009 clid->cl_boot, clid->cl_id);
3011 /* XXX check for lease expiration */
3013 status = nfserr_stale_clientid;
3014 if (STALE_CLIENTID(clid))
3019 status = nfserr_locks_held;
3020 /* XXX: we're doing a linear search through all the lockowners.
3021 * Yipes! For now we'll just hope clients aren't really using
3022 * release_lockowner much, but eventually we have to fix these
3023 * data structures. */
3024 INIT_LIST_HEAD(&matches);
3025 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3026 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3027 if (!same_owner_str(sop, owner, clid))
3029 list_for_each_entry(stp, &sop->so_stateids,
3031 if (check_for_locks(stp->st_vfs_file, sop))
3033 /* Note: so_perclient unused for lockowners,
3034 * so it's OK to fool with here. */
3035 list_add(&sop->so_perclient, &matches);
3039 /* Clients probably won't expect us to return with some (but not all)
3040 * of the lockowner state released; so don't release any until all
3041 * have been checked. */
3043 while (!list_empty(&matches)) {
3044 sop = list_entry(matches.next, struct nfs4_stateowner,
3046 /* unhash_stateowner deletes so_perclient only
3047 * for openowners. */
3048 list_del(&sop->so_perclient);
3049 release_stateowner(sop);
3052 nfs4_unlock_state();
3056 static inline struct nfs4_client_reclaim *
3059 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3063 nfs4_has_reclaimed_state(const char *name)
3065 unsigned int strhashval = clientstr_hashval(name);
3066 struct nfs4_client *clp;
3068 clp = find_confirmed_client_by_str(name, strhashval);
3073 * failure => all reset bets are off, nfserr_no_grace...
3076 nfs4_client_to_reclaim(const char *name)
3078 unsigned int strhashval;
3079 struct nfs4_client_reclaim *crp = NULL;
3081 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3082 crp = alloc_reclaim();
3085 strhashval = clientstr_hashval(name);
3086 INIT_LIST_HEAD(&crp->cr_strhash);
3087 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3088 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3089 reclaim_str_hashtbl_size++;
3094 nfs4_release_reclaim(void)
3096 struct nfs4_client_reclaim *crp = NULL;
3099 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3100 while (!list_empty(&reclaim_str_hashtbl[i])) {
3101 crp = list_entry(reclaim_str_hashtbl[i].next,
3102 struct nfs4_client_reclaim, cr_strhash);
3103 list_del(&crp->cr_strhash);
3105 reclaim_str_hashtbl_size--;
3108 BUG_ON(reclaim_str_hashtbl_size);
3112 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3113 static struct nfs4_client_reclaim *
3114 nfs4_find_reclaim_client(clientid_t *clid)
3116 unsigned int strhashval;
3117 struct nfs4_client *clp;
3118 struct nfs4_client_reclaim *crp = NULL;
3121 /* find clientid in conf_id_hashtbl */
3122 clp = find_confirmed_client(clid);
3126 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3127 clp->cl_name.len, clp->cl_name.data,
3130 /* find clp->cl_name in reclaim_str_hashtbl */
3131 strhashval = clientstr_hashval(clp->cl_recdir);
3132 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3133 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3141 * Called from OPEN. Look for clientid in reclaim list.
3144 nfs4_check_open_reclaim(clientid_t *clid)
3146 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3149 /* initialization to perform at module load time: */
3152 nfs4_state_init(void)
3156 status = nfsd4_init_slabs();
3159 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3160 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3161 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3162 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3163 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3165 for (i = 0; i < FILE_HASH_SIZE; i++) {
3166 INIT_LIST_HEAD(&file_hashtbl[i]);
3168 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3169 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3170 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3172 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3173 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3174 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3176 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3177 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3178 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3180 memset(&onestateid, ~0, sizeof(stateid_t));
3181 INIT_LIST_HEAD(&close_lru);
3182 INIT_LIST_HEAD(&client_lru);
3183 INIT_LIST_HEAD(&del_recall_lru);
3184 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3185 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3186 reclaim_str_hashtbl_size = 0;
3191 nfsd4_load_reboot_recovery_data(void)
3196 nfsd4_init_recdir(user_recovery_dirname);
3197 status = nfsd4_recdir_load();
3198 nfs4_unlock_state();
3200 printk("NFSD: Failure reading reboot recovery data\n");
3204 get_nfs4_grace_period(void)
3206 return max(user_lease_time, lease_time) * HZ;
3210 * Since the lifetime of a delegation isn't limited to that of an open, a
3211 * client may quite reasonably hang on to a delegation as long as it has
3212 * the inode cached. This becomes an obvious problem the first time a
3213 * client's inode cache approaches the size of the server's total memory.
3215 * For now we avoid this problem by imposing a hard limit on the number
3216 * of delegations, which varies according to the server's memory size.
3219 set_max_delegations(void)
3222 * Allow at most 4 delegations per megabyte of RAM. Quick
3223 * estimates suggest that in the worst case (where every delegation
3224 * is for a different inode), a delegation could take about 1.5K,
3225 * giving a worst case usage of about 6% of memory.
3227 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3230 /* initialization to perform when the nfsd service is started: */
3233 __nfs4_state_start(void)
3235 unsigned long grace_time;
3237 boot_time = get_seconds();
3238 grace_time = get_nfs4_grace_period();
3239 lease_time = user_lease_time;
3240 locks_start_grace(&nfsd4_manager);
3241 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3243 laundry_wq = create_singlethread_workqueue("nfsd4");
3244 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
3245 set_max_delegations();
3249 nfs4_state_start(void)
3253 nfsd4_load_reboot_recovery_data();
3254 __nfs4_state_start();
3260 nfs4_lease_time(void)
3266 __nfs4_state_shutdown(void)
3269 struct nfs4_client *clp = NULL;
3270 struct nfs4_delegation *dp = NULL;
3271 struct list_head *pos, *next, reaplist;
3273 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3274 while (!list_empty(&conf_id_hashtbl[i])) {
3275 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3278 while (!list_empty(&unconf_str_hashtbl[i])) {
3279 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3283 INIT_LIST_HEAD(&reaplist);
3284 spin_lock(&recall_lock);
3285 list_for_each_safe(pos, next, &del_recall_lru) {
3286 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3287 list_move(&dp->dl_recall_lru, &reaplist);
3289 spin_unlock(&recall_lock);
3290 list_for_each_safe(pos, next, &reaplist) {
3291 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3292 list_del_init(&dp->dl_recall_lru);
3293 unhash_delegation(dp);
3296 nfsd4_shutdown_recdir();
3301 nfs4_state_shutdown(void)
3303 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3304 destroy_workqueue(laundry_wq);
3305 locks_end_grace(&nfsd4_manager);
3307 nfs4_release_reclaim();
3308 __nfs4_state_shutdown();
3309 nfs4_unlock_state();
3313 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3314 * accessed when nfsd is starting.
3317 nfs4_set_recdir(char *recdir)
3319 strcpy(user_recovery_dirname, recdir);
3323 * Change the NFSv4 recovery directory to recdir.
3326 nfs4_reset_recoverydir(char *recdir)
3331 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
3335 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
3336 nfs4_set_recdir(recdir);
3344 nfs4_recoverydir(void)
3346 return user_recovery_dirname;
3350 * Called when leasetime is changed.
3352 * The only way the protocol gives us to handle on-the-fly lease changes is to
3353 * simulate a reboot. Instead of doing that, we just wait till the next time
3354 * we start to register any changes in lease time. If the administrator
3355 * really wants to change the lease time *now*, they can go ahead and bring
3356 * nfsd down and then back up again after changing the lease time.
3358 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3359 * when nfsd is starting
3362 nfs4_reset_lease(time_t leasetime)
3364 user_lease_time = leasetime;