]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/nfsd/nfs4state.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: hard page limit for DRC
[net-next-2.6.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
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.
22*
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.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
1da177e4
LT		 58
59#define NFSDDBG_FACILITY NFSDDBG_PROC
60
61/* Globals */
62static time_t lease_time = 90; /* default lease time */
d99a05ad 63static time_t user_lease_time = 90;
fd39ca9a 64static time_t boot_time;
1da177e4
LT		 65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N		 69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
71
72#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 74
1da177e4 75/* forward declarations */
fd39ca9a 76static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 77static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 78static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
79static void nfs4_set_recdir(char *recdir);
1da177e4 80
8b671b80
BF		 81/* Locking: */
82
83/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 84static DEFINE_MUTEX(client_mutex);
1da177e4 85
8b671b80
BF		 86/*
87 * Currently used for the del_recall_lru and file hash table. In an
88 * effort to decrease the scope of the client_mutex, this spinlock may
89 * eventually cover more:
90 */
91static DEFINE_SPINLOCK(recall_lock);
92
e18b890b
CL		 93static struct kmem_cache *stateowner_slab = NULL;
94static struct kmem_cache *file_slab = NULL;
95static struct kmem_cache *stateid_slab = NULL;
96static struct kmem_cache *deleg_slab = NULL;
e60d4398 97
1da177e4
LT		 98void
99nfs4_lock_state(void)
100{
353ab6e9 101 mutex_lock(&client_mutex);
1da177e4
LT		 102}
103
104void
105nfs4_unlock_state(void)
106{
353ab6e9 107 mutex_unlock(&client_mutex);
1da177e4
LT		 108}
109
110static inline u32
111opaque_hashval(const void *ptr, int nbytes)
112{
113 unsigned char *cptr = (unsigned char *) ptr;
114
115 u32 x = 0;
116 while (nbytes--) {
117 x *= 37;
118 x += *cptr++;
119 }
120 return x;
121}
122
1da177e4
LT		 123static struct list_head del_recall_lru;
124
13cd2184
N		 125static inline void
126put_nfs4_file(struct nfs4_file *fi)
127{
8b671b80
BF		 128 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
129 list_del(&fi->fi_hash);
130 spin_unlock(&recall_lock);
131 iput(fi->fi_inode);
132 kmem_cache_free(file_slab, fi);
133 }
13cd2184
N		 134}
135
136static inline void
137get_nfs4_file(struct nfs4_file *fi)
138{
8b671b80 139 atomic_inc(&fi->fi_ref);
13cd2184
N		 140}
141
ef0f3390 142static int num_delegations;
c2f1a551 143unsigned int max_delegations;
ef0f3390
N		 144
145/*
146 * Open owner state (share locks)
147 */
148
149/* hash tables for nfs4_stateowner */
150#define OWNER_HASH_BITS 8
151#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
152#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
153
154#define ownerid_hashval(id) \
155 ((id) & OWNER_HASH_MASK)
156#define ownerstr_hashval(clientid, ownername) \
157 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
158
159static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
160static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
161
162/* hash table for nfs4_file */
163#define FILE_HASH_BITS 8
164#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
165#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
166/* hash table for (open)nfs4_stateid */
167#define STATEID_HASH_BITS 10
168#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
169#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
170
171#define file_hashval(x) \
172 hash_ptr(x, FILE_HASH_BITS)
173#define stateid_hashval(owner_id, file_id) \
174 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
175
176static struct list_head file_hashtbl[FILE_HASH_SIZE];
177static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
178
1da177e4
LT		 179static struct nfs4_delegation *
180alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
181{
182 struct nfs4_delegation *dp;
183 struct nfs4_file *fp = stp->st_file;
184 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
185
186 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 187 if (fp->fi_had_conflict)
188 return NULL;
c2f1a551 189 if (num_delegations > max_delegations)
ef0f3390 190 return NULL;
5b2d21c1
N		 191 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
192 if (dp == NULL)
1da177e4 193 return dp;
ef0f3390 194 num_delegations++;
ea1da636
N		 195 INIT_LIST_HEAD(&dp->dl_perfile);
196 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 197 INIT_LIST_HEAD(&dp->dl_recall_lru);
198 dp->dl_client = clp;
13cd2184 199 get_nfs4_file(fp);
1da177e4
LT		 200 dp->dl_file = fp;
201 dp->dl_flock = NULL;
202 get_file(stp->st_vfs_file);
203 dp->dl_vfs_file = stp->st_vfs_file;
204 dp->dl_type = type;
205 dp->dl_recall.cbr_dp = NULL;
206 dp->dl_recall.cbr_ident = cb->cb_ident;
207 dp->dl_recall.cbr_trunc = 0;
208 dp->dl_stateid.si_boot = boot_time;
209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
6c02eaa1 212 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 213 dp->dl_time = 0;
214 atomic_set(&dp->dl_count, 1);
ea1da636
N		 215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 217 return dp;
218}
219
220void
221nfs4_put_delegation(struct nfs4_delegation *dp)
222{
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 225 put_nfs4_file(dp->dl_file);
5b2d21c1 226 kmem_cache_free(deleg_slab, dp);
ef0f3390 227 num_delegations--;
1da177e4
LT		 228 }
229}
230
231/* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
235 */
236static void
237nfs4_close_delegation(struct nfs4_delegation *dp)
238{
239 struct file *filp = dp->dl_vfs_file;
240
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
c907132d 245 if (dp->dl_flock)
a9933cea 246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 247 nfsd_close(filp);
1da177e4
LT		 248}
249
250/* Called under the state lock. */
251static void
252unhash_delegation(struct nfs4_delegation *dp)
253{
ea1da636
N		 254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
261}
262
263/*
264 * SETCLIENTID state
265 */
266
267/* Hash tables for nfs4_clientid state */
268#define CLIENT_HASH_BITS 4
269#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271
272#define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 274#define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 276/*
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
279 *
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
282 *
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
284 * setclientid info.
285 *
286 * client_lru holds client queue ordered by nfs4_client.cl_time
287 * for lease renewal.
288 *
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
291 */
292static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293static int reclaim_str_hashtbl_size = 0;
294static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head client_lru;
299static struct list_head close_lru;
300
2283963f
BF		 301static void unhash_generic_stateid(struct nfs4_stateid *stp)
302{
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
306}
307
308static void free_generic_stateid(struct nfs4_stateid *stp)
309{
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
312}
313
314static void release_lock_stateid(struct nfs4_stateid *stp)
315{
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
319}
320
f044ff83
BF		 321static void unhash_lockowner(struct nfs4_stateowner *sop)
322{
323 struct nfs4_stateid *stp;
324
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
332 }
333}
334
335static void release_lockowner(struct nfs4_stateowner *sop)
336{
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
339}
340
f1d110ca
BF		 341static void
342release_stateid_lockowners(struct nfs4_stateid *open_stp)
343{
344 struct nfs4_stateowner *lock_sop;
345
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 351 release_lockowner(lock_sop);
f1d110ca
BF		 352 }
353}
354
2283963f
BF		 355static void release_open_stateid(struct nfs4_stateid *stp)
356{
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
361}
362
f044ff83 363static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 364{
365 struct nfs4_stateid *stp;
366
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
f044ff83
BF		 369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 371 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
f1d110ca
BF		 375 }
376}
377
f044ff83 378static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 379{
f044ff83 380 unhash_openowner(sop);
f1d110ca
BF		 381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
383}
384
5282fd72
ME		 385static DEFINE_SPINLOCK(sessionid_lock);
386#define SESSION_HASH_SIZE 512
387static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
388
389static inline int
390hash_sessionid(struct nfs4_sessionid *sessionid)
391{
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
393
394 return sid->sequence % SESSION_HASH_SIZE;
395}
396
397static inline void
398dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
399{
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
402}
403
404/* caller must hold sessionid_lock */
405static struct nfsd4_session *
406find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
407{
408 struct nfsd4_session *elem;
409 int idx;
410
411 dump_sessionid(__func__, sessionid);
412 idx = hash_sessionid(sessionid);
413 dprintk("%s: idx is %d\n", __func__, idx);
414 /* Search in the appropriate list */
415 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
416 dump_sessionid("list traversal", &elem->se_sessionid);
417 if (!memcmp(elem->se_sessionid.data, sessionid->data,
418 NFS4_MAX_SESSIONID_LEN)) {
419 return elem;
420 }
421 }
422
423 dprintk("%s: session not found\n", __func__);
424 return NULL;
425}
426
427/* caller must hold sessionid_lock */
7116ed6b 428static void
5282fd72 429unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 430{
431 list_del(&ses->se_hash);
432 list_del(&ses->se_perclnt);
5282fd72
ME		 433}
434
435static void
436release_session(struct nfsd4_session *ses)
437{
438 spin_lock(&sessionid_lock);
439 unhash_session(ses);
440 spin_unlock(&sessionid_lock);
7116ed6b
AA		 441 nfsd4_put_session(ses);
442}
443
444void
445free_session(struct kref *kref)
446{
447 struct nfsd4_session *ses;
448
449 ses = container_of(kref, struct nfsd4_session, se_ref);
450 kfree(ses->se_slots);
451 kfree(ses);
452}
453
1da177e4
LT		 454static inline void
455renew_client(struct nfs4_client *clp)
456{
457 /*
458 * Move client to the end to the LRU list.
459 */
460 dprintk("renewing client (clientid %08x/%08x)\n",
461 clp->cl_clientid.cl_boot,
462 clp->cl_clientid.cl_id);
463 list_move_tail(&clp->cl_lru, &client_lru);
464 clp->cl_time = get_seconds();
465}
466
467/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
468static int
469STALE_CLIENTID(clientid_t *clid)
470{
471 if (clid->cl_boot == boot_time)
472 return 0;
473 dprintk("NFSD stale clientid (%08x/%08x)\n",
474 clid->cl_boot, clid->cl_id);
475 return 1;
476}
477
478/*
479 * XXX Should we use a slab cache ?
480 * This type of memory management is somewhat inefficient, but we use it
481 * anyway since SETCLIENTID is not a common operation.
482 */
35bba9a3 483static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 484{
485 struct nfs4_client *clp;
486
35bba9a3
BF		 487 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
488 if (clp == NULL)
489 return NULL;
490 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
491 if (clp->cl_name.data == NULL) {
492 kfree(clp);
493 return NULL;
1da177e4 494 }
35bba9a3
BF		 495 memcpy(clp->cl_name.data, name.data, name.len);
496 clp->cl_name.len = name.len;
1da177e4
LT		 497 return clp;
498}
499
1b1a9b31
BF		 500static void
501shutdown_callback_client(struct nfs4_client *clp)
502{
503 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
504
1b1a9b31 505 if (clnt) {
404ec117
BF		 506 /*
507 * Callback threads take a reference on the client, so there
508 * should be no outstanding callbacks at this point.
509 */
1b1a9b31
BF		 510 clp->cl_callback.cb_client = NULL;
511 rpc_shutdown_client(clnt);
512 }
513}
514
1da177e4
LT		 515static inline void
516free_client(struct nfs4_client *clp)
517{
1b1a9b31 518 shutdown_callback_client(clp);
1da177e4
LT		 519 if (clp->cl_cred.cr_group_info)
520 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 521 kfree(clp->cl_principal);
1da177e4
LT		 522 kfree(clp->cl_name.data);
523 kfree(clp);
524}
525
526void
527put_nfs4_client(struct nfs4_client *clp)
528{
529 if (atomic_dec_and_test(&clp->cl_count))
530 free_client(clp);
531}
532
533static void
534expire_client(struct nfs4_client *clp)
535{
536 struct nfs4_stateowner *sop;
537 struct nfs4_delegation *dp;
1da177e4
LT		 538 struct list_head reaplist;
539
540 dprintk("NFSD: expire_client cl_count %d\n",
541 atomic_read(&clp->cl_count));
542
1da177e4
LT		 543 INIT_LIST_HEAD(&reaplist);
544 spin_lock(&recall_lock);
ea1da636
N		 545 while (!list_empty(&clp->cl_delegations)) {
546 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 547 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
548 dp->dl_flock);
ea1da636 549 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 550 list_move(&dp->dl_recall_lru, &reaplist);
551 }
552 spin_unlock(&recall_lock);
553 while (!list_empty(&reaplist)) {
554 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
555 list_del_init(&dp->dl_recall_lru);
556 unhash_delegation(dp);
557 }
558 list_del(&clp->cl_idhash);
559 list_del(&clp->cl_strhash);
560 list_del(&clp->cl_lru);
ea1da636
N		 561 while (!list_empty(&clp->cl_openowners)) {
562 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 563 release_openowner(sop);
1da177e4 564 }
c4bf7868
ME		 565 while (!list_empty(&clp->cl_sessions)) {
566 struct nfsd4_session *ses;
567 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
568 se_perclnt);
569 release_session(ses);
570 }
1da177e4
LT		 571 put_nfs4_client(clp);
572}
573
35bba9a3
BF		 574static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
575{
1da177e4
LT		 576 struct nfs4_client *clp;
577
35bba9a3
BF		 578 clp = alloc_client(name);
579 if (clp == NULL)
580 return NULL;
a55370a3 581 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT		 582 atomic_set(&clp->cl_count, 1);
583 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT		 584 INIT_LIST_HEAD(&clp->cl_idhash);
585 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 586 INIT_LIST_HEAD(&clp->cl_openowners);
587 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 588 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 589 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 590 return clp;
591}
592
35bba9a3
BF		 593static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
594{
595 memcpy(target->cl_verifier.data, source->data,
596 sizeof(target->cl_verifier.data));
1da177e4
LT		 597}
598
35bba9a3
BF		 599static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
600{
1da177e4
LT		 601 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
602 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
603}
604
35bba9a3
BF		 605static void copy_cred(struct svc_cred *target, struct svc_cred *source)
606{
1da177e4
LT		 607 target->cr_uid = source->cr_uid;
608 target->cr_gid = source->cr_gid;
609 target->cr_group_info = source->cr_group_info;
610 get_group_info(target->cr_group_info);
611}
612
35bba9a3 613static int same_name(const char *n1, const char *n2)
599e0a22 614{
a55370a3 615 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 616}
617
618static int
599e0a22
BF		 619same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
620{
621 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 622}
623
624static int
599e0a22
BF		 625same_clid(clientid_t *cl1, clientid_t *cl2)
626{
627 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 628}
629
630/* XXX what about NGROUP */
631static int
599e0a22
BF		 632same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
633{
634 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 635}
636
5ec7b46c
BF		 637static void gen_clid(struct nfs4_client *clp)
638{
639 static u32 current_clientid = 1;
640
1da177e4
LT		 641 clp->cl_clientid.cl_boot = boot_time;
642 clp->cl_clientid.cl_id = current_clientid++;
643}
644
deda2faa
BF		 645static void gen_confirm(struct nfs4_client *clp)
646{
647 static u32 i;
648 u32 *p;
1da177e4 649
1da177e4 650 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 651 *p++ = get_seconds();
652 *p++ = i++;
1da177e4
LT		 653}
654
35bba9a3
BF		 655static int check_name(struct xdr_netobj name)
656{
1da177e4
LT		 657 if (name.len == 0)
658 return 0;
659 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 660 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 661 return 0;
662 }
663 return 1;
664}
665
fd39ca9a 666static void
1da177e4
LT		 667add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
668{
669 unsigned int idhashval;
670
671 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
672 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
673 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
674 list_add_tail(&clp->cl_lru, &client_lru);
675 clp->cl_time = get_seconds();
676}
677
fd39ca9a 678static void
1da177e4
LT		 679move_to_confirmed(struct nfs4_client *clp)
680{
681 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
682 unsigned int strhashval;
683
684 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
685 list_del_init(&clp->cl_strhash);
f116629d 686 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 687 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 688 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
689 renew_client(clp);
690}
691
692static struct nfs4_client *
693find_confirmed_client(clientid_t *clid)
694{
695 struct nfs4_client *clp;
696 unsigned int idhashval = clientid_hashval(clid->cl_id);
697
698 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 699 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 700 return clp;
701 }
702 return NULL;
703}
704
705static struct nfs4_client *
706find_unconfirmed_client(clientid_t *clid)
707{
708 struct nfs4_client *clp;
709 unsigned int idhashval = clientid_hashval(clid->cl_id);
710
711 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 712 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 713 return clp;
714 }
715 return NULL;
716}
717
a1bcecd2
AA		 718/*
719 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
720 * parameter. Matching is based on the fact the at least one of the
721 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
722 *
723 * FIXME: we need to unify the clientid namespaces for nfsv4.x
724 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
725 * and SET_CLIENTID{,_CONFIRM}
726 */
727static inline int
728match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
729{
730 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
731 return use_exchange_id == has_exchange_flags;
732}
733
28ce6054 734static struct nfs4_client *
a1bcecd2
AA		 735find_confirmed_client_by_str(const char *dname, unsigned int hashval,
736 bool use_exchange_id)
28ce6054
N		 737{
738 struct nfs4_client *clp;
739
740 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 741 if (same_name(clp->cl_recdir, dname) &&
742 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 743 return clp;
744 }
745 return NULL;
746}
747
748static struct nfs4_client *
a1bcecd2
AA		 749find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
750 bool use_exchange_id)
28ce6054
N		 751{
752 struct nfs4_client *clp;
753
754 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 755 if (same_name(clp->cl_recdir, dname) &&
756 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 757 return clp;
758 }
759 return NULL;
760}
761
1da177e4
LT		 762/* a helper function for parse_callback */
763static int
764parse_octet(unsigned int *lenp, char **addrp)
765{
766 unsigned int len = *lenp;
767 char *p = *addrp;
768 int n = -1;
769 char c;
770
771 for (;;) {
772 if (!len)
773 break;
774 len--;
775 c = *p++;
776 if (c == '.')
777 break;
778 if ((c < '0') || (c > '9')) {
779 n = -1;
780 break;
781 }
782 if (n < 0)
783 n = 0;
784 n = (n * 10) + (c - '0');
785 if (n > 255) {
786 n = -1;
787 break;
788 }
789 }
790 *lenp = len;
791 *addrp = p;
792 return n;
793}
794
795/* parse and set the setclientid ipv4 callback address */
fd39ca9a 796static int
1da177e4
LT		 797parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
798{
799 int temp = 0;
800 u32 cbaddr = 0;
801 u16 cbport = 0;
802 u32 addrlen = addr_len;
803 char *addr = addr_val;
804 int i, shift;
805
806 /* ipaddress */
807 shift = 24;
808 for(i = 4; i > 0 ; i--) {
809 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
810 return 0;
811 }
812 cbaddr |= (temp << shift);
813 if (shift > 0)
814 shift -= 8;
815 }
816 *cbaddrp = cbaddr;
817
818 /* port */
819 shift = 8;
820 for(i = 2; i > 0 ; i--) {
821 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
822 return 0;
823 }
824 cbport |= (temp << shift);
825 if (shift > 0)
826 shift -= 8;
827 }
828 *cbportp = cbport;
829 return 1;
830}
831
fd39ca9a 832static void
1da177e4
LT		 833gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
834{
835 struct nfs4_callback *cb = &clp->cl_callback;
836
837 /* Currently, we only support tcp for the callback channel */
838 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
839 goto out_err;
840
841 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
842 &cb->cb_addr, &cb->cb_port)))
843 goto out_err;
844 cb->cb_prog = se->se_callback_prog;
845 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 846 return;
847out_err:
849823c5 848 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 849 "will not receive delegations\n",
850 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
851
1da177e4
LT		 852 return;
853}
854
074fe897
AA		 855void
856nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
857{
858 struct nfsd4_compoundres *resp = rqstp->rq_resp;
859
860 resp->cstate.statp = statp;
861}
862
863/*
864 * Dereference the result pages.
865 */
866static void
867nfsd4_release_respages(struct page **respages, short resused)
868{
869 int i;
870
871 dprintk("--> %s\n", __func__);
872 for (i = 0; i < resused; i++) {
873 if (!respages[i])
874 continue;
875 put_page(respages[i]);
876 respages[i] = NULL;
877 }
878}
879
880static void
881nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
882{
883 int i;
884
885 for (i = 0; i < count; i++) {
886 topages[i] = frompages[i];
887 if (!topages[i])
888 continue;
889 get_page(topages[i]);
890 }
891}
892
893/*
894 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
895 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
896 * length of the XDR response is less than se_fmaxresp_cached
897 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
898 * of the reply (e.g. readdir).
899 *
900 * Store the base and length of the rq_req.head[0] page
901 * of the NFSv4.1 data, just past the rpc header.
902 */
903void
904nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
905{
906 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
907 struct svc_rqst *rqstp = resp->rqstp;
908 struct nfsd4_compoundargs *args = rqstp->rq_argp;
909 struct nfsd4_op *op = &args->ops[resp->opcnt];
910 struct kvec *resv = &rqstp->rq_res.head[0];
911
912 dprintk("--> %s entry %p\n", __func__, entry);
913
914 /* Don't cache a failed OP_SEQUENCE. */
915 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
916 return;
917 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
918 entry->ce_resused = rqstp->rq_resused;
919 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
920 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
921 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
922 entry->ce_resused);
923 entry->ce_status = resp->cstate.status;
924 entry->ce_datav.iov_base = resp->cstate.statp;
925 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
926 (char *)page_address(rqstp->rq_respages[0]));
927 entry->ce_opcnt = resp->opcnt;
928 /* Current request rpc header length*/
929 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
930 (char *)page_address(rqstp->rq_respages[0]);
931}
932
933/*
934 * We keep the rpc header, but take the nfs reply from the replycache.
935 */
936static int
937nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
938 struct nfsd4_cache_entry *entry)
939{
940 struct svc_rqst *rqstp = resp->rqstp;
941 struct kvec *resv = &resp->rqstp->rq_res.head[0];
942 int len;
943
944 /* Current request rpc header length*/
945 len = (char *)resp->cstate.statp -
946 (char *)page_address(rqstp->rq_respages[0]);
947 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
948 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
949 entry->ce_datav.iov_len);
950 return 0;
951 }
952 /* copy the cached reply nfsd data past the current rpc header */
953 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
954 entry->ce_datav.iov_len);
955 resv->iov_len = len + entry->ce_datav.iov_len;
956 return 1;
957}
958
959/*
960 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
961 * cached page. Replace any futher replay pages from the cache.
962 */
963__be32
964nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp)
965{
966 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
967 __be32 status;
968
969 dprintk("--> %s entry %p\n", __func__, entry);
970
971
972 if (!nfsd41_copy_replay_data(resp, entry)) {
973 /*
974 * Not enough room to use the replay rpc header, send the
975 * cached header. Release all the allocated result pages.
976 */
977 svc_free_res_pages(resp->rqstp);
978 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
979 entry->ce_resused);
980 } else {
981 /* Release all but the first allocated result page */
982
983 resp->rqstp->rq_resused--;
984 svc_free_res_pages(resp->rqstp);
985
986 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
987 &entry->ce_respages[1],
988 entry->ce_resused - 1);
989 }
990
991 resp->rqstp->rq_resused = entry->ce_resused;
992 status = entry->ce_status;
993
994 return status;
995}
996
0733d213
AA		 997/*
998 * Set the exchange_id flags returned by the server.
999 */
1000static void
1001nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1002{
1003 /* pNFS is not supported */
1004 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1005
1006 /* Referrals are supported, Migration is not. */
1007 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1008
1009 /* set the wire flags to return to client. */
1010 clid->flags = new->cl_exchange_flags;
1011}
1012
069b6ad4
AA		 1013__be32
1014nfsd4_exchange_id(struct svc_rqst *rqstp,
1015 struct nfsd4_compound_state *cstate,
1016 struct nfsd4_exchange_id *exid)
1017{
0733d213
AA		 1018 struct nfs4_client *unconf, *conf, *new;
1019 int status;
1020 unsigned int strhashval;
1021 char dname[HEXDIR_LEN];
1022 nfs4_verifier verf = exid->verifier;
1023 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1024
1025 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1026 " ip_addr=%u flags %x, spa_how %d\n",
1027 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1028 ip_addr, exid->flags, exid->spa_how);
1029
1030 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1031 return nfserr_inval;
1032
1033 /* Currently only support SP4_NONE */
1034 switch (exid->spa_how) {
1035 case SP4_NONE:
1036 break;
1037 case SP4_SSV:
1038 return nfserr_encr_alg_unsupp;
1039 default:
1040 BUG(); /* checked by xdr code */
1041 case SP4_MACH_CRED:
1042 return nfserr_serverfault; /* no excuse :-/ */
1043 }
1044
1045 status = nfs4_make_rec_clidname(dname, &exid->clname);
1046
1047 if (status)
1048 goto error;
1049
1050 strhashval = clientstr_hashval(dname);
1051
1052 nfs4_lock_state();
1053 status = nfs_ok;
1054
a1bcecd2 1055 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1056 if (conf) {
1057 if (!same_verf(&verf, &conf->cl_verifier)) {
1058 /* 18.35.4 case 8 */
1059 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1060 status = nfserr_not_same;
1061 goto out;
1062 }
1063 /* Client reboot: destroy old state */
1064 expire_client(conf);
1065 goto out_new;
1066 }
1067 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1068 /* 18.35.4 case 9 */
1069 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1070 status = nfserr_perm;
1071 goto out;
1072 }
1073 expire_client(conf);
1074 goto out_new;
1075 }
1076 if (ip_addr != conf->cl_addr &&
1077 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1078 /* Client collision. 18.35.4 case 3 */
1079 status = nfserr_clid_inuse;
1080 goto out;
1081 }
1082 /*
1083 * Set bit when the owner id and verifier map to an already
1084 * confirmed client id (18.35.3).
1085 */
1086 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1087
1088 /*
1089 * Falling into 18.35.4 case 2, possible router replay.
1090 * Leave confirmed record intact and return same result.
1091 */
1092 copy_verf(conf, &verf);
1093 new = conf;
1094 goto out_copy;
1095 } else {
1096 /* 18.35.4 case 7 */
1097 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1098 status = nfserr_noent;
1099 goto out;
1100 }
1101 }
1102
a1bcecd2 1103 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1104 if (unconf) {
1105 /*
1106 * Possible retry or client restart. Per 18.35.4 case 4,
1107 * a new unconfirmed record should be generated regardless
1108 * of whether any properties have changed.
1109 */
1110 expire_client(unconf);
1111 }
1112
1113out_new:
1114 /* Normal case */
1115 new = create_client(exid->clname, dname);
1116 if (new == NULL) {
1117 status = nfserr_resource;
1118 goto out;
1119 }
1120
1121 copy_verf(new, &verf);
1122 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1123 new->cl_addr = ip_addr;
1124 gen_clid(new);
1125 gen_confirm(new);
1126 add_to_unconfirmed(new, strhashval);
1127out_copy:
1128 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1129 exid->clientid.cl_id = new->cl_clientid.cl_id;
1130
1131 new->cl_seqid = exid->seqid = 1;
1132 nfsd4_set_ex_flags(new, exid);
1133
1134 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1135 new->cl_seqid, new->cl_exchange_flags);
1136 status = nfs_ok;
1137
1138out:
1139 nfs4_unlock_state();
1140error:
1141 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1142 return status;
069b6ad4
AA		 1143}
1144
b85d4c01
BH		 1145static int
1146check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1147{
1148 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1149 slot->sl_seqid);
1150
1151 /* The slot is in use, and no response has been sent. */
1152 if (slot->sl_inuse) {
1153 if (seqid == slot->sl_seqid)
1154 return nfserr_jukebox;
1155 else
1156 return nfserr_seq_misordered;
1157 }
1158 /* Normal */
1159 if (likely(seqid == slot->sl_seqid + 1))
1160 return nfs_ok;
1161 /* Replay */
1162 if (seqid == slot->sl_seqid)
1163 return nfserr_replay_cache;
1164 /* Wraparound */
1165 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1166 return nfs_ok;
1167 /* Misordered replay or misordered new request */
1168 return nfserr_seq_misordered;
1169}
1170
069b6ad4
AA		 1171__be32
1172nfsd4_create_session(struct svc_rqst *rqstp,
1173 struct nfsd4_compound_state *cstate,
1174 struct nfsd4_create_session *cr_ses)
1175{
1176 return -1; /* stub */
1177}
1178
1179__be32
1180nfsd4_destroy_session(struct svc_rqst *r,
1181 struct nfsd4_compound_state *cstate,
1182 struct nfsd4_destroy_session *sessionid)
1183{
1184 return -1; /* stub */
1185}
1186
1187__be32
b85d4c01 1188nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1189 struct nfsd4_compound_state *cstate,
1190 struct nfsd4_sequence *seq)
1191{
f9bb94c4 1192 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1193 struct nfsd4_session *session;
1194 struct nfsd4_slot *slot;
1195 int status;
1196
f9bb94c4
AA		 1197 if (resp->opcnt != 1)
1198 return nfserr_sequence_pos;
1199
b85d4c01
BH		 1200 spin_lock(&sessionid_lock);
1201 status = nfserr_badsession;
1202 session = find_in_sessionid_hashtbl(&seq->sessionid);
1203 if (!session)
1204 goto out;
1205
1206 status = nfserr_badslot;
1207 if (seq->slotid >= session->se_fnumslots)
1208 goto out;
1209
1210 slot = &session->se_slots[seq->slotid];
1211 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1212
1213 status = check_slot_seqid(seq->seqid, slot);
1214 if (status == nfserr_replay_cache) {
1215 cstate->slot = slot;
1216 cstate->session = session;
1217 goto replay_cache;
1218 }
1219 if (status)
1220 goto out;
1221
1222 /* Success! bump slot seqid */
1223 slot->sl_inuse = true;
1224 slot->sl_seqid = seq->seqid;
1225
1226 cstate->slot = slot;
1227 cstate->session = session;
1228
1229replay_cache:
1230 /* Renew the clientid on success and on replay.
1231 * Hold a session reference until done processing the compound:
1232 * nfsd4_put_session called only if the cstate slot is set.
1233 */
1234 renew_client(session->se_client);
1235 nfsd4_get_session(session);
1236out:
1237 spin_unlock(&sessionid_lock);
1238 dprintk("%s: return %d\n", __func__, ntohl(status));
1239 return status;
069b6ad4
AA		 1240}
1241
b37ad28b 1242__be32
b591480b
BF		 1243nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1244 struct nfsd4_setclientid *setclid)
1da177e4 1245{
27459f09 1246 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1247 struct xdr_netobj clname = {
1248 .len = setclid->se_namelen,
1249 .data = setclid->se_name,
1250 };
1251 nfs4_verifier clverifier = setclid->se_verf;
1252 unsigned int strhashval;
28ce6054 1253 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1254 __be32 status;
68e76ad0 1255 char *princ;
a55370a3 1256 char dname[HEXDIR_LEN];
1da177e4 1257
1da177e4 1258 if (!check_name(clname))
73aea4ec 1259 return nfserr_inval;
1da177e4 1260
a55370a3
N		 1261 status = nfs4_make_rec_clidname(dname, &clname);
1262 if (status)
73aea4ec 1263 return status;
a55370a3 1264
1da177e4
LT		 1265 /*
1266 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1267 * We get here on a DRC miss.
1268 */
1269
a55370a3 1270 strhashval = clientstr_hashval(dname);
1da177e4 1271
1da177e4 1272 nfs4_lock_state();
a1bcecd2 1273 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1274 if (conf) {
a186e767 1275 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1276 status = nfserr_clid_inuse;
026722c2
BF		 1277 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1278 dprintk("NFSD: setclientid: string in use by client"
1279 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1280 goto out;
1281 }
1da177e4 1282 }
a186e767
BF		 1283 /*
1284 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1285 * has a description of SETCLIENTID request processing consisting
1286 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1287 */
a1bcecd2 1288 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1289 status = nfserr_resource;
1290 if (!conf) {
a186e767
BF		 1291 /*
1292 * RFC 3530 14.2.33 CASE 4:
1293 * placed first, because it is the normal case
1da177e4
LT		 1294 */
1295 if (unconf)
1296 expire_client(unconf);
a55370a3
N		 1297 new = create_client(clname, dname);
1298 if (new == NULL)
1da177e4 1299 goto out;
1da177e4 1300 gen_clid(new);
599e0a22 1301 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1302 /*
a186e767
BF		 1303 * RFC 3530 14.2.33 CASE 1:
1304 * probable callback update
1da177e4 1305 */
31f4a6c1
N		 1306 if (unconf) {
1307 /* Note this is removing unconfirmed {*x***},
1308 * which is stronger than RFC recommended {vxc**}.
1309 * This has the advantage that there is at most
1310 * one {*x***} in either list at any time.
1311 */
1312 expire_client(unconf);
1da177e4 1313 }
a55370a3
N		 1314 new = create_client(clname, dname);
1315 if (new == NULL)
1da177e4 1316 goto out;
1da177e4 1317 copy_clid(new, conf);
1da177e4
LT		 1318 } else if (!unconf) {
1319 /*
a186e767
BF		 1320 * RFC 3530 14.2.33 CASE 2:
1321 * probable client reboot; state will be removed if
1322 * confirmed.
1da177e4 1323 */
a55370a3
N		 1324 new = create_client(clname, dname);
1325 if (new == NULL)
1da177e4 1326 goto out;
1da177e4 1327 gen_clid(new);
49ba8781 1328 } else {
a186e767
BF		 1329 /*
1330 * RFC 3530 14.2.33 CASE 3:
1331 * probable client reboot; state will be removed if
1332 * confirmed.
1da177e4
LT		 1333 */
1334 expire_client(unconf);
a55370a3
N		 1335 new = create_client(clname, dname);
1336 if (new == NULL)
1da177e4 1337 goto out;
1da177e4 1338 gen_clid(new);
1da177e4 1339 }
c175b83c
BF		 1340 copy_verf(new, &clverifier);
1341 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1342 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1343 princ = svc_gss_principal(rqstp);
1344 if (princ) {
1345 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1346 if (new->cl_principal == NULL) {
1347 free_client(new);
1348 goto out;
1349 }
1350 }
c175b83c
BF		 1351 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1352 gen_confirm(new);
1353 gen_callback(new, setclid);
1354 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1355 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1356 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1357 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1358 status = nfs_ok;
1359out:
1360 nfs4_unlock_state();
1361 return status;
1362}
1363
1364
1365/*
a186e767
BF		 1366 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1367 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1368 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1369 */
b37ad28b 1370__be32
b591480b
BF		 1371nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1372 struct nfsd4_compound_state *cstate,
1373 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1374{
27459f09 1375 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1376 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1377 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1378 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1379 __be32 status;
1da177e4
LT		 1380
1381 if (STALE_CLIENTID(clid))
1382 return nfserr_stale_clientid;
1383 /*
1384 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1385 * We get here on a DRC miss.
1386 */
1387
1388 nfs4_lock_state();
21ab45a4
N		 1389
1390 conf = find_confirmed_client(clid);
1391 unconf = find_unconfirmed_client(clid);
1392
1393 status = nfserr_clid_inuse;
27459f09 1394 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1395 goto out;
27459f09 1396 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1397 goto out;
1398
a186e767
BF		 1399 /*
1400 * section 14.2.34 of RFC 3530 has a description of
1401 * SETCLIENTID_CONFIRM request processing consisting
1402 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1403 */
366e0c1d 1404 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1405 /*
1406 * RFC 3530 14.2.34 CASE 1:
1407 * callback update
1408 */
599e0a22 1409 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1410 status = nfserr_clid_inuse;
1411 else {
1a69c179
N		 1412 /* XXX: We just turn off callbacks until we can handle
1413 * change request correctly. */
cb36d634 1414 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 1415 gen_confirm(conf);
46309029 1416 nfsd4_remove_clid_dir(unconf);
1a69c179 1417 expire_client(unconf);
1da177e4 1418 status = nfs_ok;
1a69c179 1419
1da177e4 1420 }
f3aba4e5 1421 } else if (conf && !unconf) {
a186e767
BF		 1422 /*
1423 * RFC 3530 14.2.34 CASE 2:
1424 * probable retransmitted request; play it safe and
1425 * do nothing.
7c79f737 1426 */
599e0a22 1427 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1428 status = nfserr_clid_inuse;
21ab45a4 1429 else
1da177e4 1430 status = nfs_ok;
7c79f737 1431 } else if (!conf && unconf
599e0a22 1432 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1433 /*
1434 * RFC 3530 14.2.34 CASE 3:
1435 * Normal case; new or rebooted client:
7c79f737 1436 */
599e0a22 1437 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1438 status = nfserr_clid_inuse;
1439 } else {
1a69c179
N		 1440 unsigned int hash =
1441 clientstr_hashval(unconf->cl_recdir);
1442 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1443 hash, false);
1a69c179 1444 if (conf) {
c7b9a459 1445 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1446 expire_client(conf);
1447 }
1da177e4 1448 move_to_confirmed(unconf);
21ab45a4 1449 conf = unconf;
46f8a64b 1450 nfsd4_probe_callback(conf);
1a69c179 1451 status = nfs_ok;
1da177e4 1452 }
599e0a22
BF		 1453 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1454 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1455 &confirm)))) {
a186e767
BF		 1456 /*
1457 * RFC 3530 14.2.34 CASE 4:
1458 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1459 */
1da177e4 1460 status = nfserr_stale_clientid;
7c79f737 1461 } else {
08e8987c
N		 1462 /* check that we have hit one of the cases...*/
1463 status = nfserr_clid_inuse;
1464 }
1da177e4 1465out:
1da177e4
LT		 1466 nfs4_unlock_state();
1467 return status;
1468}
1469
1da177e4
LT		 1470/* OPEN Share state helper functions */
1471static inline struct nfs4_file *
1472alloc_init_file(struct inode *ino)
1473{
1474 struct nfs4_file *fp;
1475 unsigned int hashval = file_hashval(ino);
1476
e60d4398
N		 1477 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1478 if (fp) {
8b671b80 1479 atomic_set(&fp->fi_ref, 1);
1da177e4 1480 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1481 INIT_LIST_HEAD(&fp->fi_stateids);
1482 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1483 spin_lock(&recall_lock);
1da177e4 1484 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1485 spin_unlock(&recall_lock);
1da177e4
LT		 1486 fp->fi_inode = igrab(ino);
1487 fp->fi_id = current_fileid++;
47f9940c 1488 fp->fi_had_conflict = false;
1da177e4
LT		 1489 return fp;
1490 }
1491 return NULL;
1492}
1493
e60d4398 1494static void
e18b890b 1495nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1496{
e60d4398
N		 1497 if (*slab == NULL)
1498 return;
1a1d92c1 1499 kmem_cache_destroy(*slab);
e60d4398 1500 *slab = NULL;
1da177e4
LT		 1501}
1502
e8ff2a84 1503void
1da177e4
LT		 1504nfsd4_free_slabs(void)
1505{
e60d4398
N		 1506 nfsd4_free_slab(&stateowner_slab);
1507 nfsd4_free_slab(&file_slab);
5ac049ac 1508 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1509 nfsd4_free_slab(&deleg_slab);
e60d4398 1510}
1da177e4 1511
e60d4398
N		 1512static int
1513nfsd4_init_slabs(void)
1514{
1515 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1516 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1517 if (stateowner_slab == NULL)
1518 goto out_nomem;
1519 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1520 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1521 if (file_slab == NULL)
1522 goto out_nomem;
5ac049ac 1523 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1524 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1525 if (stateid_slab == NULL)
1526 goto out_nomem;
5b2d21c1 1527 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1528 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1529 if (deleg_slab == NULL)
1530 goto out_nomem;
e60d4398
N		 1531 return 0;
1532out_nomem:
1533 nfsd4_free_slabs();
1534 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1535 return -ENOMEM;
1da177e4
LT		 1536}
1537
1538void
1539nfs4_free_stateowner(struct kref *kref)
1540{
1541 struct nfs4_stateowner *sop =
1542 container_of(kref, struct nfs4_stateowner, so_ref);
1543 kfree(sop->so_owner.data);
1544 kmem_cache_free(stateowner_slab, sop);
1545}
1546
1547static inline struct nfs4_stateowner *
1548alloc_stateowner(struct xdr_netobj *owner)
1549{
1550 struct nfs4_stateowner *sop;
1551
1552 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1553 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1554 memcpy(sop->so_owner.data, owner->data, owner->len);
1555 sop->so_owner.len = owner->len;
1556 kref_init(&sop->so_ref);
1557 return sop;
1558 }
1559 kmem_cache_free(stateowner_slab, sop);
1560 }
1561 return NULL;
1562}
1563
1564static struct nfs4_stateowner *
1565alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1566 struct nfs4_stateowner *sop;
1567 struct nfs4_replay *rp;
1568 unsigned int idhashval;
1569
1570 if (!(sop = alloc_stateowner(&open->op_owner)))
1571 return NULL;
1572 idhashval = ownerid_hashval(current_ownerid);
1573 INIT_LIST_HEAD(&sop->so_idhash);
1574 INIT_LIST_HEAD(&sop->so_strhash);
1575 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1576 INIT_LIST_HEAD(&sop->so_stateids);
1577 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1578 INIT_LIST_HEAD(&sop->so_close_lru);
1579 sop->so_time = 0;
1580 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1581 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1582 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1583 sop->so_is_open_owner = 1;
1584 sop->so_id = current_ownerid++;
1585 sop->so_client = clp;
1586 sop->so_seqid = open->op_seqid;
1587 sop->so_confirmed = 0;
1588 rp = &sop->so_replay;
de1ae286 1589 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1590 rp->rp_buflen = 0;
1591 rp->rp_buf = rp->rp_ibuf;
1592 return sop;
1593}
1594
1da177e4
LT		 1595static inline void
1596init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1597 struct nfs4_stateowner *sop = open->op_stateowner;
1598 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1599
1600 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1601 INIT_LIST_HEAD(&stp->st_perstateowner);
1602 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1603 INIT_LIST_HEAD(&stp->st_perfile);
1604 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1605 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1606 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1607 stp->st_stateowner = sop;
13cd2184 1608 get_nfs4_file(fp);
1da177e4
LT		 1609 stp->st_file = fp;
1610 stp->st_stateid.si_boot = boot_time;
1611 stp->st_stateid.si_stateownerid = sop->so_id;
1612 stp->st_stateid.si_fileid = fp->fi_id;
1613 stp->st_stateid.si_generation = 0;
1614 stp->st_access_bmap = 0;
1615 stp->st_deny_bmap = 0;
1616 __set_bit(open->op_share_access, &stp->st_access_bmap);
1617 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1618 stp->st_openstp = NULL;
1da177e4
LT		 1619}
1620
fd39ca9a 1621static void
1da177e4
LT		 1622move_to_close_lru(struct nfs4_stateowner *sop)
1623{
1624 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1625
358dd55a 1626 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1627 sop->so_time = get_seconds();
1628}
1629
1da177e4 1630static int
599e0a22
BF		 1631same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1632 clientid_t *clid)
1633{
1634 return (sop->so_owner.len == owner->len) &&
1635 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1636 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1637}
1638
1639static struct nfs4_stateowner *
1640find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1641{
1642 struct nfs4_stateowner *so = NULL;
1643
1644 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1645 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1646 return so;
1647 }
1648 return NULL;
1649}
1650
1651/* search file_hashtbl[] for file */
1652static struct nfs4_file *
1653find_file(struct inode *ino)
1654{
1655 unsigned int hashval = file_hashval(ino);
1656 struct nfs4_file *fp;
1657
8b671b80 1658 spin_lock(&recall_lock);
1da177e4 1659 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1660 if (fp->fi_inode == ino) {
1661 get_nfs4_file(fp);
8b671b80 1662 spin_unlock(&recall_lock);
1da177e4 1663 return fp;
13cd2184 1664 }
1da177e4 1665 }
8b671b80 1666 spin_unlock(&recall_lock);
1da177e4
LT		 1667 return NULL;
1668}
1669
8838dc43 1670static inline int access_valid(u32 x)
ba5a6a19 1671{
8838dc43
BF		 1672 if (x < NFS4_SHARE_ACCESS_READ)
1673 return 0;
1674 if (x > NFS4_SHARE_ACCESS_BOTH)
1675 return 0;
1676 return 1;
ba5a6a19
BF		 1677}
1678
8838dc43 1679static inline int deny_valid(u32 x)
ba5a6a19 1680{
8838dc43
BF		 1681 /* Note: unlike access bits, deny bits may be zero. */
1682 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1683}
1da177e4 1684
4f83aa30
BF		 1685/*
1686 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1687 * st_{access,deny}_bmap field of the stateid, in order to track not
1688 * only what share bits are currently in force, but also what
1689 * combinations of share bits previous opens have used. This allows us
1690 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1691 * return an error if the client attempt to downgrade to a combination
1692 * of share bits not explicable by closing some of its previous opens.
1693 *
1694 * XXX: This enforcement is actually incomplete, since we don't keep
1695 * track of access/deny bit combinations; so, e.g., we allow:
1696 *
1697 * OPEN allow read, deny write
1698 * OPEN allow both, deny none
1699 * DOWNGRADE allow read, deny none
1700 *
1701 * which we should reject.
1702 */
fd39ca9a 1703static void
1da177e4
LT		 1704set_access(unsigned int *access, unsigned long bmap) {
1705 int i;
1706
1707 *access = 0;
1708 for (i = 1; i < 4; i++) {
1709 if (test_bit(i, &bmap))
1710 *access |= i;
1711 }
1712}
1713
fd39ca9a 1714static void
1da177e4
LT		 1715set_deny(unsigned int *deny, unsigned long bmap) {
1716 int i;
1717
1718 *deny = 0;
1719 for (i = 0; i < 4; i++) {
1720 if (test_bit(i, &bmap))
1721 *deny |= i ;
1722 }
1723}
1724
1725static int
1726test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1727 unsigned int access, deny;
1728
1729 set_access(&access, stp->st_access_bmap);
1730 set_deny(&deny, stp->st_deny_bmap);
1731 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1732 return 0;
1733 return 1;
1734}
1735
1736/*
1737 * Called to check deny when READ with all zero stateid or
1738 * WRITE with all zero or all one stateid
1739 */
b37ad28b 1740static __be32
1da177e4
LT		 1741nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1742{
1743 struct inode *ino = current_fh->fh_dentry->d_inode;
1744 struct nfs4_file *fp;
1745 struct nfs4_stateid *stp;
b37ad28b 1746 __be32 ret;
1da177e4
LT		 1747
1748 dprintk("NFSD: nfs4_share_conflict\n");
1749
1750 fp = find_file(ino);
13cd2184
N		 1751 if (!fp)
1752 return nfs_ok;
b700949b 1753 ret = nfserr_locked;
1da177e4 1754 /* Search for conflicting share reservations */
13cd2184
N		 1755 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1756 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1757 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1758 goto out;
1da177e4 1759 }
13cd2184
N		 1760 ret = nfs_ok;
1761out:
1762 put_nfs4_file(fp);
1763 return ret;
1da177e4
LT		 1764}
1765
1766static inline void
1767nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1768{
1769 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 1770 drop_file_write_access(filp);
1da177e4
LT		 1771 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1772 }
1773}
1774
1775/*
1776 * Recall a delegation
1777 */
1778static int
1779do_recall(void *__dp)
1780{
1781 struct nfs4_delegation *dp = __dp;
1782
47f9940c 1783 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 1784 nfsd4_cb_recall(dp);
1785 return 0;
1786}
1787
1788/*
1789 * Spawn a thread to perform a recall on the delegation represented
1790 * by the lease (file_lock)
1791 *
1792 * Called from break_lease() with lock_kernel() held.
1793 * Note: we assume break_lease will only call this *once* for any given
1794 * lease.
1795 */
1796static
1797void nfsd_break_deleg_cb(struct file_lock *fl)
1798{
1799 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1800 struct task_struct *t;
1801
1802 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1803 if (!dp)
1804 return;
1805
1806 /* We're assuming the state code never drops its reference
1807 * without first removing the lease. Since we're in this lease
1808 * callback (and since the lease code is serialized by the kernel
1809 * lock) we know the server hasn't removed the lease yet, we know
1810 * it's safe to take a reference: */
1811 atomic_inc(&dp->dl_count);
cfdcad4d 1812 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 1813
1814 spin_lock(&recall_lock);
1815 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1816 spin_unlock(&recall_lock);
1817
1818 /* only place dl_time is set. protected by lock_kernel*/
1819 dp->dl_time = get_seconds();
1820
0272e1fd
BF		 1821 /*
1822 * We don't want the locks code to timeout the lease for us;
1823 * we'll remove it ourself if the delegation isn't returned
1824 * in time.
1825 */
1826 fl->fl_break_time = 0;
1da177e4
LT		 1827
1828 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1829 if (IS_ERR(t)) {
1830 struct nfs4_client *clp = dp->dl_client;
1831
1832 printk(KERN_INFO "NFSD: Callback thread failed for "
1833 "for client (clientid %08x/%08x)\n",
1834 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 1835 put_nfs4_client(dp->dl_client);
1da177e4
LT		 1836 nfs4_put_delegation(dp);
1837 }
1838}
1839
1840/*
1841 * The file_lock is being reapd.
1842 *
1843 * Called by locks_free_lock() with lock_kernel() held.
1844 */
1845static
1846void nfsd_release_deleg_cb(struct file_lock *fl)
1847{
1848 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1849
1850 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1851
1852 if (!(fl->fl_flags & FL_LEASE) || !dp)
1853 return;
1854 dp->dl_flock = NULL;
1855}
1856
1857/*
1858 * Set the delegation file_lock back pointer.
1859 *
a9933cea 1860 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 1861 */
1862static
1863void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1864{
1865 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1866
1867 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1868 if (!dp)
1869 return;
1870 dp->dl_flock = new;
1871}
1872
1873/*
a9933cea 1874 * Called from setlease() with lock_kernel() held
1da177e4
LT		 1875 */
1876static
1877int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1878{
1879 struct nfs4_delegation *onlistd =
1880 (struct nfs4_delegation *)onlist->fl_owner;
1881 struct nfs4_delegation *tryd =
1882 (struct nfs4_delegation *)try->fl_owner;
1883
1884 if (onlist->fl_lmops != try->fl_lmops)
1885 return 0;
1886
1887 return onlistd->dl_client == tryd->dl_client;
1888}
1889
1890
1891static
1892int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1893{
1894 if (arg & F_UNLCK)
1895 return lease_modify(onlist, arg);
1896 else
1897 return -EAGAIN;
1898}
1899
fd39ca9a 1900static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 1901 .fl_break = nfsd_break_deleg_cb,
1902 .fl_release_private = nfsd_release_deleg_cb,
1903 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1904 .fl_mylease = nfsd_same_client_deleg_cb,
1905 .fl_change = nfsd_change_deleg_cb,
1906};
1907
1908
b37ad28b 1909__be32
1da177e4
LT		 1910nfsd4_process_open1(struct nfsd4_open *open)
1911{
1da177e4
LT		 1912 clientid_t *clientid = &open->op_clientid;
1913 struct nfs4_client *clp = NULL;
1914 unsigned int strhashval;
1915 struct nfs4_stateowner *sop = NULL;
1916
1da177e4 1917 if (!check_name(open->op_owner))
0f442aa2 1918 return nfserr_inval;
1da177e4
LT		 1919
1920 if (STALE_CLIENTID(&open->op_clientid))
1921 return nfserr_stale_clientid;
1922
1923 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1924 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 1925 open->op_stateowner = sop;
1926 if (!sop) {
1927 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 1928 clp = find_confirmed_client(clientid);
1929 if (clp == NULL)
0f442aa2
BF		 1930 return nfserr_expired;
1931 goto renew;
1da177e4 1932 }
0f442aa2
BF		 1933 if (!sop->so_confirmed) {
1934 /* Replace unconfirmed owners without checking for replay. */
1935 clp = sop->so_client;
f044ff83 1936 release_openowner(sop);
0f442aa2
BF		 1937 open->op_stateowner = NULL;
1938 goto renew;
1939 }
1940 if (open->op_seqid == sop->so_seqid - 1) {
1941 if (sop->so_replay.rp_buflen)
a90b061c 1942 return nfserr_replay_me;
0f442aa2
BF		 1943 /* The original OPEN failed so spectacularly
1944 * that we don't even have replay data saved!
1945 * Therefore, we have no choice but to continue
1946 * processing this OPEN; presumably, we'll
1947 * fail again for the same reason.
1948 */
1949 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1950 goto renew;
1951 }
1952 if (open->op_seqid != sop->so_seqid)
1953 return nfserr_bad_seqid;
1da177e4 1954renew:
0f442aa2
BF		 1955 if (open->op_stateowner == NULL) {
1956 sop = alloc_init_open_stateowner(strhashval, clp, open);
1957 if (sop == NULL)
1958 return nfserr_resource;
1959 open->op_stateowner = sop;
1960 }
1961 list_del_init(&sop->so_close_lru);
1da177e4 1962 renew_client(sop->so_client);
0f442aa2 1963 return nfs_ok;
1da177e4
LT		 1964}
1965
b37ad28b 1966static inline __be32
4a6e43e6
N		 1967nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1968{
1969 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1970 return nfserr_openmode;
1971 else
1972 return nfs_ok;
1973}
1974
52f4fb43
N		 1975static struct nfs4_delegation *
1976find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1977{
1978 struct nfs4_delegation *dp;
1979
ea1da636 1980 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 1981 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1982 return dp;
1983 }
1984 return NULL;
1985}
1986
b37ad28b 1987static __be32
567d9829
N		 1988nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1989 struct nfs4_delegation **dp)
1990{
1991 int flags;
b37ad28b 1992 __be32 status = nfserr_bad_stateid;
567d9829
N		 1993
1994 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1995 if (*dp == NULL)
c44c5eeb 1996 goto out;
567d9829
N		 1997 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1998 RD_STATE : WR_STATE;
1999 status = nfs4_check_delegmode(*dp, flags);
2000 if (status)
2001 *dp = NULL;
c44c5eeb
N		 2002out:
2003 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2004 return nfs_ok;
2005 if (status)
2006 return status;
2007 open->op_stateowner->so_confirmed = 1;
2008 return nfs_ok;
567d9829
N		 2009}
2010
b37ad28b 2011static __be32
1da177e4
LT		 2012nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2013{
2014 struct nfs4_stateid *local;
b37ad28b 2015 __be32 status = nfserr_share_denied;
1da177e4
LT		 2016 struct nfs4_stateowner *sop = open->op_stateowner;
2017
8beefa24 2018 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2019 /* ignore lock owners */
2020 if (local->st_stateowner->so_is_open_owner == 0)
2021 continue;
2022 /* remember if we have seen this open owner */
2023 if (local->st_stateowner == sop)
2024 *stpp = local;
2025 /* check for conflicting share reservations */
2026 if (!test_share(local, open))
2027 goto out;
2028 }
2029 status = 0;
2030out:
2031 return status;
2032}
2033
5ac049ac
N		 2034static inline struct nfs4_stateid *
2035nfs4_alloc_stateid(void)
2036{
2037 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2038}
2039
b37ad28b 2040static __be32
1da177e4 2041nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2042 struct nfs4_delegation *dp,
1da177e4
LT		 2043 struct svc_fh *cur_fh, int flags)
2044{
2045 struct nfs4_stateid *stp;
1da177e4 2046
5ac049ac 2047 stp = nfs4_alloc_stateid();
1da177e4
LT		 2048 if (stp == NULL)
2049 return nfserr_resource;
2050
567d9829
N		 2051 if (dp) {
2052 get_file(dp->dl_vfs_file);
2053 stp->st_vfs_file = dp->dl_vfs_file;
2054 } else {
b37ad28b 2055 __be32 status;
567d9829
N		 2056 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2057 &stp->st_vfs_file);
2058 if (status) {
2059 if (status == nfserr_dropit)
2060 status = nfserr_jukebox;
5ac049ac 2061 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2062 return status;
2063 }
1da177e4 2064 }
1da177e4
LT		 2065 *stpp = stp;
2066 return 0;
2067}
2068
b37ad28b 2069static inline __be32
1da177e4
LT		 2070nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2071 struct nfsd4_open *open)
2072{
2073 struct iattr iattr = {
2074 .ia_valid = ATTR_SIZE,
2075 .ia_size = 0,
2076 };
2077 if (!open->op_truncate)
2078 return 0;
2079 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2080 return nfserr_inval;
1da177e4
LT		 2081 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2082}
2083
b37ad28b 2084static __be32
1da177e4
LT		 2085nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2086{
2087 struct file *filp = stp->st_vfs_file;
7eaa36e2 2088 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2089 unsigned int share_access, new_writer;
b37ad28b 2090 __be32 status;
1da177e4
LT		 2091
2092 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2093 new_writer = (~share_access) & open->op_share_access
2094 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2095
6c26d08f 2096 if (new_writer) {
b37ad28b
AV		 2097 int err = get_write_access(inode);
2098 if (err)
2099 return nfserrno(err);
e518f056
BH		 2100 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2101 if (err)
2102 return nfserrno(err);
2103 file_take_write(filp);
6c26d08f 2104 }
1da177e4
LT		 2105 status = nfsd4_truncate(rqstp, cur_fh, open);
2106 if (status) {
6c26d08f
BF		 2107 if (new_writer)
2108 put_write_access(inode);
1da177e4
LT		 2109 return status;
2110 }
2111 /* remember the open */
6c26d08f 2112 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2113 __set_bit(open->op_share_access, &stp->st_access_bmap);
2114 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2115
2116 return nfs_ok;
2117}
2118
2119
1da177e4 2120static void
37515177 2121nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2122{
37515177
N		 2123 open->op_stateowner->so_confirmed = 1;
2124 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2125}
2126
2127/*
2128 * Attempt to hand out a delegation.
2129 */
2130static void
2131nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2132{
2133 struct nfs4_delegation *dp;
2134 struct nfs4_stateowner *sop = stp->st_stateowner;
2135 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2136 struct file_lock fl, *flp = &fl;
2137 int status, flag = 0;
2138
2139 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2140 open->op_recall = 0;
2141 switch (open->op_claim_type) {
2142 case NFS4_OPEN_CLAIM_PREVIOUS:
2143 if (!atomic_read(&cb->cb_set))
2144 open->op_recall = 1;
2145 flag = open->op_delegate_type;
2146 if (flag == NFS4_OPEN_DELEGATE_NONE)
2147 goto out;
2148 break;
2149 case NFS4_OPEN_CLAIM_NULL:
2150 /* Let's not give out any delegations till everyone's
2151 * had the chance to reclaim theirs.... */
af558e33 2152 if (locks_in_grace())
7b190fec
N		 2153 goto out;
2154 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2155 goto out;
2156 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2157 flag = NFS4_OPEN_DELEGATE_WRITE;
2158 else
2159 flag = NFS4_OPEN_DELEGATE_READ;
2160 break;
2161 default:
2162 goto out;
2163 }
1da177e4
LT		 2164
2165 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2166 if (dp == NULL) {
2167 flag = NFS4_OPEN_DELEGATE_NONE;
2168 goto out;
2169 }
2170 locks_init_lock(&fl);
2171 fl.fl_lmops = &nfsd_lease_mng_ops;
2172 fl.fl_flags = FL_LEASE;
9167f501 2173 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2174 fl.fl_end = OFFSET_MAX;
2175 fl.fl_owner = (fl_owner_t)dp;
2176 fl.fl_file = stp->st_vfs_file;
2177 fl.fl_pid = current->tgid;
2178
a9933cea 2179 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2180 * the lock_manager callbacks fl_mylease and fl_change are used
2181 */
9167f501 2182 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2183 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2184 unhash_delegation(dp);
1da177e4
LT		 2185 flag = NFS4_OPEN_DELEGATE_NONE;
2186 goto out;
2187 }
2188
2189 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2190
2191 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2192 dp->dl_stateid.si_boot,
2193 dp->dl_stateid.si_stateownerid,
2194 dp->dl_stateid.si_fileid,
2195 dp->dl_stateid.si_generation);
2196out:
7b190fec
N		 2197 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2198 && flag == NFS4_OPEN_DELEGATE_NONE
2199 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2200 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2201 open->op_delegate_type = flag;
2202}
2203
2204/*
2205 * called with nfs4_lock_state() held.
2206 */
b37ad28b 2207__be32
1da177e4
LT		 2208nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2209{
2210 struct nfs4_file *fp = NULL;
2211 struct inode *ino = current_fh->fh_dentry->d_inode;
2212 struct nfs4_stateid *stp = NULL;
567d9829 2213 struct nfs4_delegation *dp = NULL;
b37ad28b 2214 __be32 status;
1da177e4
LT		 2215
2216 status = nfserr_inval;
ba5a6a19
BF		 2217 if (!access_valid(open->op_share_access)
2218 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2219 goto out;
2220 /*
2221 * Lookup file; if found, lookup stateid and check open request,
2222 * and check for delegations in the process of being recalled.
2223 * If not found, create the nfs4_file struct
2224 */
2225 fp = find_file(ino);
2226 if (fp) {
2227 if ((status = nfs4_check_open(fp, open, &stp)))
2228 goto out;
c44c5eeb
N		 2229 status = nfs4_check_deleg(fp, open, &dp);
2230 if (status)
2231 goto out;
1da177e4 2232 } else {
c44c5eeb
N		 2233 status = nfserr_bad_stateid;
2234 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2235 goto out;
1da177e4
LT		 2236 status = nfserr_resource;
2237 fp = alloc_init_file(ino);
2238 if (fp == NULL)
2239 goto out;
2240 }
2241
2242 /*
2243 * OPEN the file, or upgrade an existing OPEN.
2244 * If truncate fails, the OPEN fails.
2245 */
2246 if (stp) {
2247 /* Stateid was found, this is an OPEN upgrade */
2248 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2249 if (status)
2250 goto out;
444c2c07 2251 update_stateid(&stp->st_stateid);
1da177e4
LT		 2252 } else {
2253 /* Stateid was not found, this is a new OPEN */
2254 int flags = 0;
9ecb6a08 2255 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2256 flags |= NFSD_MAY_READ;
1da177e4 2257 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2258 flags |= NFSD_MAY_WRITE;
567d9829
N		 2259 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2260 if (status)
1da177e4
LT		 2261 goto out;
2262 init_stateid(stp, fp, open);
2263 status = nfsd4_truncate(rqstp, current_fh, open);
2264 if (status) {
2283963f 2265 release_open_stateid(stp);
1da177e4
LT		 2266 goto out;
2267 }
2268 }
2269 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2270
2271 /*
2272 * Attempt to hand out a delegation. No error return, because the
2273 * OPEN succeeds even if we fail.
2274 */
2275 nfs4_open_delegation(current_fh, open, stp);
2276
2277 status = nfs_ok;
2278
2279 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2280 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2281 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2282out:
13cd2184
N		 2283 if (fp)
2284 put_nfs4_file(fp);
37515177
N		 2285 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2286 nfs4_set_claim_prev(open);
1da177e4
LT		 2287 /*
2288 * To finish the open response, we just need to set the rflags.
2289 */
2290 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2291 if (!open->op_stateowner->so_confirmed)
2292 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2293
2294 return status;
2295}
2296
b37ad28b 2297__be32
b591480b
BF		 2298nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2299 clientid_t *clid)
1da177e4
LT		 2300{
2301 struct nfs4_client *clp;
b37ad28b 2302 __be32 status;
1da177e4
LT		 2303
2304 nfs4_lock_state();
2305 dprintk("process_renew(%08x/%08x): starting\n",
2306 clid->cl_boot, clid->cl_id);
2307 status = nfserr_stale_clientid;
2308 if (STALE_CLIENTID(clid))
2309 goto out;
2310 clp = find_confirmed_client(clid);
2311 status = nfserr_expired;
2312 if (clp == NULL) {
2313 /* We assume the client took too long to RENEW. */
2314 dprintk("nfsd4_renew: clientid not found!\n");
2315 goto out;
2316 }
2317 renew_client(clp);
2318 status = nfserr_cb_path_down;
ea1da636 2319 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2320 && !atomic_read(&clp->cl_callback.cb_set))
2321 goto out;
2322 status = nfs_ok;
2323out:
2324 nfs4_unlock_state();
2325 return status;
2326}
2327
af558e33
BF		 2328struct lock_manager nfsd4_manager = {
2329};
2330
a76b4319 2331static void
af558e33 2332nfsd4_end_grace(void)
a76b4319
N		 2333{
2334 dprintk("NFSD: end of grace period\n");
c7b9a459 2335 nfsd4_recdir_purge_old();
af558e33 2336 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2337}
2338
fd39ca9a 2339static time_t
1da177e4
LT		 2340nfs4_laundromat(void)
2341{
2342 struct nfs4_client *clp;
2343 struct nfs4_stateowner *sop;
2344 struct nfs4_delegation *dp;
2345 struct list_head *pos, *next, reaplist;
2346 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2347 time_t t, clientid_val = NFSD_LEASE_TIME;
2348 time_t u, test_val = NFSD_LEASE_TIME;
2349
2350 nfs4_lock_state();
2351
2352 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2353 if (locks_in_grace())
2354 nfsd4_end_grace();
1da177e4
LT		 2355 list_for_each_safe(pos, next, &client_lru) {
2356 clp = list_entry(pos, struct nfs4_client, cl_lru);
2357 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2358 t = clp->cl_time - cutoff;
2359 if (clientid_val > t)
2360 clientid_val = t;
2361 break;
2362 }
2363 dprintk("NFSD: purging unused client (clientid %08x)\n",
2364 clp->cl_clientid.cl_id);
c7b9a459 2365 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2366 expire_client(clp);
2367 }
2368 INIT_LIST_HEAD(&reaplist);
2369 spin_lock(&recall_lock);
2370 list_for_each_safe(pos, next, &del_recall_lru) {
2371 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2372 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2373 u = dp->dl_time - cutoff;
2374 if (test_val > u)
2375 test_val = u;
2376 break;
2377 }
2378 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2379 dp, dp->dl_flock);
2380 list_move(&dp->dl_recall_lru, &reaplist);
2381 }
2382 spin_unlock(&recall_lock);
2383 list_for_each_safe(pos, next, &reaplist) {
2384 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2385 list_del_init(&dp->dl_recall_lru);
2386 unhash_delegation(dp);
2387 }
2388 test_val = NFSD_LEASE_TIME;
2389 list_for_each_safe(pos, next, &close_lru) {
2390 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2391 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2392 u = sop->so_time - cutoff;
2393 if (test_val > u)
2394 test_val = u;
2395 break;
2396 }
2397 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2398 sop->so_id);
f044ff83 2399 release_openowner(sop);
1da177e4
LT		 2400 }
2401 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2402 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2403 nfs4_unlock_state();
2404 return clientid_val;
2405}
2406
a254b246
HH		 2407static struct workqueue_struct *laundry_wq;
2408static void laundromat_main(struct work_struct *);
2409static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2410
2411static void
c4028958 2412laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2413{
2414 time_t t;
2415
2416 t = nfs4_laundromat();
2417 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2418 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2419}
2420
fd39ca9a 2421static struct nfs4_stateowner *
f8816512
N		 2422search_close_lru(u32 st_id, int flags)
2423{
1da177e4
LT		 2424 struct nfs4_stateowner *local = NULL;
2425
1da177e4
LT		 2426 if (flags & CLOSE_STATE) {
2427 list_for_each_entry(local, &close_lru, so_close_lru) {
2428 if (local->so_id == st_id)
2429 return local;
2430 }
2431 }
2432 return NULL;
2433}
2434
2435static inline int
2436nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2437{
7eaa36e2 2438 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2439}
2440
2441static int
2442STALE_STATEID(stateid_t *stateid)
2443{
2444 if (stateid->si_boot == boot_time)
2445 return 0;
849823c5 2446 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT		 2447 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2448 stateid->si_generation);
2449 return 1;
2450}
2451
2452static inline int
2453access_permit_read(unsigned long access_bmap)
2454{
2455 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2456 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2457 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2458}
2459
2460static inline int
2461access_permit_write(unsigned long access_bmap)
2462{
2463 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2464 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2465}
2466
2467static
b37ad28b 2468__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2469{
b37ad28b 2470 __be32 status = nfserr_openmode;
1da177e4
LT		 2471
2472 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2473 goto out;
2474 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2475 goto out;
2476 status = nfs_ok;
2477out:
2478 return status;
2479}
2480
b37ad28b 2481static inline __be32
1da177e4
LT		 2482check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2483{
203a8c8e 2484 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2485 return nfs_ok;
af558e33 2486 else if (locks_in_grace()) {
1da177e4
LT		 2487 /* Answer in remaining cases depends on existance of
2488 * conflicting state; so we must wait out the grace period. */
2489 return nfserr_grace;
2490 } else if (flags & WR_STATE)
2491 return nfs4_share_conflict(current_fh,
2492 NFS4_SHARE_DENY_WRITE);
2493 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2494 return nfs4_share_conflict(current_fh,
2495 NFS4_SHARE_DENY_READ);
2496}
2497
2498/*
2499 * Allow READ/WRITE during grace period on recovered state only for files
2500 * that are not able to provide mandatory locking.
2501 */
2502static inline int
18f82731 2503grace_disallows_io(struct inode *inode)
1da177e4 2504{
203a8c8e 2505 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2506}
2507
0836f587
BF		 2508static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2509{
2510 /* If the client sends us a stateid from the future, it's buggy: */
2511 if (in->si_generation > ref->si_generation)
2512 return nfserr_bad_stateid;
2513 /*
2514 * The following, however, can happen. For example, if the
2515 * client sends an open and some IO at the same time, the open
2516 * may bump si_generation while the IO is still in flight.
2517 * Thanks to hard links and renames, the client never knows what
2518 * file an open will affect. So it could avoid that situation
2519 * only by serializing all opens and IO from the same open
2520 * owner. To recover from the old_stateid error, the client
2521 * will just have to retry the IO:
2522 */
2523 if (in->si_generation < ref->si_generation)
2524 return nfserr_old_stateid;
2525 return nfs_ok;
2526}
2527
3e633079
BF		 2528static int is_delegation_stateid(stateid_t *stateid)
2529{
2530 return stateid->si_fileid == 0;
2531}
2532
1da177e4
LT		 2533/*
2534* Checks for stateid operations
2535*/
b37ad28b 2536__be32
1da177e4
LT		 2537nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2538{
2539 struct nfs4_stateid *stp = NULL;
2540 struct nfs4_delegation *dp = NULL;
1da177e4 2541 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2542 __be32 status;
1da177e4 2543
1da177e4
LT		 2544 if (filpp)
2545 *filpp = NULL;
2546
18f82731 2547 if (grace_disallows_io(ino))
1da177e4
LT		 2548 return nfserr_grace;
2549
2550 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2551 return check_special_stateids(current_fh, stateid, flags);
2552
1da177e4
LT		 2553 status = nfserr_stale_stateid;
2554 if (STALE_STATEID(stateid))
2555 goto out;
2556
1da177e4 2557 status = nfserr_bad_stateid;
3e633079 2558 if (is_delegation_stateid(stateid)) {
a4455be0 2559 dp = find_delegation_stateid(ino, stateid);
819a8f53 2560 if (!dp)
1da177e4 2561 goto out;
fd03b099 2562 status = check_stateid_generation(stateid, &dp->dl_stateid);
0c2a498f
BF		 2563 if (status)
2564 goto out;
dc9bf700
BF		 2565 status = nfs4_check_delegmode(dp, flags);
2566 if (status)
2567 goto out;
2568 renew_client(dp->dl_client);
dc9bf700
BF		 2569 if (filpp)
2570 *filpp = dp->dl_vfs_file;
1da177e4 2571 } else { /* open or lock stateid */
a4455be0 2572 stp = find_stateid(stateid, flags);
819a8f53 2573 if (!stp)
1da177e4 2574 goto out;
6150ef0d 2575 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2576 goto out;
2577 if (!stp->st_stateowner->so_confirmed)
2578 goto out;
fd03b099 2579 status = check_stateid_generation(stateid, &stp->st_stateid);
0c2a498f
BF		 2580 if (status)
2581 goto out;
a4455be0
BF		 2582 status = nfs4_check_openmode(stp, flags);
2583 if (status)
1da177e4
LT		 2584 goto out;
2585 renew_client(stp->st_stateowner->so_client);
2586 if (filpp)
2587 *filpp = stp->st_vfs_file;
1da177e4
LT		 2588 }
2589 status = nfs_ok;
2590out:
2591 return status;
2592}
2593
4c4cd222
N		 2594static inline int
2595setlkflg (int type)
2596{
2597 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2598 RD_STATE : WR_STATE;
2599}
1da177e4
LT		 2600
2601/*
2602 * Checks for sequence id mutating operations.
2603 */
b37ad28b 2604static __be32
4c4cd222 2605nfs4_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)
1da177e4 2606{
1da177e4
LT		 2607 struct nfs4_stateid *stp;
2608 struct nfs4_stateowner *sop;
0836f587 2609 __be32 status;
1da177e4
LT		 2610
2611 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2612 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2613 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2614 stateid->si_generation);
3a4f98bb 2615
1da177e4
LT		 2616 *stpp = NULL;
2617 *sopp = NULL;
2618
1da177e4 2619 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2620 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2621 return nfserr_bad_stateid;
1da177e4
LT		 2622 }
2623
1da177e4 2624 if (STALE_STATEID(stateid))
3a4f98bb 2625 return nfserr_stale_stateid;
1da177e4
LT		 2626 /*
2627 * We return BAD_STATEID if filehandle doesn't match stateid,
2628 * the confirmed flag is incorrecly set, or the generation
2629 * number is incorrect.
1da177e4 2630 */
f8816512
N		 2631 stp = find_stateid(stateid, flags);
2632 if (stp == NULL) {
2633 /*
2634 * Also, we should make sure this isn't just the result of
2635 * a replayed close:
2636 */
2637 sop = search_close_lru(stateid->si_stateownerid, flags);
2638 if (sop == NULL)
2639 return nfserr_bad_stateid;
2640 *sopp = sop;
2641 goto check_replay;
2642 }
1da177e4 2643
39325bd0
BF		 2644 *stpp = stp;
2645 *sopp = sop = stp->st_stateowner;
2646
4c4cd222 2647 if (lock) {
4c4cd222 2648 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2649 struct nfs4_client *clp = sop->so_client;
4c4cd222 2650 int lkflg = 0;
b37ad28b 2651 __be32 status;
4c4cd222
N		 2652
2653 lkflg = setlkflg(lock->lk_type);
2654
2655 if (lock->lk_is_new) {
599e0a22
BF		 2656 if (!sop->so_is_open_owner)
2657 return nfserr_bad_stateid;
2658 if (!same_clid(&clp->cl_clientid, lockclid))
4c4cd222 2659 return nfserr_bad_stateid;
599e0a22
BF		 2660 /* stp is the open stateid */
2661 status = nfs4_check_openmode(stp, lkflg);
2662 if (status)
2663 return status;
2664 } else {
2665 /* stp is the lock stateid */
2666 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2667 if (status)
2668 return status;
4c4cd222 2669 }
1da177e4
LT		 2670 }
2671
f3362737 2672 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2673 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2674 return nfserr_bad_stateid;
1da177e4
LT		 2675 }
2676
1da177e4
LT		 2677 /*
2678 * We now validate the seqid and stateid generation numbers.
2679 * For the moment, we ignore the possibility of
2680 * generation number wraparound.
2681 */
bd9aac52 2682 if (seqid != sop->so_seqid)
1da177e4
LT		 2683 goto check_replay;
2684
3a4f98bb 2685 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2686 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2687 " unconfirmed stateowner!\n");
2688 return nfserr_bad_stateid;
1da177e4 2689 }
3a4f98bb 2690 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2691 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2692 " confirmed yet!\n");
2693 return nfserr_bad_stateid;
1da177e4 2694 }
0836f587
BF		 2695 status = check_stateid_generation(stateid, &stp->st_stateid);
2696 if (status)
2697 return status;
52fd004e 2698 renew_client(sop->so_client);
3a4f98bb 2699 return nfs_ok;
1da177e4 2700
1da177e4 2701check_replay:
bd9aac52 2702 if (seqid == sop->so_seqid - 1) {
849823c5 2703 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2704 /* indicate replay to calling function */
a90b061c 2705 return nfserr_replay_me;
1da177e4 2706 }
2fdada03 2707 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2708 sop->so_seqid, seqid);
2709 *sopp = NULL;
2710 return nfserr_bad_seqid;
1da177e4
LT		 2711}
2712
b37ad28b 2713__be32
ca364317 2714nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2715 struct nfsd4_open_confirm *oc)
1da177e4 2716{
b37ad28b 2717 __be32 status;
1da177e4
LT		 2718 struct nfs4_stateowner *sop;
2719 struct nfs4_stateid *stp;
2720
2721 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2722 (int)cstate->current_fh.fh_dentry->d_name.len,
2723 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2724
ca364317 2725 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 2726 if (status)
2727 return status;
1da177e4
LT		 2728
2729 nfs4_lock_state();
2730
ca364317
BF		 2731 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2732 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 2733 CONFIRM | OPEN_STATE,
1da177e4
LT		 2734 &oc->oc_stateowner, &stp, NULL)))
2735 goto out;
2736
2737 sop = oc->oc_stateowner;
2738 sop->so_confirmed = 1;
2739 update_stateid(&stp->st_stateid);
2740 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2741 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2742 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2743 stp->st_stateid.si_boot,
2744 stp->st_stateid.si_stateownerid,
2745 stp->st_stateid.si_fileid,
2746 stp->st_stateid.si_generation);
c7b9a459
N		 2747
2748 nfsd4_create_clid_dir(sop->so_client);
1da177e4 2749out:
f2327d9a 2750 if (oc->oc_stateowner) {
1da177e4 2751 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 2752 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 2753 }
1da177e4
LT		 2754 nfs4_unlock_state();
2755 return status;
2756}
2757
2758
2759/*
2760 * unset all bits in union bitmap (bmap) that
2761 * do not exist in share (from successful OPEN_DOWNGRADE)
2762 */
2763static void
2764reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2765{
2766 int i;
2767 for (i = 1; i < 4; i++) {
2768 if ((i & access) != i)
2769 __clear_bit(i, bmap);
2770 }
2771}
2772
2773static void
2774reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2775{
2776 int i;
2777 for (i = 0; i < 4; i++) {
2778 if ((i & deny) != i)
2779 __clear_bit(i, bmap);
2780 }
2781}
2782
b37ad28b 2783__be32
ca364317
BF		 2784nfsd4_open_downgrade(struct svc_rqst *rqstp,
2785 struct nfsd4_compound_state *cstate,
a4f1706a 2786 struct nfsd4_open_downgrade *od)
1da177e4 2787{
b37ad28b 2788 __be32 status;
1da177e4
LT		 2789 struct nfs4_stateid *stp;
2790 unsigned int share_access;
2791
2792 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 2793 (int)cstate->current_fh.fh_dentry->d_name.len,
2794 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2795
ba5a6a19
BF		 2796 if (!access_valid(od->od_share_access)
2797 || !deny_valid(od->od_share_deny))
1da177e4
LT		 2798 return nfserr_inval;
2799
2800 nfs4_lock_state();
ca364317
BF		 2801 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2802 od->od_seqid,
1da177e4 2803 &od->od_stateid,
f3362737 2804 OPEN_STATE,
1da177e4
LT		 2805 &od->od_stateowner, &stp, NULL)))
2806 goto out;
2807
2808 status = nfserr_inval;
2809 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2810 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2811 stp->st_access_bmap, od->od_share_access);
2812 goto out;
2813 }
2814 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2815 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2816 stp->st_deny_bmap, od->od_share_deny);
2817 goto out;
2818 }
2819 set_access(&share_access, stp->st_access_bmap);
2820 nfs4_file_downgrade(stp->st_vfs_file,
2821 share_access & ~od->od_share_access);
2822
2823 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2824 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2825
2826 update_stateid(&stp->st_stateid);
2827 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2828 status = nfs_ok;
2829out:
f2327d9a 2830 if (od->od_stateowner) {
1da177e4 2831 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 2832 cstate->replay_owner = od->od_stateowner;
f2327d9a 2833 }
1da177e4
LT		 2834 nfs4_unlock_state();
2835 return status;
2836}
2837
2838/*
2839 * nfs4_unlock_state() called after encode
2840 */
b37ad28b 2841__be32
ca364317 2842nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2843 struct nfsd4_close *close)
1da177e4 2844{
b37ad28b 2845 __be32 status;
1da177e4
LT		 2846 struct nfs4_stateid *stp;
2847
2848 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 2849 (int)cstate->current_fh.fh_dentry->d_name.len,
2850 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 2851
2852 nfs4_lock_state();
2853 /* check close_lru for replay */
ca364317
BF		 2854 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2855 close->cl_seqid,
1da177e4 2856 &close->cl_stateid,
f3362737 2857 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 2858 &close->cl_stateowner, &stp, NULL)))
2859 goto out;
1da177e4
LT		 2860 status = nfs_ok;
2861 update_stateid(&stp->st_stateid);
2862 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2863
04ef5954 2864 /* release_stateid() calls nfsd_close() if needed */
2283963f 2865 release_open_stateid(stp);
04ef5954
BF		 2866
2867 /* place unused nfs4_stateowners on so_close_lru list to be
2868 * released by the laundromat service after the lease period
2869 * to enable us to handle CLOSE replay
2870 */
2871 if (list_empty(&close->cl_stateowner->so_stateids))
2872 move_to_close_lru(close->cl_stateowner);
1da177e4 2873out:
f2327d9a 2874 if (close->cl_stateowner) {
1da177e4 2875 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 2876 cstate->replay_owner = close->cl_stateowner;
f2327d9a 2877 }
1da177e4
LT		 2878 nfs4_unlock_state();
2879 return status;
2880}
2881
b37ad28b 2882__be32
ca364317
BF		 2883nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2884 struct nfsd4_delegreturn *dr)
1da177e4 2885{
203a8c8e
BF		 2886 struct nfs4_delegation *dp;
2887 stateid_t *stateid = &dr->dr_stateid;
2888 struct inode *inode;
b37ad28b 2889 __be32 status;
1da177e4 2890
ca364317 2891 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 2892 return status;
2893 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 2894
2895 nfs4_lock_state();
203a8c8e
BF		 2896 status = nfserr_bad_stateid;
2897 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2898 goto out;
2899 status = nfserr_stale_stateid;
2900 if (STALE_STATEID(stateid))
2901 goto out;
7e0f7cf5 2902 status = nfserr_bad_stateid;
203a8c8e
BF		 2903 if (!is_delegation_stateid(stateid))
2904 goto out;
203a8c8e
BF		 2905 dp = find_delegation_stateid(inode, stateid);
2906 if (!dp)
2907 goto out;
2908 status = check_stateid_generation(stateid, &dp->dl_stateid);
2909 if (status)
2910 goto out;
2911 renew_client(dp->dl_client);
2912
2913 unhash_delegation(dp);
1da177e4 2914out:
203a8c8e
BF		 2915 nfs4_unlock_state();
2916
1da177e4
LT		 2917 return status;
2918}
2919
2920
2921/*
2922 * Lock owner state (byte-range locks)
2923 */
2924#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2925#define LOCK_HASH_BITS 8
2926#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2927#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2928
87df4de8
BH		 2929static inline u64
2930end_offset(u64 start, u64 len)
2931{
2932 u64 end;
2933
2934 end = start + len;
2935 return end >= start ? end: NFS4_MAX_UINT64;
2936}
2937
2938/* last octet in a range */
2939static inline u64
2940last_byte_offset(u64 start, u64 len)
2941{
2942 u64 end;
2943
2944 BUG_ON(!len);
2945 end = start + len;
2946 return end > start ? end - 1: NFS4_MAX_UINT64;
2947}
2948
1da177e4
LT		 2949#define lockownerid_hashval(id) \
2950 ((id) & LOCK_HASH_MASK)
2951
2952static inline unsigned int
2953lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2954 struct xdr_netobj *ownername)
2955{
2956 return (file_hashval(inode) + cl_id
2957 + opaque_hashval(ownername->data, ownername->len))
2958 & LOCK_HASH_MASK;
2959}
2960
2961static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2962static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2963static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2964
fd39ca9a 2965static struct nfs4_stateid *
1da177e4
LT		 2966find_stateid(stateid_t *stid, int flags)
2967{
9346eff0 2968 struct nfs4_stateid *local;
1da177e4
LT		 2969 u32 st_id = stid->si_stateownerid;
2970 u32 f_id = stid->si_fileid;
2971 unsigned int hashval;
2972
2973 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 2974 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 2975 hashval = stateid_hashval(st_id, f_id);
2976 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2977 if ((local->st_stateid.si_stateownerid == st_id) &&
2978 (local->st_stateid.si_fileid == f_id))
2979 return local;
2980 }
2981 }
9346eff0
KK		 2982
2983 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 2984 hashval = stateid_hashval(st_id, f_id);
2985 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2986 if ((local->st_stateid.si_stateownerid == st_id) &&
2987 (local->st_stateid.si_fileid == f_id))
2988 return local;
2989 }
849823c5 2990 }
1da177e4
LT		 2991 return NULL;
2992}
2993
2994static struct nfs4_delegation *
2995find_delegation_stateid(struct inode *ino, stateid_t *stid)
2996{
13cd2184
N		 2997 struct nfs4_file *fp;
2998 struct nfs4_delegation *dl;
1da177e4
LT		 2999
3000 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3001 stid->si_boot, stid->si_stateownerid,
3002 stid->si_fileid, stid->si_generation);
3003
1da177e4 3004 fp = find_file(ino);
13cd2184
N		 3005 if (!fp)
3006 return NULL;
3007 dl = find_delegation_file(fp, stid);
3008 put_nfs4_file(fp);
3009 return dl;
1da177e4
LT		 3010}
3011
3012/*
3013 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3014 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3015 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3016 * locking, this prevents us from being completely protocol-compliant. The
3017 * real solution to this problem is to start using unsigned file offsets in
3018 * the VFS, but this is a very deep change!
3019 */
3020static inline void
3021nfs4_transform_lock_offset(struct file_lock *lock)
3022{
3023 if (lock->fl_start < 0)
3024 lock->fl_start = OFFSET_MAX;
3025 if (lock->fl_end < 0)
3026 lock->fl_end = OFFSET_MAX;
3027}
3028
d5b9026a
N		 3029/* Hack!: For now, we're defining this just so we can use a pointer to it
3030 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3031static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3032};
1da177e4
LT		 3033
3034static inline void
3035nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3036{
d5b9026a
N		 3037 struct nfs4_stateowner *sop;
3038 unsigned int hval;
1da177e4 3039
d5b9026a
N		 3040 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3041 sop = (struct nfs4_stateowner *) fl->fl_owner;
3042 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3043 kref_get(&sop->so_ref);
3044 deny->ld_sop = sop;
3045 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3046 } else {
3047 deny->ld_sop = NULL;
3048 deny->ld_clientid.cl_boot = 0;
3049 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3050 }
3051 deny->ld_start = fl->fl_start;
87df4de8
BH		 3052 deny->ld_length = NFS4_MAX_UINT64;
3053 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3054 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3055 deny->ld_type = NFS4_READ_LT;
3056 if (fl->fl_type != F_RDLCK)
3057 deny->ld_type = NFS4_WRITE_LT;
3058}
3059
1da177e4
LT		 3060static struct nfs4_stateowner *
3061find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3062 struct xdr_netobj *owner)
3063{
3064 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3065 struct nfs4_stateowner *op;
3066
3067 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3068 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3069 return op;
3070 }
3071 return NULL;
3072}
3073
3074/*
3075 * Alloc a lock owner structure.
3076 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3077 * occured.
3078 *
3079 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3080 */
3081
3082static struct nfs4_stateowner *
3083alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3084 struct nfs4_stateowner *sop;
3085 struct nfs4_replay *rp;
3086 unsigned int idhashval;
3087
3088 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3089 return NULL;
3090 idhashval = lockownerid_hashval(current_ownerid);
3091 INIT_LIST_HEAD(&sop->so_idhash);
3092 INIT_LIST_HEAD(&sop->so_strhash);
3093 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3094 INIT_LIST_HEAD(&sop->so_stateids);
3095 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3096 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3097 sop->so_time = 0;
3098 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3099 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3100 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3101 sop->so_is_open_owner = 0;
3102 sop->so_id = current_ownerid++;
3103 sop->so_client = clp;
b59e3c0e
NB		 3104 /* It is the openowner seqid that will be incremented in encode in the
3105 * case of new lockowners; so increment the lock seqid manually: */
3106 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3107 sop->so_confirmed = 1;
3108 rp = &sop->so_replay;
de1ae286 3109 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3110 rp->rp_buflen = 0;
3111 rp->rp_buf = rp->rp_ibuf;
3112 return sop;
3113}
3114
fd39ca9a 3115static struct nfs4_stateid *
1da177e4
LT		 3116alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3117{
3118 struct nfs4_stateid *stp;
3119 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3120
5ac049ac
N		 3121 stp = nfs4_alloc_stateid();
3122 if (stp == NULL)
1da177e4
LT		 3123 goto out;
3124 INIT_LIST_HEAD(&stp->st_hash);
3125 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3126 INIT_LIST_HEAD(&stp->st_perstateowner);
3127 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3128 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3129 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3130 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3131 stp->st_stateowner = sop;
13cd2184 3132 get_nfs4_file(fp);
1da177e4
LT		 3133 stp->st_file = fp;
3134 stp->st_stateid.si_boot = boot_time;
3135 stp->st_stateid.si_stateownerid = sop->so_id;
3136 stp->st_stateid.si_fileid = fp->fi_id;
3137 stp->st_stateid.si_generation = 0;
3138 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3139 stp->st_access_bmap = open_stp->st_access_bmap;
3140 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3141 stp->st_openstp = open_stp;
1da177e4
LT		 3142
3143out:
3144 return stp;
3145}
3146
fd39ca9a 3147static int
1da177e4
LT		 3148check_lock_length(u64 offset, u64 length)
3149{
87df4de8 3150 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3151 LOFF_OVERFLOW(offset, length)));
3152}
3153
3154/*
3155 * LOCK operation
3156 */
b37ad28b 3157__be32
ca364317 3158nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3159 struct nfsd4_lock *lock)
1da177e4 3160{
3e9e3dbe 3161 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3162 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3163 struct nfs4_stateid *lock_stp;
3164 struct file *filp;
3165 struct file_lock file_lock;
8dc7c311 3166 struct file_lock conflock;
b37ad28b 3167 __be32 status = 0;
1da177e4 3168 unsigned int strhashval;
150b3934 3169 unsigned int cmd;
b8dd7b9a 3170 int err;
1da177e4
LT		 3171
3172 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3173 (long long) lock->lk_offset,
3174 (long long) lock->lk_length);
3175
1da177e4
LT		 3176 if (check_lock_length(lock->lk_offset, lock->lk_length))
3177 return nfserr_inval;
3178
ca364317 3179 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3180 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3181 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3182 return status;
3183 }
3184
1da177e4
LT		 3185 nfs4_lock_state();
3186
3187 if (lock->lk_is_new) {
893f8770
N		 3188 /*
3189 * Client indicates that this is a new lockowner.
3190 * Use open owner and open stateid to create lock owner and
3191 * lock stateid.
3192 */
1da177e4
LT		 3193 struct nfs4_stateid *open_stp = NULL;
3194 struct nfs4_file *fp;
3195
3196 status = nfserr_stale_clientid;
849823c5 3197 if (STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3198 goto out;
1da177e4 3199
1da177e4 3200 /* validate and update open stateid and open seqid */
ca364317 3201 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3202 lock->lk_new_open_seqid,
3203 &lock->lk_new_open_stateid,
f3362737 3204 OPEN_STATE,
3a65588a 3205 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3206 lock);
37515177 3207 if (status)
1da177e4 3208 goto out;
3a65588a 3209 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3210 /* create lockowner and lock stateid */
3211 fp = open_stp->st_file;
3212 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3213 open_sop->so_client->cl_clientid.cl_id,
3214 &lock->v.new.owner);
3e9e3dbe
N		 3215 /* XXX: Do we need to check for duplicate stateowners on
3216 * the same file, or should they just be allowed (and
3217 * create new stateids)? */
1da177e4 3218 status = nfserr_resource;
b59e3c0e
NB		 3219 lock_sop = alloc_init_lock_stateowner(strhashval,
3220 open_sop->so_client, open_stp, lock);
3221 if (lock_sop == NULL)
1da177e4 3222 goto out;
b59e3c0e 3223 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3224 if (lock_stp == NULL)
1da177e4 3225 goto out;
1da177e4
LT		 3226 } else {
3227 /* lock (lock owner + lock stateid) already exists */
ca364317 3228 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3229 lock->lk_old_lock_seqid,
3230 &lock->lk_old_lock_stateid,
f3362737 3231 LOCK_STATE,
3a65588a 3232 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3233 if (status)
3234 goto out;
3a65588a 3235 lock_sop = lock->lk_replay_owner;
1da177e4 3236 }
3a65588a 3237 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3238 filp = lock_stp->st_vfs_file;
3239
0dd395dc 3240 status = nfserr_grace;
af558e33 3241 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3242 goto out;
3243 status = nfserr_no_grace;
af558e33 3244 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3245 goto out;
3246
1da177e4
LT		 3247 locks_init_lock(&file_lock);
3248 switch (lock->lk_type) {
3249 case NFS4_READ_LT:
3250 case NFS4_READW_LT:
3251 file_lock.fl_type = F_RDLCK;
150b3934 3252 cmd = F_SETLK;
1da177e4
LT		 3253 break;
3254 case NFS4_WRITE_LT:
3255 case NFS4_WRITEW_LT:
3256 file_lock.fl_type = F_WRLCK;
150b3934 3257 cmd = F_SETLK;
1da177e4
LT		 3258 break;
3259 default:
3260 status = nfserr_inval;
3261 goto out;
3262 }
b59e3c0e 3263 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3264 file_lock.fl_pid = current->tgid;
3265 file_lock.fl_file = filp;
3266 file_lock.fl_flags = FL_POSIX;
d5b9026a 3267 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3268
3269 file_lock.fl_start = lock->lk_offset;
87df4de8 3270 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3271 nfs4_transform_lock_offset(&file_lock);
3272
3273 /*
3274 * Try to lock the file in the VFS.
3275 * Note: locks.c uses the BKL to protect the inode's lock list.
3276 */
3277
fd85b817 3278 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3279 switch (-err) {
1da177e4
LT		 3280 case 0: /* success! */
3281 update_stateid(&lock_stp->st_stateid);
3282 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3283 sizeof(stateid_t));
b8dd7b9a 3284 status = 0;
eb76b3fd
AA		 3285 break;
3286 case (EAGAIN): /* conflock holds conflicting lock */
3287 status = nfserr_denied;
3288 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3289 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3290 break;
1da177e4
LT		 3291 case (EDEADLK):
3292 status = nfserr_deadlock;
eb76b3fd 3293 break;
1da177e4 3294 default:
fd85b817 3295 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3296 status = nfserr_resource;
3297 break;
1da177e4 3298 }
1da177e4 3299out:
8a280510 3300 if (status && lock->lk_is_new && lock_sop)
f044ff83 3301 release_lockowner(lock_sop);
3a65588a
BF		 3302 if (lock->lk_replay_owner) {
3303 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3304 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3305 }
1da177e4
LT		 3306 nfs4_unlock_state();
3307 return status;
3308}
3309
55ef1274
BF		 3310/*
3311 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3312 * so we do a temporary open here just to get an open file to pass to
3313 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3314 * inode operation.)
3315 */
3316static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3317{
3318 struct file *file;
3319 int err;
3320
3321 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3322 if (err)
3323 return err;
3324 err = vfs_test_lock(file, lock);
3325 nfsd_close(file);
3326 return err;
3327}
3328
1da177e4
LT		 3329/*
3330 * LOCKT operation
3331 */
b37ad28b 3332__be32
ca364317
BF		 3333nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3334 struct nfsd4_lockt *lockt)
1da177e4
LT		 3335{
3336 struct inode *inode;
1da177e4 3337 struct file_lock file_lock;
fd85b817 3338 int error;
b37ad28b 3339 __be32 status;
1da177e4 3340
af558e33 3341 if (locks_in_grace())
1da177e4
LT		 3342 return nfserr_grace;
3343
3344 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3345 return nfserr_inval;
3346
3347 lockt->lt_stateowner = NULL;
3348 nfs4_lock_state();
3349
3350 status = nfserr_stale_clientid;
849823c5 3351 if (STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3352 goto out;
1da177e4 3353
ca364317 3354 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3355 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3356 if (status == nfserr_symlink)
3357 status = nfserr_inval;
3358 goto out;
3359 }
3360
ca364317 3361 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3362 locks_init_lock(&file_lock);
3363 switch (lockt->lt_type) {
3364 case NFS4_READ_LT:
3365 case NFS4_READW_LT:
3366 file_lock.fl_type = F_RDLCK;
3367 break;
3368 case NFS4_WRITE_LT:
3369 case NFS4_WRITEW_LT:
3370 file_lock.fl_type = F_WRLCK;
3371 break;
3372 default:
2fdada03 3373 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3374 status = nfserr_inval;
3375 goto out;
3376 }
3377
3378 lockt->lt_stateowner = find_lockstateowner_str(inode,
3379 &lockt->lt_clientid, &lockt->lt_owner);
3380 if (lockt->lt_stateowner)
3381 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3382 file_lock.fl_pid = current->tgid;
3383 file_lock.fl_flags = FL_POSIX;
3384
3385 file_lock.fl_start = lockt->lt_offset;
87df4de8 3386 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3387
3388 nfs4_transform_lock_offset(&file_lock);
3389
1da177e4 3390 status = nfs_ok;
55ef1274 3391 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3392 if (error) {
3393 status = nfserrno(error);
3394 goto out;
3395 }
9d6a8c5c 3396 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3397 status = nfserr_denied;
9d6a8c5c 3398 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3399 }
3400out:
3401 nfs4_unlock_state();
3402 return status;
3403}
3404
b37ad28b 3405__be32
ca364317 3406nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3407 struct nfsd4_locku *locku)
1da177e4
LT		 3408{
3409 struct nfs4_stateid *stp;
3410 struct file *filp = NULL;
3411 struct file_lock file_lock;
b37ad28b 3412 __be32 status;
b8dd7b9a 3413 int err;
1da177e4
LT		 3414
3415 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3416 (long long) locku->lu_offset,
3417 (long long) locku->lu_length);
3418
3419 if (check_lock_length(locku->lu_offset, locku->lu_length))
3420 return nfserr_inval;
3421
3422 nfs4_lock_state();
3423
ca364317 3424 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3425 locku->lu_seqid,
3426 &locku->lu_stateid,
f3362737 3427 LOCK_STATE,
1da177e4
LT		 3428 &locku->lu_stateowner, &stp, NULL)))
3429 goto out;
3430
3431 filp = stp->st_vfs_file;
3432 BUG_ON(!filp);
3433 locks_init_lock(&file_lock);
3434 file_lock.fl_type = F_UNLCK;
3435 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3436 file_lock.fl_pid = current->tgid;
3437 file_lock.fl_file = filp;
3438 file_lock.fl_flags = FL_POSIX;
d5b9026a 3439 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3440 file_lock.fl_start = locku->lu_offset;
3441
87df4de8 3442 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3443 nfs4_transform_lock_offset(&file_lock);
3444
3445 /*
3446 * Try to unlock the file in the VFS.
3447 */
fd85b817 3448 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3449 if (err) {
fd85b817 3450 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3451 goto out_nfserr;
3452 }
3453 /*
3454 * OK, unlock succeeded; the only thing left to do is update the stateid.
3455 */
3456 update_stateid(&stp->st_stateid);
3457 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3458
3459out:
f2327d9a 3460 if (locku->lu_stateowner) {
1da177e4 3461 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3462 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3463 }
1da177e4
LT		 3464 nfs4_unlock_state();
3465 return status;
3466
3467out_nfserr:
b8dd7b9a 3468 status = nfserrno(err);
1da177e4
LT		 3469 goto out;
3470}
3471
3472/*
3473 * returns
3474 * 1: locks held by lockowner
3475 * 0: no locks held by lockowner
3476 */
3477static int
3478check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3479{
3480 struct file_lock **flpp;
7eaa36e2 3481 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3482 int status = 0;
3483
3484 lock_kernel();
3485 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3486 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3487 status = 1;
3488 goto out;
796dadfd 3489 }
1da177e4
LT		 3490 }
3491out:
3492 unlock_kernel();
3493 return status;
3494}
3495
b37ad28b 3496__be32
b591480b
BF		 3497nfsd4_release_lockowner(struct svc_rqst *rqstp,
3498 struct nfsd4_compound_state *cstate,
3499 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3500{
3501 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3502 struct nfs4_stateowner *sop;
3503 struct nfs4_stateid *stp;
1da177e4 3504 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3505 struct list_head matches;
3506 int i;
b37ad28b 3507 __be32 status;
1da177e4
LT		 3508
3509 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3510 clid->cl_boot, clid->cl_id);
3511
3512 /* XXX check for lease expiration */
3513
3514 status = nfserr_stale_clientid;
849823c5 3515 if (STALE_CLIENTID(clid))
1da177e4 3516 return status;
1da177e4
LT		 3517
3518 nfs4_lock_state();
3519
3e9e3dbe
N		 3520 status = nfserr_locks_held;
3521 /* XXX: we're doing a linear search through all the lockowners.
3522 * Yipes! For now we'll just hope clients aren't really using
3523 * release_lockowner much, but eventually we have to fix these
3524 * data structures. */
3525 INIT_LIST_HEAD(&matches);
3526 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3527 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3528 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3529 continue;
3530 list_for_each_entry(stp, &sop->so_stateids,
3531 st_perstateowner) {
3532 if (check_for_locks(stp->st_vfs_file, sop))
3533 goto out;
3534 /* Note: so_perclient unused for lockowners,
3535 * so it's OK to fool with here. */
3536 list_add(&sop->so_perclient, &matches);
3537 }
1da177e4 3538 }
3e9e3dbe
N		 3539 }
3540 /* Clients probably won't expect us to return with some (but not all)
3541 * of the lockowner state released; so don't release any until all
3542 * have been checked. */
3543 status = nfs_ok;
0fa822e4
N		 3544 while (!list_empty(&matches)) {
3545 sop = list_entry(matches.next, struct nfs4_stateowner,
3546 so_perclient);
3547 /* unhash_stateowner deletes so_perclient only
3548 * for openowners. */
3549 list_del(&sop->so_perclient);
f044ff83 3550 release_lockowner(sop);
1da177e4
LT		 3551 }
3552out:
3553 nfs4_unlock_state();
3554 return status;
3555}
3556
3557static inline struct nfs4_client_reclaim *
a55370a3 3558alloc_reclaim(void)
1da177e4 3559{
a55370a3 3560 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3561}
3562
c7b9a459 3563int
a1bcecd2 3564nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3565{
3566 unsigned int strhashval = clientstr_hashval(name);
3567 struct nfs4_client *clp;
3568
a1bcecd2 3569 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3570 return clp ? 1 : 0;
3571}
3572
1da177e4
LT		 3573/*
3574 * failure => all reset bets are off, nfserr_no_grace...
3575 */
190e4fbf
N		 3576int
3577nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3578{
3579 unsigned int strhashval;
3580 struct nfs4_client_reclaim *crp = NULL;
3581
a55370a3
N		 3582 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3583 crp = alloc_reclaim();
1da177e4
LT		 3584 if (!crp)
3585 return 0;
a55370a3 3586 strhashval = clientstr_hashval(name);
1da177e4
LT		 3587 INIT_LIST_HEAD(&crp->cr_strhash);
3588 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3589 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3590 reclaim_str_hashtbl_size++;
3591 return 1;
3592}
3593
3594static void
3595nfs4_release_reclaim(void)
3596{
3597 struct nfs4_client_reclaim *crp = NULL;
3598 int i;
3599
1da177e4
LT		 3600 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3601 while (!list_empty(&reclaim_str_hashtbl[i])) {
3602 crp = list_entry(reclaim_str_hashtbl[i].next,
3603 struct nfs4_client_reclaim, cr_strhash);
3604 list_del(&crp->cr_strhash);
1da177e4
LT		 3605 kfree(crp);
3606 reclaim_str_hashtbl_size--;
3607 }
3608 }
3609 BUG_ON(reclaim_str_hashtbl_size);
3610}
3611
3612/*
3613 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3614static struct nfs4_client_reclaim *
1da177e4
LT		 3615nfs4_find_reclaim_client(clientid_t *clid)
3616{
3617 unsigned int strhashval;
3618 struct nfs4_client *clp;
3619 struct nfs4_client_reclaim *crp = NULL;
3620
3621
3622 /* find clientid in conf_id_hashtbl */
3623 clp = find_confirmed_client(clid);
3624 if (clp == NULL)
3625 return NULL;
3626
a55370a3
N		 3627 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3628 clp->cl_name.len, clp->cl_name.data,
3629 clp->cl_recdir);
1da177e4
LT		 3630
3631 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3632 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3633 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3634 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3635 return crp;
3636 }
3637 }
3638 return NULL;
3639}
3640
3641/*
3642* Called from OPEN. Look for clientid in reclaim list.
3643*/
b37ad28b 3644__be32
1da177e4
LT		 3645nfs4_check_open_reclaim(clientid_t *clid)
3646{
dfc83565 3647 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3648}
3649
ac4d8ff2 3650/* initialization to perform at module load time: */
1da177e4 3651
e8ff2a84 3652int
ac4d8ff2 3653nfs4_state_init(void)
1da177e4 3654{
e8ff2a84 3655 int i, status;
1da177e4 3656
e8ff2a84
BF		 3657 status = nfsd4_init_slabs();
3658 if (status)
3659 return status;
1da177e4
LT		 3660 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3661 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3662 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3663 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3664 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3665 }
5282fd72
ME		 3666 for (i = 0; i < SESSION_HASH_SIZE; i++)
3667 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3668 for (i = 0; i < FILE_HASH_SIZE; i++) {
3669 INIT_LIST_HEAD(&file_hashtbl[i]);
3670 }
3671 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3672 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3673 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3674 }
3675 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3676 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3677 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3678 }
3679 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3680 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3681 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3682 }
1da177e4 3683 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3684 INIT_LIST_HEAD(&close_lru);
3685 INIT_LIST_HEAD(&client_lru);
3686 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 3687 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3688 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3689 reclaim_str_hashtbl_size = 0;
e8ff2a84 3690 return 0;
ac4d8ff2
N		 3691}
3692
190e4fbf
N		 3693static void
3694nfsd4_load_reboot_recovery_data(void)
3695{
3696 int status;
3697
0964a3d3
N		 3698 nfs4_lock_state();
3699 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3700 status = nfsd4_recdir_load();
0964a3d3 3701 nfs4_unlock_state();
190e4fbf
N		 3702 if (status)
3703 printk("NFSD: Failure reading reboot recovery data\n");
3704}
3705
9a8db97e
ME		 3706unsigned long
3707get_nfs4_grace_period(void)
3708{
3709 return max(user_lease_time, lease_time) * HZ;
3710}
3711
c2f1a551
MS		 3712/*
3713 * Since the lifetime of a delegation isn't limited to that of an open, a
3714 * client may quite reasonably hang on to a delegation as long as it has
3715 * the inode cached. This becomes an obvious problem the first time a
3716 * client's inode cache approaches the size of the server's total memory.
3717 *
3718 * For now we avoid this problem by imposing a hard limit on the number
3719 * of delegations, which varies according to the server's memory size.
3720 */
3721static void
3722set_max_delegations(void)
3723{
3724 /*
3725 * Allow at most 4 delegations per megabyte of RAM. Quick
3726 * estimates suggest that in the worst case (where every delegation
3727 * is for a different inode), a delegation could take about 1.5K,
3728 * giving a worst case usage of about 6% of memory.
3729 */
3730 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3731}
3732
ac4d8ff2
N		 3733/* initialization to perform when the nfsd service is started: */
3734
3735static void
3736__nfs4_state_start(void)
3737{
9a8db97e 3738 unsigned long grace_time;
ac4d8ff2 3739
1da177e4 3740 boot_time = get_seconds();
af558e33 3741 grace_time = get_nfs4_grace_period();
d99a05ad 3742 lease_time = user_lease_time;
af558e33 3743 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 3744 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3745 grace_time/HZ);
58da282b 3746 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 3747 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 3748 set_max_delegations();
1da177e4
LT		 3749}
3750
e8ff2a84 3751void
76a3550e 3752nfs4_state_start(void)
1da177e4 3753{
1da177e4 3754 if (nfs4_init)
e8ff2a84 3755 return;
190e4fbf 3756 nfsd4_load_reboot_recovery_data();
76a3550e 3757 __nfs4_state_start();
1da177e4 3758 nfs4_init = 1;
e8ff2a84 3759 return;
1da177e4
LT		 3760}
3761
1da177e4
LT		 3762time_t
3763nfs4_lease_time(void)
3764{
3765 return lease_time;
3766}
3767
3768static void
3769__nfs4_state_shutdown(void)
3770{
3771 int i;
3772 struct nfs4_client *clp = NULL;
3773 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 3774 struct list_head *pos, *next, reaplist;
3775
1da177e4
LT		 3776 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3777 while (!list_empty(&conf_id_hashtbl[i])) {
3778 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3779 expire_client(clp);
3780 }
3781 while (!list_empty(&unconf_str_hashtbl[i])) {
3782 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3783 expire_client(clp);
3784 }
3785 }
3786 INIT_LIST_HEAD(&reaplist);
3787 spin_lock(&recall_lock);
3788 list_for_each_safe(pos, next, &del_recall_lru) {
3789 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3790 list_move(&dp->dl_recall_lru, &reaplist);
3791 }
3792 spin_unlock(&recall_lock);
3793 list_for_each_safe(pos, next, &reaplist) {
3794 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3795 list_del_init(&dp->dl_recall_lru);
3796 unhash_delegation(dp);
3797 }
3798
190e4fbf 3799 nfsd4_shutdown_recdir();
1da177e4 3800 nfs4_init = 0;
1da177e4
LT		 3801}
3802
3803void
3804nfs4_state_shutdown(void)
3805{
5e8d5c29
N		 3806 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3807 destroy_workqueue(laundry_wq);
2c5e7615 3808 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 3809 nfs4_lock_state();
3810 nfs4_release_reclaim();
3811 __nfs4_state_shutdown();
1da177e4
LT		 3812 nfs4_unlock_state();
3813}
3814
3dd98a3b
JL		 3815/*
3816 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3817 * accessed when nfsd is starting.
3818 */
0964a3d3
N		 3819static void
3820nfs4_set_recdir(char *recdir)
3821{
0964a3d3 3822 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 3823}
3824
3825/*
3826 * Change the NFSv4 recovery directory to recdir.
3827 */
3828int
3829nfs4_reset_recoverydir(char *recdir)
3830{
3831 int status;
a63bb996 3832 struct path path;
0964a3d3 3833
a63bb996 3834 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 3835 if (status)
3836 return status;
3837 status = -ENOTDIR;
a63bb996 3838 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 3839 nfs4_set_recdir(recdir);
3840 status = 0;
3841 }
a63bb996 3842 path_put(&path);
0964a3d3
N		 3843 return status;
3844}
3845
3dd98a3b
JL		 3846char *
3847nfs4_recoverydir(void)
3848{
3849 return user_recovery_dirname;
3850}
3851
1da177e4
LT		 3852/*
3853 * Called when leasetime is changed.
3854 *
d99a05ad
N		 3855 * The only way the protocol gives us to handle on-the-fly lease changes is to
3856 * simulate a reboot. Instead of doing that, we just wait till the next time
3857 * we start to register any changes in lease time. If the administrator
3858 * really wants to change the lease time *now*, they can go ahead and bring
3859 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 3860 *
3861 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3862 * when nfsd is starting
1da177e4
LT		 3863 */
3864void
3865nfs4_reset_lease(time_t leasetime)
3866{
d99a05ad 3867 user_lease_time = leasetime;
1da177e4 3868}
Clone of davem's net-next-2.6 tree