USB: kill BKL in skeleton driver
[linux-2.6/mini2440.git] / fs / nfsd / nfs4state.c
blobaf360705e551afc5f9bf9954db7a2790aebca6cc
1 /*
2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * All rights reserved.
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
12 * are met:
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/mount.h>
45 #include <linux/workqueue.h>
46 #include <linux/smp_lock.h>
47 #include <linux/kthread.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfsd/state.h>
50 #include <linux/nfsd/xdr4.h>
51 #include <linux/namei.h>
52 #include <linux/mutex.h>
54 #define NFSDDBG_FACILITY NFSDDBG_PROC
56 /* Globals */
57 static time_t lease_time = 90; /* default lease time */
58 static time_t user_lease_time = 90;
59 static time_t boot_time;
60 static int in_grace = 1;
61 static u32 current_clientid = 1;
62 static u32 current_ownerid = 1;
63 static u32 current_fileid = 1;
64 static u32 current_delegid = 1;
65 static u32 nfs4_init;
66 static stateid_t zerostateid; /* bits all 0 */
67 static stateid_t onestateid; /* bits all 1 */
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
74 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
75 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
76 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
77 static void nfs4_set_recdir(char *recdir);
79 /* Locking:
81 * client_mutex:
82 * protects clientid_hashtbl[], clientstr_hashtbl[],
83 * unconfstr_hashtbl[], uncofid_hashtbl[].
85 static DEFINE_MUTEX(client_mutex);
87 static struct kmem_cache *stateowner_slab = NULL;
88 static struct kmem_cache *file_slab = NULL;
89 static struct kmem_cache *stateid_slab = NULL;
90 static struct kmem_cache *deleg_slab = NULL;
92 void
93 nfs4_lock_state(void)
95 mutex_lock(&client_mutex);
98 void
99 nfs4_unlock_state(void)
101 mutex_unlock(&client_mutex);
104 static inline u32
105 opaque_hashval(const void *ptr, int nbytes)
107 unsigned char *cptr = (unsigned char *) ptr;
109 u32 x = 0;
110 while (nbytes--) {
111 x *= 37;
112 x += *cptr++;
114 return x;
117 /* forward declarations */
118 static void release_stateowner(struct nfs4_stateowner *sop);
119 static void release_stateid(struct nfs4_stateid *stp, int flags);
122 * Delegation state
125 /* recall_lock protects the del_recall_lru */
126 static DEFINE_SPINLOCK(recall_lock);
127 static struct list_head del_recall_lru;
129 static void
130 free_nfs4_file(struct kref *kref)
132 struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
133 list_del(&fp->fi_hash);
134 iput(fp->fi_inode);
135 kmem_cache_free(file_slab, fp);
138 static inline void
139 put_nfs4_file(struct nfs4_file *fi)
141 kref_put(&fi->fi_ref, free_nfs4_file);
144 static inline void
145 get_nfs4_file(struct nfs4_file *fi)
147 kref_get(&fi->fi_ref);
150 static int num_delegations;
153 * Open owner state (share locks)
156 /* hash tables for nfs4_stateowner */
157 #define OWNER_HASH_BITS 8
158 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
159 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
161 #define ownerid_hashval(id) \
162 ((id) & OWNER_HASH_MASK)
163 #define ownerstr_hashval(clientid, ownername) \
164 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
166 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
167 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
169 /* hash table for nfs4_file */
170 #define FILE_HASH_BITS 8
171 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
172 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
173 /* hash table for (open)nfs4_stateid */
174 #define STATEID_HASH_BITS 10
175 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
176 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
178 #define file_hashval(x) \
179 hash_ptr(x, FILE_HASH_BITS)
180 #define stateid_hashval(owner_id, file_id) \
181 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
183 static struct list_head file_hashtbl[FILE_HASH_SIZE];
184 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
186 static struct nfs4_delegation *
187 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
189 struct nfs4_delegation *dp;
190 struct nfs4_file *fp = stp->st_file;
191 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
193 dprintk("NFSD alloc_init_deleg\n");
194 if (num_delegations > STATEID_HASH_SIZE * 4)
195 return NULL;
196 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
197 if (dp == NULL)
198 return dp;
199 num_delegations++;
200 INIT_LIST_HEAD(&dp->dl_perfile);
201 INIT_LIST_HEAD(&dp->dl_perclnt);
202 INIT_LIST_HEAD(&dp->dl_recall_lru);
203 dp->dl_client = clp;
204 get_nfs4_file(fp);
205 dp->dl_file = fp;
206 dp->dl_flock = NULL;
207 get_file(stp->st_vfs_file);
208 dp->dl_vfs_file = stp->st_vfs_file;
209 dp->dl_type = type;
210 dp->dl_recall.cbr_dp = NULL;
211 dp->dl_recall.cbr_ident = cb->cb_ident;
212 dp->dl_recall.cbr_trunc = 0;
213 dp->dl_stateid.si_boot = boot_time;
214 dp->dl_stateid.si_stateownerid = current_delegid++;
215 dp->dl_stateid.si_fileid = 0;
216 dp->dl_stateid.si_generation = 0;
217 dp->dl_fhlen = current_fh->fh_handle.fh_size;
218 memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
219 current_fh->fh_handle.fh_size);
220 dp->dl_time = 0;
221 atomic_set(&dp->dl_count, 1);
222 list_add(&dp->dl_perfile, &fp->fi_delegations);
223 list_add(&dp->dl_perclnt, &clp->cl_delegations);
224 return dp;
227 void
228 nfs4_put_delegation(struct nfs4_delegation *dp)
230 if (atomic_dec_and_test(&dp->dl_count)) {
231 dprintk("NFSD: freeing dp %p\n",dp);
232 put_nfs4_file(dp->dl_file);
233 kmem_cache_free(deleg_slab, dp);
234 num_delegations--;
238 /* Remove the associated file_lock first, then remove the delegation.
239 * lease_modify() is called to remove the FS_LEASE file_lock from
240 * the i_flock list, eventually calling nfsd's lock_manager
241 * fl_release_callback.
243 static void
244 nfs4_close_delegation(struct nfs4_delegation *dp)
246 struct file *filp = dp->dl_vfs_file;
248 dprintk("NFSD: close_delegation dp %p\n",dp);
249 dp->dl_vfs_file = NULL;
250 /* The following nfsd_close may not actually close the file,
251 * but we want to remove the lease in any case. */
252 if (dp->dl_flock)
253 setlease(filp, F_UNLCK, &dp->dl_flock);
254 nfsd_close(filp);
257 /* Called under the state lock. */
258 static void
259 unhash_delegation(struct nfs4_delegation *dp)
261 list_del_init(&dp->dl_perfile);
262 list_del_init(&dp->dl_perclnt);
263 spin_lock(&recall_lock);
264 list_del_init(&dp->dl_recall_lru);
265 spin_unlock(&recall_lock);
266 nfs4_close_delegation(dp);
267 nfs4_put_delegation(dp);
271 * SETCLIENTID state
274 /* Hash tables for nfs4_clientid state */
275 #define CLIENT_HASH_BITS 4
276 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
277 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
279 #define clientid_hashval(id) \
280 ((id) & CLIENT_HASH_MASK)
281 #define clientstr_hashval(name) \
282 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
284 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
285 * used in reboot/reset lease grace period processing
287 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
288 * setclientid_confirmed info.
290 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
291 * setclientid info.
293 * client_lru holds client queue ordered by nfs4_client.cl_time
294 * for lease renewal.
296 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
297 * for last close replay.
299 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
300 static int reclaim_str_hashtbl_size = 0;
301 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
302 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
303 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
304 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
305 static struct list_head client_lru;
306 static struct list_head close_lru;
308 static inline void
309 renew_client(struct nfs4_client *clp)
312 * Move client to the end to the LRU list.
314 dprintk("renewing client (clientid %08x/%08x)\n",
315 clp->cl_clientid.cl_boot,
316 clp->cl_clientid.cl_id);
317 list_move_tail(&clp->cl_lru, &client_lru);
318 clp->cl_time = get_seconds();
321 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
322 static int
323 STALE_CLIENTID(clientid_t *clid)
325 if (clid->cl_boot == boot_time)
326 return 0;
327 dprintk("NFSD stale clientid (%08x/%08x)\n",
328 clid->cl_boot, clid->cl_id);
329 return 1;
333 * XXX Should we use a slab cache ?
334 * This type of memory management is somewhat inefficient, but we use it
335 * anyway since SETCLIENTID is not a common operation.
337 static inline struct nfs4_client *
338 alloc_client(struct xdr_netobj name)
340 struct nfs4_client *clp;
342 if ((clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
343 if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
344 memcpy(clp->cl_name.data, name.data, name.len);
345 clp->cl_name.len = name.len;
347 else {
348 kfree(clp);
349 clp = NULL;
352 return clp;
355 static inline void
356 free_client(struct nfs4_client *clp)
358 if (clp->cl_cred.cr_group_info)
359 put_group_info(clp->cl_cred.cr_group_info);
360 kfree(clp->cl_name.data);
361 kfree(clp);
364 void
365 put_nfs4_client(struct nfs4_client *clp)
367 if (atomic_dec_and_test(&clp->cl_count))
368 free_client(clp);
371 static void
372 shutdown_callback_client(struct nfs4_client *clp)
374 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
376 /* shutdown rpc client, ending any outstanding recall rpcs */
377 if (clnt) {
378 clp->cl_callback.cb_client = NULL;
379 rpc_shutdown_client(clnt);
380 rpciod_down();
384 static void
385 expire_client(struct nfs4_client *clp)
387 struct nfs4_stateowner *sop;
388 struct nfs4_delegation *dp;
389 struct list_head reaplist;
391 dprintk("NFSD: expire_client cl_count %d\n",
392 atomic_read(&clp->cl_count));
394 shutdown_callback_client(clp);
396 INIT_LIST_HEAD(&reaplist);
397 spin_lock(&recall_lock);
398 while (!list_empty(&clp->cl_delegations)) {
399 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
400 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
401 dp->dl_flock);
402 list_del_init(&dp->dl_perclnt);
403 list_move(&dp->dl_recall_lru, &reaplist);
405 spin_unlock(&recall_lock);
406 while (!list_empty(&reaplist)) {
407 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
408 list_del_init(&dp->dl_recall_lru);
409 unhash_delegation(dp);
411 list_del(&clp->cl_idhash);
412 list_del(&clp->cl_strhash);
413 list_del(&clp->cl_lru);
414 while (!list_empty(&clp->cl_openowners)) {
415 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
416 release_stateowner(sop);
418 put_nfs4_client(clp);
421 static struct nfs4_client *
422 create_client(struct xdr_netobj name, char *recdir) {
423 struct nfs4_client *clp;
425 if (!(clp = alloc_client(name)))
426 goto out;
427 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
428 atomic_set(&clp->cl_count, 1);
429 atomic_set(&clp->cl_callback.cb_set, 0);
430 INIT_LIST_HEAD(&clp->cl_idhash);
431 INIT_LIST_HEAD(&clp->cl_strhash);
432 INIT_LIST_HEAD(&clp->cl_openowners);
433 INIT_LIST_HEAD(&clp->cl_delegations);
434 INIT_LIST_HEAD(&clp->cl_lru);
435 out:
436 return clp;
439 static void
440 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
441 memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
444 static void
445 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
446 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
447 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
450 static void
451 copy_cred(struct svc_cred *target, struct svc_cred *source) {
453 target->cr_uid = source->cr_uid;
454 target->cr_gid = source->cr_gid;
455 target->cr_group_info = source->cr_group_info;
456 get_group_info(target->cr_group_info);
459 static inline int
460 same_name(const char *n1, const char *n2) {
461 return 0 == memcmp(n1, n2, HEXDIR_LEN);
464 static int
465 cmp_verf(nfs4_verifier *v1, nfs4_verifier *v2) {
466 return(!memcmp(v1->data,v2->data,sizeof(v1->data)));
469 static int
470 cmp_clid(clientid_t * cl1, clientid_t * cl2) {
471 return((cl1->cl_boot == cl2->cl_boot) &&
472 (cl1->cl_id == cl2->cl_id));
475 /* XXX what about NGROUP */
476 static int
477 cmp_creds(struct svc_cred *cr1, struct svc_cred *cr2){
478 return(cr1->cr_uid == cr2->cr_uid);
482 static void
483 gen_clid(struct nfs4_client *clp) {
484 clp->cl_clientid.cl_boot = boot_time;
485 clp->cl_clientid.cl_id = current_clientid++;
488 static void
489 gen_confirm(struct nfs4_client *clp) {
490 struct timespec tv;
491 u32 * p;
493 tv = CURRENT_TIME;
494 p = (u32 *)clp->cl_confirm.data;
495 *p++ = tv.tv_sec;
496 *p++ = tv.tv_nsec;
499 static int
500 check_name(struct xdr_netobj name) {
502 if (name.len == 0)
503 return 0;
504 if (name.len > NFS4_OPAQUE_LIMIT) {
505 printk("NFSD: check_name: name too long(%d)!\n", name.len);
506 return 0;
508 return 1;
511 static void
512 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
514 unsigned int idhashval;
516 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
517 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
518 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
519 list_add_tail(&clp->cl_lru, &client_lru);
520 clp->cl_time = get_seconds();
523 static void
524 move_to_confirmed(struct nfs4_client *clp)
526 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
527 unsigned int strhashval;
529 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
530 list_del_init(&clp->cl_strhash);
531 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
532 strhashval = clientstr_hashval(clp->cl_recdir);
533 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
534 renew_client(clp);
537 static struct nfs4_client *
538 find_confirmed_client(clientid_t *clid)
540 struct nfs4_client *clp;
541 unsigned int idhashval = clientid_hashval(clid->cl_id);
543 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
544 if (cmp_clid(&clp->cl_clientid, clid))
545 return clp;
547 return NULL;
550 static struct nfs4_client *
551 find_unconfirmed_client(clientid_t *clid)
553 struct nfs4_client *clp;
554 unsigned int idhashval = clientid_hashval(clid->cl_id);
556 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
557 if (cmp_clid(&clp->cl_clientid, clid))
558 return clp;
560 return NULL;
563 static struct nfs4_client *
564 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
566 struct nfs4_client *clp;
568 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
569 if (same_name(clp->cl_recdir, dname))
570 return clp;
572 return NULL;
575 static struct nfs4_client *
576 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
578 struct nfs4_client *clp;
580 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
581 if (same_name(clp->cl_recdir, dname))
582 return clp;
584 return NULL;
587 /* a helper function for parse_callback */
588 static int
589 parse_octet(unsigned int *lenp, char **addrp)
591 unsigned int len = *lenp;
592 char *p = *addrp;
593 int n = -1;
594 char c;
596 for (;;) {
597 if (!len)
598 break;
599 len--;
600 c = *p++;
601 if (c == '.')
602 break;
603 if ((c < '0') || (c > '9')) {
604 n = -1;
605 break;
607 if (n < 0)
608 n = 0;
609 n = (n * 10) + (c - '0');
610 if (n > 255) {
611 n = -1;
612 break;
615 *lenp = len;
616 *addrp = p;
617 return n;
620 /* parse and set the setclientid ipv4 callback address */
621 static int
622 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
624 int temp = 0;
625 u32 cbaddr = 0;
626 u16 cbport = 0;
627 u32 addrlen = addr_len;
628 char *addr = addr_val;
629 int i, shift;
631 /* ipaddress */
632 shift = 24;
633 for(i = 4; i > 0 ; i--) {
634 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
635 return 0;
637 cbaddr |= (temp << shift);
638 if (shift > 0)
639 shift -= 8;
641 *cbaddrp = cbaddr;
643 /* port */
644 shift = 8;
645 for(i = 2; i > 0 ; i--) {
646 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
647 return 0;
649 cbport |= (temp << shift);
650 if (shift > 0)
651 shift -= 8;
653 *cbportp = cbport;
654 return 1;
657 static void
658 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
660 struct nfs4_callback *cb = &clp->cl_callback;
662 /* Currently, we only support tcp for the callback channel */
663 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
664 goto out_err;
666 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
667 &cb->cb_addr, &cb->cb_port)))
668 goto out_err;
669 cb->cb_prog = se->se_callback_prog;
670 cb->cb_ident = se->se_callback_ident;
671 return;
672 out_err:
673 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
674 "will not receive delegations\n",
675 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
677 return;
681 * RFC 3010 has a complex implmentation description of processing a
682 * SETCLIENTID request consisting of 5 bullets, labeled as
683 * CASE0 - CASE4 below.
685 * NOTES:
686 * callback information will be processed in a future patch
688 * an unconfirmed record is added when:
689 * NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
690 * CASE 1: confirmed record found with matching name, principal,
691 * verifier, and clientid.
692 * CASE 2: confirmed record found with matching name, principal,
693 * and there is no unconfirmed record with matching
694 * name and principal
696 * an unconfirmed record is replaced when:
697 * CASE 3: confirmed record found with matching name, principal,
698 * and an unconfirmed record is found with matching
699 * name, principal, and with clientid and
700 * confirm that does not match the confirmed record.
701 * CASE 4: there is no confirmed record with matching name and
702 * principal. there is an unconfirmed record with
703 * matching name, principal.
705 * an unconfirmed record is deleted when:
706 * CASE 1: an unconfirmed record that matches input name, verifier,
707 * and confirmed clientid.
708 * CASE 4: any unconfirmed records with matching name and principal
709 * that exist after an unconfirmed record has been replaced
710 * as described above.
713 __be32
714 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
715 struct nfsd4_setclientid *setclid)
717 struct sockaddr_in *sin = svc_addr_in(rqstp);
718 struct xdr_netobj clname = {
719 .len = setclid->se_namelen,
720 .data = setclid->se_name,
722 nfs4_verifier clverifier = setclid->se_verf;
723 unsigned int strhashval;
724 struct nfs4_client *conf, *unconf, *new;
725 __be32 status;
726 char dname[HEXDIR_LEN];
728 if (!check_name(clname))
729 return nfserr_inval;
731 status = nfs4_make_rec_clidname(dname, &clname);
732 if (status)
733 return status;
736 * XXX The Duplicate Request Cache (DRC) has been checked (??)
737 * We get here on a DRC miss.
740 strhashval = clientstr_hashval(dname);
742 nfs4_lock_state();
743 conf = find_confirmed_client_by_str(dname, strhashval);
744 if (conf) {
746 * CASE 0:
747 * clname match, confirmed, different principal
748 * or different ip_address
750 status = nfserr_clid_inuse;
751 if (!cmp_creds(&conf->cl_cred, &rqstp->rq_cred)
752 || conf->cl_addr != sin->sin_addr.s_addr) {
753 dprintk("NFSD: setclientid: string in use by client"
754 "at %u.%u.%u.%u\n", NIPQUAD(conf->cl_addr));
755 goto out;
758 unconf = find_unconfirmed_client_by_str(dname, strhashval);
759 status = nfserr_resource;
760 if (!conf) {
762 * CASE 4:
763 * placed first, because it is the normal case.
765 if (unconf)
766 expire_client(unconf);
767 new = create_client(clname, dname);
768 if (new == NULL)
769 goto out;
770 copy_verf(new, &clverifier);
771 new->cl_addr = sin->sin_addr.s_addr;
772 copy_cred(&new->cl_cred,&rqstp->rq_cred);
773 gen_clid(new);
774 gen_confirm(new);
775 gen_callback(new, setclid);
776 add_to_unconfirmed(new, strhashval);
777 } else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
779 * CASE 1:
780 * cl_name match, confirmed, principal match
781 * verifier match: probable callback update
783 * remove any unconfirmed nfs4_client with
784 * matching cl_name, cl_verifier, and cl_clientid
786 * create and insert an unconfirmed nfs4_client with same
787 * cl_name, cl_verifier, and cl_clientid as existing
788 * nfs4_client, but with the new callback info and a
789 * new cl_confirm
791 if (unconf) {
792 /* Note this is removing unconfirmed {*x***},
793 * which is stronger than RFC recommended {vxc**}.
794 * This has the advantage that there is at most
795 * one {*x***} in either list at any time.
797 expire_client(unconf);
799 new = create_client(clname, dname);
800 if (new == NULL)
801 goto out;
802 copy_verf(new,&conf->cl_verifier);
803 new->cl_addr = sin->sin_addr.s_addr;
804 copy_cred(&new->cl_cred,&rqstp->rq_cred);
805 copy_clid(new, conf);
806 gen_confirm(new);
807 gen_callback(new, setclid);
808 add_to_unconfirmed(new,strhashval);
809 } else if (!unconf) {
811 * CASE 2:
812 * clname match, confirmed, principal match
813 * verfier does not match
814 * no unconfirmed. create a new unconfirmed nfs4_client
815 * using input clverifier, clname, and callback info
816 * and generate a new cl_clientid and cl_confirm.
818 new = create_client(clname, dname);
819 if (new == NULL)
820 goto out;
821 copy_verf(new,&clverifier);
822 new->cl_addr = sin->sin_addr.s_addr;
823 copy_cred(&new->cl_cred,&rqstp->rq_cred);
824 gen_clid(new);
825 gen_confirm(new);
826 gen_callback(new, setclid);
827 add_to_unconfirmed(new, strhashval);
828 } else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
830 * CASE3:
831 * confirmed found (name, principal match)
832 * confirmed verifier does not match input clverifier
834 * unconfirmed found (name match)
835 * confirmed->cl_confirm != unconfirmed->cl_confirm
837 * remove unconfirmed.
839 * create an unconfirmed nfs4_client
840 * with same cl_name as existing confirmed nfs4_client,
841 * but with new callback info, new cl_clientid,
842 * new cl_verifier and a new cl_confirm
844 expire_client(unconf);
845 new = create_client(clname, dname);
846 if (new == NULL)
847 goto out;
848 copy_verf(new,&clverifier);
849 new->cl_addr = sin->sin_addr.s_addr;
850 copy_cred(&new->cl_cred,&rqstp->rq_cred);
851 gen_clid(new);
852 gen_confirm(new);
853 gen_callback(new, setclid);
854 add_to_unconfirmed(new, strhashval);
855 } else {
856 /* No cases hit !!! */
857 status = nfserr_inval;
858 goto out;
861 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
862 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
863 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
864 status = nfs_ok;
865 out:
866 nfs4_unlock_state();
867 return status;
872 * RFC 3010 has a complex implmentation description of processing a
873 * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
874 * processing on a DRC miss, labeled as CASE1 - CASE4 below.
876 * NOTE: callback information will be processed here in a future patch
878 __be32
879 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
880 struct nfsd4_compound_state *cstate,
881 struct nfsd4_setclientid_confirm *setclientid_confirm)
883 struct sockaddr_in *sin = svc_addr_in(rqstp);
884 struct nfs4_client *conf, *unconf;
885 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
886 clientid_t * clid = &setclientid_confirm->sc_clientid;
887 __be32 status;
889 if (STALE_CLIENTID(clid))
890 return nfserr_stale_clientid;
892 * XXX The Duplicate Request Cache (DRC) has been checked (??)
893 * We get here on a DRC miss.
896 nfs4_lock_state();
898 conf = find_confirmed_client(clid);
899 unconf = find_unconfirmed_client(clid);
901 status = nfserr_clid_inuse;
902 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
903 goto out;
904 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
905 goto out;
907 if ((conf && unconf) &&
908 (cmp_verf(&unconf->cl_confirm, &confirm)) &&
909 (cmp_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
910 (same_name(conf->cl_recdir,unconf->cl_recdir)) &&
911 (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
912 /* CASE 1:
913 * unconf record that matches input clientid and input confirm.
914 * conf record that matches input clientid.
915 * conf and unconf records match names, verifiers
917 if (!cmp_creds(&conf->cl_cred, &unconf->cl_cred))
918 status = nfserr_clid_inuse;
919 else {
920 /* XXX: We just turn off callbacks until we can handle
921 * change request correctly. */
922 atomic_set(&conf->cl_callback.cb_set, 0);
923 gen_confirm(conf);
924 nfsd4_remove_clid_dir(unconf);
925 expire_client(unconf);
926 status = nfs_ok;
929 } else if ((conf && !unconf) ||
930 ((conf && unconf) &&
931 (!cmp_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
932 !same_name(conf->cl_recdir, unconf->cl_recdir)))) {
933 /* CASE 2:
934 * conf record that matches input clientid.
935 * if unconf record matches input clientid, then
936 * unconf->cl_name or unconf->cl_verifier don't match the
937 * conf record.
939 if (!cmp_creds(&conf->cl_cred,&rqstp->rq_cred))
940 status = nfserr_clid_inuse;
941 else
942 status = nfs_ok;
943 } else if (!conf && unconf
944 && cmp_verf(&unconf->cl_confirm, &confirm)) {
945 /* CASE 3:
946 * conf record not found.
947 * unconf record found.
948 * unconf->cl_confirm matches input confirm
950 if (!cmp_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
951 status = nfserr_clid_inuse;
952 } else {
953 unsigned int hash =
954 clientstr_hashval(unconf->cl_recdir);
955 conf = find_confirmed_client_by_str(unconf->cl_recdir,
956 hash);
957 if (conf) {
958 nfsd4_remove_clid_dir(conf);
959 expire_client(conf);
961 move_to_confirmed(unconf);
962 conf = unconf;
963 status = nfs_ok;
965 } else if ((!conf || (conf && !cmp_verf(&conf->cl_confirm, &confirm)))
966 && (!unconf || (unconf && !cmp_verf(&unconf->cl_confirm,
967 &confirm)))) {
968 /* CASE 4:
969 * conf record not found, or if conf, conf->cl_confirm does not
970 * match input confirm.
971 * unconf record not found, or if unconf, unconf->cl_confirm
972 * does not match input confirm.
974 status = nfserr_stale_clientid;
975 } else {
976 /* check that we have hit one of the cases...*/
977 status = nfserr_clid_inuse;
979 out:
980 if (!status)
981 nfsd4_probe_callback(conf);
982 nfs4_unlock_state();
983 return status;
986 /* OPEN Share state helper functions */
987 static inline struct nfs4_file *
988 alloc_init_file(struct inode *ino)
990 struct nfs4_file *fp;
991 unsigned int hashval = file_hashval(ino);
993 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
994 if (fp) {
995 kref_init(&fp->fi_ref);
996 INIT_LIST_HEAD(&fp->fi_hash);
997 INIT_LIST_HEAD(&fp->fi_stateids);
998 INIT_LIST_HEAD(&fp->fi_delegations);
999 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1000 fp->fi_inode = igrab(ino);
1001 fp->fi_id = current_fileid++;
1002 return fp;
1004 return NULL;
1007 static void
1008 nfsd4_free_slab(struct kmem_cache **slab)
1010 if (*slab == NULL)
1011 return;
1012 kmem_cache_destroy(*slab);
1013 *slab = NULL;
1016 static void
1017 nfsd4_free_slabs(void)
1019 nfsd4_free_slab(&stateowner_slab);
1020 nfsd4_free_slab(&file_slab);
1021 nfsd4_free_slab(&stateid_slab);
1022 nfsd4_free_slab(&deleg_slab);
1025 static int
1026 nfsd4_init_slabs(void)
1028 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1029 sizeof(struct nfs4_stateowner), 0, 0, NULL, NULL);
1030 if (stateowner_slab == NULL)
1031 goto out_nomem;
1032 file_slab = kmem_cache_create("nfsd4_files",
1033 sizeof(struct nfs4_file), 0, 0, NULL, NULL);
1034 if (file_slab == NULL)
1035 goto out_nomem;
1036 stateid_slab = kmem_cache_create("nfsd4_stateids",
1037 sizeof(struct nfs4_stateid), 0, 0, NULL, NULL);
1038 if (stateid_slab == NULL)
1039 goto out_nomem;
1040 deleg_slab = kmem_cache_create("nfsd4_delegations",
1041 sizeof(struct nfs4_delegation), 0, 0, NULL, NULL);
1042 if (deleg_slab == NULL)
1043 goto out_nomem;
1044 return 0;
1045 out_nomem:
1046 nfsd4_free_slabs();
1047 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1048 return -ENOMEM;
1051 void
1052 nfs4_free_stateowner(struct kref *kref)
1054 struct nfs4_stateowner *sop =
1055 container_of(kref, struct nfs4_stateowner, so_ref);
1056 kfree(sop->so_owner.data);
1057 kmem_cache_free(stateowner_slab, sop);
1060 static inline struct nfs4_stateowner *
1061 alloc_stateowner(struct xdr_netobj *owner)
1063 struct nfs4_stateowner *sop;
1065 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1066 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1067 memcpy(sop->so_owner.data, owner->data, owner->len);
1068 sop->so_owner.len = owner->len;
1069 kref_init(&sop->so_ref);
1070 return sop;
1072 kmem_cache_free(stateowner_slab, sop);
1074 return NULL;
1077 static struct nfs4_stateowner *
1078 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1079 struct nfs4_stateowner *sop;
1080 struct nfs4_replay *rp;
1081 unsigned int idhashval;
1083 if (!(sop = alloc_stateowner(&open->op_owner)))
1084 return NULL;
1085 idhashval = ownerid_hashval(current_ownerid);
1086 INIT_LIST_HEAD(&sop->so_idhash);
1087 INIT_LIST_HEAD(&sop->so_strhash);
1088 INIT_LIST_HEAD(&sop->so_perclient);
1089 INIT_LIST_HEAD(&sop->so_stateids);
1090 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1091 INIT_LIST_HEAD(&sop->so_close_lru);
1092 sop->so_time = 0;
1093 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1094 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1095 list_add(&sop->so_perclient, &clp->cl_openowners);
1096 sop->so_is_open_owner = 1;
1097 sop->so_id = current_ownerid++;
1098 sop->so_client = clp;
1099 sop->so_seqid = open->op_seqid;
1100 sop->so_confirmed = 0;
1101 rp = &sop->so_replay;
1102 rp->rp_status = nfserr_serverfault;
1103 rp->rp_buflen = 0;
1104 rp->rp_buf = rp->rp_ibuf;
1105 return sop;
1108 static void
1109 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1111 struct nfs4_stateowner *lock_sop;
1113 while (!list_empty(&open_stp->st_lockowners)) {
1114 lock_sop = list_entry(open_stp->st_lockowners.next,
1115 struct nfs4_stateowner, so_perstateid);
1116 /* list_del(&open_stp->st_lockowners); */
1117 BUG_ON(lock_sop->so_is_open_owner);
1118 release_stateowner(lock_sop);
1122 static void
1123 unhash_stateowner(struct nfs4_stateowner *sop)
1125 struct nfs4_stateid *stp;
1127 list_del(&sop->so_idhash);
1128 list_del(&sop->so_strhash);
1129 if (sop->so_is_open_owner)
1130 list_del(&sop->so_perclient);
1131 list_del(&sop->so_perstateid);
1132 while (!list_empty(&sop->so_stateids)) {
1133 stp = list_entry(sop->so_stateids.next,
1134 struct nfs4_stateid, st_perstateowner);
1135 if (sop->so_is_open_owner)
1136 release_stateid(stp, OPEN_STATE);
1137 else
1138 release_stateid(stp, LOCK_STATE);
1142 static void
1143 release_stateowner(struct nfs4_stateowner *sop)
1145 unhash_stateowner(sop);
1146 list_del(&sop->so_close_lru);
1147 nfs4_put_stateowner(sop);
1150 static inline void
1151 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1152 struct nfs4_stateowner *sop = open->op_stateowner;
1153 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1155 INIT_LIST_HEAD(&stp->st_hash);
1156 INIT_LIST_HEAD(&stp->st_perstateowner);
1157 INIT_LIST_HEAD(&stp->st_lockowners);
1158 INIT_LIST_HEAD(&stp->st_perfile);
1159 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1160 list_add(&stp->st_perstateowner, &sop->so_stateids);
1161 list_add(&stp->st_perfile, &fp->fi_stateids);
1162 stp->st_stateowner = sop;
1163 get_nfs4_file(fp);
1164 stp->st_file = fp;
1165 stp->st_stateid.si_boot = boot_time;
1166 stp->st_stateid.si_stateownerid = sop->so_id;
1167 stp->st_stateid.si_fileid = fp->fi_id;
1168 stp->st_stateid.si_generation = 0;
1169 stp->st_access_bmap = 0;
1170 stp->st_deny_bmap = 0;
1171 __set_bit(open->op_share_access, &stp->st_access_bmap);
1172 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1173 stp->st_openstp = NULL;
1176 static void
1177 release_stateid(struct nfs4_stateid *stp, int flags)
1179 struct file *filp = stp->st_vfs_file;
1181 list_del(&stp->st_hash);
1182 list_del(&stp->st_perfile);
1183 list_del(&stp->st_perstateowner);
1184 if (flags & OPEN_STATE) {
1185 release_stateid_lockowners(stp);
1186 stp->st_vfs_file = NULL;
1187 nfsd_close(filp);
1188 } else if (flags & LOCK_STATE)
1189 locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1190 put_nfs4_file(stp->st_file);
1191 kmem_cache_free(stateid_slab, stp);
1194 static void
1195 move_to_close_lru(struct nfs4_stateowner *sop)
1197 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1199 list_move_tail(&sop->so_close_lru, &close_lru);
1200 sop->so_time = get_seconds();
1203 static int
1204 cmp_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner, clientid_t *clid) {
1205 return ((sop->so_owner.len == owner->len) &&
1206 !memcmp(sop->so_owner.data, owner->data, owner->len) &&
1207 (sop->so_client->cl_clientid.cl_id == clid->cl_id));
1210 static struct nfs4_stateowner *
1211 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1213 struct nfs4_stateowner *so = NULL;
1215 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1216 if (cmp_owner_str(so, &open->op_owner, &open->op_clientid))
1217 return so;
1219 return NULL;
1222 /* search file_hashtbl[] for file */
1223 static struct nfs4_file *
1224 find_file(struct inode *ino)
1226 unsigned int hashval = file_hashval(ino);
1227 struct nfs4_file *fp;
1229 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1230 if (fp->fi_inode == ino) {
1231 get_nfs4_file(fp);
1232 return fp;
1235 return NULL;
1238 static int access_valid(u32 x)
1240 return (x > 0 && x < 4);
1243 static int deny_valid(u32 x)
1245 return (x >= 0 && x < 5);
1248 static void
1249 set_access(unsigned int *access, unsigned long bmap) {
1250 int i;
1252 *access = 0;
1253 for (i = 1; i < 4; i++) {
1254 if (test_bit(i, &bmap))
1255 *access |= i;
1259 static void
1260 set_deny(unsigned int *deny, unsigned long bmap) {
1261 int i;
1263 *deny = 0;
1264 for (i = 0; i < 4; i++) {
1265 if (test_bit(i, &bmap))
1266 *deny |= i ;
1270 static int
1271 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1272 unsigned int access, deny;
1274 set_access(&access, stp->st_access_bmap);
1275 set_deny(&deny, stp->st_deny_bmap);
1276 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1277 return 0;
1278 return 1;
1282 * Called to check deny when READ with all zero stateid or
1283 * WRITE with all zero or all one stateid
1285 static __be32
1286 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1288 struct inode *ino = current_fh->fh_dentry->d_inode;
1289 struct nfs4_file *fp;
1290 struct nfs4_stateid *stp;
1291 __be32 ret;
1293 dprintk("NFSD: nfs4_share_conflict\n");
1295 fp = find_file(ino);
1296 if (!fp)
1297 return nfs_ok;
1298 ret = nfserr_locked;
1299 /* Search for conflicting share reservations */
1300 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1301 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1302 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1303 goto out;
1305 ret = nfs_ok;
1306 out:
1307 put_nfs4_file(fp);
1308 return ret;
1311 static inline void
1312 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1314 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1315 put_write_access(filp->f_path.dentry->d_inode);
1316 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1321 * Recall a delegation
1323 static int
1324 do_recall(void *__dp)
1326 struct nfs4_delegation *dp = __dp;
1328 daemonize("nfsv4-recall");
1330 nfsd4_cb_recall(dp);
1331 return 0;
1335 * Spawn a thread to perform a recall on the delegation represented
1336 * by the lease (file_lock)
1338 * Called from break_lease() with lock_kernel() held.
1339 * Note: we assume break_lease will only call this *once* for any given
1340 * lease.
1342 static
1343 void nfsd_break_deleg_cb(struct file_lock *fl)
1345 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1346 struct task_struct *t;
1348 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1349 if (!dp)
1350 return;
1352 /* We're assuming the state code never drops its reference
1353 * without first removing the lease. Since we're in this lease
1354 * callback (and since the lease code is serialized by the kernel
1355 * lock) we know the server hasn't removed the lease yet, we know
1356 * it's safe to take a reference: */
1357 atomic_inc(&dp->dl_count);
1359 spin_lock(&recall_lock);
1360 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1361 spin_unlock(&recall_lock);
1363 /* only place dl_time is set. protected by lock_kernel*/
1364 dp->dl_time = get_seconds();
1366 /* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */
1367 fl->fl_break_time = jiffies + NFSD_LEASE_TIME * HZ;
1369 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1370 if (IS_ERR(t)) {
1371 struct nfs4_client *clp = dp->dl_client;
1373 printk(KERN_INFO "NFSD: Callback thread failed for "
1374 "for client (clientid %08x/%08x)\n",
1375 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1376 nfs4_put_delegation(dp);
1381 * The file_lock is being reapd.
1383 * Called by locks_free_lock() with lock_kernel() held.
1385 static
1386 void nfsd_release_deleg_cb(struct file_lock *fl)
1388 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1390 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1392 if (!(fl->fl_flags & FL_LEASE) || !dp)
1393 return;
1394 dp->dl_flock = NULL;
1398 * Set the delegation file_lock back pointer.
1400 * Called from __setlease() with lock_kernel() held.
1402 static
1403 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1405 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1407 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1408 if (!dp)
1409 return;
1410 dp->dl_flock = new;
1414 * Called from __setlease() with lock_kernel() held
1416 static
1417 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1419 struct nfs4_delegation *onlistd =
1420 (struct nfs4_delegation *)onlist->fl_owner;
1421 struct nfs4_delegation *tryd =
1422 (struct nfs4_delegation *)try->fl_owner;
1424 if (onlist->fl_lmops != try->fl_lmops)
1425 return 0;
1427 return onlistd->dl_client == tryd->dl_client;
1431 static
1432 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1434 if (arg & F_UNLCK)
1435 return lease_modify(onlist, arg);
1436 else
1437 return -EAGAIN;
1440 static struct lock_manager_operations nfsd_lease_mng_ops = {
1441 .fl_break = nfsd_break_deleg_cb,
1442 .fl_release_private = nfsd_release_deleg_cb,
1443 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1444 .fl_mylease = nfsd_same_client_deleg_cb,
1445 .fl_change = nfsd_change_deleg_cb,
1449 __be32
1450 nfsd4_process_open1(struct nfsd4_open *open)
1452 clientid_t *clientid = &open->op_clientid;
1453 struct nfs4_client *clp = NULL;
1454 unsigned int strhashval;
1455 struct nfs4_stateowner *sop = NULL;
1457 if (!check_name(open->op_owner))
1458 return nfserr_inval;
1460 if (STALE_CLIENTID(&open->op_clientid))
1461 return nfserr_stale_clientid;
1463 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1464 sop = find_openstateowner_str(strhashval, open);
1465 open->op_stateowner = sop;
1466 if (!sop) {
1467 /* Make sure the client's lease hasn't expired. */
1468 clp = find_confirmed_client(clientid);
1469 if (clp == NULL)
1470 return nfserr_expired;
1471 goto renew;
1473 if (!sop->so_confirmed) {
1474 /* Replace unconfirmed owners without checking for replay. */
1475 clp = sop->so_client;
1476 release_stateowner(sop);
1477 open->op_stateowner = NULL;
1478 goto renew;
1480 if (open->op_seqid == sop->so_seqid - 1) {
1481 if (sop->so_replay.rp_buflen)
1482 return nfserr_replay_me;
1483 /* The original OPEN failed so spectacularly
1484 * that we don't even have replay data saved!
1485 * Therefore, we have no choice but to continue
1486 * processing this OPEN; presumably, we'll
1487 * fail again for the same reason.
1489 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1490 goto renew;
1492 if (open->op_seqid != sop->so_seqid)
1493 return nfserr_bad_seqid;
1494 renew:
1495 if (open->op_stateowner == NULL) {
1496 sop = alloc_init_open_stateowner(strhashval, clp, open);
1497 if (sop == NULL)
1498 return nfserr_resource;
1499 open->op_stateowner = sop;
1501 list_del_init(&sop->so_close_lru);
1502 renew_client(sop->so_client);
1503 return nfs_ok;
1506 static inline __be32
1507 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1509 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1510 return nfserr_openmode;
1511 else
1512 return nfs_ok;
1515 static struct nfs4_delegation *
1516 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1518 struct nfs4_delegation *dp;
1520 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1521 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1522 return dp;
1524 return NULL;
1527 static __be32
1528 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1529 struct nfs4_delegation **dp)
1531 int flags;
1532 __be32 status = nfserr_bad_stateid;
1534 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1535 if (*dp == NULL)
1536 goto out;
1537 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1538 RD_STATE : WR_STATE;
1539 status = nfs4_check_delegmode(*dp, flags);
1540 if (status)
1541 *dp = NULL;
1542 out:
1543 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1544 return nfs_ok;
1545 if (status)
1546 return status;
1547 open->op_stateowner->so_confirmed = 1;
1548 return nfs_ok;
1551 static __be32
1552 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1554 struct nfs4_stateid *local;
1555 __be32 status = nfserr_share_denied;
1556 struct nfs4_stateowner *sop = open->op_stateowner;
1558 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1559 /* ignore lock owners */
1560 if (local->st_stateowner->so_is_open_owner == 0)
1561 continue;
1562 /* remember if we have seen this open owner */
1563 if (local->st_stateowner == sop)
1564 *stpp = local;
1565 /* check for conflicting share reservations */
1566 if (!test_share(local, open))
1567 goto out;
1569 status = 0;
1570 out:
1571 return status;
1574 static inline struct nfs4_stateid *
1575 nfs4_alloc_stateid(void)
1577 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1580 static __be32
1581 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1582 struct nfs4_delegation *dp,
1583 struct svc_fh *cur_fh, int flags)
1585 struct nfs4_stateid *stp;
1587 stp = nfs4_alloc_stateid();
1588 if (stp == NULL)
1589 return nfserr_resource;
1591 if (dp) {
1592 get_file(dp->dl_vfs_file);
1593 stp->st_vfs_file = dp->dl_vfs_file;
1594 } else {
1595 __be32 status;
1596 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1597 &stp->st_vfs_file);
1598 if (status) {
1599 if (status == nfserr_dropit)
1600 status = nfserr_jukebox;
1601 kmem_cache_free(stateid_slab, stp);
1602 return status;
1605 *stpp = stp;
1606 return 0;
1609 static inline __be32
1610 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1611 struct nfsd4_open *open)
1613 struct iattr iattr = {
1614 .ia_valid = ATTR_SIZE,
1615 .ia_size = 0,
1617 if (!open->op_truncate)
1618 return 0;
1619 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1620 return nfserr_inval;
1621 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1624 static __be32
1625 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1627 struct file *filp = stp->st_vfs_file;
1628 struct inode *inode = filp->f_path.dentry->d_inode;
1629 unsigned int share_access, new_writer;
1630 __be32 status;
1632 set_access(&share_access, stp->st_access_bmap);
1633 new_writer = (~share_access) & open->op_share_access
1634 & NFS4_SHARE_ACCESS_WRITE;
1636 if (new_writer) {
1637 int err = get_write_access(inode);
1638 if (err)
1639 return nfserrno(err);
1641 status = nfsd4_truncate(rqstp, cur_fh, open);
1642 if (status) {
1643 if (new_writer)
1644 put_write_access(inode);
1645 return status;
1647 /* remember the open */
1648 filp->f_mode |= open->op_share_access;
1649 set_bit(open->op_share_access, &stp->st_access_bmap);
1650 set_bit(open->op_share_deny, &stp->st_deny_bmap);
1652 return nfs_ok;
1656 static void
1657 nfs4_set_claim_prev(struct nfsd4_open *open)
1659 open->op_stateowner->so_confirmed = 1;
1660 open->op_stateowner->so_client->cl_firststate = 1;
1664 * Attempt to hand out a delegation.
1666 static void
1667 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1669 struct nfs4_delegation *dp;
1670 struct nfs4_stateowner *sop = stp->st_stateowner;
1671 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1672 struct file_lock fl, *flp = &fl;
1673 int status, flag = 0;
1675 flag = NFS4_OPEN_DELEGATE_NONE;
1676 open->op_recall = 0;
1677 switch (open->op_claim_type) {
1678 case NFS4_OPEN_CLAIM_PREVIOUS:
1679 if (!atomic_read(&cb->cb_set))
1680 open->op_recall = 1;
1681 flag = open->op_delegate_type;
1682 if (flag == NFS4_OPEN_DELEGATE_NONE)
1683 goto out;
1684 break;
1685 case NFS4_OPEN_CLAIM_NULL:
1686 /* Let's not give out any delegations till everyone's
1687 * had the chance to reclaim theirs.... */
1688 if (nfs4_in_grace())
1689 goto out;
1690 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1691 goto out;
1692 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1693 flag = NFS4_OPEN_DELEGATE_WRITE;
1694 else
1695 flag = NFS4_OPEN_DELEGATE_READ;
1696 break;
1697 default:
1698 goto out;
1701 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1702 if (dp == NULL) {
1703 flag = NFS4_OPEN_DELEGATE_NONE;
1704 goto out;
1706 locks_init_lock(&fl);
1707 fl.fl_lmops = &nfsd_lease_mng_ops;
1708 fl.fl_flags = FL_LEASE;
1709 fl.fl_end = OFFSET_MAX;
1710 fl.fl_owner = (fl_owner_t)dp;
1711 fl.fl_file = stp->st_vfs_file;
1712 fl.fl_pid = current->tgid;
1714 /* setlease checks to see if delegation should be handed out.
1715 * the lock_manager callbacks fl_mylease and fl_change are used
1717 if ((status = setlease(stp->st_vfs_file,
1718 flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1719 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1720 unhash_delegation(dp);
1721 flag = NFS4_OPEN_DELEGATE_NONE;
1722 goto out;
1725 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1727 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1728 dp->dl_stateid.si_boot,
1729 dp->dl_stateid.si_stateownerid,
1730 dp->dl_stateid.si_fileid,
1731 dp->dl_stateid.si_generation);
1732 out:
1733 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1734 && flag == NFS4_OPEN_DELEGATE_NONE
1735 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1736 printk("NFSD: WARNING: refusing delegation reclaim\n");
1737 open->op_delegate_type = flag;
1741 * called with nfs4_lock_state() held.
1743 __be32
1744 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1746 struct nfs4_file *fp = NULL;
1747 struct inode *ino = current_fh->fh_dentry->d_inode;
1748 struct nfs4_stateid *stp = NULL;
1749 struct nfs4_delegation *dp = NULL;
1750 __be32 status;
1752 status = nfserr_inval;
1753 if (!access_valid(open->op_share_access)
1754 || !deny_valid(open->op_share_deny))
1755 goto out;
1757 * Lookup file; if found, lookup stateid and check open request,
1758 * and check for delegations in the process of being recalled.
1759 * If not found, create the nfs4_file struct
1761 fp = find_file(ino);
1762 if (fp) {
1763 if ((status = nfs4_check_open(fp, open, &stp)))
1764 goto out;
1765 status = nfs4_check_deleg(fp, open, &dp);
1766 if (status)
1767 goto out;
1768 } else {
1769 status = nfserr_bad_stateid;
1770 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1771 goto out;
1772 status = nfserr_resource;
1773 fp = alloc_init_file(ino);
1774 if (fp == NULL)
1775 goto out;
1779 * OPEN the file, or upgrade an existing OPEN.
1780 * If truncate fails, the OPEN fails.
1782 if (stp) {
1783 /* Stateid was found, this is an OPEN upgrade */
1784 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1785 if (status)
1786 goto out;
1787 update_stateid(&stp->st_stateid);
1788 } else {
1789 /* Stateid was not found, this is a new OPEN */
1790 int flags = 0;
1791 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1792 flags |= MAY_READ;
1793 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1794 flags |= MAY_WRITE;
1795 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1796 if (status)
1797 goto out;
1798 init_stateid(stp, fp, open);
1799 status = nfsd4_truncate(rqstp, current_fh, open);
1800 if (status) {
1801 release_stateid(stp, OPEN_STATE);
1802 goto out;
1805 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1808 * Attempt to hand out a delegation. No error return, because the
1809 * OPEN succeeds even if we fail.
1811 nfs4_open_delegation(current_fh, open, stp);
1813 status = nfs_ok;
1815 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1816 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1817 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1818 out:
1819 if (fp)
1820 put_nfs4_file(fp);
1821 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1822 nfs4_set_claim_prev(open);
1824 * To finish the open response, we just need to set the rflags.
1826 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1827 if (!open->op_stateowner->so_confirmed)
1828 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1830 return status;
1833 static struct workqueue_struct *laundry_wq;
1834 static void laundromat_main(struct work_struct *);
1835 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1837 __be32
1838 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1839 clientid_t *clid)
1841 struct nfs4_client *clp;
1842 __be32 status;
1844 nfs4_lock_state();
1845 dprintk("process_renew(%08x/%08x): starting\n",
1846 clid->cl_boot, clid->cl_id);
1847 status = nfserr_stale_clientid;
1848 if (STALE_CLIENTID(clid))
1849 goto out;
1850 clp = find_confirmed_client(clid);
1851 status = nfserr_expired;
1852 if (clp == NULL) {
1853 /* We assume the client took too long to RENEW. */
1854 dprintk("nfsd4_renew: clientid not found!\n");
1855 goto out;
1857 renew_client(clp);
1858 status = nfserr_cb_path_down;
1859 if (!list_empty(&clp->cl_delegations)
1860 && !atomic_read(&clp->cl_callback.cb_set))
1861 goto out;
1862 status = nfs_ok;
1863 out:
1864 nfs4_unlock_state();
1865 return status;
1868 static void
1869 end_grace(void)
1871 dprintk("NFSD: end of grace period\n");
1872 nfsd4_recdir_purge_old();
1873 in_grace = 0;
1876 static time_t
1877 nfs4_laundromat(void)
1879 struct nfs4_client *clp;
1880 struct nfs4_stateowner *sop;
1881 struct nfs4_delegation *dp;
1882 struct list_head *pos, *next, reaplist;
1883 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1884 time_t t, clientid_val = NFSD_LEASE_TIME;
1885 time_t u, test_val = NFSD_LEASE_TIME;
1887 nfs4_lock_state();
1889 dprintk("NFSD: laundromat service - starting\n");
1890 if (in_grace)
1891 end_grace();
1892 list_for_each_safe(pos, next, &client_lru) {
1893 clp = list_entry(pos, struct nfs4_client, cl_lru);
1894 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1895 t = clp->cl_time - cutoff;
1896 if (clientid_val > t)
1897 clientid_val = t;
1898 break;
1900 dprintk("NFSD: purging unused client (clientid %08x)\n",
1901 clp->cl_clientid.cl_id);
1902 nfsd4_remove_clid_dir(clp);
1903 expire_client(clp);
1905 INIT_LIST_HEAD(&reaplist);
1906 spin_lock(&recall_lock);
1907 list_for_each_safe(pos, next, &del_recall_lru) {
1908 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1909 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1910 u = dp->dl_time - cutoff;
1911 if (test_val > u)
1912 test_val = u;
1913 break;
1915 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1916 dp, dp->dl_flock);
1917 list_move(&dp->dl_recall_lru, &reaplist);
1919 spin_unlock(&recall_lock);
1920 list_for_each_safe(pos, next, &reaplist) {
1921 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1922 list_del_init(&dp->dl_recall_lru);
1923 unhash_delegation(dp);
1925 test_val = NFSD_LEASE_TIME;
1926 list_for_each_safe(pos, next, &close_lru) {
1927 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1928 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1929 u = sop->so_time - cutoff;
1930 if (test_val > u)
1931 test_val = u;
1932 break;
1934 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1935 sop->so_id);
1936 release_stateowner(sop);
1938 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1939 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1940 nfs4_unlock_state();
1941 return clientid_val;
1944 void
1945 laundromat_main(struct work_struct *not_used)
1947 time_t t;
1949 t = nfs4_laundromat();
1950 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1951 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1954 static struct nfs4_stateowner *
1955 search_close_lru(u32 st_id, int flags)
1957 struct nfs4_stateowner *local = NULL;
1959 if (flags & CLOSE_STATE) {
1960 list_for_each_entry(local, &close_lru, so_close_lru) {
1961 if (local->so_id == st_id)
1962 return local;
1965 return NULL;
1968 static inline int
1969 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1971 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1974 static int
1975 STALE_STATEID(stateid_t *stateid)
1977 if (stateid->si_boot == boot_time)
1978 return 0;
1979 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1980 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1981 stateid->si_generation);
1982 return 1;
1985 static inline int
1986 access_permit_read(unsigned long access_bmap)
1988 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1989 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1990 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1993 static inline int
1994 access_permit_write(unsigned long access_bmap)
1996 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1997 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2000 static
2001 __be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2003 __be32 status = nfserr_openmode;
2005 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2006 goto out;
2007 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2008 goto out;
2009 status = nfs_ok;
2010 out:
2011 return status;
2014 static inline __be32
2015 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2017 /* Trying to call delegreturn with a special stateid? Yuch: */
2018 if (!(flags & (RD_STATE | WR_STATE)))
2019 return nfserr_bad_stateid;
2020 else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2021 return nfs_ok;
2022 else if (nfs4_in_grace()) {
2023 /* Answer in remaining cases depends on existance of
2024 * conflicting state; so we must wait out the grace period. */
2025 return nfserr_grace;
2026 } else if (flags & WR_STATE)
2027 return nfs4_share_conflict(current_fh,
2028 NFS4_SHARE_DENY_WRITE);
2029 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2030 return nfs4_share_conflict(current_fh,
2031 NFS4_SHARE_DENY_READ);
2035 * Allow READ/WRITE during grace period on recovered state only for files
2036 * that are not able to provide mandatory locking.
2038 static inline int
2039 io_during_grace_disallowed(struct inode *inode, int flags)
2041 return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
2042 && MANDATORY_LOCK(inode);
2046 * Checks for stateid operations
2048 __be32
2049 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2051 struct nfs4_stateid *stp = NULL;
2052 struct nfs4_delegation *dp = NULL;
2053 stateid_t *stidp;
2054 struct inode *ino = current_fh->fh_dentry->d_inode;
2055 __be32 status;
2057 dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2058 stateid->si_boot, stateid->si_stateownerid,
2059 stateid->si_fileid, stateid->si_generation);
2060 if (filpp)
2061 *filpp = NULL;
2063 if (io_during_grace_disallowed(ino, flags))
2064 return nfserr_grace;
2066 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2067 return check_special_stateids(current_fh, stateid, flags);
2069 /* STALE STATEID */
2070 status = nfserr_stale_stateid;
2071 if (STALE_STATEID(stateid))
2072 goto out;
2074 /* BAD STATEID */
2075 status = nfserr_bad_stateid;
2076 if (!stateid->si_fileid) { /* delegation stateid */
2077 if(!(dp = find_delegation_stateid(ino, stateid))) {
2078 dprintk("NFSD: delegation stateid not found\n");
2079 goto out;
2081 stidp = &dp->dl_stateid;
2082 } else { /* open or lock stateid */
2083 if (!(stp = find_stateid(stateid, flags))) {
2084 dprintk("NFSD: open or lock stateid not found\n");
2085 goto out;
2087 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2088 goto out;
2089 if (!stp->st_stateowner->so_confirmed)
2090 goto out;
2091 stidp = &stp->st_stateid;
2093 if (stateid->si_generation > stidp->si_generation)
2094 goto out;
2096 /* OLD STATEID */
2097 status = nfserr_old_stateid;
2098 if (stateid->si_generation < stidp->si_generation)
2099 goto out;
2100 if (stp) {
2101 if ((status = nfs4_check_openmode(stp,flags)))
2102 goto out;
2103 renew_client(stp->st_stateowner->so_client);
2104 if (filpp)
2105 *filpp = stp->st_vfs_file;
2106 } else if (dp) {
2107 if ((status = nfs4_check_delegmode(dp, flags)))
2108 goto out;
2109 renew_client(dp->dl_client);
2110 if (flags & DELEG_RET)
2111 unhash_delegation(dp);
2112 if (filpp)
2113 *filpp = dp->dl_vfs_file;
2115 status = nfs_ok;
2116 out:
2117 return status;
2120 static inline int
2121 setlkflg (int type)
2123 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2124 RD_STATE : WR_STATE;
2128 * Checks for sequence id mutating operations.
2130 static __be32
2131 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)
2133 struct nfs4_stateid *stp;
2134 struct nfs4_stateowner *sop;
2136 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2137 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2138 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2139 stateid->si_generation);
2141 *stpp = NULL;
2142 *sopp = NULL;
2144 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2145 printk("NFSD: preprocess_seqid_op: magic stateid!\n");
2146 return nfserr_bad_stateid;
2149 if (STALE_STATEID(stateid))
2150 return nfserr_stale_stateid;
2152 * We return BAD_STATEID if filehandle doesn't match stateid,
2153 * the confirmed flag is incorrecly set, or the generation
2154 * number is incorrect.
2156 stp = find_stateid(stateid, flags);
2157 if (stp == NULL) {
2159 * Also, we should make sure this isn't just the result of
2160 * a replayed close:
2162 sop = search_close_lru(stateid->si_stateownerid, flags);
2163 if (sop == NULL)
2164 return nfserr_bad_stateid;
2165 *sopp = sop;
2166 goto check_replay;
2169 if (lock) {
2170 struct nfs4_stateowner *sop = stp->st_stateowner;
2171 clientid_t *lockclid = &lock->v.new.clientid;
2172 struct nfs4_client *clp = sop->so_client;
2173 int lkflg = 0;
2174 __be32 status;
2176 lkflg = setlkflg(lock->lk_type);
2178 if (lock->lk_is_new) {
2179 if (!sop->so_is_open_owner)
2180 return nfserr_bad_stateid;
2181 if (!cmp_clid(&clp->cl_clientid, lockclid))
2182 return nfserr_bad_stateid;
2183 /* stp is the open stateid */
2184 status = nfs4_check_openmode(stp, lkflg);
2185 if (status)
2186 return status;
2187 } else {
2188 /* stp is the lock stateid */
2189 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2190 if (status)
2191 return status;
2196 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2197 printk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2198 return nfserr_bad_stateid;
2201 *stpp = stp;
2202 *sopp = sop = stp->st_stateowner;
2205 * We now validate the seqid and stateid generation numbers.
2206 * For the moment, we ignore the possibility of
2207 * generation number wraparound.
2209 if (seqid != sop->so_seqid)
2210 goto check_replay;
2212 if (sop->so_confirmed && flags & CONFIRM) {
2213 printk("NFSD: preprocess_seqid_op: expected"
2214 " unconfirmed stateowner!\n");
2215 return nfserr_bad_stateid;
2217 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2218 printk("NFSD: preprocess_seqid_op: stateowner not"
2219 " confirmed yet!\n");
2220 return nfserr_bad_stateid;
2222 if (stateid->si_generation > stp->st_stateid.si_generation) {
2223 printk("NFSD: preprocess_seqid_op: future stateid?!\n");
2224 return nfserr_bad_stateid;
2227 if (stateid->si_generation < stp->st_stateid.si_generation) {
2228 printk("NFSD: preprocess_seqid_op: old stateid!\n");
2229 return nfserr_old_stateid;
2231 renew_client(sop->so_client);
2232 return nfs_ok;
2234 check_replay:
2235 if (seqid == sop->so_seqid - 1) {
2236 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2237 /* indicate replay to calling function */
2238 return nfserr_replay_me;
2240 printk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2241 sop->so_seqid, seqid);
2242 *sopp = NULL;
2243 return nfserr_bad_seqid;
2246 __be32
2247 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2248 struct nfsd4_open_confirm *oc)
2250 __be32 status;
2251 struct nfs4_stateowner *sop;
2252 struct nfs4_stateid *stp;
2254 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2255 (int)cstate->current_fh.fh_dentry->d_name.len,
2256 cstate->current_fh.fh_dentry->d_name.name);
2258 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
2259 if (status)
2260 return status;
2262 nfs4_lock_state();
2264 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2265 oc->oc_seqid, &oc->oc_req_stateid,
2266 CHECK_FH | CONFIRM | OPEN_STATE,
2267 &oc->oc_stateowner, &stp, NULL)))
2268 goto out;
2270 sop = oc->oc_stateowner;
2271 sop->so_confirmed = 1;
2272 update_stateid(&stp->st_stateid);
2273 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2274 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2275 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2276 stp->st_stateid.si_boot,
2277 stp->st_stateid.si_stateownerid,
2278 stp->st_stateid.si_fileid,
2279 stp->st_stateid.si_generation);
2281 nfsd4_create_clid_dir(sop->so_client);
2282 out:
2283 if (oc->oc_stateowner) {
2284 nfs4_get_stateowner(oc->oc_stateowner);
2285 cstate->replay_owner = oc->oc_stateowner;
2287 nfs4_unlock_state();
2288 return status;
2293 * unset all bits in union bitmap (bmap) that
2294 * do not exist in share (from successful OPEN_DOWNGRADE)
2296 static void
2297 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2299 int i;
2300 for (i = 1; i < 4; i++) {
2301 if ((i & access) != i)
2302 __clear_bit(i, bmap);
2306 static void
2307 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2309 int i;
2310 for (i = 0; i < 4; i++) {
2311 if ((i & deny) != i)
2312 __clear_bit(i, bmap);
2316 __be32
2317 nfsd4_open_downgrade(struct svc_rqst *rqstp,
2318 struct nfsd4_compound_state *cstate,
2319 struct nfsd4_open_downgrade *od)
2321 __be32 status;
2322 struct nfs4_stateid *stp;
2323 unsigned int share_access;
2325 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2326 (int)cstate->current_fh.fh_dentry->d_name.len,
2327 cstate->current_fh.fh_dentry->d_name.name);
2329 if (!access_valid(od->od_share_access)
2330 || !deny_valid(od->od_share_deny))
2331 return nfserr_inval;
2333 nfs4_lock_state();
2334 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2335 od->od_seqid,
2336 &od->od_stateid,
2337 CHECK_FH | OPEN_STATE,
2338 &od->od_stateowner, &stp, NULL)))
2339 goto out;
2341 status = nfserr_inval;
2342 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2343 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2344 stp->st_access_bmap, od->od_share_access);
2345 goto out;
2347 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2348 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2349 stp->st_deny_bmap, od->od_share_deny);
2350 goto out;
2352 set_access(&share_access, stp->st_access_bmap);
2353 nfs4_file_downgrade(stp->st_vfs_file,
2354 share_access & ~od->od_share_access);
2356 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2357 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2359 update_stateid(&stp->st_stateid);
2360 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2361 status = nfs_ok;
2362 out:
2363 if (od->od_stateowner) {
2364 nfs4_get_stateowner(od->od_stateowner);
2365 cstate->replay_owner = od->od_stateowner;
2367 nfs4_unlock_state();
2368 return status;
2372 * nfs4_unlock_state() called after encode
2374 __be32
2375 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2376 struct nfsd4_close *close)
2378 __be32 status;
2379 struct nfs4_stateid *stp;
2381 dprintk("NFSD: nfsd4_close on file %.*s\n",
2382 (int)cstate->current_fh.fh_dentry->d_name.len,
2383 cstate->current_fh.fh_dentry->d_name.name);
2385 nfs4_lock_state();
2386 /* check close_lru for replay */
2387 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2388 close->cl_seqid,
2389 &close->cl_stateid,
2390 CHECK_FH | OPEN_STATE | CLOSE_STATE,
2391 &close->cl_stateowner, &stp, NULL)))
2392 goto out;
2393 status = nfs_ok;
2394 update_stateid(&stp->st_stateid);
2395 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2397 /* release_stateid() calls nfsd_close() if needed */
2398 release_stateid(stp, OPEN_STATE);
2400 /* place unused nfs4_stateowners on so_close_lru list to be
2401 * released by the laundromat service after the lease period
2402 * to enable us to handle CLOSE replay
2404 if (list_empty(&close->cl_stateowner->so_stateids))
2405 move_to_close_lru(close->cl_stateowner);
2406 out:
2407 if (close->cl_stateowner) {
2408 nfs4_get_stateowner(close->cl_stateowner);
2409 cstate->replay_owner = close->cl_stateowner;
2411 nfs4_unlock_state();
2412 return status;
2415 __be32
2416 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2417 struct nfsd4_delegreturn *dr)
2419 __be32 status;
2421 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
2422 goto out;
2424 nfs4_lock_state();
2425 status = nfs4_preprocess_stateid_op(&cstate->current_fh,
2426 &dr->dr_stateid, DELEG_RET, NULL);
2427 nfs4_unlock_state();
2428 out:
2429 return status;
2434 * Lock owner state (byte-range locks)
2436 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2437 #define LOCK_HASH_BITS 8
2438 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2439 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2441 #define lockownerid_hashval(id) \
2442 ((id) & LOCK_HASH_MASK)
2444 static inline unsigned int
2445 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2446 struct xdr_netobj *ownername)
2448 return (file_hashval(inode) + cl_id
2449 + opaque_hashval(ownername->data, ownername->len))
2450 & LOCK_HASH_MASK;
2453 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2454 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2455 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2457 static struct nfs4_stateid *
2458 find_stateid(stateid_t *stid, int flags)
2460 struct nfs4_stateid *local = NULL;
2461 u32 st_id = stid->si_stateownerid;
2462 u32 f_id = stid->si_fileid;
2463 unsigned int hashval;
2465 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2466 if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2467 hashval = stateid_hashval(st_id, f_id);
2468 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2469 if ((local->st_stateid.si_stateownerid == st_id) &&
2470 (local->st_stateid.si_fileid == f_id))
2471 return local;
2474 if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2475 hashval = stateid_hashval(st_id, f_id);
2476 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2477 if ((local->st_stateid.si_stateownerid == st_id) &&
2478 (local->st_stateid.si_fileid == f_id))
2479 return local;
2482 return NULL;
2485 static struct nfs4_delegation *
2486 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2488 struct nfs4_file *fp;
2489 struct nfs4_delegation *dl;
2491 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2492 stid->si_boot, stid->si_stateownerid,
2493 stid->si_fileid, stid->si_generation);
2495 fp = find_file(ino);
2496 if (!fp)
2497 return NULL;
2498 dl = find_delegation_file(fp, stid);
2499 put_nfs4_file(fp);
2500 return dl;
2504 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2505 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2506 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2507 * locking, this prevents us from being completely protocol-compliant. The
2508 * real solution to this problem is to start using unsigned file offsets in
2509 * the VFS, but this is a very deep change!
2511 static inline void
2512 nfs4_transform_lock_offset(struct file_lock *lock)
2514 if (lock->fl_start < 0)
2515 lock->fl_start = OFFSET_MAX;
2516 if (lock->fl_end < 0)
2517 lock->fl_end = OFFSET_MAX;
2520 /* Hack!: For now, we're defining this just so we can use a pointer to it
2521 * as a unique cookie to identify our (NFSv4's) posix locks. */
2522 static struct lock_manager_operations nfsd_posix_mng_ops = {
2525 static inline void
2526 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2528 struct nfs4_stateowner *sop;
2529 unsigned int hval;
2531 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2532 sop = (struct nfs4_stateowner *) fl->fl_owner;
2533 hval = lockownerid_hashval(sop->so_id);
2534 kref_get(&sop->so_ref);
2535 deny->ld_sop = sop;
2536 deny->ld_clientid = sop->so_client->cl_clientid;
2537 } else {
2538 deny->ld_sop = NULL;
2539 deny->ld_clientid.cl_boot = 0;
2540 deny->ld_clientid.cl_id = 0;
2542 deny->ld_start = fl->fl_start;
2543 deny->ld_length = ~(u64)0;
2544 if (fl->fl_end != ~(u64)0)
2545 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2546 deny->ld_type = NFS4_READ_LT;
2547 if (fl->fl_type != F_RDLCK)
2548 deny->ld_type = NFS4_WRITE_LT;
2551 static struct nfs4_stateowner *
2552 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2553 struct xdr_netobj *owner)
2555 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2556 struct nfs4_stateowner *op;
2558 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2559 if (cmp_owner_str(op, owner, clid))
2560 return op;
2562 return NULL;
2566 * Alloc a lock owner structure.
2567 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2568 * occured.
2570 * strhashval = lock_ownerstr_hashval
2573 static struct nfs4_stateowner *
2574 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2575 struct nfs4_stateowner *sop;
2576 struct nfs4_replay *rp;
2577 unsigned int idhashval;
2579 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2580 return NULL;
2581 idhashval = lockownerid_hashval(current_ownerid);
2582 INIT_LIST_HEAD(&sop->so_idhash);
2583 INIT_LIST_HEAD(&sop->so_strhash);
2584 INIT_LIST_HEAD(&sop->so_perclient);
2585 INIT_LIST_HEAD(&sop->so_stateids);
2586 INIT_LIST_HEAD(&sop->so_perstateid);
2587 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2588 sop->so_time = 0;
2589 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2590 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2591 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2592 sop->so_is_open_owner = 0;
2593 sop->so_id = current_ownerid++;
2594 sop->so_client = clp;
2595 /* It is the openowner seqid that will be incremented in encode in the
2596 * case of new lockowners; so increment the lock seqid manually: */
2597 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2598 sop->so_confirmed = 1;
2599 rp = &sop->so_replay;
2600 rp->rp_status = nfserr_serverfault;
2601 rp->rp_buflen = 0;
2602 rp->rp_buf = rp->rp_ibuf;
2603 return sop;
2606 static struct nfs4_stateid *
2607 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2609 struct nfs4_stateid *stp;
2610 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2612 stp = nfs4_alloc_stateid();
2613 if (stp == NULL)
2614 goto out;
2615 INIT_LIST_HEAD(&stp->st_hash);
2616 INIT_LIST_HEAD(&stp->st_perfile);
2617 INIT_LIST_HEAD(&stp->st_perstateowner);
2618 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2619 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2620 list_add(&stp->st_perfile, &fp->fi_stateids);
2621 list_add(&stp->st_perstateowner, &sop->so_stateids);
2622 stp->st_stateowner = sop;
2623 get_nfs4_file(fp);
2624 stp->st_file = fp;
2625 stp->st_stateid.si_boot = boot_time;
2626 stp->st_stateid.si_stateownerid = sop->so_id;
2627 stp->st_stateid.si_fileid = fp->fi_id;
2628 stp->st_stateid.si_generation = 0;
2629 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2630 stp->st_access_bmap = open_stp->st_access_bmap;
2631 stp->st_deny_bmap = open_stp->st_deny_bmap;
2632 stp->st_openstp = open_stp;
2634 out:
2635 return stp;
2638 static int
2639 check_lock_length(u64 offset, u64 length)
2641 return ((length == 0) || ((length != ~(u64)0) &&
2642 LOFF_OVERFLOW(offset, length)));
2646 * LOCK operation
2648 __be32
2649 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2650 struct nfsd4_lock *lock)
2652 struct nfs4_stateowner *open_sop = NULL;
2653 struct nfs4_stateowner *lock_sop = NULL;
2654 struct nfs4_stateid *lock_stp;
2655 struct file *filp;
2656 struct file_lock file_lock;
2657 struct file_lock conflock;
2658 __be32 status = 0;
2659 unsigned int strhashval;
2660 int err;
2662 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2663 (long long) lock->lk_offset,
2664 (long long) lock->lk_length);
2666 if (check_lock_length(lock->lk_offset, lock->lk_length))
2667 return nfserr_inval;
2669 if ((status = fh_verify(rqstp, &cstate->current_fh,
2670 S_IFREG, MAY_LOCK))) {
2671 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2672 return status;
2675 nfs4_lock_state();
2677 if (lock->lk_is_new) {
2679 * Client indicates that this is a new lockowner.
2680 * Use open owner and open stateid to create lock owner and
2681 * lock stateid.
2683 struct nfs4_stateid *open_stp = NULL;
2684 struct nfs4_file *fp;
2686 status = nfserr_stale_clientid;
2687 if (STALE_CLIENTID(&lock->lk_new_clientid))
2688 goto out;
2690 /* validate and update open stateid and open seqid */
2691 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2692 lock->lk_new_open_seqid,
2693 &lock->lk_new_open_stateid,
2694 CHECK_FH | OPEN_STATE,
2695 &lock->lk_replay_owner, &open_stp,
2696 lock);
2697 if (status)
2698 goto out;
2699 open_sop = lock->lk_replay_owner;
2700 /* create lockowner and lock stateid */
2701 fp = open_stp->st_file;
2702 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2703 open_sop->so_client->cl_clientid.cl_id,
2704 &lock->v.new.owner);
2705 /* XXX: Do we need to check for duplicate stateowners on
2706 * the same file, or should they just be allowed (and
2707 * create new stateids)? */
2708 status = nfserr_resource;
2709 lock_sop = alloc_init_lock_stateowner(strhashval,
2710 open_sop->so_client, open_stp, lock);
2711 if (lock_sop == NULL)
2712 goto out;
2713 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2714 if (lock_stp == NULL)
2715 goto out;
2716 } else {
2717 /* lock (lock owner + lock stateid) already exists */
2718 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2719 lock->lk_old_lock_seqid,
2720 &lock->lk_old_lock_stateid,
2721 CHECK_FH | LOCK_STATE,
2722 &lock->lk_replay_owner, &lock_stp, lock);
2723 if (status)
2724 goto out;
2725 lock_sop = lock->lk_replay_owner;
2727 /* lock->lk_replay_owner and lock_stp have been created or found */
2728 filp = lock_stp->st_vfs_file;
2730 status = nfserr_grace;
2731 if (nfs4_in_grace() && !lock->lk_reclaim)
2732 goto out;
2733 status = nfserr_no_grace;
2734 if (!nfs4_in_grace() && lock->lk_reclaim)
2735 goto out;
2737 locks_init_lock(&file_lock);
2738 switch (lock->lk_type) {
2739 case NFS4_READ_LT:
2740 case NFS4_READW_LT:
2741 file_lock.fl_type = F_RDLCK;
2742 break;
2743 case NFS4_WRITE_LT:
2744 case NFS4_WRITEW_LT:
2745 file_lock.fl_type = F_WRLCK;
2746 break;
2747 default:
2748 status = nfserr_inval;
2749 goto out;
2751 file_lock.fl_owner = (fl_owner_t)lock_sop;
2752 file_lock.fl_pid = current->tgid;
2753 file_lock.fl_file = filp;
2754 file_lock.fl_flags = FL_POSIX;
2755 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2757 file_lock.fl_start = lock->lk_offset;
2758 if ((lock->lk_length == ~(u64)0) ||
2759 LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2760 file_lock.fl_end = ~(u64)0;
2761 else
2762 file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2763 nfs4_transform_lock_offset(&file_lock);
2766 * Try to lock the file in the VFS.
2767 * Note: locks.c uses the BKL to protect the inode's lock list.
2770 /* XXX?: Just to divert the locks_release_private at the start of
2771 * locks_copy_lock: */
2772 conflock.fl_ops = NULL;
2773 conflock.fl_lmops = NULL;
2774 err = posix_lock_file_conf(filp, &file_lock, &conflock);
2775 switch (-err) {
2776 case 0: /* success! */
2777 update_stateid(&lock_stp->st_stateid);
2778 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2779 sizeof(stateid_t));
2780 status = 0;
2781 break;
2782 case (EAGAIN): /* conflock holds conflicting lock */
2783 status = nfserr_denied;
2784 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2785 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2786 break;
2787 case (EDEADLK):
2788 status = nfserr_deadlock;
2789 break;
2790 default:
2791 dprintk("NFSD: nfsd4_lock: posix_lock_file_conf() failed! status %d\n",err);
2792 status = nfserr_resource;
2793 break;
2795 out:
2796 if (status && lock->lk_is_new && lock_sop)
2797 release_stateowner(lock_sop);
2798 if (lock->lk_replay_owner) {
2799 nfs4_get_stateowner(lock->lk_replay_owner);
2800 cstate->replay_owner = lock->lk_replay_owner;
2802 nfs4_unlock_state();
2803 return status;
2807 * LOCKT operation
2809 __be32
2810 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2811 struct nfsd4_lockt *lockt)
2813 struct inode *inode;
2814 struct file file;
2815 struct file_lock file_lock;
2816 struct file_lock conflock;
2817 __be32 status;
2819 if (nfs4_in_grace())
2820 return nfserr_grace;
2822 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2823 return nfserr_inval;
2825 lockt->lt_stateowner = NULL;
2826 nfs4_lock_state();
2828 status = nfserr_stale_clientid;
2829 if (STALE_CLIENTID(&lockt->lt_clientid))
2830 goto out;
2832 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
2833 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2834 if (status == nfserr_symlink)
2835 status = nfserr_inval;
2836 goto out;
2839 inode = cstate->current_fh.fh_dentry->d_inode;
2840 locks_init_lock(&file_lock);
2841 switch (lockt->lt_type) {
2842 case NFS4_READ_LT:
2843 case NFS4_READW_LT:
2844 file_lock.fl_type = F_RDLCK;
2845 break;
2846 case NFS4_WRITE_LT:
2847 case NFS4_WRITEW_LT:
2848 file_lock.fl_type = F_WRLCK;
2849 break;
2850 default:
2851 printk("NFSD: nfs4_lockt: bad lock type!\n");
2852 status = nfserr_inval;
2853 goto out;
2856 lockt->lt_stateowner = find_lockstateowner_str(inode,
2857 &lockt->lt_clientid, &lockt->lt_owner);
2858 if (lockt->lt_stateowner)
2859 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2860 file_lock.fl_pid = current->tgid;
2861 file_lock.fl_flags = FL_POSIX;
2862 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2864 file_lock.fl_start = lockt->lt_offset;
2865 if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2866 file_lock.fl_end = ~(u64)0;
2867 else
2868 file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2870 nfs4_transform_lock_offset(&file_lock);
2872 /* posix_test_lock uses the struct file _only_ to resolve the inode.
2873 * since LOCKT doesn't require an OPEN, and therefore a struct
2874 * file may not exist, pass posix_test_lock a struct file with
2875 * only the dentry:inode set.
2877 memset(&file, 0, sizeof (struct file));
2878 file.f_path.dentry = cstate->current_fh.fh_dentry;
2880 status = nfs_ok;
2881 if (posix_test_lock(&file, &file_lock, &conflock)) {
2882 status = nfserr_denied;
2883 nfs4_set_lock_denied(&conflock, &lockt->lt_denied);
2885 out:
2886 nfs4_unlock_state();
2887 return status;
2890 __be32
2891 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2892 struct nfsd4_locku *locku)
2894 struct nfs4_stateid *stp;
2895 struct file *filp = NULL;
2896 struct file_lock file_lock;
2897 __be32 status;
2898 int err;
2900 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2901 (long long) locku->lu_offset,
2902 (long long) locku->lu_length);
2904 if (check_lock_length(locku->lu_offset, locku->lu_length))
2905 return nfserr_inval;
2907 nfs4_lock_state();
2909 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2910 locku->lu_seqid,
2911 &locku->lu_stateid,
2912 CHECK_FH | LOCK_STATE,
2913 &locku->lu_stateowner, &stp, NULL)))
2914 goto out;
2916 filp = stp->st_vfs_file;
2917 BUG_ON(!filp);
2918 locks_init_lock(&file_lock);
2919 file_lock.fl_type = F_UNLCK;
2920 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2921 file_lock.fl_pid = current->tgid;
2922 file_lock.fl_file = filp;
2923 file_lock.fl_flags = FL_POSIX;
2924 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2925 file_lock.fl_start = locku->lu_offset;
2927 if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2928 file_lock.fl_end = ~(u64)0;
2929 else
2930 file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2931 nfs4_transform_lock_offset(&file_lock);
2934 * Try to unlock the file in the VFS.
2936 err = posix_lock_file(filp, &file_lock);
2937 if (err) {
2938 dprintk("NFSD: nfs4_locku: posix_lock_file failed!\n");
2939 goto out_nfserr;
2942 * OK, unlock succeeded; the only thing left to do is update the stateid.
2944 update_stateid(&stp->st_stateid);
2945 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2947 out:
2948 if (locku->lu_stateowner) {
2949 nfs4_get_stateowner(locku->lu_stateowner);
2950 cstate->replay_owner = locku->lu_stateowner;
2952 nfs4_unlock_state();
2953 return status;
2955 out_nfserr:
2956 status = nfserrno(err);
2957 goto out;
2961 * returns
2962 * 1: locks held by lockowner
2963 * 0: no locks held by lockowner
2965 static int
2966 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2968 struct file_lock **flpp;
2969 struct inode *inode = filp->f_path.dentry->d_inode;
2970 int status = 0;
2972 lock_kernel();
2973 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2974 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2975 status = 1;
2976 goto out;
2979 out:
2980 unlock_kernel();
2981 return status;
2984 __be32
2985 nfsd4_release_lockowner(struct svc_rqst *rqstp,
2986 struct nfsd4_compound_state *cstate,
2987 struct nfsd4_release_lockowner *rlockowner)
2989 clientid_t *clid = &rlockowner->rl_clientid;
2990 struct nfs4_stateowner *sop;
2991 struct nfs4_stateid *stp;
2992 struct xdr_netobj *owner = &rlockowner->rl_owner;
2993 struct list_head matches;
2994 int i;
2995 __be32 status;
2997 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2998 clid->cl_boot, clid->cl_id);
3000 /* XXX check for lease expiration */
3002 status = nfserr_stale_clientid;
3003 if (STALE_CLIENTID(clid))
3004 return status;
3006 nfs4_lock_state();
3008 status = nfserr_locks_held;
3009 /* XXX: we're doing a linear search through all the lockowners.
3010 * Yipes! For now we'll just hope clients aren't really using
3011 * release_lockowner much, but eventually we have to fix these
3012 * data structures. */
3013 INIT_LIST_HEAD(&matches);
3014 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3015 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
3016 if (!cmp_owner_str(sop, owner, clid))
3017 continue;
3018 list_for_each_entry(stp, &sop->so_stateids,
3019 st_perstateowner) {
3020 if (check_for_locks(stp->st_vfs_file, sop))
3021 goto out;
3022 /* Note: so_perclient unused for lockowners,
3023 * so it's OK to fool with here. */
3024 list_add(&sop->so_perclient, &matches);
3028 /* Clients probably won't expect us to return with some (but not all)
3029 * of the lockowner state released; so don't release any until all
3030 * have been checked. */
3031 status = nfs_ok;
3032 while (!list_empty(&matches)) {
3033 sop = list_entry(matches.next, struct nfs4_stateowner,
3034 so_perclient);
3035 /* unhash_stateowner deletes so_perclient only
3036 * for openowners. */
3037 list_del(&sop->so_perclient);
3038 release_stateowner(sop);
3040 out:
3041 nfs4_unlock_state();
3042 return status;
3045 static inline struct nfs4_client_reclaim *
3046 alloc_reclaim(void)
3048 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3052 nfs4_has_reclaimed_state(const char *name)
3054 unsigned int strhashval = clientstr_hashval(name);
3055 struct nfs4_client *clp;
3057 clp = find_confirmed_client_by_str(name, strhashval);
3058 return clp ? 1 : 0;
3062 * failure => all reset bets are off, nfserr_no_grace...
3065 nfs4_client_to_reclaim(const char *name)
3067 unsigned int strhashval;
3068 struct nfs4_client_reclaim *crp = NULL;
3070 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3071 crp = alloc_reclaim();
3072 if (!crp)
3073 return 0;
3074 strhashval = clientstr_hashval(name);
3075 INIT_LIST_HEAD(&crp->cr_strhash);
3076 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3077 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3078 reclaim_str_hashtbl_size++;
3079 return 1;
3082 static void
3083 nfs4_release_reclaim(void)
3085 struct nfs4_client_reclaim *crp = NULL;
3086 int i;
3088 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3089 while (!list_empty(&reclaim_str_hashtbl[i])) {
3090 crp = list_entry(reclaim_str_hashtbl[i].next,
3091 struct nfs4_client_reclaim, cr_strhash);
3092 list_del(&crp->cr_strhash);
3093 kfree(crp);
3094 reclaim_str_hashtbl_size--;
3097 BUG_ON(reclaim_str_hashtbl_size);
3101 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3102 static struct nfs4_client_reclaim *
3103 nfs4_find_reclaim_client(clientid_t *clid)
3105 unsigned int strhashval;
3106 struct nfs4_client *clp;
3107 struct nfs4_client_reclaim *crp = NULL;
3110 /* find clientid in conf_id_hashtbl */
3111 clp = find_confirmed_client(clid);
3112 if (clp == NULL)
3113 return NULL;
3115 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3116 clp->cl_name.len, clp->cl_name.data,
3117 clp->cl_recdir);
3119 /* find clp->cl_name in reclaim_str_hashtbl */
3120 strhashval = clientstr_hashval(clp->cl_recdir);
3121 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3122 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3123 return crp;
3126 return NULL;
3130 * Called from OPEN. Look for clientid in reclaim list.
3132 __be32
3133 nfs4_check_open_reclaim(clientid_t *clid)
3135 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3138 /* initialization to perform at module load time: */
3140 void
3141 nfs4_state_init(void)
3143 int i;
3145 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3146 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3147 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3148 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3149 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3151 for (i = 0; i < FILE_HASH_SIZE; i++) {
3152 INIT_LIST_HEAD(&file_hashtbl[i]);
3154 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3155 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3156 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3158 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3159 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3160 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3162 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3163 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3164 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3166 memset(&onestateid, ~0, sizeof(stateid_t));
3167 INIT_LIST_HEAD(&close_lru);
3168 INIT_LIST_HEAD(&client_lru);
3169 INIT_LIST_HEAD(&del_recall_lru);
3170 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3171 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3172 reclaim_str_hashtbl_size = 0;
3175 static void
3176 nfsd4_load_reboot_recovery_data(void)
3178 int status;
3180 nfs4_lock_state();
3181 nfsd4_init_recdir(user_recovery_dirname);
3182 status = nfsd4_recdir_load();
3183 nfs4_unlock_state();
3184 if (status)
3185 printk("NFSD: Failure reading reboot recovery data\n");
3188 /* initialization to perform when the nfsd service is started: */
3190 static void
3191 __nfs4_state_start(void)
3193 time_t grace_time;
3195 boot_time = get_seconds();
3196 grace_time = max(user_lease_time, lease_time);
3197 lease_time = user_lease_time;
3198 in_grace = 1;
3199 printk("NFSD: starting %ld-second grace period\n", grace_time);
3200 laundry_wq = create_singlethread_workqueue("nfsd4");
3201 queue_delayed_work(laundry_wq, &laundromat_work, grace_time*HZ);
3205 nfs4_state_start(void)
3207 int status;
3209 if (nfs4_init)
3210 return 0;
3211 status = nfsd4_init_slabs();
3212 if (status)
3213 return status;
3214 nfsd4_load_reboot_recovery_data();
3215 __nfs4_state_start();
3216 nfs4_init = 1;
3217 return 0;
3221 nfs4_in_grace(void)
3223 return in_grace;
3226 time_t
3227 nfs4_lease_time(void)
3229 return lease_time;
3232 static void
3233 __nfs4_state_shutdown(void)
3235 int i;
3236 struct nfs4_client *clp = NULL;
3237 struct nfs4_delegation *dp = NULL;
3238 struct list_head *pos, *next, reaplist;
3240 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3241 while (!list_empty(&conf_id_hashtbl[i])) {
3242 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3243 expire_client(clp);
3245 while (!list_empty(&unconf_str_hashtbl[i])) {
3246 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3247 expire_client(clp);
3250 INIT_LIST_HEAD(&reaplist);
3251 spin_lock(&recall_lock);
3252 list_for_each_safe(pos, next, &del_recall_lru) {
3253 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3254 list_move(&dp->dl_recall_lru, &reaplist);
3256 spin_unlock(&recall_lock);
3257 list_for_each_safe(pos, next, &reaplist) {
3258 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3259 list_del_init(&dp->dl_recall_lru);
3260 unhash_delegation(dp);
3263 nfsd4_shutdown_recdir();
3264 nfs4_init = 0;
3267 void
3268 nfs4_state_shutdown(void)
3270 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3271 destroy_workqueue(laundry_wq);
3272 nfs4_lock_state();
3273 nfs4_release_reclaim();
3274 __nfs4_state_shutdown();
3275 nfsd4_free_slabs();
3276 nfs4_unlock_state();
3279 static void
3280 nfs4_set_recdir(char *recdir)
3282 nfs4_lock_state();
3283 strcpy(user_recovery_dirname, recdir);
3284 nfs4_unlock_state();
3288 * Change the NFSv4 recovery directory to recdir.
3291 nfs4_reset_recoverydir(char *recdir)
3293 int status;
3294 struct nameidata nd;
3296 status = path_lookup(recdir, LOOKUP_FOLLOW, &nd);
3297 if (status)
3298 return status;
3299 status = -ENOTDIR;
3300 if (S_ISDIR(nd.dentry->d_inode->i_mode)) {
3301 nfs4_set_recdir(recdir);
3302 status = 0;
3304 path_release(&nd);
3305 return status;
3309 * Called when leasetime is changed.
3311 * The only way the protocol gives us to handle on-the-fly lease changes is to
3312 * simulate a reboot. Instead of doing that, we just wait till the next time
3313 * we start to register any changes in lease time. If the administrator
3314 * really wants to change the lease time *now*, they can go ahead and bring
3315 * nfsd down and then back up again after changing the lease time.
3317 void
3318 nfs4_reset_lease(time_t leasetime)
3320 lock_kernel();
3321 user_lease_time = leasetime;
3322 unlock_kernel();