nfsd41: release_session when client is expired
[linux-2.6-block.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
7116ed6b
AA
385static void
386release_session(struct nfsd4_session *ses)
387{
388 list_del(&ses->se_hash);
389 list_del(&ses->se_perclnt);
390 nfsd4_put_session(ses);
391}
392
393void
394free_session(struct kref *kref)
395{
396 struct nfsd4_session *ses;
397
398 ses = container_of(kref, struct nfsd4_session, se_ref);
399 kfree(ses->se_slots);
400 kfree(ses);
401}
402
1da177e4
LT
403static inline void
404renew_client(struct nfs4_client *clp)
405{
406 /*
407 * Move client to the end to the LRU list.
408 */
409 dprintk("renewing client (clientid %08x/%08x)\n",
410 clp->cl_clientid.cl_boot,
411 clp->cl_clientid.cl_id);
412 list_move_tail(&clp->cl_lru, &client_lru);
413 clp->cl_time = get_seconds();
414}
415
416/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
417static int
418STALE_CLIENTID(clientid_t *clid)
419{
420 if (clid->cl_boot == boot_time)
421 return 0;
422 dprintk("NFSD stale clientid (%08x/%08x)\n",
423 clid->cl_boot, clid->cl_id);
424 return 1;
425}
426
427/*
428 * XXX Should we use a slab cache ?
429 * This type of memory management is somewhat inefficient, but we use it
430 * anyway since SETCLIENTID is not a common operation.
431 */
35bba9a3 432static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT
433{
434 struct nfs4_client *clp;
435
35bba9a3
BF
436 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
437 if (clp == NULL)
438 return NULL;
439 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
440 if (clp->cl_name.data == NULL) {
441 kfree(clp);
442 return NULL;
1da177e4 443 }
35bba9a3
BF
444 memcpy(clp->cl_name.data, name.data, name.len);
445 clp->cl_name.len = name.len;
1da177e4
LT
446 return clp;
447}
448
1b1a9b31
BF
449static void
450shutdown_callback_client(struct nfs4_client *clp)
451{
452 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
453
1b1a9b31 454 if (clnt) {
404ec117
BF
455 /*
456 * Callback threads take a reference on the client, so there
457 * should be no outstanding callbacks at this point.
458 */
1b1a9b31
BF
459 clp->cl_callback.cb_client = NULL;
460 rpc_shutdown_client(clnt);
461 }
462}
463
1da177e4
LT
464static inline void
465free_client(struct nfs4_client *clp)
466{
1b1a9b31 467 shutdown_callback_client(clp);
1da177e4
LT
468 if (clp->cl_cred.cr_group_info)
469 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 470 kfree(clp->cl_principal);
1da177e4
LT
471 kfree(clp->cl_name.data);
472 kfree(clp);
473}
474
475void
476put_nfs4_client(struct nfs4_client *clp)
477{
478 if (atomic_dec_and_test(&clp->cl_count))
479 free_client(clp);
480}
481
482static void
483expire_client(struct nfs4_client *clp)
484{
485 struct nfs4_stateowner *sop;
486 struct nfs4_delegation *dp;
1da177e4
LT
487 struct list_head reaplist;
488
489 dprintk("NFSD: expire_client cl_count %d\n",
490 atomic_read(&clp->cl_count));
491
1da177e4
LT
492 INIT_LIST_HEAD(&reaplist);
493 spin_lock(&recall_lock);
ea1da636
N
494 while (!list_empty(&clp->cl_delegations)) {
495 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT
496 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
497 dp->dl_flock);
ea1da636 498 list_del_init(&dp->dl_perclnt);
1da177e4
LT
499 list_move(&dp->dl_recall_lru, &reaplist);
500 }
501 spin_unlock(&recall_lock);
502 while (!list_empty(&reaplist)) {
503 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
504 list_del_init(&dp->dl_recall_lru);
505 unhash_delegation(dp);
506 }
507 list_del(&clp->cl_idhash);
508 list_del(&clp->cl_strhash);
509 list_del(&clp->cl_lru);
ea1da636
N
510 while (!list_empty(&clp->cl_openowners)) {
511 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 512 release_openowner(sop);
1da177e4 513 }
c4bf7868
ME
514 while (!list_empty(&clp->cl_sessions)) {
515 struct nfsd4_session *ses;
516 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
517 se_perclnt);
518 release_session(ses);
519 }
1da177e4
LT
520 put_nfs4_client(clp);
521}
522
35bba9a3
BF
523static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
524{
1da177e4
LT
525 struct nfs4_client *clp;
526
35bba9a3
BF
527 clp = alloc_client(name);
528 if (clp == NULL)
529 return NULL;
a55370a3 530 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT
531 atomic_set(&clp->cl_count, 1);
532 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT
533 INIT_LIST_HEAD(&clp->cl_idhash);
534 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N
535 INIT_LIST_HEAD(&clp->cl_openowners);
536 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 537 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 538 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT
539 return clp;
540}
541
35bba9a3
BF
542static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
543{
544 memcpy(target->cl_verifier.data, source->data,
545 sizeof(target->cl_verifier.data));
1da177e4
LT
546}
547
35bba9a3
BF
548static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
549{
1da177e4
LT
550 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
551 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
552}
553
35bba9a3
BF
554static void copy_cred(struct svc_cred *target, struct svc_cred *source)
555{
1da177e4
LT
556 target->cr_uid = source->cr_uid;
557 target->cr_gid = source->cr_gid;
558 target->cr_group_info = source->cr_group_info;
559 get_group_info(target->cr_group_info);
560}
561
35bba9a3 562static int same_name(const char *n1, const char *n2)
599e0a22 563{
a55370a3 564 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT
565}
566
567static int
599e0a22
BF
568same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
569{
570 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT
571}
572
573static int
599e0a22
BF
574same_clid(clientid_t *cl1, clientid_t *cl2)
575{
576 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT
577}
578
579/* XXX what about NGROUP */
580static int
599e0a22
BF
581same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
582{
583 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT
584}
585
5ec7b46c
BF
586static void gen_clid(struct nfs4_client *clp)
587{
588 static u32 current_clientid = 1;
589
1da177e4
LT
590 clp->cl_clientid.cl_boot = boot_time;
591 clp->cl_clientid.cl_id = current_clientid++;
592}
593
deda2faa
BF
594static void gen_confirm(struct nfs4_client *clp)
595{
596 static u32 i;
597 u32 *p;
1da177e4 598
1da177e4 599 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF
600 *p++ = get_seconds();
601 *p++ = i++;
1da177e4
LT
602}
603
35bba9a3
BF
604static int check_name(struct xdr_netobj name)
605{
1da177e4
LT
606 if (name.len == 0)
607 return 0;
608 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 609 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT
610 return 0;
611 }
612 return 1;
613}
614
fd39ca9a 615static void
1da177e4
LT
616add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
617{
618 unsigned int idhashval;
619
620 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
621 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
622 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
623 list_add_tail(&clp->cl_lru, &client_lru);
624 clp->cl_time = get_seconds();
625}
626
fd39ca9a 627static void
1da177e4
LT
628move_to_confirmed(struct nfs4_client *clp)
629{
630 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
631 unsigned int strhashval;
632
633 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
634 list_del_init(&clp->cl_strhash);
f116629d 635 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 636 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT
637 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
638 renew_client(clp);
639}
640
641static struct nfs4_client *
642find_confirmed_client(clientid_t *clid)
643{
644 struct nfs4_client *clp;
645 unsigned int idhashval = clientid_hashval(clid->cl_id);
646
647 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 648 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
649 return clp;
650 }
651 return NULL;
652}
653
654static struct nfs4_client *
655find_unconfirmed_client(clientid_t *clid)
656{
657 struct nfs4_client *clp;
658 unsigned int idhashval = clientid_hashval(clid->cl_id);
659
660 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 661 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
662 return clp;
663 }
664 return NULL;
665}
666
28ce6054
N
667static struct nfs4_client *
668find_confirmed_client_by_str(const char *dname, unsigned int hashval)
669{
670 struct nfs4_client *clp;
671
672 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
673 if (same_name(clp->cl_recdir, dname))
674 return clp;
675 }
676 return NULL;
677}
678
679static struct nfs4_client *
680find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
681{
682 struct nfs4_client *clp;
683
684 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
685 if (same_name(clp->cl_recdir, dname))
686 return clp;
687 }
688 return NULL;
689}
690
1da177e4
LT
691/* a helper function for parse_callback */
692static int
693parse_octet(unsigned int *lenp, char **addrp)
694{
695 unsigned int len = *lenp;
696 char *p = *addrp;
697 int n = -1;
698 char c;
699
700 for (;;) {
701 if (!len)
702 break;
703 len--;
704 c = *p++;
705 if (c == '.')
706 break;
707 if ((c < '0') || (c > '9')) {
708 n = -1;
709 break;
710 }
711 if (n < 0)
712 n = 0;
713 n = (n * 10) + (c - '0');
714 if (n > 255) {
715 n = -1;
716 break;
717 }
718 }
719 *lenp = len;
720 *addrp = p;
721 return n;
722}
723
724/* parse and set the setclientid ipv4 callback address */
fd39ca9a 725static int
1da177e4
LT
726parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
727{
728 int temp = 0;
729 u32 cbaddr = 0;
730 u16 cbport = 0;
731 u32 addrlen = addr_len;
732 char *addr = addr_val;
733 int i, shift;
734
735 /* ipaddress */
736 shift = 24;
737 for(i = 4; i > 0 ; i--) {
738 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
739 return 0;
740 }
741 cbaddr |= (temp << shift);
742 if (shift > 0)
743 shift -= 8;
744 }
745 *cbaddrp = cbaddr;
746
747 /* port */
748 shift = 8;
749 for(i = 2; i > 0 ; i--) {
750 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
751 return 0;
752 }
753 cbport |= (temp << shift);
754 if (shift > 0)
755 shift -= 8;
756 }
757 *cbportp = cbport;
758 return 1;
759}
760
fd39ca9a 761static void
1da177e4
LT
762gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
763{
764 struct nfs4_callback *cb = &clp->cl_callback;
765
766 /* Currently, we only support tcp for the callback channel */
767 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
768 goto out_err;
769
770 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
771 &cb->cb_addr, &cb->cb_port)))
772 goto out_err;
773 cb->cb_prog = se->se_callback_prog;
774 cb->cb_ident = se->se_callback_ident;
1da177e4
LT
775 return;
776out_err:
849823c5 777 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT
778 "will not receive delegations\n",
779 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
780
1da177e4
LT
781 return;
782}
783
b37ad28b 784__be32
b591480b
BF
785nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
786 struct nfsd4_setclientid *setclid)
1da177e4 787{
27459f09 788 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT
789 struct xdr_netobj clname = {
790 .len = setclid->se_namelen,
791 .data = setclid->se_name,
792 };
793 nfs4_verifier clverifier = setclid->se_verf;
794 unsigned int strhashval;
28ce6054 795 struct nfs4_client *conf, *unconf, *new;
b37ad28b 796 __be32 status;
68e76ad0 797 char *princ;
a55370a3 798 char dname[HEXDIR_LEN];
1da177e4 799
1da177e4 800 if (!check_name(clname))
73aea4ec 801 return nfserr_inval;
1da177e4 802
a55370a3
N
803 status = nfs4_make_rec_clidname(dname, &clname);
804 if (status)
73aea4ec 805 return status;
a55370a3 806
1da177e4
LT
807 /*
808 * XXX The Duplicate Request Cache (DRC) has been checked (??)
809 * We get here on a DRC miss.
810 */
811
a55370a3 812 strhashval = clientstr_hashval(dname);
1da177e4 813
1da177e4 814 nfs4_lock_state();
28ce6054
N
815 conf = find_confirmed_client_by_str(dname, strhashval);
816 if (conf) {
a186e767 817 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 818 status = nfserr_clid_inuse;
026722c2
BF
819 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
820 dprintk("NFSD: setclientid: string in use by client"
821 " at %pI4\n", &conf->cl_addr);
1da177e4
LT
822 goto out;
823 }
1da177e4 824 }
a186e767
BF
825 /*
826 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
827 * has a description of SETCLIENTID request processing consisting
828 * of 5 bullet points, labeled as CASE0 - CASE4 below.
829 */
28ce6054 830 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1da177e4
LT
831 status = nfserr_resource;
832 if (!conf) {
a186e767
BF
833 /*
834 * RFC 3530 14.2.33 CASE 4:
835 * placed first, because it is the normal case
1da177e4
LT
836 */
837 if (unconf)
838 expire_client(unconf);
a55370a3
N
839 new = create_client(clname, dname);
840 if (new == NULL)
1da177e4 841 goto out;
1da177e4 842 gen_clid(new);
599e0a22 843 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 844 /*
a186e767
BF
845 * RFC 3530 14.2.33 CASE 1:
846 * probable callback update
1da177e4 847 */
31f4a6c1
N
848 if (unconf) {
849 /* Note this is removing unconfirmed {*x***},
850 * which is stronger than RFC recommended {vxc**}.
851 * This has the advantage that there is at most
852 * one {*x***} in either list at any time.
853 */
854 expire_client(unconf);
1da177e4 855 }
a55370a3
N
856 new = create_client(clname, dname);
857 if (new == NULL)
1da177e4 858 goto out;
1da177e4 859 copy_clid(new, conf);
1da177e4
LT
860 } else if (!unconf) {
861 /*
a186e767
BF
862 * RFC 3530 14.2.33 CASE 2:
863 * probable client reboot; state will be removed if
864 * confirmed.
1da177e4 865 */
a55370a3
N
866 new = create_client(clname, dname);
867 if (new == NULL)
1da177e4 868 goto out;
1da177e4 869 gen_clid(new);
49ba8781 870 } else {
a186e767
BF
871 /*
872 * RFC 3530 14.2.33 CASE 3:
873 * probable client reboot; state will be removed if
874 * confirmed.
1da177e4
LT
875 */
876 expire_client(unconf);
a55370a3
N
877 new = create_client(clname, dname);
878 if (new == NULL)
1da177e4 879 goto out;
1da177e4 880 gen_clid(new);
1da177e4 881 }
c175b83c
BF
882 copy_verf(new, &clverifier);
883 new->cl_addr = sin->sin_addr.s_addr;
61054b14 884 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK
885 princ = svc_gss_principal(rqstp);
886 if (princ) {
887 new->cl_principal = kstrdup(princ, GFP_KERNEL);
888 if (new->cl_principal == NULL) {
889 free_client(new);
890 goto out;
891 }
892 }
c175b83c
BF
893 copy_cred(&new->cl_cred, &rqstp->rq_cred);
894 gen_confirm(new);
895 gen_callback(new, setclid);
896 add_to_unconfirmed(new, strhashval);
1da177e4
LT
897 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
898 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
899 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
900 status = nfs_ok;
901out:
902 nfs4_unlock_state();
903 return status;
904}
905
906
907/*
a186e767
BF
908 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
909 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
910 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 911 */
b37ad28b 912__be32
b591480b
BF
913nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
914 struct nfsd4_compound_state *cstate,
915 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 916{
27459f09 917 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 918 struct nfs4_client *conf, *unconf;
1da177e4
LT
919 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
920 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 921 __be32 status;
1da177e4
LT
922
923 if (STALE_CLIENTID(clid))
924 return nfserr_stale_clientid;
925 /*
926 * XXX The Duplicate Request Cache (DRC) has been checked (??)
927 * We get here on a DRC miss.
928 */
929
930 nfs4_lock_state();
21ab45a4
N
931
932 conf = find_confirmed_client(clid);
933 unconf = find_unconfirmed_client(clid);
934
935 status = nfserr_clid_inuse;
27459f09 936 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 937 goto out;
27459f09 938 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N
939 goto out;
940
a186e767
BF
941 /*
942 * section 14.2.34 of RFC 3530 has a description of
943 * SETCLIENTID_CONFIRM request processing consisting
944 * of 4 bullet points, labeled as CASE1 - CASE4 below.
945 */
366e0c1d 946 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF
947 /*
948 * RFC 3530 14.2.34 CASE 1:
949 * callback update
950 */
599e0a22 951 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT
952 status = nfserr_clid_inuse;
953 else {
1a69c179
N
954 /* XXX: We just turn off callbacks until we can handle
955 * change request correctly. */
cb36d634 956 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 957 gen_confirm(conf);
46309029 958 nfsd4_remove_clid_dir(unconf);
1a69c179 959 expire_client(unconf);
1da177e4 960 status = nfs_ok;
1a69c179 961
1da177e4 962 }
f3aba4e5 963 } else if (conf && !unconf) {
a186e767
BF
964 /*
965 * RFC 3530 14.2.34 CASE 2:
966 * probable retransmitted request; play it safe and
967 * do nothing.
7c79f737 968 */
599e0a22 969 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 970 status = nfserr_clid_inuse;
21ab45a4 971 else
1da177e4 972 status = nfs_ok;
7c79f737 973 } else if (!conf && unconf
599e0a22 974 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF
975 /*
976 * RFC 3530 14.2.34 CASE 3:
977 * Normal case; new or rebooted client:
7c79f737 978 */
599e0a22 979 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT
980 status = nfserr_clid_inuse;
981 } else {
1a69c179
N
982 unsigned int hash =
983 clientstr_hashval(unconf->cl_recdir);
984 conf = find_confirmed_client_by_str(unconf->cl_recdir,
985 hash);
986 if (conf) {
c7b9a459 987 nfsd4_remove_clid_dir(conf);
1a69c179
N
988 expire_client(conf);
989 }
1da177e4 990 move_to_confirmed(unconf);
21ab45a4 991 conf = unconf;
46f8a64b 992 nfsd4_probe_callback(conf);
1a69c179 993 status = nfs_ok;
1da177e4 994 }
599e0a22
BF
995 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
996 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 997 &confirm)))) {
a186e767
BF
998 /*
999 * RFC 3530 14.2.34 CASE 4:
1000 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1001 */
1da177e4 1002 status = nfserr_stale_clientid;
7c79f737 1003 } else {
08e8987c
N
1004 /* check that we have hit one of the cases...*/
1005 status = nfserr_clid_inuse;
1006 }
1da177e4 1007out:
1da177e4
LT
1008 nfs4_unlock_state();
1009 return status;
1010}
1011
1da177e4
LT
1012/* OPEN Share state helper functions */
1013static inline struct nfs4_file *
1014alloc_init_file(struct inode *ino)
1015{
1016 struct nfs4_file *fp;
1017 unsigned int hashval = file_hashval(ino);
1018
e60d4398
N
1019 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1020 if (fp) {
8b671b80 1021 atomic_set(&fp->fi_ref, 1);
1da177e4 1022 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N
1023 INIT_LIST_HEAD(&fp->fi_stateids);
1024 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1025 spin_lock(&recall_lock);
1da177e4 1026 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1027 spin_unlock(&recall_lock);
1da177e4
LT
1028 fp->fi_inode = igrab(ino);
1029 fp->fi_id = current_fileid++;
47f9940c 1030 fp->fi_had_conflict = false;
1da177e4
LT
1031 return fp;
1032 }
1033 return NULL;
1034}
1035
e60d4398 1036static void
e18b890b 1037nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1038{
e60d4398
N
1039 if (*slab == NULL)
1040 return;
1a1d92c1 1041 kmem_cache_destroy(*slab);
e60d4398 1042 *slab = NULL;
1da177e4
LT
1043}
1044
e8ff2a84 1045void
1da177e4
LT
1046nfsd4_free_slabs(void)
1047{
e60d4398
N
1048 nfsd4_free_slab(&stateowner_slab);
1049 nfsd4_free_slab(&file_slab);
5ac049ac 1050 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1051 nfsd4_free_slab(&deleg_slab);
e60d4398 1052}
1da177e4 1053
e60d4398
N
1054static int
1055nfsd4_init_slabs(void)
1056{
1057 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1058 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N
1059 if (stateowner_slab == NULL)
1060 goto out_nomem;
1061 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1062 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N
1063 if (file_slab == NULL)
1064 goto out_nomem;
5ac049ac 1065 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1066 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N
1067 if (stateid_slab == NULL)
1068 goto out_nomem;
5b2d21c1 1069 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1070 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N
1071 if (deleg_slab == NULL)
1072 goto out_nomem;
e60d4398
N
1073 return 0;
1074out_nomem:
1075 nfsd4_free_slabs();
1076 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1077 return -ENOMEM;
1da177e4
LT
1078}
1079
1080void
1081nfs4_free_stateowner(struct kref *kref)
1082{
1083 struct nfs4_stateowner *sop =
1084 container_of(kref, struct nfs4_stateowner, so_ref);
1085 kfree(sop->so_owner.data);
1086 kmem_cache_free(stateowner_slab, sop);
1087}
1088
1089static inline struct nfs4_stateowner *
1090alloc_stateowner(struct xdr_netobj *owner)
1091{
1092 struct nfs4_stateowner *sop;
1093
1094 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1095 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1096 memcpy(sop->so_owner.data, owner->data, owner->len);
1097 sop->so_owner.len = owner->len;
1098 kref_init(&sop->so_ref);
1099 return sop;
1100 }
1101 kmem_cache_free(stateowner_slab, sop);
1102 }
1103 return NULL;
1104}
1105
1106static struct nfs4_stateowner *
1107alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1108 struct nfs4_stateowner *sop;
1109 struct nfs4_replay *rp;
1110 unsigned int idhashval;
1111
1112 if (!(sop = alloc_stateowner(&open->op_owner)))
1113 return NULL;
1114 idhashval = ownerid_hashval(current_ownerid);
1115 INIT_LIST_HEAD(&sop->so_idhash);
1116 INIT_LIST_HEAD(&sop->so_strhash);
1117 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
1118 INIT_LIST_HEAD(&sop->so_stateids);
1119 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT
1120 INIT_LIST_HEAD(&sop->so_close_lru);
1121 sop->so_time = 0;
1122 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1123 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1124 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT
1125 sop->so_is_open_owner = 1;
1126 sop->so_id = current_ownerid++;
1127 sop->so_client = clp;
1128 sop->so_seqid = open->op_seqid;
1129 sop->so_confirmed = 0;
1130 rp = &sop->so_replay;
de1ae286 1131 rp->rp_status = nfserr_serverfault;
1da177e4
LT
1132 rp->rp_buflen = 0;
1133 rp->rp_buf = rp->rp_ibuf;
1134 return sop;
1135}
1136
1da177e4
LT
1137static inline void
1138init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1139 struct nfs4_stateowner *sop = open->op_stateowner;
1140 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1141
1142 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N
1143 INIT_LIST_HEAD(&stp->st_perstateowner);
1144 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT
1145 INIT_LIST_HEAD(&stp->st_perfile);
1146 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1147 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1148 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1149 stp->st_stateowner = sop;
13cd2184 1150 get_nfs4_file(fp);
1da177e4
LT
1151 stp->st_file = fp;
1152 stp->st_stateid.si_boot = boot_time;
1153 stp->st_stateid.si_stateownerid = sop->so_id;
1154 stp->st_stateid.si_fileid = fp->fi_id;
1155 stp->st_stateid.si_generation = 0;
1156 stp->st_access_bmap = 0;
1157 stp->st_deny_bmap = 0;
1158 __set_bit(open->op_share_access, &stp->st_access_bmap);
1159 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1160 stp->st_openstp = NULL;
1da177e4
LT
1161}
1162
fd39ca9a 1163static void
1da177e4
LT
1164move_to_close_lru(struct nfs4_stateowner *sop)
1165{
1166 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1167
358dd55a 1168 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT
1169 sop->so_time = get_seconds();
1170}
1171
1da177e4 1172static int
599e0a22
BF
1173same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1174 clientid_t *clid)
1175{
1176 return (sop->so_owner.len == owner->len) &&
1177 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1178 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT
1179}
1180
1181static struct nfs4_stateowner *
1182find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1183{
1184 struct nfs4_stateowner *so = NULL;
1185
1186 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1187 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT
1188 return so;
1189 }
1190 return NULL;
1191}
1192
1193/* search file_hashtbl[] for file */
1194static struct nfs4_file *
1195find_file(struct inode *ino)
1196{
1197 unsigned int hashval = file_hashval(ino);
1198 struct nfs4_file *fp;
1199
8b671b80 1200 spin_lock(&recall_lock);
1da177e4 1201 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N
1202 if (fp->fi_inode == ino) {
1203 get_nfs4_file(fp);
8b671b80 1204 spin_unlock(&recall_lock);
1da177e4 1205 return fp;
13cd2184 1206 }
1da177e4 1207 }
8b671b80 1208 spin_unlock(&recall_lock);
1da177e4
LT
1209 return NULL;
1210}
1211
8838dc43 1212static inline int access_valid(u32 x)
ba5a6a19 1213{
8838dc43
BF
1214 if (x < NFS4_SHARE_ACCESS_READ)
1215 return 0;
1216 if (x > NFS4_SHARE_ACCESS_BOTH)
1217 return 0;
1218 return 1;
ba5a6a19
BF
1219}
1220
8838dc43 1221static inline int deny_valid(u32 x)
ba5a6a19 1222{
8838dc43
BF
1223 /* Note: unlike access bits, deny bits may be zero. */
1224 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1225}
1da177e4 1226
4f83aa30
BF
1227/*
1228 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1229 * st_{access,deny}_bmap field of the stateid, in order to track not
1230 * only what share bits are currently in force, but also what
1231 * combinations of share bits previous opens have used. This allows us
1232 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1233 * return an error if the client attempt to downgrade to a combination
1234 * of share bits not explicable by closing some of its previous opens.
1235 *
1236 * XXX: This enforcement is actually incomplete, since we don't keep
1237 * track of access/deny bit combinations; so, e.g., we allow:
1238 *
1239 * OPEN allow read, deny write
1240 * OPEN allow both, deny none
1241 * DOWNGRADE allow read, deny none
1242 *
1243 * which we should reject.
1244 */
fd39ca9a 1245static void
1da177e4
LT
1246set_access(unsigned int *access, unsigned long bmap) {
1247 int i;
1248
1249 *access = 0;
1250 for (i = 1; i < 4; i++) {
1251 if (test_bit(i, &bmap))
1252 *access |= i;
1253 }
1254}
1255
fd39ca9a 1256static void
1da177e4
LT
1257set_deny(unsigned int *deny, unsigned long bmap) {
1258 int i;
1259
1260 *deny = 0;
1261 for (i = 0; i < 4; i++) {
1262 if (test_bit(i, &bmap))
1263 *deny |= i ;
1264 }
1265}
1266
1267static int
1268test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1269 unsigned int access, deny;
1270
1271 set_access(&access, stp->st_access_bmap);
1272 set_deny(&deny, stp->st_deny_bmap);
1273 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1274 return 0;
1275 return 1;
1276}
1277
1278/*
1279 * Called to check deny when READ with all zero stateid or
1280 * WRITE with all zero or all one stateid
1281 */
b37ad28b 1282static __be32
1da177e4
LT
1283nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1284{
1285 struct inode *ino = current_fh->fh_dentry->d_inode;
1286 struct nfs4_file *fp;
1287 struct nfs4_stateid *stp;
b37ad28b 1288 __be32 ret;
1da177e4
LT
1289
1290 dprintk("NFSD: nfs4_share_conflict\n");
1291
1292 fp = find_file(ino);
13cd2184
N
1293 if (!fp)
1294 return nfs_ok;
b700949b 1295 ret = nfserr_locked;
1da177e4 1296 /* Search for conflicting share reservations */
13cd2184
N
1297 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1298 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1299 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1300 goto out;
1da177e4 1301 }
13cd2184
N
1302 ret = nfs_ok;
1303out:
1304 put_nfs4_file(fp);
1305 return ret;
1da177e4
LT
1306}
1307
1308static inline void
1309nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1310{
1311 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 1312 drop_file_write_access(filp);
1da177e4
LT
1313 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1314 }
1315}
1316
1317/*
1318 * Recall a delegation
1319 */
1320static int
1321do_recall(void *__dp)
1322{
1323 struct nfs4_delegation *dp = __dp;
1324
47f9940c 1325 dp->dl_file->fi_had_conflict = true;
1da177e4
LT
1326 nfsd4_cb_recall(dp);
1327 return 0;
1328}
1329
1330/*
1331 * Spawn a thread to perform a recall on the delegation represented
1332 * by the lease (file_lock)
1333 *
1334 * Called from break_lease() with lock_kernel() held.
1335 * Note: we assume break_lease will only call this *once* for any given
1336 * lease.
1337 */
1338static
1339void nfsd_break_deleg_cb(struct file_lock *fl)
1340{
1341 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1342 struct task_struct *t;
1343
1344 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1345 if (!dp)
1346 return;
1347
1348 /* We're assuming the state code never drops its reference
1349 * without first removing the lease. Since we're in this lease
1350 * callback (and since the lease code is serialized by the kernel
1351 * lock) we know the server hasn't removed the lease yet, we know
1352 * it's safe to take a reference: */
1353 atomic_inc(&dp->dl_count);
cfdcad4d 1354 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT
1355
1356 spin_lock(&recall_lock);
1357 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1358 spin_unlock(&recall_lock);
1359
1360 /* only place dl_time is set. protected by lock_kernel*/
1361 dp->dl_time = get_seconds();
1362
0272e1fd
BF
1363 /*
1364 * We don't want the locks code to timeout the lease for us;
1365 * we'll remove it ourself if the delegation isn't returned
1366 * in time.
1367 */
1368 fl->fl_break_time = 0;
1da177e4
LT
1369
1370 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1371 if (IS_ERR(t)) {
1372 struct nfs4_client *clp = dp->dl_client;
1373
1374 printk(KERN_INFO "NFSD: Callback thread failed for "
1375 "for client (clientid %08x/%08x)\n",
1376 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 1377 put_nfs4_client(dp->dl_client);
1da177e4
LT
1378 nfs4_put_delegation(dp);
1379 }
1380}
1381
1382/*
1383 * The file_lock is being reapd.
1384 *
1385 * Called by locks_free_lock() with lock_kernel() held.
1386 */
1387static
1388void nfsd_release_deleg_cb(struct file_lock *fl)
1389{
1390 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1391
1392 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1393
1394 if (!(fl->fl_flags & FL_LEASE) || !dp)
1395 return;
1396 dp->dl_flock = NULL;
1397}
1398
1399/*
1400 * Set the delegation file_lock back pointer.
1401 *
a9933cea 1402 * Called from setlease() with lock_kernel() held.
1da177e4
LT
1403 */
1404static
1405void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1406{
1407 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1408
1409 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1410 if (!dp)
1411 return;
1412 dp->dl_flock = new;
1413}
1414
1415/*
a9933cea 1416 * Called from setlease() with lock_kernel() held
1da177e4
LT
1417 */
1418static
1419int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1420{
1421 struct nfs4_delegation *onlistd =
1422 (struct nfs4_delegation *)onlist->fl_owner;
1423 struct nfs4_delegation *tryd =
1424 (struct nfs4_delegation *)try->fl_owner;
1425
1426 if (onlist->fl_lmops != try->fl_lmops)
1427 return 0;
1428
1429 return onlistd->dl_client == tryd->dl_client;
1430}
1431
1432
1433static
1434int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1435{
1436 if (arg & F_UNLCK)
1437 return lease_modify(onlist, arg);
1438 else
1439 return -EAGAIN;
1440}
1441
fd39ca9a 1442static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT
1443 .fl_break = nfsd_break_deleg_cb,
1444 .fl_release_private = nfsd_release_deleg_cb,
1445 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1446 .fl_mylease = nfsd_same_client_deleg_cb,
1447 .fl_change = nfsd_change_deleg_cb,
1448};
1449
1450
b37ad28b 1451__be32
1da177e4
LT
1452nfsd4_process_open1(struct nfsd4_open *open)
1453{
1da177e4
LT
1454 clientid_t *clientid = &open->op_clientid;
1455 struct nfs4_client *clp = NULL;
1456 unsigned int strhashval;
1457 struct nfs4_stateowner *sop = NULL;
1458
1da177e4 1459 if (!check_name(open->op_owner))
0f442aa2 1460 return nfserr_inval;
1da177e4
LT
1461
1462 if (STALE_CLIENTID(&open->op_clientid))
1463 return nfserr_stale_clientid;
1464
1465 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1466 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF
1467 open->op_stateowner = sop;
1468 if (!sop) {
1469 /* Make sure the client's lease hasn't expired. */
1da177e4
LT
1470 clp = find_confirmed_client(clientid);
1471 if (clp == NULL)
0f442aa2
BF
1472 return nfserr_expired;
1473 goto renew;
1da177e4 1474 }
0f442aa2
BF
1475 if (!sop->so_confirmed) {
1476 /* Replace unconfirmed owners without checking for replay. */
1477 clp = sop->so_client;
f044ff83 1478 release_openowner(sop);
0f442aa2
BF
1479 open->op_stateowner = NULL;
1480 goto renew;
1481 }
1482 if (open->op_seqid == sop->so_seqid - 1) {
1483 if (sop->so_replay.rp_buflen)
a90b061c 1484 return nfserr_replay_me;
0f442aa2
BF
1485 /* The original OPEN failed so spectacularly
1486 * that we don't even have replay data saved!
1487 * Therefore, we have no choice but to continue
1488 * processing this OPEN; presumably, we'll
1489 * fail again for the same reason.
1490 */
1491 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1492 goto renew;
1493 }
1494 if (open->op_seqid != sop->so_seqid)
1495 return nfserr_bad_seqid;
1da177e4 1496renew:
0f442aa2
BF
1497 if (open->op_stateowner == NULL) {
1498 sop = alloc_init_open_stateowner(strhashval, clp, open);
1499 if (sop == NULL)
1500 return nfserr_resource;
1501 open->op_stateowner = sop;
1502 }
1503 list_del_init(&sop->so_close_lru);
1da177e4 1504 renew_client(sop->so_client);
0f442aa2 1505 return nfs_ok;
1da177e4
LT
1506}
1507
b37ad28b 1508static inline __be32
4a6e43e6
N
1509nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1510{
1511 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1512 return nfserr_openmode;
1513 else
1514 return nfs_ok;
1515}
1516
52f4fb43
N
1517static struct nfs4_delegation *
1518find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1519{
1520 struct nfs4_delegation *dp;
1521
ea1da636 1522 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N
1523 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1524 return dp;
1525 }
1526 return NULL;
1527}
1528
b37ad28b 1529static __be32
567d9829
N
1530nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1531 struct nfs4_delegation **dp)
1532{
1533 int flags;
b37ad28b 1534 __be32 status = nfserr_bad_stateid;
567d9829
N
1535
1536 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1537 if (*dp == NULL)
c44c5eeb 1538 goto out;
567d9829
N
1539 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1540 RD_STATE : WR_STATE;
1541 status = nfs4_check_delegmode(*dp, flags);
1542 if (status)
1543 *dp = NULL;
c44c5eeb
N
1544out:
1545 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1546 return nfs_ok;
1547 if (status)
1548 return status;
1549 open->op_stateowner->so_confirmed = 1;
1550 return nfs_ok;
567d9829
N
1551}
1552
b37ad28b 1553static __be32
1da177e4
LT
1554nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1555{
1556 struct nfs4_stateid *local;
b37ad28b 1557 __be32 status = nfserr_share_denied;
1da177e4
LT
1558 struct nfs4_stateowner *sop = open->op_stateowner;
1559
8beefa24 1560 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT
1561 /* ignore lock owners */
1562 if (local->st_stateowner->so_is_open_owner == 0)
1563 continue;
1564 /* remember if we have seen this open owner */
1565 if (local->st_stateowner == sop)
1566 *stpp = local;
1567 /* check for conflicting share reservations */
1568 if (!test_share(local, open))
1569 goto out;
1570 }
1571 status = 0;
1572out:
1573 return status;
1574}
1575
5ac049ac
N
1576static inline struct nfs4_stateid *
1577nfs4_alloc_stateid(void)
1578{
1579 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1580}
1581
b37ad28b 1582static __be32
1da177e4 1583nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 1584 struct nfs4_delegation *dp,
1da177e4
LT
1585 struct svc_fh *cur_fh, int flags)
1586{
1587 struct nfs4_stateid *stp;
1da177e4 1588
5ac049ac 1589 stp = nfs4_alloc_stateid();
1da177e4
LT
1590 if (stp == NULL)
1591 return nfserr_resource;
1592
567d9829
N
1593 if (dp) {
1594 get_file(dp->dl_vfs_file);
1595 stp->st_vfs_file = dp->dl_vfs_file;
1596 } else {
b37ad28b 1597 __be32 status;
567d9829
N
1598 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1599 &stp->st_vfs_file);
1600 if (status) {
1601 if (status == nfserr_dropit)
1602 status = nfserr_jukebox;
5ac049ac 1603 kmem_cache_free(stateid_slab, stp);
567d9829
N
1604 return status;
1605 }
1da177e4 1606 }
1da177e4
LT
1607 *stpp = stp;
1608 return 0;
1609}
1610
b37ad28b 1611static inline __be32
1da177e4
LT
1612nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1613 struct nfsd4_open *open)
1614{
1615 struct iattr iattr = {
1616 .ia_valid = ATTR_SIZE,
1617 .ia_size = 0,
1618 };
1619 if (!open->op_truncate)
1620 return 0;
1621 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 1622 return nfserr_inval;
1da177e4
LT
1623 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1624}
1625
b37ad28b 1626static __be32
1da177e4
LT
1627nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1628{
1629 struct file *filp = stp->st_vfs_file;
7eaa36e2 1630 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 1631 unsigned int share_access, new_writer;
b37ad28b 1632 __be32 status;
1da177e4
LT
1633
1634 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF
1635 new_writer = (~share_access) & open->op_share_access
1636 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 1637
6c26d08f 1638 if (new_writer) {
b37ad28b
AV
1639 int err = get_write_access(inode);
1640 if (err)
1641 return nfserrno(err);
e518f056
BH
1642 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
1643 if (err)
1644 return nfserrno(err);
1645 file_take_write(filp);
6c26d08f 1646 }
1da177e4
LT
1647 status = nfsd4_truncate(rqstp, cur_fh, open);
1648 if (status) {
6c26d08f
BF
1649 if (new_writer)
1650 put_write_access(inode);
1da177e4
LT
1651 return status;
1652 }
1653 /* remember the open */
6c26d08f 1654 filp->f_mode |= open->op_share_access;
b55e0ba1
BF
1655 __set_bit(open->op_share_access, &stp->st_access_bmap);
1656 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT
1657
1658 return nfs_ok;
1659}
1660
1661
1da177e4 1662static void
37515177 1663nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 1664{
37515177
N
1665 open->op_stateowner->so_confirmed = 1;
1666 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT
1667}
1668
1669/*
1670 * Attempt to hand out a delegation.
1671 */
1672static void
1673nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1674{
1675 struct nfs4_delegation *dp;
1676 struct nfs4_stateowner *sop = stp->st_stateowner;
1677 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1678 struct file_lock fl, *flp = &fl;
1679 int status, flag = 0;
1680
1681 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N
1682 open->op_recall = 0;
1683 switch (open->op_claim_type) {
1684 case NFS4_OPEN_CLAIM_PREVIOUS:
1685 if (!atomic_read(&cb->cb_set))
1686 open->op_recall = 1;
1687 flag = open->op_delegate_type;
1688 if (flag == NFS4_OPEN_DELEGATE_NONE)
1689 goto out;
1690 break;
1691 case NFS4_OPEN_CLAIM_NULL:
1692 /* Let's not give out any delegations till everyone's
1693 * had the chance to reclaim theirs.... */
af558e33 1694 if (locks_in_grace())
7b190fec
N
1695 goto out;
1696 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1697 goto out;
1698 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1699 flag = NFS4_OPEN_DELEGATE_WRITE;
1700 else
1701 flag = NFS4_OPEN_DELEGATE_READ;
1702 break;
1703 default:
1704 goto out;
1705 }
1da177e4
LT
1706
1707 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1708 if (dp == NULL) {
1709 flag = NFS4_OPEN_DELEGATE_NONE;
1710 goto out;
1711 }
1712 locks_init_lock(&fl);
1713 fl.fl_lmops = &nfsd_lease_mng_ops;
1714 fl.fl_flags = FL_LEASE;
9167f501 1715 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT
1716 fl.fl_end = OFFSET_MAX;
1717 fl.fl_owner = (fl_owner_t)dp;
1718 fl.fl_file = stp->st_vfs_file;
1719 fl.fl_pid = current->tgid;
1720
a9933cea 1721 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT
1722 * the lock_manager callbacks fl_mylease and fl_change are used
1723 */
9167f501 1724 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 1725 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 1726 unhash_delegation(dp);
1da177e4
LT
1727 flag = NFS4_OPEN_DELEGATE_NONE;
1728 goto out;
1729 }
1730
1731 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1732
1733 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1734 dp->dl_stateid.si_boot,
1735 dp->dl_stateid.si_stateownerid,
1736 dp->dl_stateid.si_fileid,
1737 dp->dl_stateid.si_generation);
1738out:
7b190fec
N
1739 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1740 && flag == NFS4_OPEN_DELEGATE_NONE
1741 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 1742 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT
1743 open->op_delegate_type = flag;
1744}
1745
1746/*
1747 * called with nfs4_lock_state() held.
1748 */
b37ad28b 1749__be32
1da177e4
LT
1750nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1751{
1752 struct nfs4_file *fp = NULL;
1753 struct inode *ino = current_fh->fh_dentry->d_inode;
1754 struct nfs4_stateid *stp = NULL;
567d9829 1755 struct nfs4_delegation *dp = NULL;
b37ad28b 1756 __be32 status;
1da177e4
LT
1757
1758 status = nfserr_inval;
ba5a6a19
BF
1759 if (!access_valid(open->op_share_access)
1760 || !deny_valid(open->op_share_deny))
1da177e4
LT
1761 goto out;
1762 /*
1763 * Lookup file; if found, lookup stateid and check open request,
1764 * and check for delegations in the process of being recalled.
1765 * If not found, create the nfs4_file struct
1766 */
1767 fp = find_file(ino);
1768 if (fp) {
1769 if ((status = nfs4_check_open(fp, open, &stp)))
1770 goto out;
c44c5eeb
N
1771 status = nfs4_check_deleg(fp, open, &dp);
1772 if (status)
1773 goto out;
1da177e4 1774 } else {
c44c5eeb
N
1775 status = nfserr_bad_stateid;
1776 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1777 goto out;
1da177e4
LT
1778 status = nfserr_resource;
1779 fp = alloc_init_file(ino);
1780 if (fp == NULL)
1781 goto out;
1782 }
1783
1784 /*
1785 * OPEN the file, or upgrade an existing OPEN.
1786 * If truncate fails, the OPEN fails.
1787 */
1788 if (stp) {
1789 /* Stateid was found, this is an OPEN upgrade */
1790 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1791 if (status)
1792 goto out;
444c2c07 1793 update_stateid(&stp->st_stateid);
1da177e4
LT
1794 } else {
1795 /* Stateid was not found, this is a new OPEN */
1796 int flags = 0;
9ecb6a08 1797 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 1798 flags |= NFSD_MAY_READ;
1da177e4 1799 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 1800 flags |= NFSD_MAY_WRITE;
567d9829
N
1801 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1802 if (status)
1da177e4
LT
1803 goto out;
1804 init_stateid(stp, fp, open);
1805 status = nfsd4_truncate(rqstp, current_fh, open);
1806 if (status) {
2283963f 1807 release_open_stateid(stp);
1da177e4
LT
1808 goto out;
1809 }
1810 }
1811 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1812
1813 /*
1814 * Attempt to hand out a delegation. No error return, because the
1815 * OPEN succeeds even if we fail.
1816 */
1817 nfs4_open_delegation(current_fh, open, stp);
1818
1819 status = nfs_ok;
1820
1821 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1822 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1823 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1824out:
13cd2184
N
1825 if (fp)
1826 put_nfs4_file(fp);
37515177
N
1827 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1828 nfs4_set_claim_prev(open);
1da177e4
LT
1829 /*
1830 * To finish the open response, we just need to set the rflags.
1831 */
1832 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1833 if (!open->op_stateowner->so_confirmed)
1834 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1835
1836 return status;
1837}
1838
b37ad28b 1839__be32
b591480b
BF
1840nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1841 clientid_t *clid)
1da177e4
LT
1842{
1843 struct nfs4_client *clp;
b37ad28b 1844 __be32 status;
1da177e4
LT
1845
1846 nfs4_lock_state();
1847 dprintk("process_renew(%08x/%08x): starting\n",
1848 clid->cl_boot, clid->cl_id);
1849 status = nfserr_stale_clientid;
1850 if (STALE_CLIENTID(clid))
1851 goto out;
1852 clp = find_confirmed_client(clid);
1853 status = nfserr_expired;
1854 if (clp == NULL) {
1855 /* We assume the client took too long to RENEW. */
1856 dprintk("nfsd4_renew: clientid not found!\n");
1857 goto out;
1858 }
1859 renew_client(clp);
1860 status = nfserr_cb_path_down;
ea1da636 1861 if (!list_empty(&clp->cl_delegations)
1da177e4
LT
1862 && !atomic_read(&clp->cl_callback.cb_set))
1863 goto out;
1864 status = nfs_ok;
1865out:
1866 nfs4_unlock_state();
1867 return status;
1868}
1869
af558e33
BF
1870struct lock_manager nfsd4_manager = {
1871};
1872
a76b4319 1873static void
af558e33 1874nfsd4_end_grace(void)
a76b4319
N
1875{
1876 dprintk("NFSD: end of grace period\n");
c7b9a459 1877 nfsd4_recdir_purge_old();
af558e33 1878 locks_end_grace(&nfsd4_manager);
a76b4319
N
1879}
1880
fd39ca9a 1881static time_t
1da177e4
LT
1882nfs4_laundromat(void)
1883{
1884 struct nfs4_client *clp;
1885 struct nfs4_stateowner *sop;
1886 struct nfs4_delegation *dp;
1887 struct list_head *pos, *next, reaplist;
1888 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1889 time_t t, clientid_val = NFSD_LEASE_TIME;
1890 time_t u, test_val = NFSD_LEASE_TIME;
1891
1892 nfs4_lock_state();
1893
1894 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF
1895 if (locks_in_grace())
1896 nfsd4_end_grace();
1da177e4
LT
1897 list_for_each_safe(pos, next, &client_lru) {
1898 clp = list_entry(pos, struct nfs4_client, cl_lru);
1899 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1900 t = clp->cl_time - cutoff;
1901 if (clientid_val > t)
1902 clientid_val = t;
1903 break;
1904 }
1905 dprintk("NFSD: purging unused client (clientid %08x)\n",
1906 clp->cl_clientid.cl_id);
c7b9a459 1907 nfsd4_remove_clid_dir(clp);
1da177e4
LT
1908 expire_client(clp);
1909 }
1910 INIT_LIST_HEAD(&reaplist);
1911 spin_lock(&recall_lock);
1912 list_for_each_safe(pos, next, &del_recall_lru) {
1913 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1914 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1915 u = dp->dl_time - cutoff;
1916 if (test_val > u)
1917 test_val = u;
1918 break;
1919 }
1920 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1921 dp, dp->dl_flock);
1922 list_move(&dp->dl_recall_lru, &reaplist);
1923 }
1924 spin_unlock(&recall_lock);
1925 list_for_each_safe(pos, next, &reaplist) {
1926 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1927 list_del_init(&dp->dl_recall_lru);
1928 unhash_delegation(dp);
1929 }
1930 test_val = NFSD_LEASE_TIME;
1931 list_for_each_safe(pos, next, &close_lru) {
1932 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1933 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1934 u = sop->so_time - cutoff;
1935 if (test_val > u)
1936 test_val = u;
1937 break;
1938 }
1939 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1940 sop->so_id);
f044ff83 1941 release_openowner(sop);
1da177e4
LT
1942 }
1943 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1944 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1945 nfs4_unlock_state();
1946 return clientid_val;
1947}
1948
a254b246
HH
1949static struct workqueue_struct *laundry_wq;
1950static void laundromat_main(struct work_struct *);
1951static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1952
1953static void
c4028958 1954laundromat_main(struct work_struct *not_used)
1da177e4
LT
1955{
1956 time_t t;
1957
1958 t = nfs4_laundromat();
1959 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 1960 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT
1961}
1962
fd39ca9a 1963static struct nfs4_stateowner *
f8816512
N
1964search_close_lru(u32 st_id, int flags)
1965{
1da177e4
LT
1966 struct nfs4_stateowner *local = NULL;
1967
1da177e4
LT
1968 if (flags & CLOSE_STATE) {
1969 list_for_each_entry(local, &close_lru, so_close_lru) {
1970 if (local->so_id == st_id)
1971 return local;
1972 }
1973 }
1974 return NULL;
1975}
1976
1977static inline int
1978nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1979{
7eaa36e2 1980 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT
1981}
1982
1983static int
1984STALE_STATEID(stateid_t *stateid)
1985{
1986 if (stateid->si_boot == boot_time)
1987 return 0;
849823c5 1988 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT
1989 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1990 stateid->si_generation);
1991 return 1;
1992}
1993
1994static inline int
1995access_permit_read(unsigned long access_bmap)
1996{
1997 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1998 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1999 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2000}
2001
2002static inline int
2003access_permit_write(unsigned long access_bmap)
2004{
2005 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2006 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2007}
2008
2009static
b37ad28b 2010__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2011{
b37ad28b 2012 __be32 status = nfserr_openmode;
1da177e4
LT
2013
2014 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2015 goto out;
2016 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2017 goto out;
2018 status = nfs_ok;
2019out:
2020 return status;
2021}
2022
b37ad28b 2023static inline __be32
1da177e4
LT
2024check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2025{
203a8c8e 2026 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2027 return nfs_ok;
af558e33 2028 else if (locks_in_grace()) {
1da177e4
LT
2029 /* Answer in remaining cases depends on existance of
2030 * conflicting state; so we must wait out the grace period. */
2031 return nfserr_grace;
2032 } else if (flags & WR_STATE)
2033 return nfs4_share_conflict(current_fh,
2034 NFS4_SHARE_DENY_WRITE);
2035 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2036 return nfs4_share_conflict(current_fh,
2037 NFS4_SHARE_DENY_READ);
2038}
2039
2040/*
2041 * Allow READ/WRITE during grace period on recovered state only for files
2042 * that are not able to provide mandatory locking.
2043 */
2044static inline int
18f82731 2045grace_disallows_io(struct inode *inode)
1da177e4 2046{
203a8c8e 2047 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT
2048}
2049
0836f587
BF
2050static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2051{
2052 /* If the client sends us a stateid from the future, it's buggy: */
2053 if (in->si_generation > ref->si_generation)
2054 return nfserr_bad_stateid;
2055 /*
2056 * The following, however, can happen. For example, if the
2057 * client sends an open and some IO at the same time, the open
2058 * may bump si_generation while the IO is still in flight.
2059 * Thanks to hard links and renames, the client never knows what
2060 * file an open will affect. So it could avoid that situation
2061 * only by serializing all opens and IO from the same open
2062 * owner. To recover from the old_stateid error, the client
2063 * will just have to retry the IO:
2064 */
2065 if (in->si_generation < ref->si_generation)
2066 return nfserr_old_stateid;
2067 return nfs_ok;
2068}
2069
3e633079
BF
2070static int is_delegation_stateid(stateid_t *stateid)
2071{
2072 return stateid->si_fileid == 0;
2073}
2074
1da177e4
LT
2075/*
2076* Checks for stateid operations
2077*/
b37ad28b 2078__be32
1da177e4
LT
2079nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2080{
2081 struct nfs4_stateid *stp = NULL;
2082 struct nfs4_delegation *dp = NULL;
1da177e4 2083 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2084 __be32 status;
1da177e4 2085
1da177e4
LT
2086 if (filpp)
2087 *filpp = NULL;
2088
18f82731 2089 if (grace_disallows_io(ino))
1da177e4
LT
2090 return nfserr_grace;
2091
2092 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2093 return check_special_stateids(current_fh, stateid, flags);
2094
1da177e4
LT
2095 status = nfserr_stale_stateid;
2096 if (STALE_STATEID(stateid))
2097 goto out;
2098
1da177e4 2099 status = nfserr_bad_stateid;
3e633079 2100 if (is_delegation_stateid(stateid)) {
a4455be0 2101 dp = find_delegation_stateid(ino, stateid);
819a8f53 2102 if (!dp)
1da177e4 2103 goto out;
fd03b099 2104 status = check_stateid_generation(stateid, &dp->dl_stateid);
0c2a498f
BF
2105 if (status)
2106 goto out;
dc9bf700
BF
2107 status = nfs4_check_delegmode(dp, flags);
2108 if (status)
2109 goto out;
2110 renew_client(dp->dl_client);
dc9bf700
BF
2111 if (filpp)
2112 *filpp = dp->dl_vfs_file;
1da177e4 2113 } else { /* open or lock stateid */
a4455be0 2114 stp = find_stateid(stateid, flags);
819a8f53 2115 if (!stp)
1da177e4 2116 goto out;
6150ef0d 2117 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT
2118 goto out;
2119 if (!stp->st_stateowner->so_confirmed)
2120 goto out;
fd03b099 2121 status = check_stateid_generation(stateid, &stp->st_stateid);
0c2a498f
BF
2122 if (status)
2123 goto out;
a4455be0
BF
2124 status = nfs4_check_openmode(stp, flags);
2125 if (status)
1da177e4
LT
2126 goto out;
2127 renew_client(stp->st_stateowner->so_client);
2128 if (filpp)
2129 *filpp = stp->st_vfs_file;
1da177e4
LT
2130 }
2131 status = nfs_ok;
2132out:
2133 return status;
2134}
2135
4c4cd222
N
2136static inline int
2137setlkflg (int type)
2138{
2139 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2140 RD_STATE : WR_STATE;
2141}
1da177e4
LT
2142
2143/*
2144 * Checks for sequence id mutating operations.
2145 */
b37ad28b 2146static __be32
4c4cd222 2147nfs4_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 2148{
1da177e4
LT
2149 struct nfs4_stateid *stp;
2150 struct nfs4_stateowner *sop;
0836f587 2151 __be32 status;
1da177e4
LT
2152
2153 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2154 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2155 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2156 stateid->si_generation);
3a4f98bb 2157
1da177e4
LT
2158 *stpp = NULL;
2159 *sopp = NULL;
2160
1da177e4 2161 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2162 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2163 return nfserr_bad_stateid;
1da177e4
LT
2164 }
2165
1da177e4 2166 if (STALE_STATEID(stateid))
3a4f98bb 2167 return nfserr_stale_stateid;
1da177e4
LT
2168 /*
2169 * We return BAD_STATEID if filehandle doesn't match stateid,
2170 * the confirmed flag is incorrecly set, or the generation
2171 * number is incorrect.
1da177e4 2172 */
f8816512
N
2173 stp = find_stateid(stateid, flags);
2174 if (stp == NULL) {
2175 /*
2176 * Also, we should make sure this isn't just the result of
2177 * a replayed close:
2178 */
2179 sop = search_close_lru(stateid->si_stateownerid, flags);
2180 if (sop == NULL)
2181 return nfserr_bad_stateid;
2182 *sopp = sop;
2183 goto check_replay;
2184 }
1da177e4 2185
39325bd0
BF
2186 *stpp = stp;
2187 *sopp = sop = stp->st_stateowner;
2188
4c4cd222 2189 if (lock) {
4c4cd222 2190 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2191 struct nfs4_client *clp = sop->so_client;
4c4cd222 2192 int lkflg = 0;
b37ad28b 2193 __be32 status;
4c4cd222
N
2194
2195 lkflg = setlkflg(lock->lk_type);
2196
2197 if (lock->lk_is_new) {
599e0a22
BF
2198 if (!sop->so_is_open_owner)
2199 return nfserr_bad_stateid;
2200 if (!same_clid(&clp->cl_clientid, lockclid))
4c4cd222 2201 return nfserr_bad_stateid;
599e0a22
BF
2202 /* stp is the open stateid */
2203 status = nfs4_check_openmode(stp, lkflg);
2204 if (status)
2205 return status;
2206 } else {
2207 /* stp is the lock stateid */
2208 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2209 if (status)
2210 return status;
4c4cd222 2211 }
1da177e4
LT
2212 }
2213
f3362737 2214 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2215 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2216 return nfserr_bad_stateid;
1da177e4
LT
2217 }
2218
1da177e4
LT
2219 /*
2220 * We now validate the seqid and stateid generation numbers.
2221 * For the moment, we ignore the possibility of
2222 * generation number wraparound.
2223 */
bd9aac52 2224 if (seqid != sop->so_seqid)
1da177e4
LT
2225 goto check_replay;
2226
3a4f98bb 2227 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2228 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N
2229 " unconfirmed stateowner!\n");
2230 return nfserr_bad_stateid;
1da177e4 2231 }
3a4f98bb 2232 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2233 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N
2234 " confirmed yet!\n");
2235 return nfserr_bad_stateid;
1da177e4 2236 }
0836f587
BF
2237 status = check_stateid_generation(stateid, &stp->st_stateid);
2238 if (status)
2239 return status;
52fd004e 2240 renew_client(sop->so_client);
3a4f98bb 2241 return nfs_ok;
1da177e4 2242
1da177e4 2243check_replay:
bd9aac52 2244 if (seqid == sop->so_seqid - 1) {
849823c5 2245 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2246 /* indicate replay to calling function */
a90b061c 2247 return nfserr_replay_me;
1da177e4 2248 }
2fdada03 2249 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N
2250 sop->so_seqid, seqid);
2251 *sopp = NULL;
2252 return nfserr_bad_seqid;
1da177e4
LT
2253}
2254
b37ad28b 2255__be32
ca364317 2256nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2257 struct nfsd4_open_confirm *oc)
1da177e4 2258{
b37ad28b 2259 __be32 status;
1da177e4
LT
2260 struct nfs4_stateowner *sop;
2261 struct nfs4_stateid *stp;
2262
2263 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF
2264 (int)cstate->current_fh.fh_dentry->d_name.len,
2265 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2266
ca364317 2267 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF
2268 if (status)
2269 return status;
1da177e4
LT
2270
2271 nfs4_lock_state();
2272
ca364317
BF
2273 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2274 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 2275 CONFIRM | OPEN_STATE,
1da177e4
LT
2276 &oc->oc_stateowner, &stp, NULL)))
2277 goto out;
2278
2279 sop = oc->oc_stateowner;
2280 sop->so_confirmed = 1;
2281 update_stateid(&stp->st_stateid);
2282 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2283 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2284 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2285 stp->st_stateid.si_boot,
2286 stp->st_stateid.si_stateownerid,
2287 stp->st_stateid.si_fileid,
2288 stp->st_stateid.si_generation);
c7b9a459
N
2289
2290 nfsd4_create_clid_dir(sop->so_client);
1da177e4 2291out:
f2327d9a 2292 if (oc->oc_stateowner) {
1da177e4 2293 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 2294 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 2295 }
1da177e4
LT
2296 nfs4_unlock_state();
2297 return status;
2298}
2299
2300
2301/*
2302 * unset all bits in union bitmap (bmap) that
2303 * do not exist in share (from successful OPEN_DOWNGRADE)
2304 */
2305static void
2306reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2307{
2308 int i;
2309 for (i = 1; i < 4; i++) {
2310 if ((i & access) != i)
2311 __clear_bit(i, bmap);
2312 }
2313}
2314
2315static void
2316reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2317{
2318 int i;
2319 for (i = 0; i < 4; i++) {
2320 if ((i & deny) != i)
2321 __clear_bit(i, bmap);
2322 }
2323}
2324
b37ad28b 2325__be32
ca364317
BF
2326nfsd4_open_downgrade(struct svc_rqst *rqstp,
2327 struct nfsd4_compound_state *cstate,
a4f1706a 2328 struct nfsd4_open_downgrade *od)
1da177e4 2329{
b37ad28b 2330 __be32 status;
1da177e4
LT
2331 struct nfs4_stateid *stp;
2332 unsigned int share_access;
2333
2334 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF
2335 (int)cstate->current_fh.fh_dentry->d_name.len,
2336 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2337
ba5a6a19
BF
2338 if (!access_valid(od->od_share_access)
2339 || !deny_valid(od->od_share_deny))
1da177e4
LT
2340 return nfserr_inval;
2341
2342 nfs4_lock_state();
ca364317
BF
2343 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2344 od->od_seqid,
1da177e4 2345 &od->od_stateid,
f3362737 2346 OPEN_STATE,
1da177e4
LT
2347 &od->od_stateowner, &stp, NULL)))
2348 goto out;
2349
2350 status = nfserr_inval;
2351 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2352 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2353 stp->st_access_bmap, od->od_share_access);
2354 goto out;
2355 }
2356 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2357 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2358 stp->st_deny_bmap, od->od_share_deny);
2359 goto out;
2360 }
2361 set_access(&share_access, stp->st_access_bmap);
2362 nfs4_file_downgrade(stp->st_vfs_file,
2363 share_access & ~od->od_share_access);
2364
2365 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2366 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2367
2368 update_stateid(&stp->st_stateid);
2369 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2370 status = nfs_ok;
2371out:
f2327d9a 2372 if (od->od_stateowner) {
1da177e4 2373 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 2374 cstate->replay_owner = od->od_stateowner;
f2327d9a 2375 }
1da177e4
LT
2376 nfs4_unlock_state();
2377 return status;
2378}
2379
2380/*
2381 * nfs4_unlock_state() called after encode
2382 */
b37ad28b 2383__be32
ca364317 2384nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2385 struct nfsd4_close *close)
1da177e4 2386{
b37ad28b 2387 __be32 status;
1da177e4
LT
2388 struct nfs4_stateid *stp;
2389
2390 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF
2391 (int)cstate->current_fh.fh_dentry->d_name.len,
2392 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT
2393
2394 nfs4_lock_state();
2395 /* check close_lru for replay */
ca364317
BF
2396 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2397 close->cl_seqid,
1da177e4 2398 &close->cl_stateid,
f3362737 2399 OPEN_STATE | CLOSE_STATE,
1da177e4
LT
2400 &close->cl_stateowner, &stp, NULL)))
2401 goto out;
1da177e4
LT
2402 status = nfs_ok;
2403 update_stateid(&stp->st_stateid);
2404 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2405
04ef5954 2406 /* release_stateid() calls nfsd_close() if needed */
2283963f 2407 release_open_stateid(stp);
04ef5954
BF
2408
2409 /* place unused nfs4_stateowners on so_close_lru list to be
2410 * released by the laundromat service after the lease period
2411 * to enable us to handle CLOSE replay
2412 */
2413 if (list_empty(&close->cl_stateowner->so_stateids))
2414 move_to_close_lru(close->cl_stateowner);
1da177e4 2415out:
f2327d9a 2416 if (close->cl_stateowner) {
1da177e4 2417 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 2418 cstate->replay_owner = close->cl_stateowner;
f2327d9a 2419 }
1da177e4
LT
2420 nfs4_unlock_state();
2421 return status;
2422}
2423
b37ad28b 2424__be32
ca364317
BF
2425nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2426 struct nfsd4_delegreturn *dr)
1da177e4 2427{
203a8c8e
BF
2428 struct nfs4_delegation *dp;
2429 stateid_t *stateid = &dr->dr_stateid;
2430 struct inode *inode;
b37ad28b 2431 __be32 status;
1da177e4 2432
ca364317 2433 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF
2434 return status;
2435 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
2436
2437 nfs4_lock_state();
203a8c8e
BF
2438 status = nfserr_bad_stateid;
2439 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2440 goto out;
2441 status = nfserr_stale_stateid;
2442 if (STALE_STATEID(stateid))
2443 goto out;
7e0f7cf5 2444 status = nfserr_bad_stateid;
203a8c8e
BF
2445 if (!is_delegation_stateid(stateid))
2446 goto out;
203a8c8e
BF
2447 dp = find_delegation_stateid(inode, stateid);
2448 if (!dp)
2449 goto out;
2450 status = check_stateid_generation(stateid, &dp->dl_stateid);
2451 if (status)
2452 goto out;
2453 renew_client(dp->dl_client);
2454
2455 unhash_delegation(dp);
1da177e4 2456out:
203a8c8e
BF
2457 nfs4_unlock_state();
2458
1da177e4
LT
2459 return status;
2460}
2461
2462
2463/*
2464 * Lock owner state (byte-range locks)
2465 */
2466#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2467#define LOCK_HASH_BITS 8
2468#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2469#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2470
87df4de8
BH
2471static inline u64
2472end_offset(u64 start, u64 len)
2473{
2474 u64 end;
2475
2476 end = start + len;
2477 return end >= start ? end: NFS4_MAX_UINT64;
2478}
2479
2480/* last octet in a range */
2481static inline u64
2482last_byte_offset(u64 start, u64 len)
2483{
2484 u64 end;
2485
2486 BUG_ON(!len);
2487 end = start + len;
2488 return end > start ? end - 1: NFS4_MAX_UINT64;
2489}
2490
1da177e4
LT
2491#define lockownerid_hashval(id) \
2492 ((id) & LOCK_HASH_MASK)
2493
2494static inline unsigned int
2495lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2496 struct xdr_netobj *ownername)
2497{
2498 return (file_hashval(inode) + cl_id
2499 + opaque_hashval(ownername->data, ownername->len))
2500 & LOCK_HASH_MASK;
2501}
2502
2503static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2504static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2505static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2506
fd39ca9a 2507static struct nfs4_stateid *
1da177e4
LT
2508find_stateid(stateid_t *stid, int flags)
2509{
9346eff0 2510 struct nfs4_stateid *local;
1da177e4
LT
2511 u32 st_id = stid->si_stateownerid;
2512 u32 f_id = stid->si_fileid;
2513 unsigned int hashval;
2514
2515 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 2516 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
2517 hashval = stateid_hashval(st_id, f_id);
2518 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2519 if ((local->st_stateid.si_stateownerid == st_id) &&
2520 (local->st_stateid.si_fileid == f_id))
2521 return local;
2522 }
2523 }
9346eff0
KK
2524
2525 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
2526 hashval = stateid_hashval(st_id, f_id);
2527 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2528 if ((local->st_stateid.si_stateownerid == st_id) &&
2529 (local->st_stateid.si_fileid == f_id))
2530 return local;
2531 }
849823c5 2532 }
1da177e4
LT
2533 return NULL;
2534}
2535
2536static struct nfs4_delegation *
2537find_delegation_stateid(struct inode *ino, stateid_t *stid)
2538{
13cd2184
N
2539 struct nfs4_file *fp;
2540 struct nfs4_delegation *dl;
1da177e4
LT
2541
2542 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2543 stid->si_boot, stid->si_stateownerid,
2544 stid->si_fileid, stid->si_generation);
2545
1da177e4 2546 fp = find_file(ino);
13cd2184
N
2547 if (!fp)
2548 return NULL;
2549 dl = find_delegation_file(fp, stid);
2550 put_nfs4_file(fp);
2551 return dl;
1da177e4
LT
2552}
2553
2554/*
2555 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2556 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2557 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2558 * locking, this prevents us from being completely protocol-compliant. The
2559 * real solution to this problem is to start using unsigned file offsets in
2560 * the VFS, but this is a very deep change!
2561 */
2562static inline void
2563nfs4_transform_lock_offset(struct file_lock *lock)
2564{
2565 if (lock->fl_start < 0)
2566 lock->fl_start = OFFSET_MAX;
2567 if (lock->fl_end < 0)
2568 lock->fl_end = OFFSET_MAX;
2569}
2570
d5b9026a
N
2571/* Hack!: For now, we're defining this just so we can use a pointer to it
2572 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 2573static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 2574};
1da177e4
LT
2575
2576static inline void
2577nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2578{
d5b9026a
N
2579 struct nfs4_stateowner *sop;
2580 unsigned int hval;
1da177e4 2581
d5b9026a
N
2582 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2583 sop = (struct nfs4_stateowner *) fl->fl_owner;
2584 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT
2585 kref_get(&sop->so_ref);
2586 deny->ld_sop = sop;
2587 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N
2588 } else {
2589 deny->ld_sop = NULL;
2590 deny->ld_clientid.cl_boot = 0;
2591 deny->ld_clientid.cl_id = 0;
1da177e4
LT
2592 }
2593 deny->ld_start = fl->fl_start;
87df4de8
BH
2594 deny->ld_length = NFS4_MAX_UINT64;
2595 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT
2596 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2597 deny->ld_type = NFS4_READ_LT;
2598 if (fl->fl_type != F_RDLCK)
2599 deny->ld_type = NFS4_WRITE_LT;
2600}
2601
1da177e4
LT
2602static struct nfs4_stateowner *
2603find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2604 struct xdr_netobj *owner)
2605{
2606 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2607 struct nfs4_stateowner *op;
2608
2609 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 2610 if (same_owner_str(op, owner, clid))
1da177e4
LT
2611 return op;
2612 }
2613 return NULL;
2614}
2615
2616/*
2617 * Alloc a lock owner structure.
2618 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2619 * occured.
2620 *
2621 * strhashval = lock_ownerstr_hashval
1da177e4
LT
2622 */
2623
2624static struct nfs4_stateowner *
2625alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2626 struct nfs4_stateowner *sop;
2627 struct nfs4_replay *rp;
2628 unsigned int idhashval;
2629
2630 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2631 return NULL;
2632 idhashval = lockownerid_hashval(current_ownerid);
2633 INIT_LIST_HEAD(&sop->so_idhash);
2634 INIT_LIST_HEAD(&sop->so_strhash);
2635 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
2636 INIT_LIST_HEAD(&sop->so_stateids);
2637 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT
2638 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2639 sop->so_time = 0;
2640 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2641 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 2642 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT
2643 sop->so_is_open_owner = 0;
2644 sop->so_id = current_ownerid++;
2645 sop->so_client = clp;
b59e3c0e
NB
2646 /* It is the openowner seqid that will be incremented in encode in the
2647 * case of new lockowners; so increment the lock seqid manually: */
2648 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT
2649 sop->so_confirmed = 1;
2650 rp = &sop->so_replay;
de1ae286 2651 rp->rp_status = nfserr_serverfault;
1da177e4
LT
2652 rp->rp_buflen = 0;
2653 rp->rp_buf = rp->rp_ibuf;
2654 return sop;
2655}
2656
fd39ca9a 2657static struct nfs4_stateid *
1da177e4
LT
2658alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2659{
2660 struct nfs4_stateid *stp;
2661 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2662
5ac049ac
N
2663 stp = nfs4_alloc_stateid();
2664 if (stp == NULL)
1da177e4
LT
2665 goto out;
2666 INIT_LIST_HEAD(&stp->st_hash);
2667 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N
2668 INIT_LIST_HEAD(&stp->st_perstateowner);
2669 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 2670 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 2671 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 2672 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 2673 stp->st_stateowner = sop;
13cd2184 2674 get_nfs4_file(fp);
1da177e4
LT
2675 stp->st_file = fp;
2676 stp->st_stateid.si_boot = boot_time;
2677 stp->st_stateid.si_stateownerid = sop->so_id;
2678 stp->st_stateid.si_fileid = fp->fi_id;
2679 stp->st_stateid.si_generation = 0;
2680 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2681 stp->st_access_bmap = open_stp->st_access_bmap;
2682 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 2683 stp->st_openstp = open_stp;
1da177e4
LT
2684
2685out:
2686 return stp;
2687}
2688
fd39ca9a 2689static int
1da177e4
LT
2690check_lock_length(u64 offset, u64 length)
2691{
87df4de8 2692 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT
2693 LOFF_OVERFLOW(offset, length)));
2694}
2695
2696/*
2697 * LOCK operation
2698 */
b37ad28b 2699__be32
ca364317 2700nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2701 struct nfsd4_lock *lock)
1da177e4 2702{
3e9e3dbe 2703 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 2704 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT
2705 struct nfs4_stateid *lock_stp;
2706 struct file *filp;
2707 struct file_lock file_lock;
8dc7c311 2708 struct file_lock conflock;
b37ad28b 2709 __be32 status = 0;
1da177e4 2710 unsigned int strhashval;
150b3934 2711 unsigned int cmd;
b8dd7b9a 2712 int err;
1da177e4
LT
2713
2714 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2715 (long long) lock->lk_offset,
2716 (long long) lock->lk_length);
2717
1da177e4
LT
2718 if (check_lock_length(lock->lk_offset, lock->lk_length))
2719 return nfserr_inval;
2720
ca364317 2721 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 2722 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA
2723 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2724 return status;
2725 }
2726
1da177e4
LT
2727 nfs4_lock_state();
2728
2729 if (lock->lk_is_new) {
893f8770
N
2730 /*
2731 * Client indicates that this is a new lockowner.
2732 * Use open owner and open stateid to create lock owner and
2733 * lock stateid.
2734 */
1da177e4
LT
2735 struct nfs4_stateid *open_stp = NULL;
2736 struct nfs4_file *fp;
2737
2738 status = nfserr_stale_clientid;
849823c5 2739 if (STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 2740 goto out;
1da177e4 2741
1da177e4 2742 /* validate and update open stateid and open seqid */
ca364317 2743 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2744 lock->lk_new_open_seqid,
2745 &lock->lk_new_open_stateid,
f3362737 2746 OPEN_STATE,
3a65588a 2747 &lock->lk_replay_owner, &open_stp,
b59e3c0e 2748 lock);
37515177 2749 if (status)
1da177e4 2750 goto out;
3a65588a 2751 open_sop = lock->lk_replay_owner;
1da177e4
LT
2752 /* create lockowner and lock stateid */
2753 fp = open_stp->st_file;
2754 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2755 open_sop->so_client->cl_clientid.cl_id,
2756 &lock->v.new.owner);
3e9e3dbe
N
2757 /* XXX: Do we need to check for duplicate stateowners on
2758 * the same file, or should they just be allowed (and
2759 * create new stateids)? */
1da177e4 2760 status = nfserr_resource;
b59e3c0e
NB
2761 lock_sop = alloc_init_lock_stateowner(strhashval,
2762 open_sop->so_client, open_stp, lock);
2763 if (lock_sop == NULL)
1da177e4 2764 goto out;
b59e3c0e 2765 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 2766 if (lock_stp == NULL)
1da177e4 2767 goto out;
1da177e4
LT
2768 } else {
2769 /* lock (lock owner + lock stateid) already exists */
ca364317 2770 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2771 lock->lk_old_lock_seqid,
2772 &lock->lk_old_lock_stateid,
f3362737 2773 LOCK_STATE,
3a65588a 2774 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT
2775 if (status)
2776 goto out;
3a65588a 2777 lock_sop = lock->lk_replay_owner;
1da177e4 2778 }
3a65588a 2779 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT
2780 filp = lock_stp->st_vfs_file;
2781
0dd395dc 2782 status = nfserr_grace;
af558e33 2783 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N
2784 goto out;
2785 status = nfserr_no_grace;
af558e33 2786 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N
2787 goto out;
2788
1da177e4
LT
2789 locks_init_lock(&file_lock);
2790 switch (lock->lk_type) {
2791 case NFS4_READ_LT:
2792 case NFS4_READW_LT:
2793 file_lock.fl_type = F_RDLCK;
150b3934 2794 cmd = F_SETLK;
1da177e4
LT
2795 break;
2796 case NFS4_WRITE_LT:
2797 case NFS4_WRITEW_LT:
2798 file_lock.fl_type = F_WRLCK;
150b3934 2799 cmd = F_SETLK;
1da177e4
LT
2800 break;
2801 default:
2802 status = nfserr_inval;
2803 goto out;
2804 }
b59e3c0e 2805 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT
2806 file_lock.fl_pid = current->tgid;
2807 file_lock.fl_file = filp;
2808 file_lock.fl_flags = FL_POSIX;
d5b9026a 2809 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
2810
2811 file_lock.fl_start = lock->lk_offset;
87df4de8 2812 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT
2813 nfs4_transform_lock_offset(&file_lock);
2814
2815 /*
2816 * Try to lock the file in the VFS.
2817 * Note: locks.c uses the BKL to protect the inode's lock list.
2818 */
2819
fd85b817 2820 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 2821 switch (-err) {
1da177e4
LT
2822 case 0: /* success! */
2823 update_stateid(&lock_stp->st_stateid);
2824 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2825 sizeof(stateid_t));
b8dd7b9a 2826 status = 0;
eb76b3fd
AA
2827 break;
2828 case (EAGAIN): /* conflock holds conflicting lock */
2829 status = nfserr_denied;
2830 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2831 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2832 break;
1da177e4
LT
2833 case (EDEADLK):
2834 status = nfserr_deadlock;
eb76b3fd 2835 break;
1da177e4 2836 default:
fd85b817 2837 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA
2838 status = nfserr_resource;
2839 break;
1da177e4 2840 }
1da177e4 2841out:
8a280510 2842 if (status && lock->lk_is_new && lock_sop)
f044ff83 2843 release_lockowner(lock_sop);
3a65588a
BF
2844 if (lock->lk_replay_owner) {
2845 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 2846 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 2847 }
1da177e4
LT
2848 nfs4_unlock_state();
2849 return status;
2850}
2851
55ef1274
BF
2852/*
2853 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
2854 * so we do a temporary open here just to get an open file to pass to
2855 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
2856 * inode operation.)
2857 */
2858static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
2859{
2860 struct file *file;
2861 int err;
2862
2863 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
2864 if (err)
2865 return err;
2866 err = vfs_test_lock(file, lock);
2867 nfsd_close(file);
2868 return err;
2869}
2870
1da177e4
LT
2871/*
2872 * LOCKT operation
2873 */
b37ad28b 2874__be32
ca364317
BF
2875nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2876 struct nfsd4_lockt *lockt)
1da177e4
LT
2877{
2878 struct inode *inode;
1da177e4 2879 struct file_lock file_lock;
fd85b817 2880 int error;
b37ad28b 2881 __be32 status;
1da177e4 2882
af558e33 2883 if (locks_in_grace())
1da177e4
LT
2884 return nfserr_grace;
2885
2886 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2887 return nfserr_inval;
2888
2889 lockt->lt_stateowner = NULL;
2890 nfs4_lock_state();
2891
2892 status = nfserr_stale_clientid;
849823c5 2893 if (STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 2894 goto out;
1da177e4 2895
ca364317 2896 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 2897 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT
2898 if (status == nfserr_symlink)
2899 status = nfserr_inval;
2900 goto out;
2901 }
2902
ca364317 2903 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
2904 locks_init_lock(&file_lock);
2905 switch (lockt->lt_type) {
2906 case NFS4_READ_LT:
2907 case NFS4_READW_LT:
2908 file_lock.fl_type = F_RDLCK;
2909 break;
2910 case NFS4_WRITE_LT:
2911 case NFS4_WRITEW_LT:
2912 file_lock.fl_type = F_WRLCK;
2913 break;
2914 default:
2fdada03 2915 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT
2916 status = nfserr_inval;
2917 goto out;
2918 }
2919
2920 lockt->lt_stateowner = find_lockstateowner_str(inode,
2921 &lockt->lt_clientid, &lockt->lt_owner);
2922 if (lockt->lt_stateowner)
2923 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2924 file_lock.fl_pid = current->tgid;
2925 file_lock.fl_flags = FL_POSIX;
2926
2927 file_lock.fl_start = lockt->lt_offset;
87df4de8 2928 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT
2929
2930 nfs4_transform_lock_offset(&file_lock);
2931
1da177e4 2932 status = nfs_ok;
55ef1274 2933 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME
2934 if (error) {
2935 status = nfserrno(error);
2936 goto out;
2937 }
9d6a8c5c 2938 if (file_lock.fl_type != F_UNLCK) {
1da177e4 2939 status = nfserr_denied;
9d6a8c5c 2940 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT
2941 }
2942out:
2943 nfs4_unlock_state();
2944 return status;
2945}
2946
b37ad28b 2947__be32
ca364317 2948nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2949 struct nfsd4_locku *locku)
1da177e4
LT
2950{
2951 struct nfs4_stateid *stp;
2952 struct file *filp = NULL;
2953 struct file_lock file_lock;
b37ad28b 2954 __be32 status;
b8dd7b9a 2955 int err;
1da177e4
LT
2956
2957 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2958 (long long) locku->lu_offset,
2959 (long long) locku->lu_length);
2960
2961 if (check_lock_length(locku->lu_offset, locku->lu_length))
2962 return nfserr_inval;
2963
2964 nfs4_lock_state();
2965
ca364317 2966 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2967 locku->lu_seqid,
2968 &locku->lu_stateid,
f3362737 2969 LOCK_STATE,
1da177e4
LT
2970 &locku->lu_stateowner, &stp, NULL)))
2971 goto out;
2972
2973 filp = stp->st_vfs_file;
2974 BUG_ON(!filp);
2975 locks_init_lock(&file_lock);
2976 file_lock.fl_type = F_UNLCK;
2977 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2978 file_lock.fl_pid = current->tgid;
2979 file_lock.fl_file = filp;
2980 file_lock.fl_flags = FL_POSIX;
d5b9026a 2981 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
2982 file_lock.fl_start = locku->lu_offset;
2983
87df4de8 2984 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT
2985 nfs4_transform_lock_offset(&file_lock);
2986
2987 /*
2988 * Try to unlock the file in the VFS.
2989 */
fd85b817 2990 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 2991 if (err) {
fd85b817 2992 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT
2993 goto out_nfserr;
2994 }
2995 /*
2996 * OK, unlock succeeded; the only thing left to do is update the stateid.
2997 */
2998 update_stateid(&stp->st_stateid);
2999 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3000
3001out:
f2327d9a 3002 if (locku->lu_stateowner) {
1da177e4 3003 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3004 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3005 }
1da177e4
LT
3006 nfs4_unlock_state();
3007 return status;
3008
3009out_nfserr:
b8dd7b9a 3010 status = nfserrno(err);
1da177e4
LT
3011 goto out;
3012}
3013
3014/*
3015 * returns
3016 * 1: locks held by lockowner
3017 * 0: no locks held by lockowner
3018 */
3019static int
3020check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3021{
3022 struct file_lock **flpp;
7eaa36e2 3023 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT
3024 int status = 0;
3025
3026 lock_kernel();
3027 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3028 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT
3029 status = 1;
3030 goto out;
796dadfd 3031 }
1da177e4
LT
3032 }
3033out:
3034 unlock_kernel();
3035 return status;
3036}
3037
b37ad28b 3038__be32
b591480b
BF
3039nfsd4_release_lockowner(struct svc_rqst *rqstp,
3040 struct nfsd4_compound_state *cstate,
3041 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT
3042{
3043 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N
3044 struct nfs4_stateowner *sop;
3045 struct nfs4_stateid *stp;
1da177e4 3046 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N
3047 struct list_head matches;
3048 int i;
b37ad28b 3049 __be32 status;
1da177e4
LT
3050
3051 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3052 clid->cl_boot, clid->cl_id);
3053
3054 /* XXX check for lease expiration */
3055
3056 status = nfserr_stale_clientid;
849823c5 3057 if (STALE_CLIENTID(clid))
1da177e4 3058 return status;
1da177e4
LT
3059
3060 nfs4_lock_state();
3061
3e9e3dbe
N
3062 status = nfserr_locks_held;
3063 /* XXX: we're doing a linear search through all the lockowners.
3064 * Yipes! For now we'll just hope clients aren't really using
3065 * release_lockowner much, but eventually we have to fix these
3066 * data structures. */
3067 INIT_LIST_HEAD(&matches);
3068 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3069 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3070 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N
3071 continue;
3072 list_for_each_entry(stp, &sop->so_stateids,
3073 st_perstateowner) {
3074 if (check_for_locks(stp->st_vfs_file, sop))
3075 goto out;
3076 /* Note: so_perclient unused for lockowners,
3077 * so it's OK to fool with here. */
3078 list_add(&sop->so_perclient, &matches);
3079 }
1da177e4 3080 }
3e9e3dbe
N
3081 }
3082 /* Clients probably won't expect us to return with some (but not all)
3083 * of the lockowner state released; so don't release any until all
3084 * have been checked. */
3085 status = nfs_ok;
0fa822e4
N
3086 while (!list_empty(&matches)) {
3087 sop = list_entry(matches.next, struct nfs4_stateowner,
3088 so_perclient);
3089 /* unhash_stateowner deletes so_perclient only
3090 * for openowners. */
3091 list_del(&sop->so_perclient);
f044ff83 3092 release_lockowner(sop);
1da177e4
LT
3093 }
3094out:
3095 nfs4_unlock_state();
3096 return status;
3097}
3098
3099static inline struct nfs4_client_reclaim *
a55370a3 3100alloc_reclaim(void)
1da177e4 3101{
a55370a3 3102 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT
3103}
3104
c7b9a459
N
3105int
3106nfs4_has_reclaimed_state(const char *name)
3107{
3108 unsigned int strhashval = clientstr_hashval(name);
3109 struct nfs4_client *clp;
3110
3111 clp = find_confirmed_client_by_str(name, strhashval);
3112 return clp ? 1 : 0;
3113}
3114
1da177e4
LT
3115/*
3116 * failure => all reset bets are off, nfserr_no_grace...
3117 */
190e4fbf
N
3118int
3119nfs4_client_to_reclaim(const char *name)
1da177e4
LT
3120{
3121 unsigned int strhashval;
3122 struct nfs4_client_reclaim *crp = NULL;
3123
a55370a3
N
3124 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3125 crp = alloc_reclaim();
1da177e4
LT
3126 if (!crp)
3127 return 0;
a55370a3 3128 strhashval = clientstr_hashval(name);
1da177e4
LT
3129 INIT_LIST_HEAD(&crp->cr_strhash);
3130 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3131 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT
3132 reclaim_str_hashtbl_size++;
3133 return 1;
3134}
3135
3136static void
3137nfs4_release_reclaim(void)
3138{
3139 struct nfs4_client_reclaim *crp = NULL;
3140 int i;
3141
1da177e4
LT
3142 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3143 while (!list_empty(&reclaim_str_hashtbl[i])) {
3144 crp = list_entry(reclaim_str_hashtbl[i].next,
3145 struct nfs4_client_reclaim, cr_strhash);
3146 list_del(&crp->cr_strhash);
1da177e4
LT
3147 kfree(crp);
3148 reclaim_str_hashtbl_size--;
3149 }
3150 }
3151 BUG_ON(reclaim_str_hashtbl_size);
3152}
3153
3154/*
3155 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3156static struct nfs4_client_reclaim *
1da177e4
LT
3157nfs4_find_reclaim_client(clientid_t *clid)
3158{
3159 unsigned int strhashval;
3160 struct nfs4_client *clp;
3161 struct nfs4_client_reclaim *crp = NULL;
3162
3163
3164 /* find clientid in conf_id_hashtbl */
3165 clp = find_confirmed_client(clid);
3166 if (clp == NULL)
3167 return NULL;
3168
a55370a3
N
3169 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3170 clp->cl_name.len, clp->cl_name.data,
3171 clp->cl_recdir);
1da177e4
LT
3172
3173 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3174 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3175 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3176 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT
3177 return crp;
3178 }
3179 }
3180 return NULL;
3181}
3182
3183/*
3184* Called from OPEN. Look for clientid in reclaim list.
3185*/
b37ad28b 3186__be32
1da177e4
LT
3187nfs4_check_open_reclaim(clientid_t *clid)
3188{
dfc83565 3189 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT
3190}
3191
ac4d8ff2 3192/* initialization to perform at module load time: */
1da177e4 3193
e8ff2a84 3194int
ac4d8ff2 3195nfs4_state_init(void)
1da177e4 3196{
e8ff2a84 3197 int i, status;
1da177e4 3198
e8ff2a84
BF
3199 status = nfsd4_init_slabs();
3200 if (status)
3201 return status;
1da177e4
LT
3202 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3203 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3204 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3205 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3206 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3207 }
3208 for (i = 0; i < FILE_HASH_SIZE; i++) {
3209 INIT_LIST_HEAD(&file_hashtbl[i]);
3210 }
3211 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3212 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3213 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3214 }
3215 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3216 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3217 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3218 }
3219 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3220 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3221 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3222 }
1da177e4 3223 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT
3224 INIT_LIST_HEAD(&close_lru);
3225 INIT_LIST_HEAD(&client_lru);
3226 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N
3227 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3228 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3229 reclaim_str_hashtbl_size = 0;
e8ff2a84 3230 return 0;
ac4d8ff2
N
3231}
3232
190e4fbf
N
3233static void
3234nfsd4_load_reboot_recovery_data(void)
3235{
3236 int status;
3237
0964a3d3
N
3238 nfs4_lock_state();
3239 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3240 status = nfsd4_recdir_load();
0964a3d3 3241 nfs4_unlock_state();
190e4fbf
N
3242 if (status)
3243 printk("NFSD: Failure reading reboot recovery data\n");
3244}
3245
9a8db97e
ME
3246unsigned long
3247get_nfs4_grace_period(void)
3248{
3249 return max(user_lease_time, lease_time) * HZ;
3250}
3251
c2f1a551
MS
3252/*
3253 * Since the lifetime of a delegation isn't limited to that of an open, a
3254 * client may quite reasonably hang on to a delegation as long as it has
3255 * the inode cached. This becomes an obvious problem the first time a
3256 * client's inode cache approaches the size of the server's total memory.
3257 *
3258 * For now we avoid this problem by imposing a hard limit on the number
3259 * of delegations, which varies according to the server's memory size.
3260 */
3261static void
3262set_max_delegations(void)
3263{
3264 /*
3265 * Allow at most 4 delegations per megabyte of RAM. Quick
3266 * estimates suggest that in the worst case (where every delegation
3267 * is for a different inode), a delegation could take about 1.5K,
3268 * giving a worst case usage of about 6% of memory.
3269 */
3270 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3271}
3272
ac4d8ff2
N
3273/* initialization to perform when the nfsd service is started: */
3274
3275static void
3276__nfs4_state_start(void)
3277{
9a8db97e 3278 unsigned long grace_time;
ac4d8ff2 3279
1da177e4 3280 boot_time = get_seconds();
af558e33 3281 grace_time = get_nfs4_grace_period();
d99a05ad 3282 lease_time = user_lease_time;
af558e33 3283 locks_start_grace(&nfsd4_manager);
9a8db97e
ME
3284 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3285 grace_time/HZ);
58da282b 3286 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 3287 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 3288 set_max_delegations();
1da177e4
LT
3289}
3290
e8ff2a84 3291void
76a3550e 3292nfs4_state_start(void)
1da177e4 3293{
1da177e4 3294 if (nfs4_init)
e8ff2a84 3295 return;
190e4fbf 3296 nfsd4_load_reboot_recovery_data();
76a3550e 3297 __nfs4_state_start();
1da177e4 3298 nfs4_init = 1;
e8ff2a84 3299 return;
1da177e4
LT
3300}
3301
1da177e4
LT
3302time_t
3303nfs4_lease_time(void)
3304{
3305 return lease_time;
3306}
3307
3308static void
3309__nfs4_state_shutdown(void)
3310{
3311 int i;
3312 struct nfs4_client *clp = NULL;
3313 struct nfs4_delegation *dp = NULL;
1da177e4
LT
3314 struct list_head *pos, *next, reaplist;
3315
1da177e4
LT
3316 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3317 while (!list_empty(&conf_id_hashtbl[i])) {
3318 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3319 expire_client(clp);
3320 }
3321 while (!list_empty(&unconf_str_hashtbl[i])) {
3322 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3323 expire_client(clp);
3324 }
3325 }
3326 INIT_LIST_HEAD(&reaplist);
3327 spin_lock(&recall_lock);
3328 list_for_each_safe(pos, next, &del_recall_lru) {
3329 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3330 list_move(&dp->dl_recall_lru, &reaplist);
3331 }
3332 spin_unlock(&recall_lock);
3333 list_for_each_safe(pos, next, &reaplist) {
3334 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3335 list_del_init(&dp->dl_recall_lru);
3336 unhash_delegation(dp);
3337 }
3338
190e4fbf 3339 nfsd4_shutdown_recdir();
1da177e4 3340 nfs4_init = 0;
1da177e4
LT
3341}
3342
3343void
3344nfs4_state_shutdown(void)
3345{
5e8d5c29
N
3346 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3347 destroy_workqueue(laundry_wq);
2c5e7615 3348 locks_end_grace(&nfsd4_manager);
1da177e4
LT
3349 nfs4_lock_state();
3350 nfs4_release_reclaim();
3351 __nfs4_state_shutdown();
1da177e4
LT
3352 nfs4_unlock_state();
3353}
3354
3dd98a3b
JL
3355/*
3356 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3357 * accessed when nfsd is starting.
3358 */
0964a3d3
N
3359static void
3360nfs4_set_recdir(char *recdir)
3361{
0964a3d3 3362 strcpy(user_recovery_dirname, recdir);
0964a3d3
N
3363}
3364
3365/*
3366 * Change the NFSv4 recovery directory to recdir.
3367 */
3368int
3369nfs4_reset_recoverydir(char *recdir)
3370{
3371 int status;
a63bb996 3372 struct path path;
0964a3d3 3373
a63bb996 3374 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N
3375 if (status)
3376 return status;
3377 status = -ENOTDIR;
a63bb996 3378 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N
3379 nfs4_set_recdir(recdir);
3380 status = 0;
3381 }
a63bb996 3382 path_put(&path);
0964a3d3
N
3383 return status;
3384}
3385
3dd98a3b
JL
3386char *
3387nfs4_recoverydir(void)
3388{
3389 return user_recovery_dirname;
3390}
3391
1da177e4
LT
3392/*
3393 * Called when leasetime is changed.
3394 *
d99a05ad
N
3395 * The only way the protocol gives us to handle on-the-fly lease changes is to
3396 * simulate a reboot. Instead of doing that, we just wait till the next time
3397 * we start to register any changes in lease time. If the administrator
3398 * really wants to change the lease time *now*, they can go ahead and bring
3399 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL
3400 *
3401 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3402 * when nfsd is starting
1da177e4
LT
3403 */
3404void
3405nfs4_reset_lease(time_t leasetime)
3406{
d99a05ad 3407 user_lease_time = leasetime;
1da177e4 3408}