]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/nfsd/nfs4state.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: reclaim DRC memory on session free
[net-next-2.6.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
1da177e4
LT		 58
59#define NFSDDBG_FACILITY NFSDDBG_PROC
60
61/* Globals */
62static time_t lease_time = 90; /* default lease time */
d99a05ad 63static time_t user_lease_time = 90;
fd39ca9a 64static time_t boot_time;
1da177e4
LT		 65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N		 69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 71static u64 current_sessionid = 1;
fd39ca9a
N		 72
73#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
74#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 75
1da177e4 76/* forward declarations */
fd39ca9a 77static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 78static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 79static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80static void nfs4_set_recdir(char *recdir);
1da177e4 81
8b671b80
BF		 82/* Locking: */
83
84/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 85static DEFINE_MUTEX(client_mutex);
1da177e4 86
8b671b80
BF		 87/*
88 * Currently used for the del_recall_lru and file hash table. In an
89 * effort to decrease the scope of the client_mutex, this spinlock may
90 * eventually cover more:
91 */
92static DEFINE_SPINLOCK(recall_lock);
93
e18b890b
CL		 94static struct kmem_cache *stateowner_slab = NULL;
95static struct kmem_cache *file_slab = NULL;
96static struct kmem_cache *stateid_slab = NULL;
97static struct kmem_cache *deleg_slab = NULL;
e60d4398 98
1da177e4
LT		 99void
100nfs4_lock_state(void)
101{
353ab6e9 102 mutex_lock(&client_mutex);
1da177e4
LT		 103}
104
105void
106nfs4_unlock_state(void)
107{
353ab6e9 108 mutex_unlock(&client_mutex);
1da177e4
LT		 109}
110
111static inline u32
112opaque_hashval(const void *ptr, int nbytes)
113{
114 unsigned char *cptr = (unsigned char *) ptr;
115
116 u32 x = 0;
117 while (nbytes--) {
118 x *= 37;
119 x += *cptr++;
120 }
121 return x;
122}
123
1da177e4
LT		 124static struct list_head del_recall_lru;
125
13cd2184
N		 126static inline void
127put_nfs4_file(struct nfs4_file *fi)
128{
8b671b80
BF		 129 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
130 list_del(&fi->fi_hash);
131 spin_unlock(&recall_lock);
132 iput(fi->fi_inode);
133 kmem_cache_free(file_slab, fi);
134 }
13cd2184
N		 135}
136
137static inline void
138get_nfs4_file(struct nfs4_file *fi)
139{
8b671b80 140 atomic_inc(&fi->fi_ref);
13cd2184
N		 141}
142
ef0f3390 143static int num_delegations;
c2f1a551 144unsigned int max_delegations;
ef0f3390
N		 145
146/*
147 * Open owner state (share locks)
148 */
149
150/* hash tables for nfs4_stateowner */
151#define OWNER_HASH_BITS 8
152#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
153#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
154
155#define ownerid_hashval(id) \
156 ((id) & OWNER_HASH_MASK)
157#define ownerstr_hashval(clientid, ownername) \
158 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
159
160static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
161static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
162
163/* hash table for nfs4_file */
164#define FILE_HASH_BITS 8
165#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
166#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
167/* hash table for (open)nfs4_stateid */
168#define STATEID_HASH_BITS 10
169#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
170#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
171
172#define file_hashval(x) \
173 hash_ptr(x, FILE_HASH_BITS)
174#define stateid_hashval(owner_id, file_id) \
175 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
176
177static struct list_head file_hashtbl[FILE_HASH_SIZE];
178static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
179
1da177e4
LT		 180static struct nfs4_delegation *
181alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
182{
183 struct nfs4_delegation *dp;
184 struct nfs4_file *fp = stp->st_file;
c237dc03 185 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 186
187 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 188 if (fp->fi_had_conflict)
189 return NULL;
c2f1a551 190 if (num_delegations > max_delegations)
ef0f3390 191 return NULL;
5b2d21c1
N		 192 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
193 if (dp == NULL)
1da177e4 194 return dp;
ef0f3390 195 num_delegations++;
ea1da636
N		 196 INIT_LIST_HEAD(&dp->dl_perfile);
197 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 198 INIT_LIST_HEAD(&dp->dl_recall_lru);
199 dp->dl_client = clp;
13cd2184 200 get_nfs4_file(fp);
1da177e4
LT		 201 dp->dl_file = fp;
202 dp->dl_flock = NULL;
203 get_file(stp->st_vfs_file);
204 dp->dl_vfs_file = stp->st_vfs_file;
205 dp->dl_type = type;
b53d40c5 206 dp->dl_ident = cb->cb_ident;
78155ed7 207 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 208 dp->dl_stateid.si_stateownerid = current_delegid++;
209 dp->dl_stateid.si_fileid = 0;
210 dp->dl_stateid.si_generation = 0;
6c02eaa1 211 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 212 dp->dl_time = 0;
213 atomic_set(&dp->dl_count, 1);
ea1da636
N		 214 list_add(&dp->dl_perfile, &fp->fi_delegations);
215 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 216 return dp;
217}
218
219void
220nfs4_put_delegation(struct nfs4_delegation *dp)
221{
222 if (atomic_dec_and_test(&dp->dl_count)) {
223 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 224 put_nfs4_file(dp->dl_file);
5b2d21c1 225 kmem_cache_free(deleg_slab, dp);
ef0f3390 226 num_delegations--;
1da177e4
LT		 227 }
228}
229
230/* Remove the associated file_lock first, then remove the delegation.
231 * lease_modify() is called to remove the FS_LEASE file_lock from
232 * the i_flock list, eventually calling nfsd's lock_manager
233 * fl_release_callback.
234 */
235static void
236nfs4_close_delegation(struct nfs4_delegation *dp)
237{
238 struct file *filp = dp->dl_vfs_file;
239
240 dprintk("NFSD: close_delegation dp %p\n",dp);
241 dp->dl_vfs_file = NULL;
242 /* The following nfsd_close may not actually close the file,
243 * but we want to remove the lease in any case. */
c907132d 244 if (dp->dl_flock)
a9933cea 245 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 246 nfsd_close(filp);
1da177e4
LT		 247}
248
249/* Called under the state lock. */
250static void
251unhash_delegation(struct nfs4_delegation *dp)
252{
ea1da636
N		 253 list_del_init(&dp->dl_perfile);
254 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 255 spin_lock(&recall_lock);
256 list_del_init(&dp->dl_recall_lru);
257 spin_unlock(&recall_lock);
258 nfs4_close_delegation(dp);
259 nfs4_put_delegation(dp);
260}
261
262/*
263 * SETCLIENTID state
264 */
265
266/* Hash tables for nfs4_clientid state */
267#define CLIENT_HASH_BITS 4
268#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
269#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
270
271#define clientid_hashval(id) \
272 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 273#define clientstr_hashval(name) \
274 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 275/*
276 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
277 * used in reboot/reset lease grace period processing
278 *
279 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
280 * setclientid_confirmed info.
281 *
282 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
283 * setclientid info.
284 *
285 * client_lru holds client queue ordered by nfs4_client.cl_time
286 * for lease renewal.
287 *
288 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
289 * for last close replay.
290 */
291static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
292static int reclaim_str_hashtbl_size = 0;
293static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
294static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head client_lru;
298static struct list_head close_lru;
299
2283963f
BF		 300static void unhash_generic_stateid(struct nfs4_stateid *stp)
301{
302 list_del(&stp->st_hash);
303 list_del(&stp->st_perfile);
304 list_del(&stp->st_perstateowner);
305}
306
307static void free_generic_stateid(struct nfs4_stateid *stp)
308{
309 put_nfs4_file(stp->st_file);
310 kmem_cache_free(stateid_slab, stp);
311}
312
313static void release_lock_stateid(struct nfs4_stateid *stp)
314{
315 unhash_generic_stateid(stp);
316 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
317 free_generic_stateid(stp);
318}
319
f044ff83
BF		 320static void unhash_lockowner(struct nfs4_stateowner *sop)
321{
322 struct nfs4_stateid *stp;
323
324 list_del(&sop->so_idhash);
325 list_del(&sop->so_strhash);
326 list_del(&sop->so_perstateid);
327 while (!list_empty(&sop->so_stateids)) {
328 stp = list_first_entry(&sop->so_stateids,
329 struct nfs4_stateid, st_perstateowner);
330 release_lock_stateid(stp);
331 }
332}
333
334static void release_lockowner(struct nfs4_stateowner *sop)
335{
336 unhash_lockowner(sop);
337 nfs4_put_stateowner(sop);
338}
339
f1d110ca
BF		 340static void
341release_stateid_lockowners(struct nfs4_stateid *open_stp)
342{
343 struct nfs4_stateowner *lock_sop;
344
345 while (!list_empty(&open_stp->st_lockowners)) {
346 lock_sop = list_entry(open_stp->st_lockowners.next,
347 struct nfs4_stateowner, so_perstateid);
348 /* list_del(&open_stp->st_lockowners); */
349 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 350 release_lockowner(lock_sop);
f1d110ca
BF		 351 }
352}
353
2283963f
BF		 354static void release_open_stateid(struct nfs4_stateid *stp)
355{
356 unhash_generic_stateid(stp);
357 release_stateid_lockowners(stp);
358 nfsd_close(stp->st_vfs_file);
359 free_generic_stateid(stp);
360}
361
f044ff83 362static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 363{
364 struct nfs4_stateid *stp;
365
366 list_del(&sop->so_idhash);
367 list_del(&sop->so_strhash);
f044ff83
BF		 368 list_del(&sop->so_perclient);
369 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 370 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 371 stp = list_first_entry(&sop->so_stateids,
372 struct nfs4_stateid, st_perstateowner);
373 release_open_stateid(stp);
f1d110ca
BF		 374 }
375}
376
f044ff83 377static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 378{
f044ff83 379 unhash_openowner(sop);
f1d110ca
BF		 380 list_del(&sop->so_close_lru);
381 nfs4_put_stateowner(sop);
382}
383
5282fd72
ME		 384static DEFINE_SPINLOCK(sessionid_lock);
385#define SESSION_HASH_SIZE 512
386static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
387
388static inline int
389hash_sessionid(struct nfs4_sessionid *sessionid)
390{
391 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
392
393 return sid->sequence % SESSION_HASH_SIZE;
394}
395
396static inline void
397dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
398{
399 u32 *ptr = (u32 *)(&sessionid->data[0]);
400 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
401}
402
ec6b5d7b
AA		 403static void
404gen_sessionid(struct nfsd4_session *ses)
405{
406 struct nfs4_client *clp = ses->se_client;
407 struct nfsd4_sessionid *sid;
408
409 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
410 sid->clientid = clp->cl_clientid;
411 sid->sequence = current_sessionid++;
412 sid->reserved = 0;
413}
414
415/*
416 * Give the client the number of slots it requests bound by
417 * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
418 *
419 * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
420 * should (up to a point) re-negotiate active sessions and reduce their
421 * slot usage to make rooom for new connections. For now we just fail the
422 * create session.
423 */
424static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
425{
426 int status = 0, np = fchan->maxreqs * NFSD_PAGES_PER_SLOT;
427
5d77ddfb
AA		 428 if (fchan->maxreqs < 1)
429 return nfserr_inval;
430 else if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
431 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
432
4bd9b0f4
AA		 433 spin_lock(&nfsd_drc_lock);
434 if (np + nfsd_drc_pages_used > nfsd_drc_max_pages)
435 np = nfsd_drc_max_pages - nfsd_drc_pages_used;
436 nfsd_drc_pages_used += np;
437 spin_unlock(&nfsd_drc_lock);
ec6b5d7b
AA		 438
439 if (np <= 0) {
440 status = nfserr_resource;
441 fchan->maxreqs = 0;
442 } else
443 fchan->maxreqs = np / NFSD_PAGES_PER_SLOT;
444
445 return status;
446}
447
448/*
449 * fchan holds the client values on input, and the server values on output
450 */
451static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 452 struct nfsd4_channel_attrs *session_fchan,
453 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 454{
455 int status = 0;
456 __u32 maxcount = svc_max_payload(rqstp);
457
458 /* headerpadsz set to zero in encode routine */
459
460 /* Use the client's max request and max response size if possible */
461 if (fchan->maxreq_sz > maxcount)
462 fchan->maxreq_sz = maxcount;
6c18ba9f 463 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 464
465 if (fchan->maxresp_sz > maxcount)
466 fchan->maxresp_sz = maxcount;
6c18ba9f 467 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b
AA		 468
469 /* Set the max response cached size our default which is
470 * a multiple of PAGE_SIZE and small */
6c18ba9f
AB		 471 session_fchan->maxresp_cached = NFSD_PAGES_PER_SLOT * PAGE_SIZE;
472 fchan->maxresp_cached = session_fchan->maxresp_cached;
ec6b5d7b
AA		 473
474 /* Use the client's maxops if possible */
475 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
476 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 477 session_fchan->maxops = fchan->maxops;
ec6b5d7b
AA		 478
479 /* try to use the client requested number of slots */
480 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
481 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
482
483 /* FIXME: Error means no more DRC pages so the server should
484 * recover pages from existing sessions. For now fail session
485 * creation.
486 */
487 status = set_forechannel_maxreqs(fchan);
488
6c18ba9f 489 session_fchan->maxreqs = fchan->maxreqs;
ec6b5d7b
AA		 490 return status;
491}
492
493static int
494alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
495 struct nfsd4_create_session *cses)
496{
497 struct nfsd4_session *new, tmp;
498 int idx, status = nfserr_resource, slotsize;
499
500 memset(&tmp, 0, sizeof(tmp));
501
502 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 503 tmp.se_bchannel = cses->back_channel;
504 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
505 &cses->fore_channel);
ec6b5d7b
AA		 506 if (status)
507 goto out;
508
509 /* allocate struct nfsd4_session and slot table in one piece */
6c18ba9f 510 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot);
ec6b5d7b
AA		 511 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
512 if (!new)
513 goto out;
514
515 memcpy(new, &tmp, sizeof(*new));
516
517 new->se_client = clp;
518 gen_sessionid(new);
519 idx = hash_sessionid(&new->se_sessionid);
520 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
521 NFS4_MAX_SESSIONID_LEN);
522
523 new->se_flags = cses->flags;
524 kref_init(&new->se_ref);
525 spin_lock(&sessionid_lock);
526 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
527 list_add(&new->se_perclnt, &clp->cl_sessions);
528 spin_unlock(&sessionid_lock);
529
530 status = nfs_ok;
531out:
532 return status;
533}
534
5282fd72
ME		 535/* caller must hold sessionid_lock */
536static struct nfsd4_session *
537find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
538{
539 struct nfsd4_session *elem;
540 int idx;
541
542 dump_sessionid(__func__, sessionid);
543 idx = hash_sessionid(sessionid);
544 dprintk("%s: idx is %d\n", __func__, idx);
545 /* Search in the appropriate list */
546 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
547 dump_sessionid("list traversal", &elem->se_sessionid);
548 if (!memcmp(elem->se_sessionid.data, sessionid->data,
549 NFS4_MAX_SESSIONID_LEN)) {
550 return elem;
551 }
552 }
553
554 dprintk("%s: session not found\n", __func__);
555 return NULL;
556}
557
558/* caller must hold sessionid_lock */
7116ed6b 559static void
5282fd72 560unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 561{
562 list_del(&ses->se_hash);
563 list_del(&ses->se_perclnt);
5282fd72
ME		 564}
565
566static void
567release_session(struct nfsd4_session *ses)
568{
569 spin_lock(&sessionid_lock);
570 unhash_session(ses);
571 spin_unlock(&sessionid_lock);
7116ed6b
AA		 572 nfsd4_put_session(ses);
573}
574
14778a13
AA		 575static void nfsd4_release_respages(struct page **respages, short resused);
576
7116ed6b
AA		 577void
578free_session(struct kref *kref)
579{
580 struct nfsd4_session *ses;
14778a13 581 int i;
7116ed6b
AA		 582
583 ses = container_of(kref, struct nfsd4_session, se_ref);
6c18ba9f 584 for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
14778a13
AA		 585 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
586 nfsd4_release_respages(e->ce_respages, e->ce_resused);
587 }
be98d1bb
AA		 588 spin_lock(&nfsd_drc_lock);
589 nfsd_drc_pages_used -= ses->se_fchannel.maxreqs * NFSD_PAGES_PER_SLOT;
590 spin_unlock(&nfsd_drc_lock);
7116ed6b
AA		 591 kfree(ses);
592}
593
1da177e4
LT		 594static inline void
595renew_client(struct nfs4_client *clp)
596{
597 /*
598 * Move client to the end to the LRU list.
599 */
600 dprintk("renewing client (clientid %08x/%08x)\n",
601 clp->cl_clientid.cl_boot,
602 clp->cl_clientid.cl_id);
603 list_move_tail(&clp->cl_lru, &client_lru);
604 clp->cl_time = get_seconds();
605}
606
607/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
608static int
609STALE_CLIENTID(clientid_t *clid)
610{
611 if (clid->cl_boot == boot_time)
612 return 0;
60adfc50
AA		 613 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
614 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 615 return 1;
616}
617
618/*
619 * XXX Should we use a slab cache ?
620 * This type of memory management is somewhat inefficient, but we use it
621 * anyway since SETCLIENTID is not a common operation.
622 */
35bba9a3 623static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 624{
625 struct nfs4_client *clp;
626
35bba9a3
BF		 627 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
628 if (clp == NULL)
629 return NULL;
630 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
631 if (clp->cl_name.data == NULL) {
632 kfree(clp);
633 return NULL;
1da177e4 634 }
35bba9a3
BF		 635 memcpy(clp->cl_name.data, name.data, name.len);
636 clp->cl_name.len = name.len;
1da177e4
LT		 637 return clp;
638}
639
1b1a9b31
BF		 640static void
641shutdown_callback_client(struct nfs4_client *clp)
642{
c237dc03 643 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 644
1b1a9b31 645 if (clnt) {
404ec117
BF		 646 /*
647 * Callback threads take a reference on the client, so there
648 * should be no outstanding callbacks at this point.
649 */
c237dc03 650 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 651 rpc_shutdown_client(clnt);
652 }
c237dc03
BF		 653 if (clp->cl_cb_conn.cb_cred) {
654 put_rpccred(clp->cl_cb_conn.cb_cred);
655 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 656 }
1b1a9b31
BF		 657}
658
1da177e4
LT		 659static inline void
660free_client(struct nfs4_client *clp)
661{
1b1a9b31 662 shutdown_callback_client(clp);
38eb76a5
AA		 663 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
664 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 665 if (clp->cl_cred.cr_group_info)
666 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 667 kfree(clp->cl_principal);
1da177e4
LT		 668 kfree(clp->cl_name.data);
669 kfree(clp);
670}
671
672void
673put_nfs4_client(struct nfs4_client *clp)
674{
675 if (atomic_dec_and_test(&clp->cl_count))
676 free_client(clp);
677}
678
679static void
680expire_client(struct nfs4_client *clp)
681{
682 struct nfs4_stateowner *sop;
683 struct nfs4_delegation *dp;
1da177e4
LT		 684 struct list_head reaplist;
685
686 dprintk("NFSD: expire_client cl_count %d\n",
687 atomic_read(&clp->cl_count));
688
1da177e4
LT		 689 INIT_LIST_HEAD(&reaplist);
690 spin_lock(&recall_lock);
ea1da636
N		 691 while (!list_empty(&clp->cl_delegations)) {
692 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 693 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
694 dp->dl_flock);
ea1da636 695 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 696 list_move(&dp->dl_recall_lru, &reaplist);
697 }
698 spin_unlock(&recall_lock);
699 while (!list_empty(&reaplist)) {
700 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
701 list_del_init(&dp->dl_recall_lru);
702 unhash_delegation(dp);
703 }
704 list_del(&clp->cl_idhash);
705 list_del(&clp->cl_strhash);
706 list_del(&clp->cl_lru);
ea1da636
N		 707 while (!list_empty(&clp->cl_openowners)) {
708 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 709 release_openowner(sop);
1da177e4 710 }
c4bf7868
ME		 711 while (!list_empty(&clp->cl_sessions)) {
712 struct nfsd4_session *ses;
713 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
714 se_perclnt);
715 release_session(ses);
716 }
1da177e4
LT		 717 put_nfs4_client(clp);
718}
719
35bba9a3
BF		 720static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
721{
1da177e4
LT		 722 struct nfs4_client *clp;
723
35bba9a3
BF		 724 clp = alloc_client(name);
725 if (clp == NULL)
726 return NULL;
a55370a3 727 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 728 atomic_set(&clp->cl_count, 1);
c237dc03 729 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 730 INIT_LIST_HEAD(&clp->cl_idhash);
731 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 732 INIT_LIST_HEAD(&clp->cl_openowners);
733 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 734 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 735 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 736 return clp;
737}
738
35bba9a3
BF		 739static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
740{
741 memcpy(target->cl_verifier.data, source->data,
742 sizeof(target->cl_verifier.data));
1da177e4
LT		 743}
744
35bba9a3
BF		 745static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
746{
1da177e4
LT		 747 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
748 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
749}
750
35bba9a3
BF		 751static void copy_cred(struct svc_cred *target, struct svc_cred *source)
752{
1da177e4
LT		 753 target->cr_uid = source->cr_uid;
754 target->cr_gid = source->cr_gid;
755 target->cr_group_info = source->cr_group_info;
756 get_group_info(target->cr_group_info);
757}
758
35bba9a3 759static int same_name(const char *n1, const char *n2)
599e0a22 760{
a55370a3 761 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 762}
763
764static int
599e0a22
BF		 765same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
766{
767 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 768}
769
770static int
599e0a22
BF		 771same_clid(clientid_t *cl1, clientid_t *cl2)
772{
773 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 774}
775
776/* XXX what about NGROUP */
777static int
599e0a22
BF		 778same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
779{
780 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 781}
782
5ec7b46c
BF		 783static void gen_clid(struct nfs4_client *clp)
784{
785 static u32 current_clientid = 1;
786
1da177e4
LT		 787 clp->cl_clientid.cl_boot = boot_time;
788 clp->cl_clientid.cl_id = current_clientid++;
789}
790
deda2faa
BF		 791static void gen_confirm(struct nfs4_client *clp)
792{
793 static u32 i;
794 u32 *p;
1da177e4 795
1da177e4 796 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 797 *p++ = get_seconds();
798 *p++ = i++;
1da177e4
LT		 799}
800
35bba9a3
BF		 801static int check_name(struct xdr_netobj name)
802{
1da177e4
LT		 803 if (name.len == 0)
804 return 0;
805 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 806 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 807 return 0;
808 }
809 return 1;
810}
811
fd39ca9a 812static void
1da177e4
LT		 813add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
814{
815 unsigned int idhashval;
816
817 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
818 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
819 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
820 list_add_tail(&clp->cl_lru, &client_lru);
821 clp->cl_time = get_seconds();
822}
823
fd39ca9a 824static void
1da177e4
LT		 825move_to_confirmed(struct nfs4_client *clp)
826{
827 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
828 unsigned int strhashval;
829
830 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
831 list_del_init(&clp->cl_strhash);
f116629d 832 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 833 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 834 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
835 renew_client(clp);
836}
837
838static struct nfs4_client *
839find_confirmed_client(clientid_t *clid)
840{
841 struct nfs4_client *clp;
842 unsigned int idhashval = clientid_hashval(clid->cl_id);
843
844 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 845 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 846 return clp;
847 }
848 return NULL;
849}
850
851static struct nfs4_client *
852find_unconfirmed_client(clientid_t *clid)
853{
854 struct nfs4_client *clp;
855 unsigned int idhashval = clientid_hashval(clid->cl_id);
856
857 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 858 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 859 return clp;
860 }
861 return NULL;
862}
863
a1bcecd2
AA		 864/*
865 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
866 * parameter. Matching is based on the fact the at least one of the
867 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
868 *
869 * FIXME: we need to unify the clientid namespaces for nfsv4.x
870 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
871 * and SET_CLIENTID{,_CONFIRM}
872 */
873static inline int
874match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
875{
876 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
877 return use_exchange_id == has_exchange_flags;
878}
879
28ce6054 880static struct nfs4_client *
a1bcecd2
AA		 881find_confirmed_client_by_str(const char *dname, unsigned int hashval,
882 bool use_exchange_id)
28ce6054
N		 883{
884 struct nfs4_client *clp;
885
886 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 887 if (same_name(clp->cl_recdir, dname) &&
888 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 889 return clp;
890 }
891 return NULL;
892}
893
894static struct nfs4_client *
a1bcecd2
AA		 895find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
896 bool use_exchange_id)
28ce6054
N		 897{
898 struct nfs4_client *clp;
899
900 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 901 if (same_name(clp->cl_recdir, dname) &&
902 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 903 return clp;
904 }
905 return NULL;
906}
907
1da177e4
LT		 908/* a helper function for parse_callback */
909static int
910parse_octet(unsigned int *lenp, char **addrp)
911{
912 unsigned int len = *lenp;
913 char *p = *addrp;
914 int n = -1;
915 char c;
916
917 for (;;) {
918 if (!len)
919 break;
920 len--;
921 c = *p++;
922 if (c == '.')
923 break;
924 if ((c < '0') || (c > '9')) {
925 n = -1;
926 break;
927 }
928 if (n < 0)
929 n = 0;
930 n = (n * 10) + (c - '0');
931 if (n > 255) {
932 n = -1;
933 break;
934 }
935 }
936 *lenp = len;
937 *addrp = p;
938 return n;
939}
940
941/* parse and set the setclientid ipv4 callback address */
fd39ca9a 942static int
1da177e4
LT		 943parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
944{
945 int temp = 0;
946 u32 cbaddr = 0;
947 u16 cbport = 0;
948 u32 addrlen = addr_len;
949 char *addr = addr_val;
950 int i, shift;
951
952 /* ipaddress */
953 shift = 24;
954 for(i = 4; i > 0 ; i--) {
955 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
956 return 0;
957 }
958 cbaddr |= (temp << shift);
959 if (shift > 0)
960 shift -= 8;
961 }
962 *cbaddrp = cbaddr;
963
964 /* port */
965 shift = 8;
966 for(i = 2; i > 0 ; i--) {
967 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
968 return 0;
969 }
970 cbport |= (temp << shift);
971 if (shift > 0)
972 shift -= 8;
973 }
974 *cbportp = cbport;
975 return 1;
976}
977
fd39ca9a 978static void
1da177e4
LT		 979gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
980{
c237dc03 981 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT		 982
983 /* Currently, we only support tcp for the callback channel */
984 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
985 goto out_err;
986
987 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
988 &cb->cb_addr, &cb->cb_port)))
989 goto out_err;
ab52ae6d 990 cb->cb_minorversion = 0;
1da177e4
LT		 991 cb->cb_prog = se->se_callback_prog;
992 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 993 return;
994out_err:
849823c5 995 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 996 "will not receive delegations\n",
997 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
998
1da177e4
LT		 999 return;
1000}
1001
074fe897
AA		 1002void
1003nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
1004{
1005 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1006
1007 resp->cstate.statp = statp;
1008}
1009
1010/*
1011 * Dereference the result pages.
1012 */
1013static void
1014nfsd4_release_respages(struct page **respages, short resused)
1015{
1016 int i;
1017
1018 dprintk("--> %s\n", __func__);
1019 for (i = 0; i < resused; i++) {
1020 if (!respages[i])
1021 continue;
1022 put_page(respages[i]);
1023 respages[i] = NULL;
1024 }
1025}
1026
1027static void
1028nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1029{
1030 int i;
1031
1032 for (i = 0; i < count; i++) {
1033 topages[i] = frompages[i];
1034 if (!topages[i])
1035 continue;
1036 get_page(topages[i]);
1037 }
1038}
1039
1040/*
1041 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1042 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1043 * length of the XDR response is less than se_fmaxresp_cached
1044 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1045 * of the reply (e.g. readdir).
1046 *
1047 * Store the base and length of the rq_req.head[0] page
1048 * of the NFSv4.1 data, just past the rpc header.
1049 */
1050void
1051nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1052{
1053 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1054 struct svc_rqst *rqstp = resp->rqstp;
1055 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1056 struct nfsd4_op *op = &args->ops[resp->opcnt];
1057 struct kvec *resv = &rqstp->rq_res.head[0];
1058
1059 dprintk("--> %s entry %p\n", __func__, entry);
1060
1061 /* Don't cache a failed OP_SEQUENCE. */
1062 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1063 return;
bf864a31 1064
074fe897 1065 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1066 entry->ce_opcnt = resp->opcnt;
1067 entry->ce_status = resp->cstate.status;
1068
1069 /*
1070 * Don't need a page to cache just the sequence operation - the slot
1071 * does this for us!
1072 */
1073
1074 if (nfsd4_not_cached(resp)) {
1075 entry->ce_resused = 0;
1076 entry->ce_rpchdrlen = 0;
1077 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1078 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1079 return;
1080 }
074fe897
AA		 1081 entry->ce_resused = rqstp->rq_resused;
1082 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1083 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1084 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1085 entry->ce_resused);
074fe897
AA		 1086 entry->ce_datav.iov_base = resp->cstate.statp;
1087 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1088 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1089 /* Current request rpc header length*/
1090 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1091 (char *)page_address(rqstp->rq_respages[0]);
1092}
1093
1094/*
1095 * We keep the rpc header, but take the nfs reply from the replycache.
1096 */
1097static int
1098nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1099 struct nfsd4_cache_entry *entry)
1100{
1101 struct svc_rqst *rqstp = resp->rqstp;
1102 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1103 int len;
1104
1105 /* Current request rpc header length*/
1106 len = (char *)resp->cstate.statp -
1107 (char *)page_address(rqstp->rq_respages[0]);
1108 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1109 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1110 entry->ce_datav.iov_len);
1111 return 0;
1112 }
1113 /* copy the cached reply nfsd data past the current rpc header */
1114 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1115 entry->ce_datav.iov_len);
1116 resv->iov_len = len + entry->ce_datav.iov_len;
1117 return 1;
1118}
1119
1120/*
1121 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1122 * cached page. Replace any futher replay pages from the cache.
1123 */
1124__be32
bf864a31
AA		 1125nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1126 struct nfsd4_sequence *seq)
074fe897
AA		 1127{
1128 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1129 __be32 status;
1130
1131 dprintk("--> %s entry %p\n", __func__, entry);
1132
bf864a31
AA		 1133 /*
1134 * If this is just the sequence operation, we did not keep
1135 * a page in the cache entry because we can just use the
1136 * slot info stored in struct nfsd4_sequence that was checked
1137 * against the slot in nfsd4_sequence().
1138 *
1139 * This occurs when seq->cachethis is FALSE, or when the client
1140 * session inactivity timer fires and a solo sequence operation
1141 * is sent (lease renewal).
1142 */
1143 if (seq && nfsd4_not_cached(resp)) {
6c18ba9f 1144 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
bf864a31
AA		 1145 return nfs_ok;
1146 }
074fe897
AA		 1147
1148 if (!nfsd41_copy_replay_data(resp, entry)) {
1149 /*
1150 * Not enough room to use the replay rpc header, send the
1151 * cached header. Release all the allocated result pages.
1152 */
1153 svc_free_res_pages(resp->rqstp);
1154 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1155 entry->ce_resused);
1156 } else {
1157 /* Release all but the first allocated result page */
1158
1159 resp->rqstp->rq_resused--;
1160 svc_free_res_pages(resp->rqstp);
1161
1162 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1163 &entry->ce_respages[1],
1164 entry->ce_resused - 1);
1165 }
1166
1167 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1168 resp->opcnt = entry->ce_opcnt;
1169 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1170 status = entry->ce_status;
1171
1172 return status;
1173}
1174
0733d213
AA		 1175/*
1176 * Set the exchange_id flags returned by the server.
1177 */
1178static void
1179nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1180{
1181 /* pNFS is not supported */
1182 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1183
1184 /* Referrals are supported, Migration is not. */
1185 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1186
1187 /* set the wire flags to return to client. */
1188 clid->flags = new->cl_exchange_flags;
1189}
1190
069b6ad4
AA		 1191__be32
1192nfsd4_exchange_id(struct svc_rqst *rqstp,
1193 struct nfsd4_compound_state *cstate,
1194 struct nfsd4_exchange_id *exid)
1195{
0733d213
AA		 1196 struct nfs4_client *unconf, *conf, *new;
1197 int status;
1198 unsigned int strhashval;
1199 char dname[HEXDIR_LEN];
1200 nfs4_verifier verf = exid->verifier;
1201 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1202
1203 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1204 " ip_addr=%u flags %x, spa_how %d\n",
1205 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1206 ip_addr, exid->flags, exid->spa_how);
1207
1208 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1209 return nfserr_inval;
1210
1211 /* Currently only support SP4_NONE */
1212 switch (exid->spa_how) {
1213 case SP4_NONE:
1214 break;
1215 case SP4_SSV:
1216 return nfserr_encr_alg_unsupp;
1217 default:
1218 BUG(); /* checked by xdr code */
1219 case SP4_MACH_CRED:
1220 return nfserr_serverfault; /* no excuse :-/ */
1221 }
1222
1223 status = nfs4_make_rec_clidname(dname, &exid->clname);
1224
1225 if (status)
1226 goto error;
1227
1228 strhashval = clientstr_hashval(dname);
1229
1230 nfs4_lock_state();
1231 status = nfs_ok;
1232
a1bcecd2 1233 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1234 if (conf) {
1235 if (!same_verf(&verf, &conf->cl_verifier)) {
1236 /* 18.35.4 case 8 */
1237 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1238 status = nfserr_not_same;
1239 goto out;
1240 }
1241 /* Client reboot: destroy old state */
1242 expire_client(conf);
1243 goto out_new;
1244 }
1245 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1246 /* 18.35.4 case 9 */
1247 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1248 status = nfserr_perm;
1249 goto out;
1250 }
1251 expire_client(conf);
1252 goto out_new;
1253 }
0733d213
AA		 1254 /*
1255 * Set bit when the owner id and verifier map to an already
1256 * confirmed client id (18.35.3).
1257 */
1258 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1259
1260 /*
1261 * Falling into 18.35.4 case 2, possible router replay.
1262 * Leave confirmed record intact and return same result.
1263 */
1264 copy_verf(conf, &verf);
1265 new = conf;
1266 goto out_copy;
6ddbbbfe
MS		 1267 }
1268
1269 /* 18.35.4 case 7 */
1270 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1271 status = nfserr_noent;
1272 goto out;
0733d213
AA		 1273 }
1274
a1bcecd2 1275 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1276 if (unconf) {
1277 /*
1278 * Possible retry or client restart. Per 18.35.4 case 4,
1279 * a new unconfirmed record should be generated regardless
1280 * of whether any properties have changed.
1281 */
1282 expire_client(unconf);
1283 }
1284
1285out_new:
1286 /* Normal case */
1287 new = create_client(exid->clname, dname);
1288 if (new == NULL) {
1289 status = nfserr_resource;
1290 goto out;
1291 }
1292
1293 copy_verf(new, &verf);
1294 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1295 new->cl_addr = ip_addr;
1296 gen_clid(new);
1297 gen_confirm(new);
1298 add_to_unconfirmed(new, strhashval);
1299out_copy:
1300 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1301 exid->clientid.cl_id = new->cl_clientid.cl_id;
1302
38eb76a5
AA		 1303 new->cl_slot.sl_seqid = 0;
1304 exid->seqid = 1;
0733d213
AA		 1305 nfsd4_set_ex_flags(new, exid);
1306
1307 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1308 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1309 status = nfs_ok;
1310
1311out:
1312 nfs4_unlock_state();
1313error:
1314 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1315 return status;
069b6ad4
AA		 1316}
1317
b85d4c01
BH		 1318static int
1319check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1320{
1321 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1322 slot->sl_seqid);
1323
1324 /* The slot is in use, and no response has been sent. */
1325 if (slot->sl_inuse) {
1326 if (seqid == slot->sl_seqid)
1327 return nfserr_jukebox;
1328 else
1329 return nfserr_seq_misordered;
1330 }
1331 /* Normal */
1332 if (likely(seqid == slot->sl_seqid + 1))
1333 return nfs_ok;
1334 /* Replay */
1335 if (seqid == slot->sl_seqid)
1336 return nfserr_replay_cache;
1337 /* Wraparound */
1338 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1339 return nfs_ok;
1340 /* Misordered replay or misordered new request */
1341 return nfserr_seq_misordered;
1342}
1343
069b6ad4
AA		 1344__be32
1345nfsd4_create_session(struct svc_rqst *rqstp,
1346 struct nfsd4_compound_state *cstate,
1347 struct nfsd4_create_session *cr_ses)
1348{
ec6b5d7b 1349 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1350 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1351 struct nfs4_client *conf, *unconf;
38eb76a5 1352 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1353 int status = 0;
1354
1355 nfs4_lock_state();
1356 unconf = find_unconfirmed_client(&cr_ses->clientid);
1357 conf = find_confirmed_client(&cr_ses->clientid);
1358
1359 if (conf) {
38eb76a5
AA		 1360 slot = &conf->cl_slot;
1361 status = check_slot_seqid(cr_ses->seqid, slot);
1362 if (status == nfserr_replay_cache) {
ec6b5d7b 1363 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1364 slot->sl_seqid);
1365 cstate->slot = slot;
1366 cstate->status = status;
1367 /* Return the cached reply status */
bf864a31 1368 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1369 goto out;
1370 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1371 status = nfserr_seq_misordered;
1372 dprintk("Sequence misordered!\n");
1373 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1374 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1375 goto out;
1376 }
38eb76a5 1377 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1378 } else if (unconf) {
1379 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1380 (ip_addr != unconf->cl_addr)) {
1381 status = nfserr_clid_inuse;
1382 goto out;
1383 }
1384
38eb76a5
AA		 1385 slot = &unconf->cl_slot;
1386 status = check_slot_seqid(cr_ses->seqid, slot);
1387 if (status) {
1388 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1389 status = nfserr_seq_misordered;
1390 goto out;
1391 }
1392
38eb76a5 1393 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1394 move_to_confirmed(unconf);
1395
1396 /*
1397 * We do not support RDMA or persistent sessions
1398 */
1399 cr_ses->flags &= ~SESSION4_PERSIST;
1400 cr_ses->flags &= ~SESSION4_RDMA;
1401
1402 conf = unconf;
1403 } else {
1404 status = nfserr_stale_clientid;
1405 goto out;
1406 }
1407
1408 status = alloc_init_session(rqstp, conf, cr_ses);
1409 if (status)
1410 goto out;
1411
ec6b5d7b
AA		 1412 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1413 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1414 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1415
38eb76a5
AA		 1416 slot->sl_inuse = true;
1417 cstate->slot = slot;
bf864a31
AA		 1418 /* Ensure a page is used for the cache */
1419 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1420out:
1421 nfs4_unlock_state();
1422 dprintk("%s returns %d\n", __func__, ntohl(status));
1423 return status;
069b6ad4
AA		 1424}
1425
1426__be32
1427nfsd4_destroy_session(struct svc_rqst *r,
1428 struct nfsd4_compound_state *cstate,
1429 struct nfsd4_destroy_session *sessionid)
1430{
e10e0cfc
BH		 1431 struct nfsd4_session *ses;
1432 u32 status = nfserr_badsession;
1433
1434 /* Notes:
1435 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1436 * - Should we return nfserr_back_chan_busy if waiting for
1437 * callbacks on to-be-destroyed session?
1438 * - Do we need to clear any callback info from previous session?
1439 */
1440
1441 dump_sessionid(__func__, &sessionid->sessionid);
1442 spin_lock(&sessionid_lock);
1443 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1444 if (!ses) {
1445 spin_unlock(&sessionid_lock);
1446 goto out;
1447 }
1448
1449 unhash_session(ses);
1450 spin_unlock(&sessionid_lock);
1451
1452 /* wait for callbacks */
1453 shutdown_callback_client(ses->se_client);
1454 nfsd4_put_session(ses);
1455 status = nfs_ok;
1456out:
1457 dprintk("%s returns %d\n", __func__, ntohl(status));
1458 return status;
069b6ad4
AA		 1459}
1460
1461__be32
b85d4c01 1462nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1463 struct nfsd4_compound_state *cstate,
1464 struct nfsd4_sequence *seq)
1465{
f9bb94c4 1466 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1467 struct nfsd4_session *session;
1468 struct nfsd4_slot *slot;
1469 int status;
1470
f9bb94c4
AA		 1471 if (resp->opcnt != 1)
1472 return nfserr_sequence_pos;
1473
b85d4c01
BH		 1474 spin_lock(&sessionid_lock);
1475 status = nfserr_badsession;
1476 session = find_in_sessionid_hashtbl(&seq->sessionid);
1477 if (!session)
1478 goto out;
1479
1480 status = nfserr_badslot;
6c18ba9f 1481 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1482 goto out;
1483
1484 slot = &session->se_slots[seq->slotid];
1485 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1486
1487 status = check_slot_seqid(seq->seqid, slot);
1488 if (status == nfserr_replay_cache) {
1489 cstate->slot = slot;
1490 cstate->session = session;
da3846a2 1491 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1492 * for nfsd4_svc_encode_compoundres processing */
1493 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1494 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1495 goto replay_cache;
1496 }
1497 if (status)
1498 goto out;
1499
1500 /* Success! bump slot seqid */
1501 slot->sl_inuse = true;
1502 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1503 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1504 /* Always set the cache entry cachethis for solo sequence */
1505 if (nfsd4_is_solo_sequence(resp))
1506 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1507
1508 cstate->slot = slot;
1509 cstate->session = session;
1510
1511replay_cache:
1512 /* Renew the clientid on success and on replay.
1513 * Hold a session reference until done processing the compound:
1514 * nfsd4_put_session called only if the cstate slot is set.
1515 */
1516 renew_client(session->se_client);
1517 nfsd4_get_session(session);
1518out:
1519 spin_unlock(&sessionid_lock);
1520 dprintk("%s: return %d\n", __func__, ntohl(status));
1521 return status;
069b6ad4
AA		 1522}
1523
b37ad28b 1524__be32
b591480b
BF		 1525nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1526 struct nfsd4_setclientid *setclid)
1da177e4 1527{
27459f09 1528 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1529 struct xdr_netobj clname = {
1530 .len = setclid->se_namelen,
1531 .data = setclid->se_name,
1532 };
1533 nfs4_verifier clverifier = setclid->se_verf;
1534 unsigned int strhashval;
28ce6054 1535 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1536 __be32 status;
68e76ad0 1537 char *princ;
a55370a3 1538 char dname[HEXDIR_LEN];
1da177e4 1539
1da177e4 1540 if (!check_name(clname))
73aea4ec 1541 return nfserr_inval;
1da177e4 1542
a55370a3
N		 1543 status = nfs4_make_rec_clidname(dname, &clname);
1544 if (status)
73aea4ec 1545 return status;
a55370a3 1546
1da177e4
LT		 1547 /*
1548 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1549 * We get here on a DRC miss.
1550 */
1551
a55370a3 1552 strhashval = clientstr_hashval(dname);
1da177e4 1553
1da177e4 1554 nfs4_lock_state();
a1bcecd2 1555 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1556 if (conf) {
a186e767 1557 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1558 status = nfserr_clid_inuse;
026722c2
BF		 1559 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1560 dprintk("NFSD: setclientid: string in use by client"
1561 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1562 goto out;
1563 }
1da177e4 1564 }
a186e767
BF		 1565 /*
1566 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1567 * has a description of SETCLIENTID request processing consisting
1568 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1569 */
a1bcecd2 1570 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1571 status = nfserr_resource;
1572 if (!conf) {
a186e767
BF		 1573 /*
1574 * RFC 3530 14.2.33 CASE 4:
1575 * placed first, because it is the normal case
1da177e4
LT		 1576 */
1577 if (unconf)
1578 expire_client(unconf);
a55370a3
N		 1579 new = create_client(clname, dname);
1580 if (new == NULL)
1da177e4 1581 goto out;
1da177e4 1582 gen_clid(new);
599e0a22 1583 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1584 /*
a186e767
BF		 1585 * RFC 3530 14.2.33 CASE 1:
1586 * probable callback update
1da177e4 1587 */
31f4a6c1
N		 1588 if (unconf) {
1589 /* Note this is removing unconfirmed {*x***},
1590 * which is stronger than RFC recommended {vxc**}.
1591 * This has the advantage that there is at most
1592 * one {*x***} in either list at any time.
1593 */
1594 expire_client(unconf);
1da177e4 1595 }
a55370a3
N		 1596 new = create_client(clname, dname);
1597 if (new == NULL)
1da177e4 1598 goto out;
1da177e4 1599 copy_clid(new, conf);
1da177e4
LT		 1600 } else if (!unconf) {
1601 /*
a186e767
BF		 1602 * RFC 3530 14.2.33 CASE 2:
1603 * probable client reboot; state will be removed if
1604 * confirmed.
1da177e4 1605 */
a55370a3
N		 1606 new = create_client(clname, dname);
1607 if (new == NULL)
1da177e4 1608 goto out;
1da177e4 1609 gen_clid(new);
49ba8781 1610 } else {
a186e767
BF		 1611 /*
1612 * RFC 3530 14.2.33 CASE 3:
1613 * probable client reboot; state will be removed if
1614 * confirmed.
1da177e4
LT		 1615 */
1616 expire_client(unconf);
a55370a3
N		 1617 new = create_client(clname, dname);
1618 if (new == NULL)
1da177e4 1619 goto out;
1da177e4 1620 gen_clid(new);
1da177e4 1621 }
c175b83c
BF		 1622 copy_verf(new, &clverifier);
1623 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1624 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1625 princ = svc_gss_principal(rqstp);
1626 if (princ) {
1627 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1628 if (new->cl_principal == NULL) {
1629 free_client(new);
1630 goto out;
1631 }
1632 }
c175b83c
BF		 1633 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1634 gen_confirm(new);
1635 gen_callback(new, setclid);
1636 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1637 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1638 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1639 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1640 status = nfs_ok;
1641out:
1642 nfs4_unlock_state();
1643 return status;
1644}
1645
1646
1647/*
a186e767
BF		 1648 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1649 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1650 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1651 */
b37ad28b 1652__be32
b591480b
BF		 1653nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1654 struct nfsd4_compound_state *cstate,
1655 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1656{
27459f09 1657 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1658 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1659 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1660 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1661 __be32 status;
1da177e4
LT		 1662
1663 if (STALE_CLIENTID(clid))
1664 return nfserr_stale_clientid;
1665 /*
1666 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1667 * We get here on a DRC miss.
1668 */
1669
1670 nfs4_lock_state();
21ab45a4
N		 1671
1672 conf = find_confirmed_client(clid);
1673 unconf = find_unconfirmed_client(clid);
1674
1675 status = nfserr_clid_inuse;
27459f09 1676 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1677 goto out;
27459f09 1678 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1679 goto out;
1680
a186e767
BF		 1681 /*
1682 * section 14.2.34 of RFC 3530 has a description of
1683 * SETCLIENTID_CONFIRM request processing consisting
1684 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1685 */
366e0c1d 1686 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1687 /*
1688 * RFC 3530 14.2.34 CASE 1:
1689 * callback update
1690 */
599e0a22 1691 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1692 status = nfserr_clid_inuse;
1693 else {
1a69c179
N		 1694 /* XXX: We just turn off callbacks until we can handle
1695 * change request correctly. */
c237dc03 1696 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1697 expire_client(unconf);
1da177e4 1698 status = nfs_ok;
1a69c179 1699
1da177e4 1700 }
f3aba4e5 1701 } else if (conf && !unconf) {
a186e767
BF		 1702 /*
1703 * RFC 3530 14.2.34 CASE 2:
1704 * probable retransmitted request; play it safe and
1705 * do nothing.
7c79f737 1706 */
599e0a22 1707 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1708 status = nfserr_clid_inuse;
21ab45a4 1709 else
1da177e4 1710 status = nfs_ok;
7c79f737 1711 } else if (!conf && unconf
599e0a22 1712 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1713 /*
1714 * RFC 3530 14.2.34 CASE 3:
1715 * Normal case; new or rebooted client:
7c79f737 1716 */
599e0a22 1717 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1718 status = nfserr_clid_inuse;
1719 } else {
1a69c179
N		 1720 unsigned int hash =
1721 clientstr_hashval(unconf->cl_recdir);
1722 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1723 hash, false);
1a69c179 1724 if (conf) {
c7b9a459 1725 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1726 expire_client(conf);
1727 }
1da177e4 1728 move_to_confirmed(unconf);
21ab45a4 1729 conf = unconf;
46f8a64b 1730 nfsd4_probe_callback(conf);
1a69c179 1731 status = nfs_ok;
1da177e4 1732 }
599e0a22
BF		 1733 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1734 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1735 &confirm)))) {
a186e767
BF		 1736 /*
1737 * RFC 3530 14.2.34 CASE 4:
1738 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1739 */
1da177e4 1740 status = nfserr_stale_clientid;
7c79f737 1741 } else {
08e8987c
N		 1742 /* check that we have hit one of the cases...*/
1743 status = nfserr_clid_inuse;
1744 }
1da177e4 1745out:
1da177e4
LT		 1746 nfs4_unlock_state();
1747 return status;
1748}
1749
1da177e4
LT		 1750/* OPEN Share state helper functions */
1751static inline struct nfs4_file *
1752alloc_init_file(struct inode *ino)
1753{
1754 struct nfs4_file *fp;
1755 unsigned int hashval = file_hashval(ino);
1756
e60d4398
N		 1757 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1758 if (fp) {
8b671b80 1759 atomic_set(&fp->fi_ref, 1);
1da177e4 1760 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1761 INIT_LIST_HEAD(&fp->fi_stateids);
1762 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1763 spin_lock(&recall_lock);
1da177e4 1764 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1765 spin_unlock(&recall_lock);
1da177e4
LT		 1766 fp->fi_inode = igrab(ino);
1767 fp->fi_id = current_fileid++;
47f9940c 1768 fp->fi_had_conflict = false;
1da177e4
LT		 1769 return fp;
1770 }
1771 return NULL;
1772}
1773
e60d4398 1774static void
e18b890b 1775nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1776{
e60d4398
N		 1777 if (*slab == NULL)
1778 return;
1a1d92c1 1779 kmem_cache_destroy(*slab);
e60d4398 1780 *slab = NULL;
1da177e4
LT		 1781}
1782
e8ff2a84 1783void
1da177e4
LT		 1784nfsd4_free_slabs(void)
1785{
e60d4398
N		 1786 nfsd4_free_slab(&stateowner_slab);
1787 nfsd4_free_slab(&file_slab);
5ac049ac 1788 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1789 nfsd4_free_slab(&deleg_slab);
e60d4398 1790}
1da177e4 1791
e60d4398
N		 1792static int
1793nfsd4_init_slabs(void)
1794{
1795 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1796 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1797 if (stateowner_slab == NULL)
1798 goto out_nomem;
1799 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1800 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1801 if (file_slab == NULL)
1802 goto out_nomem;
5ac049ac 1803 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1804 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1805 if (stateid_slab == NULL)
1806 goto out_nomem;
5b2d21c1 1807 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1808 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1809 if (deleg_slab == NULL)
1810 goto out_nomem;
e60d4398
N		 1811 return 0;
1812out_nomem:
1813 nfsd4_free_slabs();
1814 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1815 return -ENOMEM;
1da177e4
LT		 1816}
1817
1818void
1819nfs4_free_stateowner(struct kref *kref)
1820{
1821 struct nfs4_stateowner *sop =
1822 container_of(kref, struct nfs4_stateowner, so_ref);
1823 kfree(sop->so_owner.data);
1824 kmem_cache_free(stateowner_slab, sop);
1825}
1826
1827static inline struct nfs4_stateowner *
1828alloc_stateowner(struct xdr_netobj *owner)
1829{
1830 struct nfs4_stateowner *sop;
1831
1832 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1833 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1834 memcpy(sop->so_owner.data, owner->data, owner->len);
1835 sop->so_owner.len = owner->len;
1836 kref_init(&sop->so_ref);
1837 return sop;
1838 }
1839 kmem_cache_free(stateowner_slab, sop);
1840 }
1841 return NULL;
1842}
1843
1844static struct nfs4_stateowner *
1845alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1846 struct nfs4_stateowner *sop;
1847 struct nfs4_replay *rp;
1848 unsigned int idhashval;
1849
1850 if (!(sop = alloc_stateowner(&open->op_owner)))
1851 return NULL;
1852 idhashval = ownerid_hashval(current_ownerid);
1853 INIT_LIST_HEAD(&sop->so_idhash);
1854 INIT_LIST_HEAD(&sop->so_strhash);
1855 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1856 INIT_LIST_HEAD(&sop->so_stateids);
1857 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1858 INIT_LIST_HEAD(&sop->so_close_lru);
1859 sop->so_time = 0;
1860 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1861 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1862 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1863 sop->so_is_open_owner = 1;
1864 sop->so_id = current_ownerid++;
1865 sop->so_client = clp;
1866 sop->so_seqid = open->op_seqid;
1867 sop->so_confirmed = 0;
1868 rp = &sop->so_replay;
de1ae286 1869 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1870 rp->rp_buflen = 0;
1871 rp->rp_buf = rp->rp_ibuf;
1872 return sop;
1873}
1874
1da177e4
LT		 1875static inline void
1876init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1877 struct nfs4_stateowner *sop = open->op_stateowner;
1878 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1879
1880 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1881 INIT_LIST_HEAD(&stp->st_perstateowner);
1882 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1883 INIT_LIST_HEAD(&stp->st_perfile);
1884 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1885 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1886 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1887 stp->st_stateowner = sop;
13cd2184 1888 get_nfs4_file(fp);
1da177e4 1889 stp->st_file = fp;
78155ed7 1890 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1891 stp->st_stateid.si_stateownerid = sop->so_id;
1892 stp->st_stateid.si_fileid = fp->fi_id;
1893 stp->st_stateid.si_generation = 0;
1894 stp->st_access_bmap = 0;
1895 stp->st_deny_bmap = 0;
84459a11
AA		 1896 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1897 &stp->st_access_bmap);
1da177e4 1898 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1899 stp->st_openstp = NULL;
1da177e4
LT		 1900}
1901
fd39ca9a 1902static void
1da177e4
LT		 1903move_to_close_lru(struct nfs4_stateowner *sop)
1904{
1905 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1906
358dd55a 1907 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1908 sop->so_time = get_seconds();
1909}
1910
1da177e4 1911static int
599e0a22
BF		 1912same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1913 clientid_t *clid)
1914{
1915 return (sop->so_owner.len == owner->len) &&
1916 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1917 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1918}
1919
1920static struct nfs4_stateowner *
1921find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1922{
1923 struct nfs4_stateowner *so = NULL;
1924
1925 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1926 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1927 return so;
1928 }
1929 return NULL;
1930}
1931
1932/* search file_hashtbl[] for file */
1933static struct nfs4_file *
1934find_file(struct inode *ino)
1935{
1936 unsigned int hashval = file_hashval(ino);
1937 struct nfs4_file *fp;
1938
8b671b80 1939 spin_lock(&recall_lock);
1da177e4 1940 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1941 if (fp->fi_inode == ino) {
1942 get_nfs4_file(fp);
8b671b80 1943 spin_unlock(&recall_lock);
1da177e4 1944 return fp;
13cd2184 1945 }
1da177e4 1946 }
8b671b80 1947 spin_unlock(&recall_lock);
1da177e4
LT		 1948 return NULL;
1949}
1950
d87a8ade 1951static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1952{
d87a8ade 1953 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1954 return 0;
d87a8ade
AA		 1955 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1956 return 0;
1957 x &= ~NFS4_SHARE_ACCESS_MASK;
1958 if (minorversion && x) {
1959 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1960 return 0;
1961 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1962 return 0;
1963 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1964 }
1965 if (x)
8838dc43
BF		 1966 return 0;
1967 return 1;
ba5a6a19
BF		 1968}
1969
8838dc43 1970static inline int deny_valid(u32 x)
ba5a6a19 1971{
8838dc43
BF		 1972 /* Note: unlike access bits, deny bits may be zero. */
1973 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1974}
1da177e4 1975
4f83aa30
BF		 1976/*
1977 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1978 * st_{access,deny}_bmap field of the stateid, in order to track not
1979 * only what share bits are currently in force, but also what
1980 * combinations of share bits previous opens have used. This allows us
1981 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1982 * return an error if the client attempt to downgrade to a combination
1983 * of share bits not explicable by closing some of its previous opens.
1984 *
1985 * XXX: This enforcement is actually incomplete, since we don't keep
1986 * track of access/deny bit combinations; so, e.g., we allow:
1987 *
1988 * OPEN allow read, deny write
1989 * OPEN allow both, deny none
1990 * DOWNGRADE allow read, deny none
1991 *
1992 * which we should reject.
1993 */
fd39ca9a 1994static void
1da177e4
LT		 1995set_access(unsigned int *access, unsigned long bmap) {
1996 int i;
1997
1998 *access = 0;
1999 for (i = 1; i < 4; i++) {
2000 if (test_bit(i, &bmap))
2001 *access |= i;
2002 }
2003}
2004
fd39ca9a 2005static void
1da177e4
LT		 2006set_deny(unsigned int *deny, unsigned long bmap) {
2007 int i;
2008
2009 *deny = 0;
2010 for (i = 0; i < 4; i++) {
2011 if (test_bit(i, &bmap))
2012 *deny |= i ;
2013 }
2014}
2015
2016static int
2017test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2018 unsigned int access, deny;
2019
2020 set_access(&access, stp->st_access_bmap);
2021 set_deny(&deny, stp->st_deny_bmap);
2022 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2023 return 0;
2024 return 1;
2025}
2026
2027/*
2028 * Called to check deny when READ with all zero stateid or
2029 * WRITE with all zero or all one stateid
2030 */
b37ad28b 2031static __be32
1da177e4
LT		 2032nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2033{
2034 struct inode *ino = current_fh->fh_dentry->d_inode;
2035 struct nfs4_file *fp;
2036 struct nfs4_stateid *stp;
b37ad28b 2037 __be32 ret;
1da177e4
LT		 2038
2039 dprintk("NFSD: nfs4_share_conflict\n");
2040
2041 fp = find_file(ino);
13cd2184
N		 2042 if (!fp)
2043 return nfs_ok;
b700949b 2044 ret = nfserr_locked;
1da177e4 2045 /* Search for conflicting share reservations */
13cd2184
N		 2046 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2047 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2048 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2049 goto out;
1da177e4 2050 }
13cd2184
N		 2051 ret = nfs_ok;
2052out:
2053 put_nfs4_file(fp);
2054 return ret;
1da177e4
LT		 2055}
2056
2057static inline void
2058nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2059{
2060 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2061 drop_file_write_access(filp);
1da177e4
LT		 2062 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2063 }
2064}
2065
1da177e4
LT		 2066/*
2067 * Spawn a thread to perform a recall on the delegation represented
2068 * by the lease (file_lock)
2069 *
2070 * Called from break_lease() with lock_kernel() held.
2071 * Note: we assume break_lease will only call this *once* for any given
2072 * lease.
2073 */
2074static
2075void nfsd_break_deleg_cb(struct file_lock *fl)
2076{
63e4863f 2077 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2078
2079 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2080 if (!dp)
2081 return;
2082
2083 /* We're assuming the state code never drops its reference
2084 * without first removing the lease. Since we're in this lease
2085 * callback (and since the lease code is serialized by the kernel
2086 * lock) we know the server hasn't removed the lease yet, we know
2087 * it's safe to take a reference: */
2088 atomic_inc(&dp->dl_count);
cfdcad4d 2089 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2090
2091 spin_lock(&recall_lock);
2092 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2093 spin_unlock(&recall_lock);
2094
2095 /* only place dl_time is set. protected by lock_kernel*/
2096 dp->dl_time = get_seconds();
2097
0272e1fd
BF		 2098 /*
2099 * We don't want the locks code to timeout the lease for us;
2100 * we'll remove it ourself if the delegation isn't returned
2101 * in time.
2102 */
2103 fl->fl_break_time = 0;
1da177e4 2104
63e4863f
BF		 2105 dp->dl_file->fi_had_conflict = true;
2106 nfsd4_cb_recall(dp);
1da177e4
LT		 2107}
2108
2109/*
2110 * The file_lock is being reapd.
2111 *
2112 * Called by locks_free_lock() with lock_kernel() held.
2113 */
2114static
2115void nfsd_release_deleg_cb(struct file_lock *fl)
2116{
2117 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2118
2119 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2120
2121 if (!(fl->fl_flags & FL_LEASE) || !dp)
2122 return;
2123 dp->dl_flock = NULL;
2124}
2125
2126/*
2127 * Set the delegation file_lock back pointer.
2128 *
a9933cea 2129 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2130 */
2131static
2132void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2133{
2134 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2135
2136 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2137 if (!dp)
2138 return;
2139 dp->dl_flock = new;
2140}
2141
2142/*
a9933cea 2143 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2144 */
2145static
2146int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2147{
2148 struct nfs4_delegation *onlistd =
2149 (struct nfs4_delegation *)onlist->fl_owner;
2150 struct nfs4_delegation *tryd =
2151 (struct nfs4_delegation *)try->fl_owner;
2152
2153 if (onlist->fl_lmops != try->fl_lmops)
2154 return 0;
2155
2156 return onlistd->dl_client == tryd->dl_client;
2157}
2158
2159
2160static
2161int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2162{
2163 if (arg & F_UNLCK)
2164 return lease_modify(onlist, arg);
2165 else
2166 return -EAGAIN;
2167}
2168
fd39ca9a 2169static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2170 .fl_break = nfsd_break_deleg_cb,
2171 .fl_release_private = nfsd_release_deleg_cb,
2172 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2173 .fl_mylease = nfsd_same_client_deleg_cb,
2174 .fl_change = nfsd_change_deleg_cb,
2175};
2176
2177
b37ad28b 2178__be32
6668958f
AA		 2179nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2180 struct nfsd4_open *open)
1da177e4 2181{
1da177e4
LT		 2182 clientid_t *clientid = &open->op_clientid;
2183 struct nfs4_client *clp = NULL;
2184 unsigned int strhashval;
2185 struct nfs4_stateowner *sop = NULL;
2186
1da177e4 2187 if (!check_name(open->op_owner))
0f442aa2 2188 return nfserr_inval;
1da177e4
LT		 2189
2190 if (STALE_CLIENTID(&open->op_clientid))
2191 return nfserr_stale_clientid;
2192
2193 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2194 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2195 open->op_stateowner = sop;
2196 if (!sop) {
2197 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2198 clp = find_confirmed_client(clientid);
2199 if (clp == NULL)
0f442aa2
BF		 2200 return nfserr_expired;
2201 goto renew;
1da177e4 2202 }
6668958f
AA		 2203 /* When sessions are used, skip open sequenceid processing */
2204 if (nfsd4_has_session(cstate))
2205 goto renew;
0f442aa2
BF		 2206 if (!sop->so_confirmed) {
2207 /* Replace unconfirmed owners without checking for replay. */
2208 clp = sop->so_client;
f044ff83 2209 release_openowner(sop);
0f442aa2
BF		 2210 open->op_stateowner = NULL;
2211 goto renew;
2212 }
2213 if (open->op_seqid == sop->so_seqid - 1) {
2214 if (sop->so_replay.rp_buflen)
a90b061c 2215 return nfserr_replay_me;
0f442aa2
BF		 2216 /* The original OPEN failed so spectacularly
2217 * that we don't even have replay data saved!
2218 * Therefore, we have no choice but to continue
2219 * processing this OPEN; presumably, we'll
2220 * fail again for the same reason.
2221 */
2222 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2223 goto renew;
2224 }
2225 if (open->op_seqid != sop->so_seqid)
2226 return nfserr_bad_seqid;
1da177e4 2227renew:
0f442aa2
BF		 2228 if (open->op_stateowner == NULL) {
2229 sop = alloc_init_open_stateowner(strhashval, clp, open);
2230 if (sop == NULL)
2231 return nfserr_resource;
2232 open->op_stateowner = sop;
2233 }
2234 list_del_init(&sop->so_close_lru);
1da177e4 2235 renew_client(sop->so_client);
0f442aa2 2236 return nfs_ok;
1da177e4
LT		 2237}
2238
b37ad28b 2239static inline __be32
4a6e43e6
N		 2240nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2241{
2242 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2243 return nfserr_openmode;
2244 else
2245 return nfs_ok;
2246}
2247
52f4fb43
N		 2248static struct nfs4_delegation *
2249find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2250{
2251 struct nfs4_delegation *dp;
2252
ea1da636 2253 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2254 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2255 return dp;
2256 }
2257 return NULL;
2258}
2259
b37ad28b 2260static __be32
567d9829
N		 2261nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2262 struct nfs4_delegation **dp)
2263{
2264 int flags;
b37ad28b 2265 __be32 status = nfserr_bad_stateid;
567d9829
N		 2266
2267 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2268 if (*dp == NULL)
c44c5eeb 2269 goto out;
567d9829
N		 2270 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2271 RD_STATE : WR_STATE;
2272 status = nfs4_check_delegmode(*dp, flags);
2273 if (status)
2274 *dp = NULL;
c44c5eeb
N		 2275out:
2276 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2277 return nfs_ok;
2278 if (status)
2279 return status;
2280 open->op_stateowner->so_confirmed = 1;
2281 return nfs_ok;
567d9829
N		 2282}
2283
b37ad28b 2284static __be32
1da177e4
LT		 2285nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2286{
2287 struct nfs4_stateid *local;
b37ad28b 2288 __be32 status = nfserr_share_denied;
1da177e4
LT		 2289 struct nfs4_stateowner *sop = open->op_stateowner;
2290
8beefa24 2291 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2292 /* ignore lock owners */
2293 if (local->st_stateowner->so_is_open_owner == 0)
2294 continue;
2295 /* remember if we have seen this open owner */
2296 if (local->st_stateowner == sop)
2297 *stpp = local;
2298 /* check for conflicting share reservations */
2299 if (!test_share(local, open))
2300 goto out;
2301 }
2302 status = 0;
2303out:
2304 return status;
2305}
2306
5ac049ac
N		 2307static inline struct nfs4_stateid *
2308nfs4_alloc_stateid(void)
2309{
2310 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2311}
2312
b37ad28b 2313static __be32
1da177e4 2314nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2315 struct nfs4_delegation *dp,
1da177e4
LT		 2316 struct svc_fh *cur_fh, int flags)
2317{
2318 struct nfs4_stateid *stp;
1da177e4 2319
5ac049ac 2320 stp = nfs4_alloc_stateid();
1da177e4
LT		 2321 if (stp == NULL)
2322 return nfserr_resource;
2323
567d9829
N		 2324 if (dp) {
2325 get_file(dp->dl_vfs_file);
2326 stp->st_vfs_file = dp->dl_vfs_file;
2327 } else {
b37ad28b 2328 __be32 status;
567d9829
N		 2329 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2330 &stp->st_vfs_file);
2331 if (status) {
2332 if (status == nfserr_dropit)
2333 status = nfserr_jukebox;
5ac049ac 2334 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2335 return status;
2336 }
1da177e4 2337 }
1da177e4
LT		 2338 *stpp = stp;
2339 return 0;
2340}
2341
b37ad28b 2342static inline __be32
1da177e4
LT		 2343nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2344 struct nfsd4_open *open)
2345{
2346 struct iattr iattr = {
2347 .ia_valid = ATTR_SIZE,
2348 .ia_size = 0,
2349 };
2350 if (!open->op_truncate)
2351 return 0;
2352 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2353 return nfserr_inval;
1da177e4
LT		 2354 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2355}
2356
b37ad28b 2357static __be32
1da177e4
LT		 2358nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2359{
2360 struct file *filp = stp->st_vfs_file;
7eaa36e2 2361 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2362 unsigned int share_access, new_writer;
b37ad28b 2363 __be32 status;
1da177e4
LT		 2364
2365 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2366 new_writer = (~share_access) & open->op_share_access
2367 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2368
6c26d08f 2369 if (new_writer) {
b37ad28b
AV		 2370 int err = get_write_access(inode);
2371 if (err)
2372 return nfserrno(err);
e518f056
BH		 2373 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2374 if (err)
2375 return nfserrno(err);
2376 file_take_write(filp);
6c26d08f 2377 }
1da177e4
LT		 2378 status = nfsd4_truncate(rqstp, cur_fh, open);
2379 if (status) {
6c26d08f
BF		 2380 if (new_writer)
2381 put_write_access(inode);
1da177e4
LT		 2382 return status;
2383 }
2384 /* remember the open */
6c26d08f 2385 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2386 __set_bit(open->op_share_access, &stp->st_access_bmap);
2387 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2388
2389 return nfs_ok;
2390}
2391
2392
1da177e4 2393static void
37515177 2394nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2395{
37515177
N		 2396 open->op_stateowner->so_confirmed = 1;
2397 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2398}
2399
2400/*
2401 * Attempt to hand out a delegation.
2402 */
2403static void
2404nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2405{
2406 struct nfs4_delegation *dp;
2407 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2408 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2409 struct file_lock fl, *flp = &fl;
2410 int status, flag = 0;
2411
2412 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2413 open->op_recall = 0;
2414 switch (open->op_claim_type) {
2415 case NFS4_OPEN_CLAIM_PREVIOUS:
2416 if (!atomic_read(&cb->cb_set))
2417 open->op_recall = 1;
2418 flag = open->op_delegate_type;
2419 if (flag == NFS4_OPEN_DELEGATE_NONE)
2420 goto out;
2421 break;
2422 case NFS4_OPEN_CLAIM_NULL:
2423 /* Let's not give out any delegations till everyone's
2424 * had the chance to reclaim theirs.... */
af558e33 2425 if (locks_in_grace())
7b190fec
N		 2426 goto out;
2427 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2428 goto out;
2429 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2430 flag = NFS4_OPEN_DELEGATE_WRITE;
2431 else
2432 flag = NFS4_OPEN_DELEGATE_READ;
2433 break;
2434 default:
2435 goto out;
2436 }
1da177e4
LT		 2437
2438 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2439 if (dp == NULL) {
2440 flag = NFS4_OPEN_DELEGATE_NONE;
2441 goto out;
2442 }
2443 locks_init_lock(&fl);
2444 fl.fl_lmops = &nfsd_lease_mng_ops;
2445 fl.fl_flags = FL_LEASE;
9167f501 2446 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2447 fl.fl_end = OFFSET_MAX;
2448 fl.fl_owner = (fl_owner_t)dp;
2449 fl.fl_file = stp->st_vfs_file;
2450 fl.fl_pid = current->tgid;
2451
a9933cea 2452 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2453 * the lock_manager callbacks fl_mylease and fl_change are used
2454 */
9167f501 2455 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2456 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2457 unhash_delegation(dp);
1da177e4
LT		 2458 flag = NFS4_OPEN_DELEGATE_NONE;
2459 goto out;
2460 }
2461
2462 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2463
2464 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2465 dp->dl_stateid.si_boot,
2466 dp->dl_stateid.si_stateownerid,
2467 dp->dl_stateid.si_fileid,
2468 dp->dl_stateid.si_generation);
2469out:
7b190fec
N		 2470 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2471 && flag == NFS4_OPEN_DELEGATE_NONE
2472 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2473 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2474 open->op_delegate_type = flag;
2475}
2476
2477/*
2478 * called with nfs4_lock_state() held.
2479 */
b37ad28b 2480__be32
1da177e4
LT		 2481nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2482{
6668958f 2483 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2484 struct nfs4_file *fp = NULL;
2485 struct inode *ino = current_fh->fh_dentry->d_inode;
2486 struct nfs4_stateid *stp = NULL;
567d9829 2487 struct nfs4_delegation *dp = NULL;
b37ad28b 2488 __be32 status;
1da177e4
LT		 2489
2490 status = nfserr_inval;
d87a8ade 2491 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2492 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2493 goto out;
2494 /*
2495 * Lookup file; if found, lookup stateid and check open request,
2496 * and check for delegations in the process of being recalled.
2497 * If not found, create the nfs4_file struct
2498 */
2499 fp = find_file(ino);
2500 if (fp) {
2501 if ((status = nfs4_check_open(fp, open, &stp)))
2502 goto out;
c44c5eeb
N		 2503 status = nfs4_check_deleg(fp, open, &dp);
2504 if (status)
2505 goto out;
1da177e4 2506 } else {
c44c5eeb
N		 2507 status = nfserr_bad_stateid;
2508 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2509 goto out;
1da177e4
LT		 2510 status = nfserr_resource;
2511 fp = alloc_init_file(ino);
2512 if (fp == NULL)
2513 goto out;
2514 }
2515
2516 /*
2517 * OPEN the file, or upgrade an existing OPEN.
2518 * If truncate fails, the OPEN fails.
2519 */
2520 if (stp) {
2521 /* Stateid was found, this is an OPEN upgrade */
2522 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2523 if (status)
2524 goto out;
444c2c07 2525 update_stateid(&stp->st_stateid);
1da177e4
LT		 2526 } else {
2527 /* Stateid was not found, this is a new OPEN */
2528 int flags = 0;
9ecb6a08 2529 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2530 flags |= NFSD_MAY_READ;
1da177e4 2531 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2532 flags |= NFSD_MAY_WRITE;
567d9829
N		 2533 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2534 if (status)
1da177e4
LT		 2535 goto out;
2536 init_stateid(stp, fp, open);
2537 status = nfsd4_truncate(rqstp, current_fh, open);
2538 if (status) {
2283963f 2539 release_open_stateid(stp);
1da177e4
LT		 2540 goto out;
2541 }
6668958f
AA		 2542 if (nfsd4_has_session(&resp->cstate))
2543 update_stateid(&stp->st_stateid);
1da177e4
LT		 2544 }
2545 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2546
6668958f
AA		 2547 if (nfsd4_has_session(&resp->cstate))
2548 open->op_stateowner->so_confirmed = 1;
2549
1da177e4
LT		 2550 /*
2551 * Attempt to hand out a delegation. No error return, because the
2552 * OPEN succeeds even if we fail.
2553 */
2554 nfs4_open_delegation(current_fh, open, stp);
2555
2556 status = nfs_ok;
2557
2558 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2559 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2560 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2561out:
13cd2184
N		 2562 if (fp)
2563 put_nfs4_file(fp);
37515177
N		 2564 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2565 nfs4_set_claim_prev(open);
1da177e4
LT		 2566 /*
2567 * To finish the open response, we just need to set the rflags.
2568 */
2569 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2570 if (!open->op_stateowner->so_confirmed &&
2571 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2572 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2573
2574 return status;
2575}
2576
b37ad28b 2577__be32
b591480b
BF		 2578nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2579 clientid_t *clid)
1da177e4
LT		 2580{
2581 struct nfs4_client *clp;
b37ad28b 2582 __be32 status;
1da177e4
LT		 2583
2584 nfs4_lock_state();
2585 dprintk("process_renew(%08x/%08x): starting\n",
2586 clid->cl_boot, clid->cl_id);
2587 status = nfserr_stale_clientid;
2588 if (STALE_CLIENTID(clid))
2589 goto out;
2590 clp = find_confirmed_client(clid);
2591 status = nfserr_expired;
2592 if (clp == NULL) {
2593 /* We assume the client took too long to RENEW. */
2594 dprintk("nfsd4_renew: clientid not found!\n");
2595 goto out;
2596 }
2597 renew_client(clp);
2598 status = nfserr_cb_path_down;
ea1da636 2599 if (!list_empty(&clp->cl_delegations)
c237dc03 2600 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2601 goto out;
2602 status = nfs_ok;
2603out:
2604 nfs4_unlock_state();
2605 return status;
2606}
2607
af558e33
BF		 2608struct lock_manager nfsd4_manager = {
2609};
2610
a76b4319 2611static void
af558e33 2612nfsd4_end_grace(void)
a76b4319
N		 2613{
2614 dprintk("NFSD: end of grace period\n");
c7b9a459 2615 nfsd4_recdir_purge_old();
af558e33 2616 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2617}
2618
fd39ca9a 2619static time_t
1da177e4
LT		 2620nfs4_laundromat(void)
2621{
2622 struct nfs4_client *clp;
2623 struct nfs4_stateowner *sop;
2624 struct nfs4_delegation *dp;
2625 struct list_head *pos, *next, reaplist;
2626 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2627 time_t t, clientid_val = NFSD_LEASE_TIME;
2628 time_t u, test_val = NFSD_LEASE_TIME;
2629
2630 nfs4_lock_state();
2631
2632 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2633 if (locks_in_grace())
2634 nfsd4_end_grace();
1da177e4
LT		 2635 list_for_each_safe(pos, next, &client_lru) {
2636 clp = list_entry(pos, struct nfs4_client, cl_lru);
2637 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2638 t = clp->cl_time - cutoff;
2639 if (clientid_val > t)
2640 clientid_val = t;
2641 break;
2642 }
2643 dprintk("NFSD: purging unused client (clientid %08x)\n",
2644 clp->cl_clientid.cl_id);
c7b9a459 2645 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2646 expire_client(clp);
2647 }
2648 INIT_LIST_HEAD(&reaplist);
2649 spin_lock(&recall_lock);
2650 list_for_each_safe(pos, next, &del_recall_lru) {
2651 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2652 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2653 u = dp->dl_time - cutoff;
2654 if (test_val > u)
2655 test_val = u;
2656 break;
2657 }
2658 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2659 dp, dp->dl_flock);
2660 list_move(&dp->dl_recall_lru, &reaplist);
2661 }
2662 spin_unlock(&recall_lock);
2663 list_for_each_safe(pos, next, &reaplist) {
2664 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2665 list_del_init(&dp->dl_recall_lru);
2666 unhash_delegation(dp);
2667 }
2668 test_val = NFSD_LEASE_TIME;
2669 list_for_each_safe(pos, next, &close_lru) {
2670 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2671 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2672 u = sop->so_time - cutoff;
2673 if (test_val > u)
2674 test_val = u;
2675 break;
2676 }
2677 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2678 sop->so_id);
f044ff83 2679 release_openowner(sop);
1da177e4
LT		 2680 }
2681 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2682 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2683 nfs4_unlock_state();
2684 return clientid_val;
2685}
2686
a254b246
HH		 2687static struct workqueue_struct *laundry_wq;
2688static void laundromat_main(struct work_struct *);
2689static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2690
2691static void
c4028958 2692laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2693{
2694 time_t t;
2695
2696 t = nfs4_laundromat();
2697 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2698 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2699}
2700
fd39ca9a 2701static struct nfs4_stateowner *
f8816512
N		 2702search_close_lru(u32 st_id, int flags)
2703{
1da177e4
LT		 2704 struct nfs4_stateowner *local = NULL;
2705
1da177e4
LT		 2706 if (flags & CLOSE_STATE) {
2707 list_for_each_entry(local, &close_lru, so_close_lru) {
2708 if (local->so_id == st_id)
2709 return local;
2710 }
2711 }
2712 return NULL;
2713}
2714
2715static inline int
2716nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2717{
7eaa36e2 2718 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2719}
2720
2721static int
2722STALE_STATEID(stateid_t *stateid)
2723{
78155ed7
BN		 2724 if (time_after((unsigned long)boot_time,
2725 (unsigned long)stateid->si_boot)) {
2726 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2727 stateid->si_boot, stateid->si_stateownerid,
2728 stateid->si_fileid, stateid->si_generation);
2729 return 1;
2730 }
2731 return 0;
2732}
2733
2734static int
2735EXPIRED_STATEID(stateid_t *stateid)
2736{
2737 if (time_before((unsigned long)boot_time,
2738 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2739 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2740 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2741 stateid->si_boot, stateid->si_stateownerid,
2742 stateid->si_fileid, stateid->si_generation);
2743 return 1;
2744 }
2745 return 0;
2746}
2747
2748static __be32
2749stateid_error_map(stateid_t *stateid)
2750{
2751 if (STALE_STATEID(stateid))
2752 return nfserr_stale_stateid;
2753 if (EXPIRED_STATEID(stateid))
2754 return nfserr_expired;
2755
2756 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2757 stateid->si_boot, stateid->si_stateownerid,
2758 stateid->si_fileid, stateid->si_generation);
2759 return nfserr_bad_stateid;
1da177e4
LT		 2760}
2761
2762static inline int
2763access_permit_read(unsigned long access_bmap)
2764{
2765 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2766 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2767 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2768}
2769
2770static inline int
2771access_permit_write(unsigned long access_bmap)
2772{
2773 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2774 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2775}
2776
2777static
b37ad28b 2778__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2779{
b37ad28b 2780 __be32 status = nfserr_openmode;
1da177e4
LT		 2781
2782 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2783 goto out;
2784 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2785 goto out;
2786 status = nfs_ok;
2787out:
2788 return status;
2789}
2790
b37ad28b 2791static inline __be32
1da177e4
LT		 2792check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2793{
203a8c8e 2794 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2795 return nfs_ok;
af558e33 2796 else if (locks_in_grace()) {
1da177e4
LT		 2797 /* Answer in remaining cases depends on existance of
2798 * conflicting state; so we must wait out the grace period. */
2799 return nfserr_grace;
2800 } else if (flags & WR_STATE)
2801 return nfs4_share_conflict(current_fh,
2802 NFS4_SHARE_DENY_WRITE);
2803 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2804 return nfs4_share_conflict(current_fh,
2805 NFS4_SHARE_DENY_READ);
2806}
2807
2808/*
2809 * Allow READ/WRITE during grace period on recovered state only for files
2810 * that are not able to provide mandatory locking.
2811 */
2812static inline int
18f82731 2813grace_disallows_io(struct inode *inode)
1da177e4 2814{
203a8c8e 2815 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2816}
2817
6668958f 2818static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2819{
6668958f
AA		 2820 /*
2821 * When sessions are used the stateid generation number is ignored
2822 * when it is zero.
2823 */
2824 if ((flags & HAS_SESSION) && in->si_generation == 0)
2825 goto out;
2826
0836f587
BF		 2827 /* If the client sends us a stateid from the future, it's buggy: */
2828 if (in->si_generation > ref->si_generation)
2829 return nfserr_bad_stateid;
2830 /*
2831 * The following, however, can happen. For example, if the
2832 * client sends an open and some IO at the same time, the open
2833 * may bump si_generation while the IO is still in flight.
2834 * Thanks to hard links and renames, the client never knows what
2835 * file an open will affect. So it could avoid that situation
2836 * only by serializing all opens and IO from the same open
2837 * owner. To recover from the old_stateid error, the client
2838 * will just have to retry the IO:
2839 */
2840 if (in->si_generation < ref->si_generation)
2841 return nfserr_old_stateid;
6668958f 2842out:
0836f587
BF		 2843 return nfs_ok;
2844}
2845
3e633079
BF		 2846static int is_delegation_stateid(stateid_t *stateid)
2847{
2848 return stateid->si_fileid == 0;
2849}
2850
1da177e4
LT		 2851/*
2852* Checks for stateid operations
2853*/
b37ad28b 2854__be32
dd453dfd
BH		 2855nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2856 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2857{
2858 struct nfs4_stateid *stp = NULL;
2859 struct nfs4_delegation *dp = NULL;
dd453dfd 2860 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2861 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2862 __be32 status;
1da177e4 2863
1da177e4
LT		 2864 if (filpp)
2865 *filpp = NULL;
2866
18f82731 2867 if (grace_disallows_io(ino))
1da177e4
LT		 2868 return nfserr_grace;
2869
6668958f
AA		 2870 if (nfsd4_has_session(cstate))
2871 flags |= HAS_SESSION;
2872
1da177e4
LT		 2873 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2874 return check_special_stateids(current_fh, stateid, flags);
2875
1da177e4
LT		 2876 status = nfserr_stale_stateid;
2877 if (STALE_STATEID(stateid))
2878 goto out;
2879
1da177e4 2880 status = nfserr_bad_stateid;
3e633079 2881 if (is_delegation_stateid(stateid)) {
a4455be0 2882 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2883 if (!dp) {
2884 status = stateid_error_map(stateid);
1da177e4 2885 goto out;
78155ed7 2886 }
6668958f
AA		 2887 status = check_stateid_generation(stateid, &dp->dl_stateid,
2888 flags);
0c2a498f
BF		 2889 if (status)
2890 goto out;
dc9bf700
BF		 2891 status = nfs4_check_delegmode(dp, flags);
2892 if (status)
2893 goto out;
2894 renew_client(dp->dl_client);
dc9bf700
BF		 2895 if (filpp)
2896 *filpp = dp->dl_vfs_file;
1da177e4 2897 } else { /* open or lock stateid */
a4455be0 2898 stp = find_stateid(stateid, flags);
78155ed7
BN		 2899 if (!stp) {
2900 status = stateid_error_map(stateid);
1da177e4 2901 goto out;
78155ed7 2902 }
6150ef0d 2903 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2904 goto out;
2905 if (!stp->st_stateowner->so_confirmed)
2906 goto out;
6668958f
AA		 2907 status = check_stateid_generation(stateid, &stp->st_stateid,
2908 flags);
0c2a498f
BF		 2909 if (status)
2910 goto out;
a4455be0
BF		 2911 status = nfs4_check_openmode(stp, flags);
2912 if (status)
1da177e4
LT		 2913 goto out;
2914 renew_client(stp->st_stateowner->so_client);
2915 if (filpp)
2916 *filpp = stp->st_vfs_file;
1da177e4
LT		 2917 }
2918 status = nfs_ok;
2919out:
2920 return status;
2921}
2922
4c4cd222
N		 2923static inline int
2924setlkflg (int type)
2925{
2926 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2927 RD_STATE : WR_STATE;
2928}
1da177e4
LT		 2929
2930/*
2931 * Checks for sequence id mutating operations.
2932 */
b37ad28b 2933static __be32
dd453dfd
BH		 2934nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2935 stateid_t *stateid, int flags,
2936 struct nfs4_stateowner **sopp,
2937 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2938{
1da177e4
LT		 2939 struct nfs4_stateid *stp;
2940 struct nfs4_stateowner *sop;
dd453dfd 2941 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2942 __be32 status;
1da177e4
LT		 2943
2944 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2945 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2946 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2947 stateid->si_generation);
3a4f98bb 2948
1da177e4
LT		 2949 *stpp = NULL;
2950 *sopp = NULL;
2951
1da177e4 2952 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2953 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2954 return nfserr_bad_stateid;
1da177e4
LT		 2955 }
2956
1da177e4 2957 if (STALE_STATEID(stateid))
3a4f98bb 2958 return nfserr_stale_stateid;
6668958f
AA		 2959
2960 if (nfsd4_has_session(cstate))
2961 flags |= HAS_SESSION;
2962
1da177e4
LT		 2963 /*
2964 * We return BAD_STATEID if filehandle doesn't match stateid,
2965 * the confirmed flag is incorrecly set, or the generation
2966 * number is incorrect.
1da177e4 2967 */
f8816512
N		 2968 stp = find_stateid(stateid, flags);
2969 if (stp == NULL) {
2970 /*
2971 * Also, we should make sure this isn't just the result of
2972 * a replayed close:
2973 */
2974 sop = search_close_lru(stateid->si_stateownerid, flags);
2975 if (sop == NULL)
78155ed7 2976 return stateid_error_map(stateid);
f8816512
N		 2977 *sopp = sop;
2978 goto check_replay;
2979 }
1da177e4 2980
39325bd0
BF		 2981 *stpp = stp;
2982 *sopp = sop = stp->st_stateowner;
2983
4c4cd222 2984 if (lock) {
4c4cd222 2985 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2986 struct nfs4_client *clp = sop->so_client;
4c4cd222 2987 int lkflg = 0;
b37ad28b 2988 __be32 status;
4c4cd222
N		 2989
2990 lkflg = setlkflg(lock->lk_type);
2991
2992 if (lock->lk_is_new) {
599e0a22
BF		 2993 if (!sop->so_is_open_owner)
2994 return nfserr_bad_stateid;
60adfc50
AA		 2995 if (!(flags & HAS_SESSION) &&
2996 !same_clid(&clp->cl_clientid, lockclid))
2997 return nfserr_bad_stateid;
599e0a22
BF		 2998 /* stp is the open stateid */
2999 status = nfs4_check_openmode(stp, lkflg);
3000 if (status)
3001 return status;
3002 } else {
3003 /* stp is the lock stateid */
3004 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3005 if (status)
3006 return status;
4c4cd222 3007 }
1da177e4
LT		 3008 }
3009
f3362737 3010 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3011 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3012 return nfserr_bad_stateid;
1da177e4
LT		 3013 }
3014
1da177e4
LT		 3015 /*
3016 * We now validate the seqid and stateid generation numbers.
3017 * For the moment, we ignore the possibility of
3018 * generation number wraparound.
3019 */
6668958f 3020 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3021 goto check_replay;
3022
3a4f98bb 3023 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3024 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3025 " unconfirmed stateowner!\n");
3026 return nfserr_bad_stateid;
1da177e4 3027 }
3a4f98bb 3028 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3029 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3030 " confirmed yet!\n");
3031 return nfserr_bad_stateid;
1da177e4 3032 }
6668958f 3033 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3034 if (status)
3035 return status;
52fd004e 3036 renew_client(sop->so_client);
3a4f98bb 3037 return nfs_ok;
1da177e4 3038
1da177e4 3039check_replay:
bd9aac52 3040 if (seqid == sop->so_seqid - 1) {
849823c5 3041 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3042 /* indicate replay to calling function */
a90b061c 3043 return nfserr_replay_me;
1da177e4 3044 }
2fdada03 3045 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3046 sop->so_seqid, seqid);
3047 *sopp = NULL;
3048 return nfserr_bad_seqid;
1da177e4
LT		 3049}
3050
b37ad28b 3051__be32
ca364317 3052nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3053 struct nfsd4_open_confirm *oc)
1da177e4 3054{
b37ad28b 3055 __be32 status;
1da177e4
LT		 3056 struct nfs4_stateowner *sop;
3057 struct nfs4_stateid *stp;
3058
3059 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3060 (int)cstate->current_fh.fh_dentry->d_name.len,
3061 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3062
ca364317 3063 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3064 if (status)
3065 return status;
1da177e4
LT		 3066
3067 nfs4_lock_state();
3068
dd453dfd 3069 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3070 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3071 CONFIRM | OPEN_STATE,
1da177e4
LT		 3072 &oc->oc_stateowner, &stp, NULL)))
3073 goto out;
3074
3075 sop = oc->oc_stateowner;
3076 sop->so_confirmed = 1;
3077 update_stateid(&stp->st_stateid);
3078 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3079 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3080 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3081 stp->st_stateid.si_boot,
3082 stp->st_stateid.si_stateownerid,
3083 stp->st_stateid.si_fileid,
3084 stp->st_stateid.si_generation);
c7b9a459
N		 3085
3086 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3087out:
f2327d9a 3088 if (oc->oc_stateowner) {
1da177e4 3089 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3090 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3091 }
1da177e4
LT		 3092 nfs4_unlock_state();
3093 return status;
3094}
3095
3096
3097/*
3098 * unset all bits in union bitmap (bmap) that
3099 * do not exist in share (from successful OPEN_DOWNGRADE)
3100 */
3101static void
3102reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3103{
3104 int i;
3105 for (i = 1; i < 4; i++) {
3106 if ((i & access) != i)
3107 __clear_bit(i, bmap);
3108 }
3109}
3110
3111static void
3112reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3113{
3114 int i;
3115 for (i = 0; i < 4; i++) {
3116 if ((i & deny) != i)
3117 __clear_bit(i, bmap);
3118 }
3119}
3120
b37ad28b 3121__be32
ca364317
BF		 3122nfsd4_open_downgrade(struct svc_rqst *rqstp,
3123 struct nfsd4_compound_state *cstate,
a4f1706a 3124 struct nfsd4_open_downgrade *od)
1da177e4 3125{
b37ad28b 3126 __be32 status;
1da177e4
LT		 3127 struct nfs4_stateid *stp;
3128 unsigned int share_access;
3129
3130 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3131 (int)cstate->current_fh.fh_dentry->d_name.len,
3132 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3133
d87a8ade 3134 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3135 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3136 return nfserr_inval;
3137
3138 nfs4_lock_state();
dd453dfd 3139 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3140 od->od_seqid,
1da177e4 3141 &od->od_stateid,
f3362737 3142 OPEN_STATE,
1da177e4
LT		 3143 &od->od_stateowner, &stp, NULL)))
3144 goto out;
3145
3146 status = nfserr_inval;
3147 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3148 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3149 stp->st_access_bmap, od->od_share_access);
3150 goto out;
3151 }
3152 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3153 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3154 stp->st_deny_bmap, od->od_share_deny);
3155 goto out;
3156 }
3157 set_access(&share_access, stp->st_access_bmap);
3158 nfs4_file_downgrade(stp->st_vfs_file,
3159 share_access & ~od->od_share_access);
3160
3161 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3162 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3163
3164 update_stateid(&stp->st_stateid);
3165 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3166 status = nfs_ok;
3167out:
f2327d9a 3168 if (od->od_stateowner) {
1da177e4 3169 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3170 cstate->replay_owner = od->od_stateowner;
f2327d9a 3171 }
1da177e4
LT		 3172 nfs4_unlock_state();
3173 return status;
3174}
3175
3176/*
3177 * nfs4_unlock_state() called after encode
3178 */
b37ad28b 3179__be32
ca364317 3180nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3181 struct nfsd4_close *close)
1da177e4 3182{
b37ad28b 3183 __be32 status;
1da177e4
LT		 3184 struct nfs4_stateid *stp;
3185
3186 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3187 (int)cstate->current_fh.fh_dentry->d_name.len,
3188 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3189
3190 nfs4_lock_state();
3191 /* check close_lru for replay */
dd453dfd 3192 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3193 close->cl_seqid,
1da177e4 3194 &close->cl_stateid,
f3362737 3195 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3196 &close->cl_stateowner, &stp, NULL)))
3197 goto out;
1da177e4
LT		 3198 status = nfs_ok;
3199 update_stateid(&stp->st_stateid);
3200 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3201
04ef5954 3202 /* release_stateid() calls nfsd_close() if needed */
2283963f 3203 release_open_stateid(stp);
04ef5954
BF		 3204
3205 /* place unused nfs4_stateowners on so_close_lru list to be
3206 * released by the laundromat service after the lease period
3207 * to enable us to handle CLOSE replay
3208 */
3209 if (list_empty(&close->cl_stateowner->so_stateids))
3210 move_to_close_lru(close->cl_stateowner);
1da177e4 3211out:
f2327d9a 3212 if (close->cl_stateowner) {
1da177e4 3213 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3214 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3215 }
1da177e4
LT		 3216 nfs4_unlock_state();
3217 return status;
3218}
3219
b37ad28b 3220__be32
ca364317
BF		 3221nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3222 struct nfsd4_delegreturn *dr)
1da177e4 3223{
203a8c8e
BF		 3224 struct nfs4_delegation *dp;
3225 stateid_t *stateid = &dr->dr_stateid;
3226 struct inode *inode;
b37ad28b 3227 __be32 status;
6668958f 3228 int flags = 0;
1da177e4 3229
ca364317 3230 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3231 return status;
3232 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3233
6668958f
AA		 3234 if (nfsd4_has_session(cstate))
3235 flags |= HAS_SESSION;
1da177e4 3236 nfs4_lock_state();
203a8c8e
BF		 3237 status = nfserr_bad_stateid;
3238 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3239 goto out;
3240 status = nfserr_stale_stateid;
3241 if (STALE_STATEID(stateid))
3242 goto out;
7e0f7cf5 3243 status = nfserr_bad_stateid;
203a8c8e
BF		 3244 if (!is_delegation_stateid(stateid))
3245 goto out;
203a8c8e 3246 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3247 if (!dp) {
3248 status = stateid_error_map(stateid);
203a8c8e 3249 goto out;
78155ed7 3250 }
6668958f 3251 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3252 if (status)
3253 goto out;
3254 renew_client(dp->dl_client);
3255
3256 unhash_delegation(dp);
1da177e4 3257out:
203a8c8e
BF		 3258 nfs4_unlock_state();
3259
1da177e4
LT		 3260 return status;
3261}
3262
3263
3264/*
3265 * Lock owner state (byte-range locks)
3266 */
3267#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3268#define LOCK_HASH_BITS 8
3269#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3270#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3271
87df4de8
BH		 3272static inline u64
3273end_offset(u64 start, u64 len)
3274{
3275 u64 end;
3276
3277 end = start + len;
3278 return end >= start ? end: NFS4_MAX_UINT64;
3279}
3280
3281/* last octet in a range */
3282static inline u64
3283last_byte_offset(u64 start, u64 len)
3284{
3285 u64 end;
3286
3287 BUG_ON(!len);
3288 end = start + len;
3289 return end > start ? end - 1: NFS4_MAX_UINT64;
3290}
3291
1da177e4
LT		 3292#define lockownerid_hashval(id) \
3293 ((id) & LOCK_HASH_MASK)
3294
3295static inline unsigned int
3296lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3297 struct xdr_netobj *ownername)
3298{
3299 return (file_hashval(inode) + cl_id
3300 + opaque_hashval(ownername->data, ownername->len))
3301 & LOCK_HASH_MASK;
3302}
3303
3304static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3305static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3306static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3307
fd39ca9a 3308static struct nfs4_stateid *
1da177e4
LT		 3309find_stateid(stateid_t *stid, int flags)
3310{
9346eff0 3311 struct nfs4_stateid *local;
1da177e4
LT		 3312 u32 st_id = stid->si_stateownerid;
3313 u32 f_id = stid->si_fileid;
3314 unsigned int hashval;
3315
3316 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3317 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3318 hashval = stateid_hashval(st_id, f_id);
3319 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3320 if ((local->st_stateid.si_stateownerid == st_id) &&
3321 (local->st_stateid.si_fileid == f_id))
3322 return local;
3323 }
3324 }
9346eff0
KK		 3325
3326 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3327 hashval = stateid_hashval(st_id, f_id);
3328 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3329 if ((local->st_stateid.si_stateownerid == st_id) &&
3330 (local->st_stateid.si_fileid == f_id))
3331 return local;
3332 }
849823c5 3333 }
1da177e4
LT		 3334 return NULL;
3335}
3336
3337static struct nfs4_delegation *
3338find_delegation_stateid(struct inode *ino, stateid_t *stid)
3339{
13cd2184
N		 3340 struct nfs4_file *fp;
3341 struct nfs4_delegation *dl;
1da177e4
LT		 3342
3343 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3344 stid->si_boot, stid->si_stateownerid,
3345 stid->si_fileid, stid->si_generation);
3346
1da177e4 3347 fp = find_file(ino);
13cd2184
N		 3348 if (!fp)
3349 return NULL;
3350 dl = find_delegation_file(fp, stid);
3351 put_nfs4_file(fp);
3352 return dl;
1da177e4
LT		 3353}
3354
3355/*
3356 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3357 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3358 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3359 * locking, this prevents us from being completely protocol-compliant. The
3360 * real solution to this problem is to start using unsigned file offsets in
3361 * the VFS, but this is a very deep change!
3362 */
3363static inline void
3364nfs4_transform_lock_offset(struct file_lock *lock)
3365{
3366 if (lock->fl_start < 0)
3367 lock->fl_start = OFFSET_MAX;
3368 if (lock->fl_end < 0)
3369 lock->fl_end = OFFSET_MAX;
3370}
3371
d5b9026a
N		 3372/* Hack!: For now, we're defining this just so we can use a pointer to it
3373 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3374static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3375};
1da177e4
LT		 3376
3377static inline void
3378nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3379{
d5b9026a
N		 3380 struct nfs4_stateowner *sop;
3381 unsigned int hval;
1da177e4 3382
d5b9026a
N		 3383 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3384 sop = (struct nfs4_stateowner *) fl->fl_owner;
3385 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3386 kref_get(&sop->so_ref);
3387 deny->ld_sop = sop;
3388 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3389 } else {
3390 deny->ld_sop = NULL;
3391 deny->ld_clientid.cl_boot = 0;
3392 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3393 }
3394 deny->ld_start = fl->fl_start;
87df4de8
BH		 3395 deny->ld_length = NFS4_MAX_UINT64;
3396 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3397 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3398 deny->ld_type = NFS4_READ_LT;
3399 if (fl->fl_type != F_RDLCK)
3400 deny->ld_type = NFS4_WRITE_LT;
3401}
3402
1da177e4
LT		 3403static struct nfs4_stateowner *
3404find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3405 struct xdr_netobj *owner)
3406{
3407 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3408 struct nfs4_stateowner *op;
3409
3410 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3411 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3412 return op;
3413 }
3414 return NULL;
3415}
3416
3417/*
3418 * Alloc a lock owner structure.
3419 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3420 * occured.
3421 *
3422 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3423 */
3424
3425static struct nfs4_stateowner *
3426alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3427 struct nfs4_stateowner *sop;
3428 struct nfs4_replay *rp;
3429 unsigned int idhashval;
3430
3431 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3432 return NULL;
3433 idhashval = lockownerid_hashval(current_ownerid);
3434 INIT_LIST_HEAD(&sop->so_idhash);
3435 INIT_LIST_HEAD(&sop->so_strhash);
3436 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3437 INIT_LIST_HEAD(&sop->so_stateids);
3438 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3439 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3440 sop->so_time = 0;
3441 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3442 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3443 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3444 sop->so_is_open_owner = 0;
3445 sop->so_id = current_ownerid++;
3446 sop->so_client = clp;
b59e3c0e
NB		 3447 /* It is the openowner seqid that will be incremented in encode in the
3448 * case of new lockowners; so increment the lock seqid manually: */
3449 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3450 sop->so_confirmed = 1;
3451 rp = &sop->so_replay;
de1ae286 3452 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3453 rp->rp_buflen = 0;
3454 rp->rp_buf = rp->rp_ibuf;
3455 return sop;
3456}
3457
fd39ca9a 3458static struct nfs4_stateid *
1da177e4
LT		 3459alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3460{
3461 struct nfs4_stateid *stp;
3462 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3463
5ac049ac
N		 3464 stp = nfs4_alloc_stateid();
3465 if (stp == NULL)
1da177e4
LT		 3466 goto out;
3467 INIT_LIST_HEAD(&stp->st_hash);
3468 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3469 INIT_LIST_HEAD(&stp->st_perstateowner);
3470 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3471 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3472 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3473 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3474 stp->st_stateowner = sop;
13cd2184 3475 get_nfs4_file(fp);
1da177e4 3476 stp->st_file = fp;
78155ed7 3477 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3478 stp->st_stateid.si_stateownerid = sop->so_id;
3479 stp->st_stateid.si_fileid = fp->fi_id;
3480 stp->st_stateid.si_generation = 0;
3481 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3482 stp->st_access_bmap = open_stp->st_access_bmap;
3483 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3484 stp->st_openstp = open_stp;
1da177e4
LT		 3485
3486out:
3487 return stp;
3488}
3489
fd39ca9a 3490static int
1da177e4
LT		 3491check_lock_length(u64 offset, u64 length)
3492{
87df4de8 3493 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3494 LOFF_OVERFLOW(offset, length)));
3495}
3496
3497/*
3498 * LOCK operation
3499 */
b37ad28b 3500__be32
ca364317 3501nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3502 struct nfsd4_lock *lock)
1da177e4 3503{
3e9e3dbe 3504 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3505 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3506 struct nfs4_stateid *lock_stp;
3507 struct file *filp;
3508 struct file_lock file_lock;
8dc7c311 3509 struct file_lock conflock;
b37ad28b 3510 __be32 status = 0;
1da177e4 3511 unsigned int strhashval;
150b3934 3512 unsigned int cmd;
b8dd7b9a 3513 int err;
1da177e4
LT		 3514
3515 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3516 (long long) lock->lk_offset,
3517 (long long) lock->lk_length);
3518
1da177e4
LT		 3519 if (check_lock_length(lock->lk_offset, lock->lk_length))
3520 return nfserr_inval;
3521
ca364317 3522 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3523 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3524 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3525 return status;
3526 }
3527
1da177e4
LT		 3528 nfs4_lock_state();
3529
3530 if (lock->lk_is_new) {
893f8770
N		 3531 /*
3532 * Client indicates that this is a new lockowner.
3533 * Use open owner and open stateid to create lock owner and
3534 * lock stateid.
3535 */
1da177e4
LT		 3536 struct nfs4_stateid *open_stp = NULL;
3537 struct nfs4_file *fp;
3538
3539 status = nfserr_stale_clientid;
60adfc50
AA		 3540 if (!nfsd4_has_session(cstate) &&
3541 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3542 goto out;
1da177e4 3543
1da177e4 3544 /* validate and update open stateid and open seqid */
dd453dfd 3545 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3546 lock->lk_new_open_seqid,
3547 &lock->lk_new_open_stateid,
f3362737 3548 OPEN_STATE,
3a65588a 3549 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3550 lock);
37515177 3551 if (status)
1da177e4 3552 goto out;
3a65588a 3553 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3554 /* create lockowner and lock stateid */
3555 fp = open_stp->st_file;
3556 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3557 open_sop->so_client->cl_clientid.cl_id,
3558 &lock->v.new.owner);
3e9e3dbe
N		 3559 /* XXX: Do we need to check for duplicate stateowners on
3560 * the same file, or should they just be allowed (and
3561 * create new stateids)? */
1da177e4 3562 status = nfserr_resource;
b59e3c0e
NB		 3563 lock_sop = alloc_init_lock_stateowner(strhashval,
3564 open_sop->so_client, open_stp, lock);
3565 if (lock_sop == NULL)
1da177e4 3566 goto out;
b59e3c0e 3567 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3568 if (lock_stp == NULL)
1da177e4 3569 goto out;
1da177e4
LT		 3570 } else {
3571 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3572 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3573 lock->lk_old_lock_seqid,
3574 &lock->lk_old_lock_stateid,
f3362737 3575 LOCK_STATE,
3a65588a 3576 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3577 if (status)
3578 goto out;
3a65588a 3579 lock_sop = lock->lk_replay_owner;
1da177e4 3580 }
3a65588a 3581 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3582 filp = lock_stp->st_vfs_file;
3583
0dd395dc 3584 status = nfserr_grace;
af558e33 3585 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3586 goto out;
3587 status = nfserr_no_grace;
af558e33 3588 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3589 goto out;
3590
1da177e4
LT		 3591 locks_init_lock(&file_lock);
3592 switch (lock->lk_type) {
3593 case NFS4_READ_LT:
3594 case NFS4_READW_LT:
3595 file_lock.fl_type = F_RDLCK;
150b3934 3596 cmd = F_SETLK;
1da177e4
LT		 3597 break;
3598 case NFS4_WRITE_LT:
3599 case NFS4_WRITEW_LT:
3600 file_lock.fl_type = F_WRLCK;
150b3934 3601 cmd = F_SETLK;
1da177e4
LT		 3602 break;
3603 default:
3604 status = nfserr_inval;
3605 goto out;
3606 }
b59e3c0e 3607 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3608 file_lock.fl_pid = current->tgid;
3609 file_lock.fl_file = filp;
3610 file_lock.fl_flags = FL_POSIX;
d5b9026a 3611 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3612
3613 file_lock.fl_start = lock->lk_offset;
87df4de8 3614 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3615 nfs4_transform_lock_offset(&file_lock);
3616
3617 /*
3618 * Try to lock the file in the VFS.
3619 * Note: locks.c uses the BKL to protect the inode's lock list.
3620 */
3621
fd85b817 3622 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3623 switch (-err) {
1da177e4
LT		 3624 case 0: /* success! */
3625 update_stateid(&lock_stp->st_stateid);
3626 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3627 sizeof(stateid_t));
b8dd7b9a 3628 status = 0;
eb76b3fd
AA		 3629 break;
3630 case (EAGAIN): /* conflock holds conflicting lock */
3631 status = nfserr_denied;
3632 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3633 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3634 break;
1da177e4
LT		 3635 case (EDEADLK):
3636 status = nfserr_deadlock;
eb76b3fd 3637 break;
1da177e4 3638 default:
fd85b817 3639 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3640 status = nfserr_resource;
3641 break;
1da177e4 3642 }
1da177e4 3643out:
8a280510 3644 if (status && lock->lk_is_new && lock_sop)
f044ff83 3645 release_lockowner(lock_sop);
3a65588a
BF		 3646 if (lock->lk_replay_owner) {
3647 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3648 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3649 }
1da177e4
LT		 3650 nfs4_unlock_state();
3651 return status;
3652}
3653
55ef1274
BF		 3654/*
3655 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3656 * so we do a temporary open here just to get an open file to pass to
3657 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3658 * inode operation.)
3659 */
3660static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3661{
3662 struct file *file;
3663 int err;
3664
3665 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3666 if (err)
3667 return err;
3668 err = vfs_test_lock(file, lock);
3669 nfsd_close(file);
3670 return err;
3671}
3672
1da177e4
LT		 3673/*
3674 * LOCKT operation
3675 */
b37ad28b 3676__be32
ca364317
BF		 3677nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3678 struct nfsd4_lockt *lockt)
1da177e4
LT		 3679{
3680 struct inode *inode;
1da177e4 3681 struct file_lock file_lock;
fd85b817 3682 int error;
b37ad28b 3683 __be32 status;
1da177e4 3684
af558e33 3685 if (locks_in_grace())
1da177e4
LT		 3686 return nfserr_grace;
3687
3688 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3689 return nfserr_inval;
3690
3691 lockt->lt_stateowner = NULL;
3692 nfs4_lock_state();
3693
3694 status = nfserr_stale_clientid;
60adfc50 3695 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3696 goto out;
1da177e4 3697
ca364317 3698 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3699 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3700 if (status == nfserr_symlink)
3701 status = nfserr_inval;
3702 goto out;
3703 }
3704
ca364317 3705 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3706 locks_init_lock(&file_lock);
3707 switch (lockt->lt_type) {
3708 case NFS4_READ_LT:
3709 case NFS4_READW_LT:
3710 file_lock.fl_type = F_RDLCK;
3711 break;
3712 case NFS4_WRITE_LT:
3713 case NFS4_WRITEW_LT:
3714 file_lock.fl_type = F_WRLCK;
3715 break;
3716 default:
2fdada03 3717 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3718 status = nfserr_inval;
3719 goto out;
3720 }
3721
3722 lockt->lt_stateowner = find_lockstateowner_str(inode,
3723 &lockt->lt_clientid, &lockt->lt_owner);
3724 if (lockt->lt_stateowner)
3725 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3726 file_lock.fl_pid = current->tgid;
3727 file_lock.fl_flags = FL_POSIX;
3728
3729 file_lock.fl_start = lockt->lt_offset;
87df4de8 3730 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3731
3732 nfs4_transform_lock_offset(&file_lock);
3733
1da177e4 3734 status = nfs_ok;
55ef1274 3735 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3736 if (error) {
3737 status = nfserrno(error);
3738 goto out;
3739 }
9d6a8c5c 3740 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3741 status = nfserr_denied;
9d6a8c5c 3742 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3743 }
3744out:
3745 nfs4_unlock_state();
3746 return status;
3747}
3748
b37ad28b 3749__be32
ca364317 3750nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3751 struct nfsd4_locku *locku)
1da177e4
LT		 3752{
3753 struct nfs4_stateid *stp;
3754 struct file *filp = NULL;
3755 struct file_lock file_lock;
b37ad28b 3756 __be32 status;
b8dd7b9a 3757 int err;
1da177e4
LT		 3758
3759 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3760 (long long) locku->lu_offset,
3761 (long long) locku->lu_length);
3762
3763 if (check_lock_length(locku->lu_offset, locku->lu_length))
3764 return nfserr_inval;
3765
3766 nfs4_lock_state();
3767
dd453dfd 3768 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3769 locku->lu_seqid,
3770 &locku->lu_stateid,
f3362737 3771 LOCK_STATE,
1da177e4
LT		 3772 &locku->lu_stateowner, &stp, NULL)))
3773 goto out;
3774
3775 filp = stp->st_vfs_file;
3776 BUG_ON(!filp);
3777 locks_init_lock(&file_lock);
3778 file_lock.fl_type = F_UNLCK;
3779 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3780 file_lock.fl_pid = current->tgid;
3781 file_lock.fl_file = filp;
3782 file_lock.fl_flags = FL_POSIX;
d5b9026a 3783 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3784 file_lock.fl_start = locku->lu_offset;
3785
87df4de8 3786 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3787 nfs4_transform_lock_offset(&file_lock);
3788
3789 /*
3790 * Try to unlock the file in the VFS.
3791 */
fd85b817 3792 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3793 if (err) {
fd85b817 3794 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3795 goto out_nfserr;
3796 }
3797 /*
3798 * OK, unlock succeeded; the only thing left to do is update the stateid.
3799 */
3800 update_stateid(&stp->st_stateid);
3801 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3802
3803out:
f2327d9a 3804 if (locku->lu_stateowner) {
1da177e4 3805 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3806 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3807 }
1da177e4
LT		 3808 nfs4_unlock_state();
3809 return status;
3810
3811out_nfserr:
b8dd7b9a 3812 status = nfserrno(err);
1da177e4
LT		 3813 goto out;
3814}
3815
3816/*
3817 * returns
3818 * 1: locks held by lockowner
3819 * 0: no locks held by lockowner
3820 */
3821static int
3822check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3823{
3824 struct file_lock **flpp;
7eaa36e2 3825 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3826 int status = 0;
3827
3828 lock_kernel();
3829 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3830 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3831 status = 1;
3832 goto out;
796dadfd 3833 }
1da177e4
LT		 3834 }
3835out:
3836 unlock_kernel();
3837 return status;
3838}
3839
b37ad28b 3840__be32
b591480b
BF		 3841nfsd4_release_lockowner(struct svc_rqst *rqstp,
3842 struct nfsd4_compound_state *cstate,
3843 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3844{
3845 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3846 struct nfs4_stateowner *sop;
3847 struct nfs4_stateid *stp;
1da177e4 3848 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3849 struct list_head matches;
3850 int i;
b37ad28b 3851 __be32 status;
1da177e4
LT		 3852
3853 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3854 clid->cl_boot, clid->cl_id);
3855
3856 /* XXX check for lease expiration */
3857
3858 status = nfserr_stale_clientid;
849823c5 3859 if (STALE_CLIENTID(clid))
1da177e4 3860 return status;
1da177e4
LT		 3861
3862 nfs4_lock_state();
3863
3e9e3dbe
N		 3864 status = nfserr_locks_held;
3865 /* XXX: we're doing a linear search through all the lockowners.
3866 * Yipes! For now we'll just hope clients aren't really using
3867 * release_lockowner much, but eventually we have to fix these
3868 * data structures. */
3869 INIT_LIST_HEAD(&matches);
3870 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3871 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3872 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3873 continue;
3874 list_for_each_entry(stp, &sop->so_stateids,
3875 st_perstateowner) {
3876 if (check_for_locks(stp->st_vfs_file, sop))
3877 goto out;
3878 /* Note: so_perclient unused for lockowners,
3879 * so it's OK to fool with here. */
3880 list_add(&sop->so_perclient, &matches);
3881 }
1da177e4 3882 }
3e9e3dbe
N		 3883 }
3884 /* Clients probably won't expect us to return with some (but not all)
3885 * of the lockowner state released; so don't release any until all
3886 * have been checked. */
3887 status = nfs_ok;
0fa822e4
N		 3888 while (!list_empty(&matches)) {
3889 sop = list_entry(matches.next, struct nfs4_stateowner,
3890 so_perclient);
3891 /* unhash_stateowner deletes so_perclient only
3892 * for openowners. */
3893 list_del(&sop->so_perclient);
f044ff83 3894 release_lockowner(sop);
1da177e4
LT		 3895 }
3896out:
3897 nfs4_unlock_state();
3898 return status;
3899}
3900
3901static inline struct nfs4_client_reclaim *
a55370a3 3902alloc_reclaim(void)
1da177e4 3903{
a55370a3 3904 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3905}
3906
c7b9a459 3907int
a1bcecd2 3908nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3909{
3910 unsigned int strhashval = clientstr_hashval(name);
3911 struct nfs4_client *clp;
3912
a1bcecd2 3913 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3914 return clp ? 1 : 0;
3915}
3916
1da177e4
LT		 3917/*
3918 * failure => all reset bets are off, nfserr_no_grace...
3919 */
190e4fbf
N		 3920int
3921nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3922{
3923 unsigned int strhashval;
3924 struct nfs4_client_reclaim *crp = NULL;
3925
a55370a3
N		 3926 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3927 crp = alloc_reclaim();
1da177e4
LT		 3928 if (!crp)
3929 return 0;
a55370a3 3930 strhashval = clientstr_hashval(name);
1da177e4
LT		 3931 INIT_LIST_HEAD(&crp->cr_strhash);
3932 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3933 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3934 reclaim_str_hashtbl_size++;
3935 return 1;
3936}
3937
3938static void
3939nfs4_release_reclaim(void)
3940{
3941 struct nfs4_client_reclaim *crp = NULL;
3942 int i;
3943
1da177e4
LT		 3944 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3945 while (!list_empty(&reclaim_str_hashtbl[i])) {
3946 crp = list_entry(reclaim_str_hashtbl[i].next,
3947 struct nfs4_client_reclaim, cr_strhash);
3948 list_del(&crp->cr_strhash);
1da177e4
LT		 3949 kfree(crp);
3950 reclaim_str_hashtbl_size--;
3951 }
3952 }
3953 BUG_ON(reclaim_str_hashtbl_size);
3954}
3955
3956/*
3957 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3958static struct nfs4_client_reclaim *
1da177e4
LT		 3959nfs4_find_reclaim_client(clientid_t *clid)
3960{
3961 unsigned int strhashval;
3962 struct nfs4_client *clp;
3963 struct nfs4_client_reclaim *crp = NULL;
3964
3965
3966 /* find clientid in conf_id_hashtbl */
3967 clp = find_confirmed_client(clid);
3968 if (clp == NULL)
3969 return NULL;
3970
a55370a3
N		 3971 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3972 clp->cl_name.len, clp->cl_name.data,
3973 clp->cl_recdir);
1da177e4
LT		 3974
3975 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3976 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3977 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3978 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3979 return crp;
3980 }
3981 }
3982 return NULL;
3983}
3984
3985/*
3986* Called from OPEN. Look for clientid in reclaim list.
3987*/
b37ad28b 3988__be32
1da177e4
LT		 3989nfs4_check_open_reclaim(clientid_t *clid)
3990{
dfc83565 3991 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3992}
3993
ac4d8ff2 3994/* initialization to perform at module load time: */
1da177e4 3995
e8ff2a84 3996int
ac4d8ff2 3997nfs4_state_init(void)
1da177e4 3998{
e8ff2a84 3999 int i, status;
1da177e4 4000
e8ff2a84
BF		 4001 status = nfsd4_init_slabs();
4002 if (status)
4003 return status;
1da177e4
LT		 4004 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4005 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4006 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4007 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4008 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4009 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4010 }
5282fd72
ME		 4011 for (i = 0; i < SESSION_HASH_SIZE; i++)
4012 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4013 for (i = 0; i < FILE_HASH_SIZE; i++) {
4014 INIT_LIST_HEAD(&file_hashtbl[i]);
4015 }
4016 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4017 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4018 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4019 }
4020 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4021 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4022 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4023 }
4024 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4025 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4026 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4027 }
1da177e4 4028 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4029 INIT_LIST_HEAD(&close_lru);
4030 INIT_LIST_HEAD(&client_lru);
4031 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4032 reclaim_str_hashtbl_size = 0;
e8ff2a84 4033 return 0;
ac4d8ff2
N		 4034}
4035
190e4fbf
N		 4036static void
4037nfsd4_load_reboot_recovery_data(void)
4038{
4039 int status;
4040
0964a3d3
N		 4041 nfs4_lock_state();
4042 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4043 status = nfsd4_recdir_load();
0964a3d3 4044 nfs4_unlock_state();
190e4fbf
N		 4045 if (status)
4046 printk("NFSD: Failure reading reboot recovery data\n");
4047}
4048
9a8db97e
ME		 4049unsigned long
4050get_nfs4_grace_period(void)
4051{
4052 return max(user_lease_time, lease_time) * HZ;
4053}
4054
c2f1a551
MS		 4055/*
4056 * Since the lifetime of a delegation isn't limited to that of an open, a
4057 * client may quite reasonably hang on to a delegation as long as it has
4058 * the inode cached. This becomes an obvious problem the first time a
4059 * client's inode cache approaches the size of the server's total memory.
4060 *
4061 * For now we avoid this problem by imposing a hard limit on the number
4062 * of delegations, which varies according to the server's memory size.
4063 */
4064static void
4065set_max_delegations(void)
4066{
4067 /*
4068 * Allow at most 4 delegations per megabyte of RAM. Quick
4069 * estimates suggest that in the worst case (where every delegation
4070 * is for a different inode), a delegation could take about 1.5K,
4071 * giving a worst case usage of about 6% of memory.
4072 */
4073 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4074}
4075
ac4d8ff2
N		 4076/* initialization to perform when the nfsd service is started: */
4077
4078static void
4079__nfs4_state_start(void)
4080{
9a8db97e 4081 unsigned long grace_time;
ac4d8ff2 4082
1da177e4 4083 boot_time = get_seconds();
af558e33 4084 grace_time = get_nfs4_grace_period();
d99a05ad 4085 lease_time = user_lease_time;
af558e33 4086 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4087 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4088 grace_time/HZ);
58da282b 4089 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4090 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4091 set_max_delegations();
1da177e4
LT		 4092}
4093
e8ff2a84 4094void
76a3550e 4095nfs4_state_start(void)
1da177e4 4096{
1da177e4 4097 if (nfs4_init)
e8ff2a84 4098 return;
190e4fbf 4099 nfsd4_load_reboot_recovery_data();
76a3550e 4100 __nfs4_state_start();
1da177e4 4101 nfs4_init = 1;
e8ff2a84 4102 return;
1da177e4
LT		 4103}
4104
1da177e4
LT		 4105time_t
4106nfs4_lease_time(void)
4107{
4108 return lease_time;
4109}
4110
4111static void
4112__nfs4_state_shutdown(void)
4113{
4114 int i;
4115 struct nfs4_client *clp = NULL;
4116 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4117 struct list_head *pos, *next, reaplist;
4118
1da177e4
LT		 4119 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4120 while (!list_empty(&conf_id_hashtbl[i])) {
4121 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4122 expire_client(clp);
4123 }
4124 while (!list_empty(&unconf_str_hashtbl[i])) {
4125 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4126 expire_client(clp);
4127 }
4128 }
4129 INIT_LIST_HEAD(&reaplist);
4130 spin_lock(&recall_lock);
4131 list_for_each_safe(pos, next, &del_recall_lru) {
4132 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4133 list_move(&dp->dl_recall_lru, &reaplist);
4134 }
4135 spin_unlock(&recall_lock);
4136 list_for_each_safe(pos, next, &reaplist) {
4137 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4138 list_del_init(&dp->dl_recall_lru);
4139 unhash_delegation(dp);
4140 }
4141
190e4fbf 4142 nfsd4_shutdown_recdir();
1da177e4 4143 nfs4_init = 0;
1da177e4
LT		 4144}
4145
4146void
4147nfs4_state_shutdown(void)
4148{
5e8d5c29
N		 4149 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4150 destroy_workqueue(laundry_wq);
2c5e7615 4151 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4152 nfs4_lock_state();
4153 nfs4_release_reclaim();
4154 __nfs4_state_shutdown();
1da177e4
LT		 4155 nfs4_unlock_state();
4156}
4157
3dd98a3b
JL		 4158/*
4159 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4160 * accessed when nfsd is starting.
4161 */
0964a3d3
N		 4162static void
4163nfs4_set_recdir(char *recdir)
4164{
0964a3d3 4165 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4166}
4167
4168/*
4169 * Change the NFSv4 recovery directory to recdir.
4170 */
4171int
4172nfs4_reset_recoverydir(char *recdir)
4173{
4174 int status;
a63bb996 4175 struct path path;
0964a3d3 4176
a63bb996 4177 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4178 if (status)
4179 return status;
4180 status = -ENOTDIR;
a63bb996 4181 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4182 nfs4_set_recdir(recdir);
4183 status = 0;
4184 }
a63bb996 4185 path_put(&path);
0964a3d3
N		 4186 return status;
4187}
4188
3dd98a3b
JL		 4189char *
4190nfs4_recoverydir(void)
4191{
4192 return user_recovery_dirname;
4193}
4194
1da177e4
LT		 4195/*
4196 * Called when leasetime is changed.
4197 *
d99a05ad
N		 4198 * The only way the protocol gives us to handle on-the-fly lease changes is to
4199 * simulate a reboot. Instead of doing that, we just wait till the next time
4200 * we start to register any changes in lease time. If the administrator
4201 * really wants to change the lease time *now*, they can go ahead and bring
4202 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4203 *
4204 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4205 * when nfsd is starting
1da177e4
LT		 4206 */
4207void
4208nfs4_reset_lease(time_t leasetime)
4209{
d99a05ad 4210 user_lease_time = leasetime;
1da177e4 4211}
Clone of davem's net-next-2.6 tree