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