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