]> 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
nfsd: optimise the starting of zero threads when none are running.
[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
ec6b5d7b
AA		 433 spin_lock(&nfsd_serv->sv_lock);
434 if (np + nfsd_serv->sv_drc_pages_used > nfsd_serv->sv_drc_max_pages)
435 np = nfsd_serv->sv_drc_max_pages - nfsd_serv->sv_drc_pages_used;
436 nfsd_serv->sv_drc_pages_used += np;
437 spin_unlock(&nfsd_serv->sv_lock);
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 }
7116ed6b
AA		 588 kfree(ses);
589}
590
1da177e4
LT		 591static inline void
592renew_client(struct nfs4_client *clp)
593{
594 /*
595 * Move client to the end to the LRU list.
596 */
597 dprintk("renewing client (clientid %08x/%08x)\n",
598 clp->cl_clientid.cl_boot,
599 clp->cl_clientid.cl_id);
600 list_move_tail(&clp->cl_lru, &client_lru);
601 clp->cl_time = get_seconds();
602}
603
604/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
605static int
606STALE_CLIENTID(clientid_t *clid)
607{
608 if (clid->cl_boot == boot_time)
609 return 0;
60adfc50
AA		 610 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
611 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 612 return 1;
613}
614
615/*
616 * XXX Should we use a slab cache ?
617 * This type of memory management is somewhat inefficient, but we use it
618 * anyway since SETCLIENTID is not a common operation.
619 */
35bba9a3 620static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 621{
622 struct nfs4_client *clp;
623
35bba9a3
BF		 624 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
625 if (clp == NULL)
626 return NULL;
627 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
628 if (clp->cl_name.data == NULL) {
629 kfree(clp);
630 return NULL;
1da177e4 631 }
35bba9a3
BF		 632 memcpy(clp->cl_name.data, name.data, name.len);
633 clp->cl_name.len = name.len;
1da177e4
LT		 634 return clp;
635}
636
1b1a9b31
BF		 637static void
638shutdown_callback_client(struct nfs4_client *clp)
639{
c237dc03 640 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 641
1b1a9b31 642 if (clnt) {
404ec117
BF		 643 /*
644 * Callback threads take a reference on the client, so there
645 * should be no outstanding callbacks at this point.
646 */
c237dc03 647 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 648 rpc_shutdown_client(clnt);
649 }
c237dc03
BF		 650 if (clp->cl_cb_conn.cb_cred) {
651 put_rpccred(clp->cl_cb_conn.cb_cred);
652 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 653 }
1b1a9b31
BF		 654}
655
1da177e4
LT		 656static inline void
657free_client(struct nfs4_client *clp)
658{
1b1a9b31 659 shutdown_callback_client(clp);
38eb76a5
AA		 660 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
661 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 662 if (clp->cl_cred.cr_group_info)
663 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 664 kfree(clp->cl_principal);
1da177e4
LT		 665 kfree(clp->cl_name.data);
666 kfree(clp);
667}
668
669void
670put_nfs4_client(struct nfs4_client *clp)
671{
672 if (atomic_dec_and_test(&clp->cl_count))
673 free_client(clp);
674}
675
676static void
677expire_client(struct nfs4_client *clp)
678{
679 struct nfs4_stateowner *sop;
680 struct nfs4_delegation *dp;
1da177e4
LT		 681 struct list_head reaplist;
682
683 dprintk("NFSD: expire_client cl_count %d\n",
684 atomic_read(&clp->cl_count));
685
1da177e4
LT		 686 INIT_LIST_HEAD(&reaplist);
687 spin_lock(&recall_lock);
ea1da636
N		 688 while (!list_empty(&clp->cl_delegations)) {
689 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 690 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
691 dp->dl_flock);
ea1da636 692 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 693 list_move(&dp->dl_recall_lru, &reaplist);
694 }
695 spin_unlock(&recall_lock);
696 while (!list_empty(&reaplist)) {
697 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
698 list_del_init(&dp->dl_recall_lru);
699 unhash_delegation(dp);
700 }
701 list_del(&clp->cl_idhash);
702 list_del(&clp->cl_strhash);
703 list_del(&clp->cl_lru);
ea1da636
N		 704 while (!list_empty(&clp->cl_openowners)) {
705 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 706 release_openowner(sop);
1da177e4 707 }
c4bf7868
ME		 708 while (!list_empty(&clp->cl_sessions)) {
709 struct nfsd4_session *ses;
710 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
711 se_perclnt);
712 release_session(ses);
713 }
1da177e4
LT		 714 put_nfs4_client(clp);
715}
716
35bba9a3
BF		 717static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
718{
1da177e4
LT		 719 struct nfs4_client *clp;
720
35bba9a3
BF		 721 clp = alloc_client(name);
722 if (clp == NULL)
723 return NULL;
a55370a3 724 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 725 atomic_set(&clp->cl_count, 1);
c237dc03 726 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 727 INIT_LIST_HEAD(&clp->cl_idhash);
728 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 729 INIT_LIST_HEAD(&clp->cl_openowners);
730 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 731 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 732 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 733 return clp;
734}
735
35bba9a3
BF		 736static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
737{
738 memcpy(target->cl_verifier.data, source->data,
739 sizeof(target->cl_verifier.data));
1da177e4
LT		 740}
741
35bba9a3
BF		 742static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
743{
1da177e4
LT		 744 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
745 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
746}
747
35bba9a3
BF		 748static void copy_cred(struct svc_cred *target, struct svc_cred *source)
749{
1da177e4
LT		 750 target->cr_uid = source->cr_uid;
751 target->cr_gid = source->cr_gid;
752 target->cr_group_info = source->cr_group_info;
753 get_group_info(target->cr_group_info);
754}
755
35bba9a3 756static int same_name(const char *n1, const char *n2)
599e0a22 757{
a55370a3 758 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 759}
760
761static int
599e0a22
BF		 762same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
763{
764 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 765}
766
767static int
599e0a22
BF		 768same_clid(clientid_t *cl1, clientid_t *cl2)
769{
770 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 771}
772
773/* XXX what about NGROUP */
774static int
599e0a22
BF		 775same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
776{
777 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 778}
779
5ec7b46c
BF		 780static void gen_clid(struct nfs4_client *clp)
781{
782 static u32 current_clientid = 1;
783
1da177e4
LT		 784 clp->cl_clientid.cl_boot = boot_time;
785 clp->cl_clientid.cl_id = current_clientid++;
786}
787
deda2faa
BF		 788static void gen_confirm(struct nfs4_client *clp)
789{
790 static u32 i;
791 u32 *p;
1da177e4 792
1da177e4 793 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 794 *p++ = get_seconds();
795 *p++ = i++;
1da177e4
LT		 796}
797
35bba9a3
BF		 798static int check_name(struct xdr_netobj name)
799{
1da177e4
LT		 800 if (name.len == 0)
801 return 0;
802 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 803 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 804 return 0;
805 }
806 return 1;
807}
808
fd39ca9a 809static void
1da177e4
LT		 810add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
811{
812 unsigned int idhashval;
813
814 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
815 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
816 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
817 list_add_tail(&clp->cl_lru, &client_lru);
818 clp->cl_time = get_seconds();
819}
820
fd39ca9a 821static void
1da177e4
LT		 822move_to_confirmed(struct nfs4_client *clp)
823{
824 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
825 unsigned int strhashval;
826
827 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
828 list_del_init(&clp->cl_strhash);
f116629d 829 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 830 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 831 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
832 renew_client(clp);
833}
834
835static struct nfs4_client *
836find_confirmed_client(clientid_t *clid)
837{
838 struct nfs4_client *clp;
839 unsigned int idhashval = clientid_hashval(clid->cl_id);
840
841 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 842 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 843 return clp;
844 }
845 return NULL;
846}
847
848static struct nfs4_client *
849find_unconfirmed_client(clientid_t *clid)
850{
851 struct nfs4_client *clp;
852 unsigned int idhashval = clientid_hashval(clid->cl_id);
853
854 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 855 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 856 return clp;
857 }
858 return NULL;
859}
860
a1bcecd2
AA		 861/*
862 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
863 * parameter. Matching is based on the fact the at least one of the
864 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
865 *
866 * FIXME: we need to unify the clientid namespaces for nfsv4.x
867 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
868 * and SET_CLIENTID{,_CONFIRM}
869 */
870static inline int
871match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
872{
873 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
874 return use_exchange_id == has_exchange_flags;
875}
876
28ce6054 877static struct nfs4_client *
a1bcecd2
AA		 878find_confirmed_client_by_str(const char *dname, unsigned int hashval,
879 bool use_exchange_id)
28ce6054
N		 880{
881 struct nfs4_client *clp;
882
883 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 884 if (same_name(clp->cl_recdir, dname) &&
885 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 886 return clp;
887 }
888 return NULL;
889}
890
891static struct nfs4_client *
a1bcecd2
AA		 892find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
893 bool use_exchange_id)
28ce6054
N		 894{
895 struct nfs4_client *clp;
896
897 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 898 if (same_name(clp->cl_recdir, dname) &&
899 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 900 return clp;
901 }
902 return NULL;
903}
904
1da177e4
LT		 905/* a helper function for parse_callback */
906static int
907parse_octet(unsigned int *lenp, char **addrp)
908{
909 unsigned int len = *lenp;
910 char *p = *addrp;
911 int n = -1;
912 char c;
913
914 for (;;) {
915 if (!len)
916 break;
917 len--;
918 c = *p++;
919 if (c == '.')
920 break;
921 if ((c < '0') || (c > '9')) {
922 n = -1;
923 break;
924 }
925 if (n < 0)
926 n = 0;
927 n = (n * 10) + (c - '0');
928 if (n > 255) {
929 n = -1;
930 break;
931 }
932 }
933 *lenp = len;
934 *addrp = p;
935 return n;
936}
937
938/* parse and set the setclientid ipv4 callback address */
fd39ca9a 939static int
1da177e4
LT		 940parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
941{
942 int temp = 0;
943 u32 cbaddr = 0;
944 u16 cbport = 0;
945 u32 addrlen = addr_len;
946 char *addr = addr_val;
947 int i, shift;
948
949 /* ipaddress */
950 shift = 24;
951 for(i = 4; i > 0 ; i--) {
952 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
953 return 0;
954 }
955 cbaddr |= (temp << shift);
956 if (shift > 0)
957 shift -= 8;
958 }
959 *cbaddrp = cbaddr;
960
961 /* port */
962 shift = 8;
963 for(i = 2; i > 0 ; i--) {
964 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
965 return 0;
966 }
967 cbport |= (temp << shift);
968 if (shift > 0)
969 shift -= 8;
970 }
971 *cbportp = cbport;
972 return 1;
973}
974
fd39ca9a 975static void
1da177e4
LT		 976gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
977{
c237dc03 978 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT		 979
980 /* Currently, we only support tcp for the callback channel */
981 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
982 goto out_err;
983
984 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
985 &cb->cb_addr, &cb->cb_port)))
986 goto out_err;
987 cb->cb_prog = se->se_callback_prog;
988 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 989 return;
990out_err:
849823c5 991 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 992 "will not receive delegations\n",
993 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
994
1da177e4
LT		 995 return;
996}
997
074fe897
AA		 998void
999nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
1000{
1001 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1002
1003 resp->cstate.statp = statp;
1004}
1005
1006/*
1007 * Dereference the result pages.
1008 */
1009static void
1010nfsd4_release_respages(struct page **respages, short resused)
1011{
1012 int i;
1013
1014 dprintk("--> %s\n", __func__);
1015 for (i = 0; i < resused; i++) {
1016 if (!respages[i])
1017 continue;
1018 put_page(respages[i]);
1019 respages[i] = NULL;
1020 }
1021}
1022
1023static void
1024nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1025{
1026 int i;
1027
1028 for (i = 0; i < count; i++) {
1029 topages[i] = frompages[i];
1030 if (!topages[i])
1031 continue;
1032 get_page(topages[i]);
1033 }
1034}
1035
1036/*
1037 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1038 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1039 * length of the XDR response is less than se_fmaxresp_cached
1040 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1041 * of the reply (e.g. readdir).
1042 *
1043 * Store the base and length of the rq_req.head[0] page
1044 * of the NFSv4.1 data, just past the rpc header.
1045 */
1046void
1047nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1048{
1049 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1050 struct svc_rqst *rqstp = resp->rqstp;
1051 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1052 struct nfsd4_op *op = &args->ops[resp->opcnt];
1053 struct kvec *resv = &rqstp->rq_res.head[0];
1054
1055 dprintk("--> %s entry %p\n", __func__, entry);
1056
1057 /* Don't cache a failed OP_SEQUENCE. */
1058 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1059 return;
bf864a31 1060
074fe897 1061 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1062 entry->ce_opcnt = resp->opcnt;
1063 entry->ce_status = resp->cstate.status;
1064
1065 /*
1066 * Don't need a page to cache just the sequence operation - the slot
1067 * does this for us!
1068 */
1069
1070 if (nfsd4_not_cached(resp)) {
1071 entry->ce_resused = 0;
1072 entry->ce_rpchdrlen = 0;
1073 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1074 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1075 return;
1076 }
074fe897
AA		 1077 entry->ce_resused = rqstp->rq_resused;
1078 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1079 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1080 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1081 entry->ce_resused);
074fe897
AA		 1082 entry->ce_datav.iov_base = resp->cstate.statp;
1083 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1084 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1085 /* Current request rpc header length*/
1086 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1087 (char *)page_address(rqstp->rq_respages[0]);
1088}
1089
1090/*
1091 * We keep the rpc header, but take the nfs reply from the replycache.
1092 */
1093static int
1094nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1095 struct nfsd4_cache_entry *entry)
1096{
1097 struct svc_rqst *rqstp = resp->rqstp;
1098 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1099 int len;
1100
1101 /* Current request rpc header length*/
1102 len = (char *)resp->cstate.statp -
1103 (char *)page_address(rqstp->rq_respages[0]);
1104 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1105 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1106 entry->ce_datav.iov_len);
1107 return 0;
1108 }
1109 /* copy the cached reply nfsd data past the current rpc header */
1110 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1111 entry->ce_datav.iov_len);
1112 resv->iov_len = len + entry->ce_datav.iov_len;
1113 return 1;
1114}
1115
1116/*
1117 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1118 * cached page. Replace any futher replay pages from the cache.
1119 */
1120__be32
bf864a31
AA		 1121nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1122 struct nfsd4_sequence *seq)
074fe897
AA		 1123{
1124 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1125 __be32 status;
1126
1127 dprintk("--> %s entry %p\n", __func__, entry);
1128
bf864a31
AA		 1129 /*
1130 * If this is just the sequence operation, we did not keep
1131 * a page in the cache entry because we can just use the
1132 * slot info stored in struct nfsd4_sequence that was checked
1133 * against the slot in nfsd4_sequence().
1134 *
1135 * This occurs when seq->cachethis is FALSE, or when the client
1136 * session inactivity timer fires and a solo sequence operation
1137 * is sent (lease renewal).
1138 */
1139 if (seq && nfsd4_not_cached(resp)) {
6c18ba9f 1140 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
bf864a31
AA		 1141 return nfs_ok;
1142 }
074fe897
AA		 1143
1144 if (!nfsd41_copy_replay_data(resp, entry)) {
1145 /*
1146 * Not enough room to use the replay rpc header, send the
1147 * cached header. Release all the allocated result pages.
1148 */
1149 svc_free_res_pages(resp->rqstp);
1150 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1151 entry->ce_resused);
1152 } else {
1153 /* Release all but the first allocated result page */
1154
1155 resp->rqstp->rq_resused--;
1156 svc_free_res_pages(resp->rqstp);
1157
1158 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1159 &entry->ce_respages[1],
1160 entry->ce_resused - 1);
1161 }
1162
1163 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1164 resp->opcnt = entry->ce_opcnt;
1165 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1166 status = entry->ce_status;
1167
1168 return status;
1169}
1170
0733d213
AA		 1171/*
1172 * Set the exchange_id flags returned by the server.
1173 */
1174static void
1175nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1176{
1177 /* pNFS is not supported */
1178 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1179
1180 /* Referrals are supported, Migration is not. */
1181 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1182
1183 /* set the wire flags to return to client. */
1184 clid->flags = new->cl_exchange_flags;
1185}
1186
069b6ad4
AA		 1187__be32
1188nfsd4_exchange_id(struct svc_rqst *rqstp,
1189 struct nfsd4_compound_state *cstate,
1190 struct nfsd4_exchange_id *exid)
1191{
0733d213
AA		 1192 struct nfs4_client *unconf, *conf, *new;
1193 int status;
1194 unsigned int strhashval;
1195 char dname[HEXDIR_LEN];
1196 nfs4_verifier verf = exid->verifier;
1197 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1198
1199 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1200 " ip_addr=%u flags %x, spa_how %d\n",
1201 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1202 ip_addr, exid->flags, exid->spa_how);
1203
1204 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1205 return nfserr_inval;
1206
1207 /* Currently only support SP4_NONE */
1208 switch (exid->spa_how) {
1209 case SP4_NONE:
1210 break;
1211 case SP4_SSV:
1212 return nfserr_encr_alg_unsupp;
1213 default:
1214 BUG(); /* checked by xdr code */
1215 case SP4_MACH_CRED:
1216 return nfserr_serverfault; /* no excuse :-/ */
1217 }
1218
1219 status = nfs4_make_rec_clidname(dname, &exid->clname);
1220
1221 if (status)
1222 goto error;
1223
1224 strhashval = clientstr_hashval(dname);
1225
1226 nfs4_lock_state();
1227 status = nfs_ok;
1228
a1bcecd2 1229 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1230 if (conf) {
1231 if (!same_verf(&verf, &conf->cl_verifier)) {
1232 /* 18.35.4 case 8 */
1233 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1234 status = nfserr_not_same;
1235 goto out;
1236 }
1237 /* Client reboot: destroy old state */
1238 expire_client(conf);
1239 goto out_new;
1240 }
1241 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1242 /* 18.35.4 case 9 */
1243 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1244 status = nfserr_perm;
1245 goto out;
1246 }
1247 expire_client(conf);
1248 goto out_new;
1249 }
1250 if (ip_addr != conf->cl_addr &&
1251 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1252 /* Client collision. 18.35.4 case 3 */
1253 status = nfserr_clid_inuse;
1254 goto out;
1255 }
1256 /*
1257 * Set bit when the owner id and verifier map to an already
1258 * confirmed client id (18.35.3).
1259 */
1260 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1261
1262 /*
1263 * Falling into 18.35.4 case 2, possible router replay.
1264 * Leave confirmed record intact and return same result.
1265 */
1266 copy_verf(conf, &verf);
1267 new = conf;
1268 goto out_copy;
1269 } else {
1270 /* 18.35.4 case 7 */
1271 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1272 status = nfserr_noent;
1273 goto out;
1274 }
1275 }
1276
a1bcecd2 1277 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1278 if (unconf) {
1279 /*
1280 * Possible retry or client restart. Per 18.35.4 case 4,
1281 * a new unconfirmed record should be generated regardless
1282 * of whether any properties have changed.
1283 */
1284 expire_client(unconf);
1285 }
1286
1287out_new:
1288 /* Normal case */
1289 new = create_client(exid->clname, dname);
1290 if (new == NULL) {
1291 status = nfserr_resource;
1292 goto out;
1293 }
1294
1295 copy_verf(new, &verf);
1296 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1297 new->cl_addr = ip_addr;
1298 gen_clid(new);
1299 gen_confirm(new);
1300 add_to_unconfirmed(new, strhashval);
1301out_copy:
1302 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1303 exid->clientid.cl_id = new->cl_clientid.cl_id;
1304
38eb76a5
AA		 1305 new->cl_slot.sl_seqid = 0;
1306 exid->seqid = 1;
0733d213
AA		 1307 nfsd4_set_ex_flags(new, exid);
1308
1309 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1310 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1311 status = nfs_ok;
1312
1313out:
1314 nfs4_unlock_state();
1315error:
1316 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1317 return status;
069b6ad4
AA		 1318}
1319
b85d4c01
BH		 1320static int
1321check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1322{
1323 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1324 slot->sl_seqid);
1325
1326 /* The slot is in use, and no response has been sent. */
1327 if (slot->sl_inuse) {
1328 if (seqid == slot->sl_seqid)
1329 return nfserr_jukebox;
1330 else
1331 return nfserr_seq_misordered;
1332 }
1333 /* Normal */
1334 if (likely(seqid == slot->sl_seqid + 1))
1335 return nfs_ok;
1336 /* Replay */
1337 if (seqid == slot->sl_seqid)
1338 return nfserr_replay_cache;
1339 /* Wraparound */
1340 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1341 return nfs_ok;
1342 /* Misordered replay or misordered new request */
1343 return nfserr_seq_misordered;
1344}
1345
069b6ad4
AA		 1346__be32
1347nfsd4_create_session(struct svc_rqst *rqstp,
1348 struct nfsd4_compound_state *cstate,
1349 struct nfsd4_create_session *cr_ses)
1350{
ec6b5d7b 1351 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1352 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1353 struct nfs4_client *conf, *unconf;
38eb76a5 1354 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1355 int status = 0;
1356
1357 nfs4_lock_state();
1358 unconf = find_unconfirmed_client(&cr_ses->clientid);
1359 conf = find_confirmed_client(&cr_ses->clientid);
1360
1361 if (conf) {
38eb76a5
AA		 1362 slot = &conf->cl_slot;
1363 status = check_slot_seqid(cr_ses->seqid, slot);
1364 if (status == nfserr_replay_cache) {
ec6b5d7b 1365 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1366 slot->sl_seqid);
1367 cstate->slot = slot;
1368 cstate->status = status;
1369 /* Return the cached reply status */
bf864a31 1370 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1371 goto out;
1372 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1373 status = nfserr_seq_misordered;
1374 dprintk("Sequence misordered!\n");
1375 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1376 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1377 goto out;
1378 }
38eb76a5 1379 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1380 } else if (unconf) {
1381 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1382 (ip_addr != unconf->cl_addr)) {
1383 status = nfserr_clid_inuse;
1384 goto out;
1385 }
1386
38eb76a5
AA		 1387 slot = &unconf->cl_slot;
1388 status = check_slot_seqid(cr_ses->seqid, slot);
1389 if (status) {
1390 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1391 status = nfserr_seq_misordered;
1392 goto out;
1393 }
1394
38eb76a5 1395 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1396 move_to_confirmed(unconf);
1397
1398 /*
1399 * We do not support RDMA or persistent sessions
1400 */
1401 cr_ses->flags &= ~SESSION4_PERSIST;
1402 cr_ses->flags &= ~SESSION4_RDMA;
1403
1404 conf = unconf;
1405 } else {
1406 status = nfserr_stale_clientid;
1407 goto out;
1408 }
1409
1410 status = alloc_init_session(rqstp, conf, cr_ses);
1411 if (status)
1412 goto out;
1413
ec6b5d7b
AA		 1414 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1415 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1416 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1417
38eb76a5
AA		 1418 slot->sl_inuse = true;
1419 cstate->slot = slot;
bf864a31
AA		 1420 /* Ensure a page is used for the cache */
1421 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1422out:
1423 nfs4_unlock_state();
1424 dprintk("%s returns %d\n", __func__, ntohl(status));
1425 return status;
069b6ad4
AA		 1426}
1427
1428__be32
1429nfsd4_destroy_session(struct svc_rqst *r,
1430 struct nfsd4_compound_state *cstate,
1431 struct nfsd4_destroy_session *sessionid)
1432{
e10e0cfc
BH		 1433 struct nfsd4_session *ses;
1434 u32 status = nfserr_badsession;
1435
1436 /* Notes:
1437 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1438 * - Should we return nfserr_back_chan_busy if waiting for
1439 * callbacks on to-be-destroyed session?
1440 * - Do we need to clear any callback info from previous session?
1441 */
1442
1443 dump_sessionid(__func__, &sessionid->sessionid);
1444 spin_lock(&sessionid_lock);
1445 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1446 if (!ses) {
1447 spin_unlock(&sessionid_lock);
1448 goto out;
1449 }
1450
1451 unhash_session(ses);
1452 spin_unlock(&sessionid_lock);
1453
1454 /* wait for callbacks */
1455 shutdown_callback_client(ses->se_client);
1456 nfsd4_put_session(ses);
1457 status = nfs_ok;
1458out:
1459 dprintk("%s returns %d\n", __func__, ntohl(status));
1460 return status;
069b6ad4
AA		 1461}
1462
1463__be32
b85d4c01 1464nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1465 struct nfsd4_compound_state *cstate,
1466 struct nfsd4_sequence *seq)
1467{
f9bb94c4 1468 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1469 struct nfsd4_session *session;
1470 struct nfsd4_slot *slot;
1471 int status;
1472
f9bb94c4
AA		 1473 if (resp->opcnt != 1)
1474 return nfserr_sequence_pos;
1475
b85d4c01
BH		 1476 spin_lock(&sessionid_lock);
1477 status = nfserr_badsession;
1478 session = find_in_sessionid_hashtbl(&seq->sessionid);
1479 if (!session)
1480 goto out;
1481
1482 status = nfserr_badslot;
6c18ba9f 1483 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1484 goto out;
1485
1486 slot = &session->se_slots[seq->slotid];
1487 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1488
1489 status = check_slot_seqid(seq->seqid, slot);
1490 if (status == nfserr_replay_cache) {
1491 cstate->slot = slot;
1492 cstate->session = session;
da3846a2 1493 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1494 * for nfsd4_svc_encode_compoundres processing */
1495 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1496 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1497 goto replay_cache;
1498 }
1499 if (status)
1500 goto out;
1501
1502 /* Success! bump slot seqid */
1503 slot->sl_inuse = true;
1504 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1505 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1506 /* Always set the cache entry cachethis for solo sequence */
1507 if (nfsd4_is_solo_sequence(resp))
1508 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1509
1510 cstate->slot = slot;
1511 cstate->session = session;
1512
1513replay_cache:
1514 /* Renew the clientid on success and on replay.
1515 * Hold a session reference until done processing the compound:
1516 * nfsd4_put_session called only if the cstate slot is set.
1517 */
1518 renew_client(session->se_client);
1519 nfsd4_get_session(session);
1520out:
1521 spin_unlock(&sessionid_lock);
1522 dprintk("%s: return %d\n", __func__, ntohl(status));
1523 return status;
069b6ad4
AA		 1524}
1525
b37ad28b 1526__be32
b591480b
BF		 1527nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1528 struct nfsd4_setclientid *setclid)
1da177e4 1529{
27459f09 1530 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1531 struct xdr_netobj clname = {
1532 .len = setclid->se_namelen,
1533 .data = setclid->se_name,
1534 };
1535 nfs4_verifier clverifier = setclid->se_verf;
1536 unsigned int strhashval;
28ce6054 1537 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1538 __be32 status;
68e76ad0 1539 char *princ;
a55370a3 1540 char dname[HEXDIR_LEN];
1da177e4 1541
1da177e4 1542 if (!check_name(clname))
73aea4ec 1543 return nfserr_inval;
1da177e4 1544
a55370a3
N		 1545 status = nfs4_make_rec_clidname(dname, &clname);
1546 if (status)
73aea4ec 1547 return status;
a55370a3 1548
1da177e4
LT		 1549 /*
1550 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1551 * We get here on a DRC miss.
1552 */
1553
a55370a3 1554 strhashval = clientstr_hashval(dname);
1da177e4 1555
1da177e4 1556 nfs4_lock_state();
a1bcecd2 1557 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1558 if (conf) {
a186e767 1559 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1560 status = nfserr_clid_inuse;
026722c2
BF		 1561 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1562 dprintk("NFSD: setclientid: string in use by client"
1563 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1564 goto out;
1565 }
1da177e4 1566 }
a186e767
BF		 1567 /*
1568 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1569 * has a description of SETCLIENTID request processing consisting
1570 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1571 */
a1bcecd2 1572 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1573 status = nfserr_resource;
1574 if (!conf) {
a186e767
BF		 1575 /*
1576 * RFC 3530 14.2.33 CASE 4:
1577 * placed first, because it is the normal case
1da177e4
LT		 1578 */
1579 if (unconf)
1580 expire_client(unconf);
a55370a3
N		 1581 new = create_client(clname, dname);
1582 if (new == NULL)
1da177e4 1583 goto out;
1da177e4 1584 gen_clid(new);
599e0a22 1585 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1586 /*
a186e767
BF		 1587 * RFC 3530 14.2.33 CASE 1:
1588 * probable callback update
1da177e4 1589 */
31f4a6c1
N		 1590 if (unconf) {
1591 /* Note this is removing unconfirmed {*x***},
1592 * which is stronger than RFC recommended {vxc**}.
1593 * This has the advantage that there is at most
1594 * one {*x***} in either list at any time.
1595 */
1596 expire_client(unconf);
1da177e4 1597 }
a55370a3
N		 1598 new = create_client(clname, dname);
1599 if (new == NULL)
1da177e4 1600 goto out;
1da177e4 1601 copy_clid(new, conf);
1da177e4
LT		 1602 } else if (!unconf) {
1603 /*
a186e767
BF		 1604 * RFC 3530 14.2.33 CASE 2:
1605 * probable client reboot; state will be removed if
1606 * confirmed.
1da177e4 1607 */
a55370a3
N		 1608 new = create_client(clname, dname);
1609 if (new == NULL)
1da177e4 1610 goto out;
1da177e4 1611 gen_clid(new);
49ba8781 1612 } else {
a186e767
BF		 1613 /*
1614 * RFC 3530 14.2.33 CASE 3:
1615 * probable client reboot; state will be removed if
1616 * confirmed.
1da177e4
LT		 1617 */
1618 expire_client(unconf);
a55370a3
N		 1619 new = create_client(clname, dname);
1620 if (new == NULL)
1da177e4 1621 goto out;
1da177e4 1622 gen_clid(new);
1da177e4 1623 }
c175b83c
BF		 1624 copy_verf(new, &clverifier);
1625 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1626 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1627 princ = svc_gss_principal(rqstp);
1628 if (princ) {
1629 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1630 if (new->cl_principal == NULL) {
1631 free_client(new);
1632 goto out;
1633 }
1634 }
c175b83c
BF		 1635 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1636 gen_confirm(new);
1637 gen_callback(new, setclid);
1638 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1639 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1640 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1641 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1642 status = nfs_ok;
1643out:
1644 nfs4_unlock_state();
1645 return status;
1646}
1647
1648
1649/*
a186e767
BF		 1650 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1651 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1652 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1653 */
b37ad28b 1654__be32
b591480b
BF		 1655nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1656 struct nfsd4_compound_state *cstate,
1657 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1658{
27459f09 1659 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1660 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1661 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1662 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1663 __be32 status;
1da177e4
LT		 1664
1665 if (STALE_CLIENTID(clid))
1666 return nfserr_stale_clientid;
1667 /*
1668 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1669 * We get here on a DRC miss.
1670 */
1671
1672 nfs4_lock_state();
21ab45a4
N		 1673
1674 conf = find_confirmed_client(clid);
1675 unconf = find_unconfirmed_client(clid);
1676
1677 status = nfserr_clid_inuse;
27459f09 1678 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1679 goto out;
27459f09 1680 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1681 goto out;
1682
a186e767
BF		 1683 /*
1684 * section 14.2.34 of RFC 3530 has a description of
1685 * SETCLIENTID_CONFIRM request processing consisting
1686 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1687 */
366e0c1d 1688 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1689 /*
1690 * RFC 3530 14.2.34 CASE 1:
1691 * callback update
1692 */
599e0a22 1693 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1694 status = nfserr_clid_inuse;
1695 else {
1a69c179
N		 1696 /* XXX: We just turn off callbacks until we can handle
1697 * change request correctly. */
c237dc03 1698 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1699 expire_client(unconf);
1da177e4 1700 status = nfs_ok;
1a69c179 1701
1da177e4 1702 }
f3aba4e5 1703 } else if (conf && !unconf) {
a186e767
BF		 1704 /*
1705 * RFC 3530 14.2.34 CASE 2:
1706 * probable retransmitted request; play it safe and
1707 * do nothing.
7c79f737 1708 */
599e0a22 1709 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1710 status = nfserr_clid_inuse;
21ab45a4 1711 else
1da177e4 1712 status = nfs_ok;
7c79f737 1713 } else if (!conf && unconf
599e0a22 1714 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1715 /*
1716 * RFC 3530 14.2.34 CASE 3:
1717 * Normal case; new or rebooted client:
7c79f737 1718 */
599e0a22 1719 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1720 status = nfserr_clid_inuse;
1721 } else {
1a69c179
N		 1722 unsigned int hash =
1723 clientstr_hashval(unconf->cl_recdir);
1724 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1725 hash, false);
1a69c179 1726 if (conf) {
c7b9a459 1727 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1728 expire_client(conf);
1729 }
1da177e4 1730 move_to_confirmed(unconf);
21ab45a4 1731 conf = unconf;
46f8a64b 1732 nfsd4_probe_callback(conf);
1a69c179 1733 status = nfs_ok;
1da177e4 1734 }
599e0a22
BF		 1735 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1736 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1737 &confirm)))) {
a186e767
BF		 1738 /*
1739 * RFC 3530 14.2.34 CASE 4:
1740 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1741 */
1da177e4 1742 status = nfserr_stale_clientid;
7c79f737 1743 } else {
08e8987c
N		 1744 /* check that we have hit one of the cases...*/
1745 status = nfserr_clid_inuse;
1746 }
1da177e4 1747out:
1da177e4
LT		 1748 nfs4_unlock_state();
1749 return status;
1750}
1751
1da177e4
LT		 1752/* OPEN Share state helper functions */
1753static inline struct nfs4_file *
1754alloc_init_file(struct inode *ino)
1755{
1756 struct nfs4_file *fp;
1757 unsigned int hashval = file_hashval(ino);
1758
e60d4398
N		 1759 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1760 if (fp) {
8b671b80 1761 atomic_set(&fp->fi_ref, 1);
1da177e4 1762 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1763 INIT_LIST_HEAD(&fp->fi_stateids);
1764 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1765 spin_lock(&recall_lock);
1da177e4 1766 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1767 spin_unlock(&recall_lock);
1da177e4
LT		 1768 fp->fi_inode = igrab(ino);
1769 fp->fi_id = current_fileid++;
47f9940c 1770 fp->fi_had_conflict = false;
1da177e4
LT		 1771 return fp;
1772 }
1773 return NULL;
1774}
1775
e60d4398 1776static void
e18b890b 1777nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1778{
e60d4398
N		 1779 if (*slab == NULL)
1780 return;
1a1d92c1 1781 kmem_cache_destroy(*slab);
e60d4398 1782 *slab = NULL;
1da177e4
LT		 1783}
1784
e8ff2a84 1785void
1da177e4
LT		 1786nfsd4_free_slabs(void)
1787{
e60d4398
N		 1788 nfsd4_free_slab(&stateowner_slab);
1789 nfsd4_free_slab(&file_slab);
5ac049ac 1790 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1791 nfsd4_free_slab(&deleg_slab);
e60d4398 1792}
1da177e4 1793
e60d4398
N		 1794static int
1795nfsd4_init_slabs(void)
1796{
1797 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1798 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1799 if (stateowner_slab == NULL)
1800 goto out_nomem;
1801 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1802 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1803 if (file_slab == NULL)
1804 goto out_nomem;
5ac049ac 1805 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1806 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1807 if (stateid_slab == NULL)
1808 goto out_nomem;
5b2d21c1 1809 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1810 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1811 if (deleg_slab == NULL)
1812 goto out_nomem;
e60d4398
N		 1813 return 0;
1814out_nomem:
1815 nfsd4_free_slabs();
1816 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1817 return -ENOMEM;
1da177e4
LT		 1818}
1819
1820void
1821nfs4_free_stateowner(struct kref *kref)
1822{
1823 struct nfs4_stateowner *sop =
1824 container_of(kref, struct nfs4_stateowner, so_ref);
1825 kfree(sop->so_owner.data);
1826 kmem_cache_free(stateowner_slab, sop);
1827}
1828
1829static inline struct nfs4_stateowner *
1830alloc_stateowner(struct xdr_netobj *owner)
1831{
1832 struct nfs4_stateowner *sop;
1833
1834 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1835 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1836 memcpy(sop->so_owner.data, owner->data, owner->len);
1837 sop->so_owner.len = owner->len;
1838 kref_init(&sop->so_ref);
1839 return sop;
1840 }
1841 kmem_cache_free(stateowner_slab, sop);
1842 }
1843 return NULL;
1844}
1845
1846static struct nfs4_stateowner *
1847alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1848 struct nfs4_stateowner *sop;
1849 struct nfs4_replay *rp;
1850 unsigned int idhashval;
1851
1852 if (!(sop = alloc_stateowner(&open->op_owner)))
1853 return NULL;
1854 idhashval = ownerid_hashval(current_ownerid);
1855 INIT_LIST_HEAD(&sop->so_idhash);
1856 INIT_LIST_HEAD(&sop->so_strhash);
1857 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1858 INIT_LIST_HEAD(&sop->so_stateids);
1859 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1860 INIT_LIST_HEAD(&sop->so_close_lru);
1861 sop->so_time = 0;
1862 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1863 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1864 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1865 sop->so_is_open_owner = 1;
1866 sop->so_id = current_ownerid++;
1867 sop->so_client = clp;
1868 sop->so_seqid = open->op_seqid;
1869 sop->so_confirmed = 0;
1870 rp = &sop->so_replay;
de1ae286 1871 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1872 rp->rp_buflen = 0;
1873 rp->rp_buf = rp->rp_ibuf;
1874 return sop;
1875}
1876
1da177e4
LT		 1877static inline void
1878init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1879 struct nfs4_stateowner *sop = open->op_stateowner;
1880 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1881
1882 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1883 INIT_LIST_HEAD(&stp->st_perstateowner);
1884 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1885 INIT_LIST_HEAD(&stp->st_perfile);
1886 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1887 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1888 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1889 stp->st_stateowner = sop;
13cd2184 1890 get_nfs4_file(fp);
1da177e4 1891 stp->st_file = fp;
78155ed7 1892 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1893 stp->st_stateid.si_stateownerid = sop->so_id;
1894 stp->st_stateid.si_fileid = fp->fi_id;
1895 stp->st_stateid.si_generation = 0;
1896 stp->st_access_bmap = 0;
1897 stp->st_deny_bmap = 0;
84459a11
AA		 1898 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1899 &stp->st_access_bmap);
1da177e4 1900 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1901 stp->st_openstp = NULL;
1da177e4
LT		 1902}
1903
fd39ca9a 1904static void
1da177e4
LT		 1905move_to_close_lru(struct nfs4_stateowner *sop)
1906{
1907 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1908
358dd55a 1909 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1910 sop->so_time = get_seconds();
1911}
1912
1da177e4 1913static int
599e0a22
BF		 1914same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1915 clientid_t *clid)
1916{
1917 return (sop->so_owner.len == owner->len) &&
1918 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1919 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1920}
1921
1922static struct nfs4_stateowner *
1923find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1924{
1925 struct nfs4_stateowner *so = NULL;
1926
1927 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1928 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1929 return so;
1930 }
1931 return NULL;
1932}
1933
1934/* search file_hashtbl[] for file */
1935static struct nfs4_file *
1936find_file(struct inode *ino)
1937{
1938 unsigned int hashval = file_hashval(ino);
1939 struct nfs4_file *fp;
1940
8b671b80 1941 spin_lock(&recall_lock);
1da177e4 1942 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1943 if (fp->fi_inode == ino) {
1944 get_nfs4_file(fp);
8b671b80 1945 spin_unlock(&recall_lock);
1da177e4 1946 return fp;
13cd2184 1947 }
1da177e4 1948 }
8b671b80 1949 spin_unlock(&recall_lock);
1da177e4
LT		 1950 return NULL;
1951}
1952
d87a8ade 1953static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1954{
d87a8ade 1955 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1956 return 0;
d87a8ade
AA		 1957 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1958 return 0;
1959 x &= ~NFS4_SHARE_ACCESS_MASK;
1960 if (minorversion && x) {
1961 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1962 return 0;
1963 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1964 return 0;
1965 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1966 }
1967 if (x)
8838dc43
BF		 1968 return 0;
1969 return 1;
ba5a6a19
BF		 1970}
1971
8838dc43 1972static inline int deny_valid(u32 x)
ba5a6a19 1973{
8838dc43
BF		 1974 /* Note: unlike access bits, deny bits may be zero. */
1975 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1976}
1da177e4 1977
4f83aa30
BF		 1978/*
1979 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1980 * st_{access,deny}_bmap field of the stateid, in order to track not
1981 * only what share bits are currently in force, but also what
1982 * combinations of share bits previous opens have used. This allows us
1983 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1984 * return an error if the client attempt to downgrade to a combination
1985 * of share bits not explicable by closing some of its previous opens.
1986 *
1987 * XXX: This enforcement is actually incomplete, since we don't keep
1988 * track of access/deny bit combinations; so, e.g., we allow:
1989 *
1990 * OPEN allow read, deny write
1991 * OPEN allow both, deny none
1992 * DOWNGRADE allow read, deny none
1993 *
1994 * which we should reject.
1995 */
fd39ca9a 1996static void
1da177e4
LT		 1997set_access(unsigned int *access, unsigned long bmap) {
1998 int i;
1999
2000 *access = 0;
2001 for (i = 1; i < 4; i++) {
2002 if (test_bit(i, &bmap))
2003 *access |= i;
2004 }
2005}
2006
fd39ca9a 2007static void
1da177e4
LT		 2008set_deny(unsigned int *deny, unsigned long bmap) {
2009 int i;
2010
2011 *deny = 0;
2012 for (i = 0; i < 4; i++) {
2013 if (test_bit(i, &bmap))
2014 *deny |= i ;
2015 }
2016}
2017
2018static int
2019test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2020 unsigned int access, deny;
2021
2022 set_access(&access, stp->st_access_bmap);
2023 set_deny(&deny, stp->st_deny_bmap);
2024 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2025 return 0;
2026 return 1;
2027}
2028
2029/*
2030 * Called to check deny when READ with all zero stateid or
2031 * WRITE with all zero or all one stateid
2032 */
b37ad28b 2033static __be32
1da177e4
LT		 2034nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2035{
2036 struct inode *ino = current_fh->fh_dentry->d_inode;
2037 struct nfs4_file *fp;
2038 struct nfs4_stateid *stp;
b37ad28b 2039 __be32 ret;
1da177e4
LT		 2040
2041 dprintk("NFSD: nfs4_share_conflict\n");
2042
2043 fp = find_file(ino);
13cd2184
N		 2044 if (!fp)
2045 return nfs_ok;
b700949b 2046 ret = nfserr_locked;
1da177e4 2047 /* Search for conflicting share reservations */
13cd2184
N		 2048 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2049 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2050 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2051 goto out;
1da177e4 2052 }
13cd2184
N		 2053 ret = nfs_ok;
2054out:
2055 put_nfs4_file(fp);
2056 return ret;
1da177e4
LT		 2057}
2058
2059static inline void
2060nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2061{
2062 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2063 drop_file_write_access(filp);
1da177e4
LT		 2064 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2065 }
2066}
2067
1da177e4
LT		 2068/*
2069 * Spawn a thread to perform a recall on the delegation represented
2070 * by the lease (file_lock)
2071 *
2072 * Called from break_lease() with lock_kernel() held.
2073 * Note: we assume break_lease will only call this *once* for any given
2074 * lease.
2075 */
2076static
2077void nfsd_break_deleg_cb(struct file_lock *fl)
2078{
63e4863f 2079 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2080
2081 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2082 if (!dp)
2083 return;
2084
2085 /* We're assuming the state code never drops its reference
2086 * without first removing the lease. Since we're in this lease
2087 * callback (and since the lease code is serialized by the kernel
2088 * lock) we know the server hasn't removed the lease yet, we know
2089 * it's safe to take a reference: */
2090 atomic_inc(&dp->dl_count);
cfdcad4d 2091 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2092
2093 spin_lock(&recall_lock);
2094 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2095 spin_unlock(&recall_lock);
2096
2097 /* only place dl_time is set. protected by lock_kernel*/
2098 dp->dl_time = get_seconds();
2099
0272e1fd
BF		 2100 /*
2101 * We don't want the locks code to timeout the lease for us;
2102 * we'll remove it ourself if the delegation isn't returned
2103 * in time.
2104 */
2105 fl->fl_break_time = 0;
1da177e4 2106
63e4863f
BF		 2107 dp->dl_file->fi_had_conflict = true;
2108 nfsd4_cb_recall(dp);
1da177e4
LT		 2109}
2110
2111/*
2112 * The file_lock is being reapd.
2113 *
2114 * Called by locks_free_lock() with lock_kernel() held.
2115 */
2116static
2117void nfsd_release_deleg_cb(struct file_lock *fl)
2118{
2119 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2120
2121 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2122
2123 if (!(fl->fl_flags & FL_LEASE) || !dp)
2124 return;
2125 dp->dl_flock = NULL;
2126}
2127
2128/*
2129 * Set the delegation file_lock back pointer.
2130 *
a9933cea 2131 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2132 */
2133static
2134void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2135{
2136 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2137
2138 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2139 if (!dp)
2140 return;
2141 dp->dl_flock = new;
2142}
2143
2144/*
a9933cea 2145 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2146 */
2147static
2148int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2149{
2150 struct nfs4_delegation *onlistd =
2151 (struct nfs4_delegation *)onlist->fl_owner;
2152 struct nfs4_delegation *tryd =
2153 (struct nfs4_delegation *)try->fl_owner;
2154
2155 if (onlist->fl_lmops != try->fl_lmops)
2156 return 0;
2157
2158 return onlistd->dl_client == tryd->dl_client;
2159}
2160
2161
2162static
2163int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2164{
2165 if (arg & F_UNLCK)
2166 return lease_modify(onlist, arg);
2167 else
2168 return -EAGAIN;
2169}
2170
fd39ca9a 2171static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2172 .fl_break = nfsd_break_deleg_cb,
2173 .fl_release_private = nfsd_release_deleg_cb,
2174 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2175 .fl_mylease = nfsd_same_client_deleg_cb,
2176 .fl_change = nfsd_change_deleg_cb,
2177};
2178
2179
b37ad28b 2180__be32
6668958f
AA		 2181nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2182 struct nfsd4_open *open)
1da177e4 2183{
1da177e4
LT		 2184 clientid_t *clientid = &open->op_clientid;
2185 struct nfs4_client *clp = NULL;
2186 unsigned int strhashval;
2187 struct nfs4_stateowner *sop = NULL;
2188
1da177e4 2189 if (!check_name(open->op_owner))
0f442aa2 2190 return nfserr_inval;
1da177e4
LT		 2191
2192 if (STALE_CLIENTID(&open->op_clientid))
2193 return nfserr_stale_clientid;
2194
2195 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2196 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2197 open->op_stateowner = sop;
2198 if (!sop) {
2199 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2200 clp = find_confirmed_client(clientid);
2201 if (clp == NULL)
0f442aa2
BF		 2202 return nfserr_expired;
2203 goto renew;
1da177e4 2204 }
6668958f
AA		 2205 /* When sessions are used, skip open sequenceid processing */
2206 if (nfsd4_has_session(cstate))
2207 goto renew;
0f442aa2
BF		 2208 if (!sop->so_confirmed) {
2209 /* Replace unconfirmed owners without checking for replay. */
2210 clp = sop->so_client;
f044ff83 2211 release_openowner(sop);
0f442aa2
BF		 2212 open->op_stateowner = NULL;
2213 goto renew;
2214 }
2215 if (open->op_seqid == sop->so_seqid - 1) {
2216 if (sop->so_replay.rp_buflen)
a90b061c 2217 return nfserr_replay_me;
0f442aa2
BF		 2218 /* The original OPEN failed so spectacularly
2219 * that we don't even have replay data saved!
2220 * Therefore, we have no choice but to continue
2221 * processing this OPEN; presumably, we'll
2222 * fail again for the same reason.
2223 */
2224 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2225 goto renew;
2226 }
2227 if (open->op_seqid != sop->so_seqid)
2228 return nfserr_bad_seqid;
1da177e4 2229renew:
0f442aa2
BF		 2230 if (open->op_stateowner == NULL) {
2231 sop = alloc_init_open_stateowner(strhashval, clp, open);
2232 if (sop == NULL)
2233 return nfserr_resource;
2234 open->op_stateowner = sop;
2235 }
2236 list_del_init(&sop->so_close_lru);
1da177e4 2237 renew_client(sop->so_client);
0f442aa2 2238 return nfs_ok;
1da177e4
LT		 2239}
2240
b37ad28b 2241static inline __be32
4a6e43e6
N		 2242nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2243{
2244 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2245 return nfserr_openmode;
2246 else
2247 return nfs_ok;
2248}
2249
52f4fb43
N		 2250static struct nfs4_delegation *
2251find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2252{
2253 struct nfs4_delegation *dp;
2254
ea1da636 2255 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2256 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2257 return dp;
2258 }
2259 return NULL;
2260}
2261
b37ad28b 2262static __be32
567d9829
N		 2263nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2264 struct nfs4_delegation **dp)
2265{
2266 int flags;
b37ad28b 2267 __be32 status = nfserr_bad_stateid;
567d9829
N		 2268
2269 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2270 if (*dp == NULL)
c44c5eeb 2271 goto out;
567d9829
N		 2272 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2273 RD_STATE : WR_STATE;
2274 status = nfs4_check_delegmode(*dp, flags);
2275 if (status)
2276 *dp = NULL;
c44c5eeb
N		 2277out:
2278 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2279 return nfs_ok;
2280 if (status)
2281 return status;
2282 open->op_stateowner->so_confirmed = 1;
2283 return nfs_ok;
567d9829
N		 2284}
2285
b37ad28b 2286static __be32
1da177e4
LT		 2287nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2288{
2289 struct nfs4_stateid *local;
b37ad28b 2290 __be32 status = nfserr_share_denied;
1da177e4
LT		 2291 struct nfs4_stateowner *sop = open->op_stateowner;
2292
8beefa24 2293 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2294 /* ignore lock owners */
2295 if (local->st_stateowner->so_is_open_owner == 0)
2296 continue;
2297 /* remember if we have seen this open owner */
2298 if (local->st_stateowner == sop)
2299 *stpp = local;
2300 /* check for conflicting share reservations */
2301 if (!test_share(local, open))
2302 goto out;
2303 }
2304 status = 0;
2305out:
2306 return status;
2307}
2308
5ac049ac
N		 2309static inline struct nfs4_stateid *
2310nfs4_alloc_stateid(void)
2311{
2312 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2313}
2314
b37ad28b 2315static __be32
1da177e4 2316nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2317 struct nfs4_delegation *dp,
1da177e4
LT		 2318 struct svc_fh *cur_fh, int flags)
2319{
2320 struct nfs4_stateid *stp;
1da177e4 2321
5ac049ac 2322 stp = nfs4_alloc_stateid();
1da177e4
LT		 2323 if (stp == NULL)
2324 return nfserr_resource;
2325
567d9829
N		 2326 if (dp) {
2327 get_file(dp->dl_vfs_file);
2328 stp->st_vfs_file = dp->dl_vfs_file;
2329 } else {
b37ad28b 2330 __be32 status;
567d9829
N		 2331 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2332 &stp->st_vfs_file);
2333 if (status) {
2334 if (status == nfserr_dropit)
2335 status = nfserr_jukebox;
5ac049ac 2336 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2337 return status;
2338 }
1da177e4 2339 }
1da177e4
LT		 2340 *stpp = stp;
2341 return 0;
2342}
2343
b37ad28b 2344static inline __be32
1da177e4
LT		 2345nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2346 struct nfsd4_open *open)
2347{
2348 struct iattr iattr = {
2349 .ia_valid = ATTR_SIZE,
2350 .ia_size = 0,
2351 };
2352 if (!open->op_truncate)
2353 return 0;
2354 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2355 return nfserr_inval;
1da177e4
LT		 2356 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2357}
2358
b37ad28b 2359static __be32
1da177e4
LT		 2360nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2361{
2362 struct file *filp = stp->st_vfs_file;
7eaa36e2 2363 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2364 unsigned int share_access, new_writer;
b37ad28b 2365 __be32 status;
1da177e4
LT		 2366
2367 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2368 new_writer = (~share_access) & open->op_share_access
2369 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2370
6c26d08f 2371 if (new_writer) {
b37ad28b
AV		 2372 int err = get_write_access(inode);
2373 if (err)
2374 return nfserrno(err);
e518f056
BH		 2375 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2376 if (err)
2377 return nfserrno(err);
2378 file_take_write(filp);
6c26d08f 2379 }
1da177e4
LT		 2380 status = nfsd4_truncate(rqstp, cur_fh, open);
2381 if (status) {
6c26d08f
BF		 2382 if (new_writer)
2383 put_write_access(inode);
1da177e4
LT		 2384 return status;
2385 }
2386 /* remember the open */
6c26d08f 2387 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2388 __set_bit(open->op_share_access, &stp->st_access_bmap);
2389 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2390
2391 return nfs_ok;
2392}
2393
2394
1da177e4 2395static void
37515177 2396nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2397{
37515177
N		 2398 open->op_stateowner->so_confirmed = 1;
2399 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2400}
2401
2402/*
2403 * Attempt to hand out a delegation.
2404 */
2405static void
2406nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2407{
2408 struct nfs4_delegation *dp;
2409 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2410 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2411 struct file_lock fl, *flp = &fl;
2412 int status, flag = 0;
2413
2414 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2415 open->op_recall = 0;
2416 switch (open->op_claim_type) {
2417 case NFS4_OPEN_CLAIM_PREVIOUS:
2418 if (!atomic_read(&cb->cb_set))
2419 open->op_recall = 1;
2420 flag = open->op_delegate_type;
2421 if (flag == NFS4_OPEN_DELEGATE_NONE)
2422 goto out;
2423 break;
2424 case NFS4_OPEN_CLAIM_NULL:
2425 /* Let's not give out any delegations till everyone's
2426 * had the chance to reclaim theirs.... */
af558e33 2427 if (locks_in_grace())
7b190fec
N		 2428 goto out;
2429 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2430 goto out;
2431 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2432 flag = NFS4_OPEN_DELEGATE_WRITE;
2433 else
2434 flag = NFS4_OPEN_DELEGATE_READ;
2435 break;
2436 default:
2437 goto out;
2438 }
1da177e4
LT		 2439
2440 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2441 if (dp == NULL) {
2442 flag = NFS4_OPEN_DELEGATE_NONE;
2443 goto out;
2444 }
2445 locks_init_lock(&fl);
2446 fl.fl_lmops = &nfsd_lease_mng_ops;
2447 fl.fl_flags = FL_LEASE;
9167f501 2448 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2449 fl.fl_end = OFFSET_MAX;
2450 fl.fl_owner = (fl_owner_t)dp;
2451 fl.fl_file = stp->st_vfs_file;
2452 fl.fl_pid = current->tgid;
2453
a9933cea 2454 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2455 * the lock_manager callbacks fl_mylease and fl_change are used
2456 */
9167f501 2457 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2458 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2459 unhash_delegation(dp);
1da177e4
LT		 2460 flag = NFS4_OPEN_DELEGATE_NONE;
2461 goto out;
2462 }
2463
2464 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2465
2466 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2467 dp->dl_stateid.si_boot,
2468 dp->dl_stateid.si_stateownerid,
2469 dp->dl_stateid.si_fileid,
2470 dp->dl_stateid.si_generation);
2471out:
7b190fec
N		 2472 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2473 && flag == NFS4_OPEN_DELEGATE_NONE
2474 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2475 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2476 open->op_delegate_type = flag;
2477}
2478
2479/*
2480 * called with nfs4_lock_state() held.
2481 */
b37ad28b 2482__be32
1da177e4
LT		 2483nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2484{
6668958f 2485 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2486 struct nfs4_file *fp = NULL;
2487 struct inode *ino = current_fh->fh_dentry->d_inode;
2488 struct nfs4_stateid *stp = NULL;
567d9829 2489 struct nfs4_delegation *dp = NULL;
b37ad28b 2490 __be32 status;
1da177e4
LT		 2491
2492 status = nfserr_inval;
d87a8ade 2493 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2494 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2495 goto out;
2496 /*
2497 * Lookup file; if found, lookup stateid and check open request,
2498 * and check for delegations in the process of being recalled.
2499 * If not found, create the nfs4_file struct
2500 */
2501 fp = find_file(ino);
2502 if (fp) {
2503 if ((status = nfs4_check_open(fp, open, &stp)))
2504 goto out;
c44c5eeb
N		 2505 status = nfs4_check_deleg(fp, open, &dp);
2506 if (status)
2507 goto out;
1da177e4 2508 } else {
c44c5eeb
N		 2509 status = nfserr_bad_stateid;
2510 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2511 goto out;
1da177e4
LT		 2512 status = nfserr_resource;
2513 fp = alloc_init_file(ino);
2514 if (fp == NULL)
2515 goto out;
2516 }
2517
2518 /*
2519 * OPEN the file, or upgrade an existing OPEN.
2520 * If truncate fails, the OPEN fails.
2521 */
2522 if (stp) {
2523 /* Stateid was found, this is an OPEN upgrade */
2524 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2525 if (status)
2526 goto out;
444c2c07 2527 update_stateid(&stp->st_stateid);
1da177e4
LT		 2528 } else {
2529 /* Stateid was not found, this is a new OPEN */
2530 int flags = 0;
9ecb6a08 2531 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2532 flags |= NFSD_MAY_READ;
1da177e4 2533 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2534 flags |= NFSD_MAY_WRITE;
567d9829
N		 2535 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2536 if (status)
1da177e4
LT		 2537 goto out;
2538 init_stateid(stp, fp, open);
2539 status = nfsd4_truncate(rqstp, current_fh, open);
2540 if (status) {
2283963f 2541 release_open_stateid(stp);
1da177e4
LT		 2542 goto out;
2543 }
6668958f
AA		 2544 if (nfsd4_has_session(&resp->cstate))
2545 update_stateid(&stp->st_stateid);
1da177e4
LT		 2546 }
2547 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2548
6668958f
AA		 2549 if (nfsd4_has_session(&resp->cstate))
2550 open->op_stateowner->so_confirmed = 1;
2551
1da177e4
LT		 2552 /*
2553 * Attempt to hand out a delegation. No error return, because the
2554 * OPEN succeeds even if we fail.
2555 */
2556 nfs4_open_delegation(current_fh, open, stp);
2557
2558 status = nfs_ok;
2559
2560 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2561 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2562 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2563out:
13cd2184
N		 2564 if (fp)
2565 put_nfs4_file(fp);
37515177
N		 2566 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2567 nfs4_set_claim_prev(open);
1da177e4
LT		 2568 /*
2569 * To finish the open response, we just need to set the rflags.
2570 */
2571 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2572 if (!open->op_stateowner->so_confirmed &&
2573 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2574 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2575
2576 return status;
2577}
2578
b37ad28b 2579__be32
b591480b
BF		 2580nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2581 clientid_t *clid)
1da177e4
LT		 2582{
2583 struct nfs4_client *clp;
b37ad28b 2584 __be32 status;
1da177e4
LT		 2585
2586 nfs4_lock_state();
2587 dprintk("process_renew(%08x/%08x): starting\n",
2588 clid->cl_boot, clid->cl_id);
2589 status = nfserr_stale_clientid;
2590 if (STALE_CLIENTID(clid))
2591 goto out;
2592 clp = find_confirmed_client(clid);
2593 status = nfserr_expired;
2594 if (clp == NULL) {
2595 /* We assume the client took too long to RENEW. */
2596 dprintk("nfsd4_renew: clientid not found!\n");
2597 goto out;
2598 }
2599 renew_client(clp);
2600 status = nfserr_cb_path_down;
ea1da636 2601 if (!list_empty(&clp->cl_delegations)
c237dc03 2602 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2603 goto out;
2604 status = nfs_ok;
2605out:
2606 nfs4_unlock_state();
2607 return status;
2608}
2609
af558e33
BF		 2610struct lock_manager nfsd4_manager = {
2611};
2612
a76b4319 2613static void
af558e33 2614nfsd4_end_grace(void)
a76b4319
N		 2615{
2616 dprintk("NFSD: end of grace period\n");
c7b9a459 2617 nfsd4_recdir_purge_old();
af558e33 2618 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2619}
2620
fd39ca9a 2621static time_t
1da177e4
LT		 2622nfs4_laundromat(void)
2623{
2624 struct nfs4_client *clp;
2625 struct nfs4_stateowner *sop;
2626 struct nfs4_delegation *dp;
2627 struct list_head *pos, *next, reaplist;
2628 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2629 time_t t, clientid_val = NFSD_LEASE_TIME;
2630 time_t u, test_val = NFSD_LEASE_TIME;
2631
2632 nfs4_lock_state();
2633
2634 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2635 if (locks_in_grace())
2636 nfsd4_end_grace();
1da177e4
LT		 2637 list_for_each_safe(pos, next, &client_lru) {
2638 clp = list_entry(pos, struct nfs4_client, cl_lru);
2639 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2640 t = clp->cl_time - cutoff;
2641 if (clientid_val > t)
2642 clientid_val = t;
2643 break;
2644 }
2645 dprintk("NFSD: purging unused client (clientid %08x)\n",
2646 clp->cl_clientid.cl_id);
c7b9a459 2647 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2648 expire_client(clp);
2649 }
2650 INIT_LIST_HEAD(&reaplist);
2651 spin_lock(&recall_lock);
2652 list_for_each_safe(pos, next, &del_recall_lru) {
2653 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2654 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2655 u = dp->dl_time - cutoff;
2656 if (test_val > u)
2657 test_val = u;
2658 break;
2659 }
2660 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2661 dp, dp->dl_flock);
2662 list_move(&dp->dl_recall_lru, &reaplist);
2663 }
2664 spin_unlock(&recall_lock);
2665 list_for_each_safe(pos, next, &reaplist) {
2666 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2667 list_del_init(&dp->dl_recall_lru);
2668 unhash_delegation(dp);
2669 }
2670 test_val = NFSD_LEASE_TIME;
2671 list_for_each_safe(pos, next, &close_lru) {
2672 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2673 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2674 u = sop->so_time - cutoff;
2675 if (test_val > u)
2676 test_val = u;
2677 break;
2678 }
2679 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2680 sop->so_id);
f044ff83 2681 release_openowner(sop);
1da177e4
LT		 2682 }
2683 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2684 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2685 nfs4_unlock_state();
2686 return clientid_val;
2687}
2688
a254b246
HH		 2689static struct workqueue_struct *laundry_wq;
2690static void laundromat_main(struct work_struct *);
2691static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2692
2693static void
c4028958 2694laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2695{
2696 time_t t;
2697
2698 t = nfs4_laundromat();
2699 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2700 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2701}
2702
fd39ca9a 2703static struct nfs4_stateowner *
f8816512
N		 2704search_close_lru(u32 st_id, int flags)
2705{
1da177e4
LT		 2706 struct nfs4_stateowner *local = NULL;
2707
1da177e4
LT		 2708 if (flags & CLOSE_STATE) {
2709 list_for_each_entry(local, &close_lru, so_close_lru) {
2710 if (local->so_id == st_id)
2711 return local;
2712 }
2713 }
2714 return NULL;
2715}
2716
2717static inline int
2718nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2719{
7eaa36e2 2720 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2721}
2722
2723static int
2724STALE_STATEID(stateid_t *stateid)
2725{
78155ed7
BN		 2726 if (time_after((unsigned long)boot_time,
2727 (unsigned long)stateid->si_boot)) {
2728 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2729 stateid->si_boot, stateid->si_stateownerid,
2730 stateid->si_fileid, stateid->si_generation);
2731 return 1;
2732 }
2733 return 0;
2734}
2735
2736static int
2737EXPIRED_STATEID(stateid_t *stateid)
2738{
2739 if (time_before((unsigned long)boot_time,
2740 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2741 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2742 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2743 stateid->si_boot, stateid->si_stateownerid,
2744 stateid->si_fileid, stateid->si_generation);
2745 return 1;
2746 }
2747 return 0;
2748}
2749
2750static __be32
2751stateid_error_map(stateid_t *stateid)
2752{
2753 if (STALE_STATEID(stateid))
2754 return nfserr_stale_stateid;
2755 if (EXPIRED_STATEID(stateid))
2756 return nfserr_expired;
2757
2758 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2759 stateid->si_boot, stateid->si_stateownerid,
2760 stateid->si_fileid, stateid->si_generation);
2761 return nfserr_bad_stateid;
1da177e4
LT		 2762}
2763
2764static inline int
2765access_permit_read(unsigned long access_bmap)
2766{
2767 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2768 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2769 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2770}
2771
2772static inline int
2773access_permit_write(unsigned long access_bmap)
2774{
2775 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2776 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2777}
2778
2779static
b37ad28b 2780__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2781{
b37ad28b 2782 __be32 status = nfserr_openmode;
1da177e4
LT		 2783
2784 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2785 goto out;
2786 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2787 goto out;
2788 status = nfs_ok;
2789out:
2790 return status;
2791}
2792
b37ad28b 2793static inline __be32
1da177e4
LT		 2794check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2795{
203a8c8e 2796 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2797 return nfs_ok;
af558e33 2798 else if (locks_in_grace()) {
1da177e4
LT		 2799 /* Answer in remaining cases depends on existance of
2800 * conflicting state; so we must wait out the grace period. */
2801 return nfserr_grace;
2802 } else if (flags & WR_STATE)
2803 return nfs4_share_conflict(current_fh,
2804 NFS4_SHARE_DENY_WRITE);
2805 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2806 return nfs4_share_conflict(current_fh,
2807 NFS4_SHARE_DENY_READ);
2808}
2809
2810/*
2811 * Allow READ/WRITE during grace period on recovered state only for files
2812 * that are not able to provide mandatory locking.
2813 */
2814static inline int
18f82731 2815grace_disallows_io(struct inode *inode)
1da177e4 2816{
203a8c8e 2817 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2818}
2819
6668958f 2820static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2821{
6668958f
AA		 2822 /*
2823 * When sessions are used the stateid generation number is ignored
2824 * when it is zero.
2825 */
2826 if ((flags & HAS_SESSION) && in->si_generation == 0)
2827 goto out;
2828
0836f587
BF		 2829 /* If the client sends us a stateid from the future, it's buggy: */
2830 if (in->si_generation > ref->si_generation)
2831 return nfserr_bad_stateid;
2832 /*
2833 * The following, however, can happen. For example, if the
2834 * client sends an open and some IO at the same time, the open
2835 * may bump si_generation while the IO is still in flight.
2836 * Thanks to hard links and renames, the client never knows what
2837 * file an open will affect. So it could avoid that situation
2838 * only by serializing all opens and IO from the same open
2839 * owner. To recover from the old_stateid error, the client
2840 * will just have to retry the IO:
2841 */
2842 if (in->si_generation < ref->si_generation)
2843 return nfserr_old_stateid;
6668958f 2844out:
0836f587
BF		 2845 return nfs_ok;
2846}
2847
3e633079
BF		 2848static int is_delegation_stateid(stateid_t *stateid)
2849{
2850 return stateid->si_fileid == 0;
2851}
2852
1da177e4
LT		 2853/*
2854* Checks for stateid operations
2855*/
b37ad28b 2856__be32
dd453dfd
BH		 2857nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2858 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2859{
2860 struct nfs4_stateid *stp = NULL;
2861 struct nfs4_delegation *dp = NULL;
dd453dfd 2862 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2863 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2864 __be32 status;
1da177e4 2865
1da177e4
LT		 2866 if (filpp)
2867 *filpp = NULL;
2868
18f82731 2869 if (grace_disallows_io(ino))
1da177e4
LT		 2870 return nfserr_grace;
2871
6668958f
AA		 2872 if (nfsd4_has_session(cstate))
2873 flags |= HAS_SESSION;
2874
1da177e4
LT		 2875 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2876 return check_special_stateids(current_fh, stateid, flags);
2877
1da177e4
LT		 2878 status = nfserr_stale_stateid;
2879 if (STALE_STATEID(stateid))
2880 goto out;
2881
1da177e4 2882 status = nfserr_bad_stateid;
3e633079 2883 if (is_delegation_stateid(stateid)) {
a4455be0 2884 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2885 if (!dp) {
2886 status = stateid_error_map(stateid);
1da177e4 2887 goto out;
78155ed7 2888 }
6668958f
AA		 2889 status = check_stateid_generation(stateid, &dp->dl_stateid,
2890 flags);
0c2a498f
BF		 2891 if (status)
2892 goto out;
dc9bf700
BF		 2893 status = nfs4_check_delegmode(dp, flags);
2894 if (status)
2895 goto out;
2896 renew_client(dp->dl_client);
dc9bf700
BF		 2897 if (filpp)
2898 *filpp = dp->dl_vfs_file;
1da177e4 2899 } else { /* open or lock stateid */
a4455be0 2900 stp = find_stateid(stateid, flags);
78155ed7
BN		 2901 if (!stp) {
2902 status = stateid_error_map(stateid);
1da177e4 2903 goto out;
78155ed7 2904 }
6150ef0d 2905 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2906 goto out;
2907 if (!stp->st_stateowner->so_confirmed)
2908 goto out;
6668958f
AA		 2909 status = check_stateid_generation(stateid, &stp->st_stateid,
2910 flags);
0c2a498f
BF		 2911 if (status)
2912 goto out;
a4455be0
BF		 2913 status = nfs4_check_openmode(stp, flags);
2914 if (status)
1da177e4
LT		 2915 goto out;
2916 renew_client(stp->st_stateowner->so_client);
2917 if (filpp)
2918 *filpp = stp->st_vfs_file;
1da177e4
LT		 2919 }
2920 status = nfs_ok;
2921out:
2922 return status;
2923}
2924
4c4cd222
N		 2925static inline int
2926setlkflg (int type)
2927{
2928 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2929 RD_STATE : WR_STATE;
2930}
1da177e4
LT		 2931
2932/*
2933 * Checks for sequence id mutating operations.
2934 */
b37ad28b 2935static __be32
dd453dfd
BH		 2936nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2937 stateid_t *stateid, int flags,
2938 struct nfs4_stateowner **sopp,
2939 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2940{
1da177e4
LT		 2941 struct nfs4_stateid *stp;
2942 struct nfs4_stateowner *sop;
dd453dfd 2943 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2944 __be32 status;
1da177e4
LT		 2945
2946 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2947 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2948 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2949 stateid->si_generation);
3a4f98bb 2950
1da177e4
LT		 2951 *stpp = NULL;
2952 *sopp = NULL;
2953
1da177e4 2954 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2955 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2956 return nfserr_bad_stateid;
1da177e4
LT		 2957 }
2958
1da177e4 2959 if (STALE_STATEID(stateid))
3a4f98bb 2960 return nfserr_stale_stateid;
6668958f
AA		 2961
2962 if (nfsd4_has_session(cstate))
2963 flags |= HAS_SESSION;
2964
1da177e4
LT		 2965 /*
2966 * We return BAD_STATEID if filehandle doesn't match stateid,
2967 * the confirmed flag is incorrecly set, or the generation
2968 * number is incorrect.
1da177e4 2969 */
f8816512
N		 2970 stp = find_stateid(stateid, flags);
2971 if (stp == NULL) {
2972 /*
2973 * Also, we should make sure this isn't just the result of
2974 * a replayed close:
2975 */
2976 sop = search_close_lru(stateid->si_stateownerid, flags);
2977 if (sop == NULL)
78155ed7 2978 return stateid_error_map(stateid);
f8816512
N		 2979 *sopp = sop;
2980 goto check_replay;
2981 }
1da177e4 2982
39325bd0
BF		 2983 *stpp = stp;
2984 *sopp = sop = stp->st_stateowner;
2985
4c4cd222 2986 if (lock) {
4c4cd222 2987 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2988 struct nfs4_client *clp = sop->so_client;
4c4cd222 2989 int lkflg = 0;
b37ad28b 2990 __be32 status;
4c4cd222
N		 2991
2992 lkflg = setlkflg(lock->lk_type);
2993
2994 if (lock->lk_is_new) {
599e0a22
BF		 2995 if (!sop->so_is_open_owner)
2996 return nfserr_bad_stateid;
60adfc50
AA		 2997 if (!(flags & HAS_SESSION) &&
2998 !same_clid(&clp->cl_clientid, lockclid))
2999 return nfserr_bad_stateid;
599e0a22
BF		 3000 /* stp is the open stateid */
3001 status = nfs4_check_openmode(stp, lkflg);
3002 if (status)
3003 return status;
3004 } else {
3005 /* stp is the lock stateid */
3006 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3007 if (status)
3008 return status;
4c4cd222 3009 }
1da177e4
LT		 3010 }
3011
f3362737 3012 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3013 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3014 return nfserr_bad_stateid;
1da177e4
LT		 3015 }
3016
1da177e4
LT		 3017 /*
3018 * We now validate the seqid and stateid generation numbers.
3019 * For the moment, we ignore the possibility of
3020 * generation number wraparound.
3021 */
6668958f 3022 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3023 goto check_replay;
3024
3a4f98bb 3025 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3026 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3027 " unconfirmed stateowner!\n");
3028 return nfserr_bad_stateid;
1da177e4 3029 }
3a4f98bb 3030 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3031 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3032 " confirmed yet!\n");
3033 return nfserr_bad_stateid;
1da177e4 3034 }
6668958f 3035 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3036 if (status)
3037 return status;
52fd004e 3038 renew_client(sop->so_client);
3a4f98bb 3039 return nfs_ok;
1da177e4 3040
1da177e4 3041check_replay:
bd9aac52 3042 if (seqid == sop->so_seqid - 1) {
849823c5 3043 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3044 /* indicate replay to calling function */
a90b061c 3045 return nfserr_replay_me;
1da177e4 3046 }
2fdada03 3047 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3048 sop->so_seqid, seqid);
3049 *sopp = NULL;
3050 return nfserr_bad_seqid;
1da177e4
LT		 3051}
3052
b37ad28b 3053__be32
ca364317 3054nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3055 struct nfsd4_open_confirm *oc)
1da177e4 3056{
b37ad28b 3057 __be32 status;
1da177e4
LT		 3058 struct nfs4_stateowner *sop;
3059 struct nfs4_stateid *stp;
3060
3061 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3062 (int)cstate->current_fh.fh_dentry->d_name.len,
3063 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3064
ca364317 3065 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3066 if (status)
3067 return status;
1da177e4
LT		 3068
3069 nfs4_lock_state();
3070
dd453dfd 3071 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3072 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3073 CONFIRM | OPEN_STATE,
1da177e4
LT		 3074 &oc->oc_stateowner, &stp, NULL)))
3075 goto out;
3076
3077 sop = oc->oc_stateowner;
3078 sop->so_confirmed = 1;
3079 update_stateid(&stp->st_stateid);
3080 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3081 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3082 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3083 stp->st_stateid.si_boot,
3084 stp->st_stateid.si_stateownerid,
3085 stp->st_stateid.si_fileid,
3086 stp->st_stateid.si_generation);
c7b9a459
N		 3087
3088 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3089out:
f2327d9a 3090 if (oc->oc_stateowner) {
1da177e4 3091 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3092 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3093 }
1da177e4
LT		 3094 nfs4_unlock_state();
3095 return status;
3096}
3097
3098
3099/*
3100 * unset all bits in union bitmap (bmap) that
3101 * do not exist in share (from successful OPEN_DOWNGRADE)
3102 */
3103static void
3104reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3105{
3106 int i;
3107 for (i = 1; i < 4; i++) {
3108 if ((i & access) != i)
3109 __clear_bit(i, bmap);
3110 }
3111}
3112
3113static void
3114reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3115{
3116 int i;
3117 for (i = 0; i < 4; i++) {
3118 if ((i & deny) != i)
3119 __clear_bit(i, bmap);
3120 }
3121}
3122
b37ad28b 3123__be32
ca364317
BF		 3124nfsd4_open_downgrade(struct svc_rqst *rqstp,
3125 struct nfsd4_compound_state *cstate,
a4f1706a 3126 struct nfsd4_open_downgrade *od)
1da177e4 3127{
b37ad28b 3128 __be32 status;
1da177e4
LT		 3129 struct nfs4_stateid *stp;
3130 unsigned int share_access;
3131
3132 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3133 (int)cstate->current_fh.fh_dentry->d_name.len,
3134 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3135
d87a8ade 3136 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3137 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3138 return nfserr_inval;
3139
3140 nfs4_lock_state();
dd453dfd 3141 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3142 od->od_seqid,
1da177e4 3143 &od->od_stateid,
f3362737 3144 OPEN_STATE,
1da177e4
LT		 3145 &od->od_stateowner, &stp, NULL)))
3146 goto out;
3147
3148 status = nfserr_inval;
3149 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3150 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3151 stp->st_access_bmap, od->od_share_access);
3152 goto out;
3153 }
3154 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3155 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3156 stp->st_deny_bmap, od->od_share_deny);
3157 goto out;
3158 }
3159 set_access(&share_access, stp->st_access_bmap);
3160 nfs4_file_downgrade(stp->st_vfs_file,
3161 share_access & ~od->od_share_access);
3162
3163 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3164 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3165
3166 update_stateid(&stp->st_stateid);
3167 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3168 status = nfs_ok;
3169out:
f2327d9a 3170 if (od->od_stateowner) {
1da177e4 3171 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3172 cstate->replay_owner = od->od_stateowner;
f2327d9a 3173 }
1da177e4
LT		 3174 nfs4_unlock_state();
3175 return status;
3176}
3177
3178/*
3179 * nfs4_unlock_state() called after encode
3180 */
b37ad28b 3181__be32
ca364317 3182nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3183 struct nfsd4_close *close)
1da177e4 3184{
b37ad28b 3185 __be32 status;
1da177e4
LT		 3186 struct nfs4_stateid *stp;
3187
3188 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3189 (int)cstate->current_fh.fh_dentry->d_name.len,
3190 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3191
3192 nfs4_lock_state();
3193 /* check close_lru for replay */
dd453dfd 3194 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3195 close->cl_seqid,
1da177e4 3196 &close->cl_stateid,
f3362737 3197 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3198 &close->cl_stateowner, &stp, NULL)))
3199 goto out;
1da177e4
LT		 3200 status = nfs_ok;
3201 update_stateid(&stp->st_stateid);
3202 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3203
04ef5954 3204 /* release_stateid() calls nfsd_close() if needed */
2283963f 3205 release_open_stateid(stp);
04ef5954
BF		 3206
3207 /* place unused nfs4_stateowners on so_close_lru list to be
3208 * released by the laundromat service after the lease period
3209 * to enable us to handle CLOSE replay
3210 */
3211 if (list_empty(&close->cl_stateowner->so_stateids))
3212 move_to_close_lru(close->cl_stateowner);
1da177e4 3213out:
f2327d9a 3214 if (close->cl_stateowner) {
1da177e4 3215 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3216 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3217 }
1da177e4
LT		 3218 nfs4_unlock_state();
3219 return status;
3220}
3221
b37ad28b 3222__be32
ca364317
BF		 3223nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3224 struct nfsd4_delegreturn *dr)
1da177e4 3225{
203a8c8e
BF		 3226 struct nfs4_delegation *dp;
3227 stateid_t *stateid = &dr->dr_stateid;
3228 struct inode *inode;
b37ad28b 3229 __be32 status;
6668958f 3230 int flags = 0;
1da177e4 3231
ca364317 3232 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3233 return status;
3234 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3235
6668958f
AA		 3236 if (nfsd4_has_session(cstate))
3237 flags |= HAS_SESSION;
1da177e4 3238 nfs4_lock_state();
203a8c8e
BF		 3239 status = nfserr_bad_stateid;
3240 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3241 goto out;
3242 status = nfserr_stale_stateid;
3243 if (STALE_STATEID(stateid))
3244 goto out;
7e0f7cf5 3245 status = nfserr_bad_stateid;
203a8c8e
BF		 3246 if (!is_delegation_stateid(stateid))
3247 goto out;
203a8c8e 3248 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3249 if (!dp) {
3250 status = stateid_error_map(stateid);
203a8c8e 3251 goto out;
78155ed7 3252 }
6668958f 3253 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3254 if (status)
3255 goto out;
3256 renew_client(dp->dl_client);
3257
3258 unhash_delegation(dp);
1da177e4 3259out:
203a8c8e
BF		 3260 nfs4_unlock_state();
3261
1da177e4
LT		 3262 return status;
3263}
3264
3265
3266/*
3267 * Lock owner state (byte-range locks)
3268 */
3269#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3270#define LOCK_HASH_BITS 8
3271#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3272#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3273
87df4de8
BH		 3274static inline u64
3275end_offset(u64 start, u64 len)
3276{
3277 u64 end;
3278
3279 end = start + len;
3280 return end >= start ? end: NFS4_MAX_UINT64;
3281}
3282
3283/* last octet in a range */
3284static inline u64
3285last_byte_offset(u64 start, u64 len)
3286{
3287 u64 end;
3288
3289 BUG_ON(!len);
3290 end = start + len;
3291 return end > start ? end - 1: NFS4_MAX_UINT64;
3292}
3293
1da177e4
LT		 3294#define lockownerid_hashval(id) \
3295 ((id) & LOCK_HASH_MASK)
3296
3297static inline unsigned int
3298lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3299 struct xdr_netobj *ownername)
3300{
3301 return (file_hashval(inode) + cl_id
3302 + opaque_hashval(ownername->data, ownername->len))
3303 & LOCK_HASH_MASK;
3304}
3305
3306static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3307static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3308static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3309
fd39ca9a 3310static struct nfs4_stateid *
1da177e4
LT		 3311find_stateid(stateid_t *stid, int flags)
3312{
9346eff0 3313 struct nfs4_stateid *local;
1da177e4
LT		 3314 u32 st_id = stid->si_stateownerid;
3315 u32 f_id = stid->si_fileid;
3316 unsigned int hashval;
3317
3318 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3319 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3320 hashval = stateid_hashval(st_id, f_id);
3321 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3322 if ((local->st_stateid.si_stateownerid == st_id) &&
3323 (local->st_stateid.si_fileid == f_id))
3324 return local;
3325 }
3326 }
9346eff0
KK		 3327
3328 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3329 hashval = stateid_hashval(st_id, f_id);
3330 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3331 if ((local->st_stateid.si_stateownerid == st_id) &&
3332 (local->st_stateid.si_fileid == f_id))
3333 return local;
3334 }
849823c5 3335 }
1da177e4
LT		 3336 return NULL;
3337}
3338
3339static struct nfs4_delegation *
3340find_delegation_stateid(struct inode *ino, stateid_t *stid)
3341{
13cd2184
N		 3342 struct nfs4_file *fp;
3343 struct nfs4_delegation *dl;
1da177e4
LT		 3344
3345 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3346 stid->si_boot, stid->si_stateownerid,
3347 stid->si_fileid, stid->si_generation);
3348
1da177e4 3349 fp = find_file(ino);
13cd2184
N		 3350 if (!fp)
3351 return NULL;
3352 dl = find_delegation_file(fp, stid);
3353 put_nfs4_file(fp);
3354 return dl;
1da177e4
LT		 3355}
3356
3357/*
3358 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3359 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3360 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3361 * locking, this prevents us from being completely protocol-compliant. The
3362 * real solution to this problem is to start using unsigned file offsets in
3363 * the VFS, but this is a very deep change!
3364 */
3365static inline void
3366nfs4_transform_lock_offset(struct file_lock *lock)
3367{
3368 if (lock->fl_start < 0)
3369 lock->fl_start = OFFSET_MAX;
3370 if (lock->fl_end < 0)
3371 lock->fl_end = OFFSET_MAX;
3372}
3373
d5b9026a
N		 3374/* Hack!: For now, we're defining this just so we can use a pointer to it
3375 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3376static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3377};
1da177e4
LT		 3378
3379static inline void
3380nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3381{
d5b9026a
N		 3382 struct nfs4_stateowner *sop;
3383 unsigned int hval;
1da177e4 3384
d5b9026a
N		 3385 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3386 sop = (struct nfs4_stateowner *) fl->fl_owner;
3387 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3388 kref_get(&sop->so_ref);
3389 deny->ld_sop = sop;
3390 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3391 } else {
3392 deny->ld_sop = NULL;
3393 deny->ld_clientid.cl_boot = 0;
3394 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3395 }
3396 deny->ld_start = fl->fl_start;
87df4de8
BH		 3397 deny->ld_length = NFS4_MAX_UINT64;
3398 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3399 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3400 deny->ld_type = NFS4_READ_LT;
3401 if (fl->fl_type != F_RDLCK)
3402 deny->ld_type = NFS4_WRITE_LT;
3403}
3404
1da177e4
LT		 3405static struct nfs4_stateowner *
3406find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3407 struct xdr_netobj *owner)
3408{
3409 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3410 struct nfs4_stateowner *op;
3411
3412 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3413 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3414 return op;
3415 }
3416 return NULL;
3417}
3418
3419/*
3420 * Alloc a lock owner structure.
3421 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3422 * occured.
3423 *
3424 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3425 */
3426
3427static struct nfs4_stateowner *
3428alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3429 struct nfs4_stateowner *sop;
3430 struct nfs4_replay *rp;
3431 unsigned int idhashval;
3432
3433 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3434 return NULL;
3435 idhashval = lockownerid_hashval(current_ownerid);
3436 INIT_LIST_HEAD(&sop->so_idhash);
3437 INIT_LIST_HEAD(&sop->so_strhash);
3438 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3439 INIT_LIST_HEAD(&sop->so_stateids);
3440 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3441 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3442 sop->so_time = 0;
3443 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3444 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3445 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3446 sop->so_is_open_owner = 0;
3447 sop->so_id = current_ownerid++;
3448 sop->so_client = clp;
b59e3c0e
NB		 3449 /* It is the openowner seqid that will be incremented in encode in the
3450 * case of new lockowners; so increment the lock seqid manually: */
3451 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3452 sop->so_confirmed = 1;
3453 rp = &sop->so_replay;
de1ae286 3454 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3455 rp->rp_buflen = 0;
3456 rp->rp_buf = rp->rp_ibuf;
3457 return sop;
3458}
3459
fd39ca9a 3460static struct nfs4_stateid *
1da177e4
LT		 3461alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3462{
3463 struct nfs4_stateid *stp;
3464 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3465
5ac049ac
N		 3466 stp = nfs4_alloc_stateid();
3467 if (stp == NULL)
1da177e4
LT		 3468 goto out;
3469 INIT_LIST_HEAD(&stp->st_hash);
3470 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3471 INIT_LIST_HEAD(&stp->st_perstateowner);
3472 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3473 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3474 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3475 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3476 stp->st_stateowner = sop;
13cd2184 3477 get_nfs4_file(fp);
1da177e4 3478 stp->st_file = fp;
78155ed7 3479 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3480 stp->st_stateid.si_stateownerid = sop->so_id;
3481 stp->st_stateid.si_fileid = fp->fi_id;
3482 stp->st_stateid.si_generation = 0;
3483 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3484 stp->st_access_bmap = open_stp->st_access_bmap;
3485 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3486 stp->st_openstp = open_stp;
1da177e4
LT		 3487
3488out:
3489 return stp;
3490}
3491
fd39ca9a 3492static int
1da177e4
LT		 3493check_lock_length(u64 offset, u64 length)
3494{
87df4de8 3495 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3496 LOFF_OVERFLOW(offset, length)));
3497}
3498
3499/*
3500 * LOCK operation
3501 */
b37ad28b 3502__be32
ca364317 3503nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3504 struct nfsd4_lock *lock)
1da177e4 3505{
3e9e3dbe 3506 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3507 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3508 struct nfs4_stateid *lock_stp;
3509 struct file *filp;
3510 struct file_lock file_lock;
8dc7c311 3511 struct file_lock conflock;
b37ad28b 3512 __be32 status = 0;
1da177e4 3513 unsigned int strhashval;
150b3934 3514 unsigned int cmd;
b8dd7b9a 3515 int err;
1da177e4
LT		 3516
3517 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3518 (long long) lock->lk_offset,
3519 (long long) lock->lk_length);
3520
1da177e4
LT		 3521 if (check_lock_length(lock->lk_offset, lock->lk_length))
3522 return nfserr_inval;
3523
ca364317 3524 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3525 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3526 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3527 return status;
3528 }
3529
1da177e4
LT		 3530 nfs4_lock_state();
3531
3532 if (lock->lk_is_new) {
893f8770
N		 3533 /*
3534 * Client indicates that this is a new lockowner.
3535 * Use open owner and open stateid to create lock owner and
3536 * lock stateid.
3537 */
1da177e4
LT		 3538 struct nfs4_stateid *open_stp = NULL;
3539 struct nfs4_file *fp;
3540
3541 status = nfserr_stale_clientid;
60adfc50
AA		 3542 if (!nfsd4_has_session(cstate) &&
3543 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3544 goto out;
1da177e4 3545
1da177e4 3546 /* validate and update open stateid and open seqid */
dd453dfd 3547 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3548 lock->lk_new_open_seqid,
3549 &lock->lk_new_open_stateid,
f3362737 3550 OPEN_STATE,
3a65588a 3551 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3552 lock);
37515177 3553 if (status)
1da177e4 3554 goto out;
3a65588a 3555 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3556 /* create lockowner and lock stateid */
3557 fp = open_stp->st_file;
3558 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3559 open_sop->so_client->cl_clientid.cl_id,
3560 &lock->v.new.owner);
3e9e3dbe
N		 3561 /* XXX: Do we need to check for duplicate stateowners on
3562 * the same file, or should they just be allowed (and
3563 * create new stateids)? */
1da177e4 3564 status = nfserr_resource;
b59e3c0e
NB		 3565 lock_sop = alloc_init_lock_stateowner(strhashval,
3566 open_sop->so_client, open_stp, lock);
3567 if (lock_sop == NULL)
1da177e4 3568 goto out;
b59e3c0e 3569 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3570 if (lock_stp == NULL)
1da177e4 3571 goto out;
1da177e4
LT		 3572 } else {
3573 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3574 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3575 lock->lk_old_lock_seqid,
3576 &lock->lk_old_lock_stateid,
f3362737 3577 LOCK_STATE,
3a65588a 3578 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3579 if (status)
3580 goto out;
3a65588a 3581 lock_sop = lock->lk_replay_owner;
1da177e4 3582 }
3a65588a 3583 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3584 filp = lock_stp->st_vfs_file;
3585
0dd395dc 3586 status = nfserr_grace;
af558e33 3587 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3588 goto out;
3589 status = nfserr_no_grace;
af558e33 3590 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3591 goto out;
3592
1da177e4
LT		 3593 locks_init_lock(&file_lock);
3594 switch (lock->lk_type) {
3595 case NFS4_READ_LT:
3596 case NFS4_READW_LT:
3597 file_lock.fl_type = F_RDLCK;
150b3934 3598 cmd = F_SETLK;
1da177e4
LT		 3599 break;
3600 case NFS4_WRITE_LT:
3601 case NFS4_WRITEW_LT:
3602 file_lock.fl_type = F_WRLCK;
150b3934 3603 cmd = F_SETLK;
1da177e4
LT		 3604 break;
3605 default:
3606 status = nfserr_inval;
3607 goto out;
3608 }
b59e3c0e 3609 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3610 file_lock.fl_pid = current->tgid;
3611 file_lock.fl_file = filp;
3612 file_lock.fl_flags = FL_POSIX;
d5b9026a 3613 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3614
3615 file_lock.fl_start = lock->lk_offset;
87df4de8 3616 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3617 nfs4_transform_lock_offset(&file_lock);
3618
3619 /*
3620 * Try to lock the file in the VFS.
3621 * Note: locks.c uses the BKL to protect the inode's lock list.
3622 */
3623
fd85b817 3624 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3625 switch (-err) {
1da177e4
LT		 3626 case 0: /* success! */
3627 update_stateid(&lock_stp->st_stateid);
3628 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3629 sizeof(stateid_t));
b8dd7b9a 3630 status = 0;
eb76b3fd
AA		 3631 break;
3632 case (EAGAIN): /* conflock holds conflicting lock */
3633 status = nfserr_denied;
3634 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3635 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3636 break;
1da177e4
LT		 3637 case (EDEADLK):
3638 status = nfserr_deadlock;
eb76b3fd 3639 break;
1da177e4 3640 default:
fd85b817 3641 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3642 status = nfserr_resource;
3643 break;
1da177e4 3644 }
1da177e4 3645out:
8a280510 3646 if (status && lock->lk_is_new && lock_sop)
f044ff83 3647 release_lockowner(lock_sop);
3a65588a
BF		 3648 if (lock->lk_replay_owner) {
3649 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3650 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3651 }
1da177e4
LT		 3652 nfs4_unlock_state();
3653 return status;
3654}
3655
55ef1274
BF		 3656/*
3657 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3658 * so we do a temporary open here just to get an open file to pass to
3659 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3660 * inode operation.)
3661 */
3662static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3663{
3664 struct file *file;
3665 int err;
3666
3667 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3668 if (err)
3669 return err;
3670 err = vfs_test_lock(file, lock);
3671 nfsd_close(file);
3672 return err;
3673}
3674
1da177e4
LT		 3675/*
3676 * LOCKT operation
3677 */
b37ad28b 3678__be32
ca364317
BF		 3679nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3680 struct nfsd4_lockt *lockt)
1da177e4
LT		 3681{
3682 struct inode *inode;
1da177e4 3683 struct file_lock file_lock;
fd85b817 3684 int error;
b37ad28b 3685 __be32 status;
1da177e4 3686
af558e33 3687 if (locks_in_grace())
1da177e4
LT		 3688 return nfserr_grace;
3689
3690 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3691 return nfserr_inval;
3692
3693 lockt->lt_stateowner = NULL;
3694 nfs4_lock_state();
3695
3696 status = nfserr_stale_clientid;
60adfc50 3697 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3698 goto out;
1da177e4 3699
ca364317 3700 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3701 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3702 if (status == nfserr_symlink)
3703 status = nfserr_inval;
3704 goto out;
3705 }
3706
ca364317 3707 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3708 locks_init_lock(&file_lock);
3709 switch (lockt->lt_type) {
3710 case NFS4_READ_LT:
3711 case NFS4_READW_LT:
3712 file_lock.fl_type = F_RDLCK;
3713 break;
3714 case NFS4_WRITE_LT:
3715 case NFS4_WRITEW_LT:
3716 file_lock.fl_type = F_WRLCK;
3717 break;
3718 default:
2fdada03 3719 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3720 status = nfserr_inval;
3721 goto out;
3722 }
3723
3724 lockt->lt_stateowner = find_lockstateowner_str(inode,
3725 &lockt->lt_clientid, &lockt->lt_owner);
3726 if (lockt->lt_stateowner)
3727 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3728 file_lock.fl_pid = current->tgid;
3729 file_lock.fl_flags = FL_POSIX;
3730
3731 file_lock.fl_start = lockt->lt_offset;
87df4de8 3732 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3733
3734 nfs4_transform_lock_offset(&file_lock);
3735
1da177e4 3736 status = nfs_ok;
55ef1274 3737 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3738 if (error) {
3739 status = nfserrno(error);
3740 goto out;
3741 }
9d6a8c5c 3742 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3743 status = nfserr_denied;
9d6a8c5c 3744 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3745 }
3746out:
3747 nfs4_unlock_state();
3748 return status;
3749}
3750
b37ad28b 3751__be32
ca364317 3752nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3753 struct nfsd4_locku *locku)
1da177e4
LT		 3754{
3755 struct nfs4_stateid *stp;
3756 struct file *filp = NULL;
3757 struct file_lock file_lock;
b37ad28b 3758 __be32 status;
b8dd7b9a 3759 int err;
1da177e4
LT		 3760
3761 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3762 (long long) locku->lu_offset,
3763 (long long) locku->lu_length);
3764
3765 if (check_lock_length(locku->lu_offset, locku->lu_length))
3766 return nfserr_inval;
3767
3768 nfs4_lock_state();
3769
dd453dfd 3770 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3771 locku->lu_seqid,
3772 &locku->lu_stateid,
f3362737 3773 LOCK_STATE,
1da177e4
LT		 3774 &locku->lu_stateowner, &stp, NULL)))
3775 goto out;
3776
3777 filp = stp->st_vfs_file;
3778 BUG_ON(!filp);
3779 locks_init_lock(&file_lock);
3780 file_lock.fl_type = F_UNLCK;
3781 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3782 file_lock.fl_pid = current->tgid;
3783 file_lock.fl_file = filp;
3784 file_lock.fl_flags = FL_POSIX;
d5b9026a 3785 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3786 file_lock.fl_start = locku->lu_offset;
3787
87df4de8 3788 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3789 nfs4_transform_lock_offset(&file_lock);
3790
3791 /*
3792 * Try to unlock the file in the VFS.
3793 */
fd85b817 3794 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3795 if (err) {
fd85b817 3796 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3797 goto out_nfserr;
3798 }
3799 /*
3800 * OK, unlock succeeded; the only thing left to do is update the stateid.
3801 */
3802 update_stateid(&stp->st_stateid);
3803 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3804
3805out:
f2327d9a 3806 if (locku->lu_stateowner) {
1da177e4 3807 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3808 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3809 }
1da177e4
LT		 3810 nfs4_unlock_state();
3811 return status;
3812
3813out_nfserr:
b8dd7b9a 3814 status = nfserrno(err);
1da177e4
LT		 3815 goto out;
3816}
3817
3818/*
3819 * returns
3820 * 1: locks held by lockowner
3821 * 0: no locks held by lockowner
3822 */
3823static int
3824check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3825{
3826 struct file_lock **flpp;
7eaa36e2 3827 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3828 int status = 0;
3829
3830 lock_kernel();
3831 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3832 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3833 status = 1;
3834 goto out;
796dadfd 3835 }
1da177e4
LT		 3836 }
3837out:
3838 unlock_kernel();
3839 return status;
3840}
3841
b37ad28b 3842__be32
b591480b
BF		 3843nfsd4_release_lockowner(struct svc_rqst *rqstp,
3844 struct nfsd4_compound_state *cstate,
3845 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3846{
3847 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3848 struct nfs4_stateowner *sop;
3849 struct nfs4_stateid *stp;
1da177e4 3850 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3851 struct list_head matches;
3852 int i;
b37ad28b 3853 __be32 status;
1da177e4
LT		 3854
3855 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3856 clid->cl_boot, clid->cl_id);
3857
3858 /* XXX check for lease expiration */
3859
3860 status = nfserr_stale_clientid;
849823c5 3861 if (STALE_CLIENTID(clid))
1da177e4 3862 return status;
1da177e4
LT		 3863
3864 nfs4_lock_state();
3865
3e9e3dbe
N		 3866 status = nfserr_locks_held;
3867 /* XXX: we're doing a linear search through all the lockowners.
3868 * Yipes! For now we'll just hope clients aren't really using
3869 * release_lockowner much, but eventually we have to fix these
3870 * data structures. */
3871 INIT_LIST_HEAD(&matches);
3872 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3873 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3874 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3875 continue;
3876 list_for_each_entry(stp, &sop->so_stateids,
3877 st_perstateowner) {
3878 if (check_for_locks(stp->st_vfs_file, sop))
3879 goto out;
3880 /* Note: so_perclient unused for lockowners,
3881 * so it's OK to fool with here. */
3882 list_add(&sop->so_perclient, &matches);
3883 }
1da177e4 3884 }
3e9e3dbe
N		 3885 }
3886 /* Clients probably won't expect us to return with some (but not all)
3887 * of the lockowner state released; so don't release any until all
3888 * have been checked. */
3889 status = nfs_ok;
0fa822e4
N		 3890 while (!list_empty(&matches)) {
3891 sop = list_entry(matches.next, struct nfs4_stateowner,
3892 so_perclient);
3893 /* unhash_stateowner deletes so_perclient only
3894 * for openowners. */
3895 list_del(&sop->so_perclient);
f044ff83 3896 release_lockowner(sop);
1da177e4
LT		 3897 }
3898out:
3899 nfs4_unlock_state();
3900 return status;
3901}
3902
3903static inline struct nfs4_client_reclaim *
a55370a3 3904alloc_reclaim(void)
1da177e4 3905{
a55370a3 3906 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3907}
3908
c7b9a459 3909int
a1bcecd2 3910nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3911{
3912 unsigned int strhashval = clientstr_hashval(name);
3913 struct nfs4_client *clp;
3914
a1bcecd2 3915 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3916 return clp ? 1 : 0;
3917}
3918
1da177e4
LT		 3919/*
3920 * failure => all reset bets are off, nfserr_no_grace...
3921 */
190e4fbf
N		 3922int
3923nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3924{
3925 unsigned int strhashval;
3926 struct nfs4_client_reclaim *crp = NULL;
3927
a55370a3
N		 3928 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3929 crp = alloc_reclaim();
1da177e4
LT		 3930 if (!crp)
3931 return 0;
a55370a3 3932 strhashval = clientstr_hashval(name);
1da177e4
LT		 3933 INIT_LIST_HEAD(&crp->cr_strhash);
3934 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3935 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3936 reclaim_str_hashtbl_size++;
3937 return 1;
3938}
3939
3940static void
3941nfs4_release_reclaim(void)
3942{
3943 struct nfs4_client_reclaim *crp = NULL;
3944 int i;
3945
1da177e4
LT		 3946 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3947 while (!list_empty(&reclaim_str_hashtbl[i])) {
3948 crp = list_entry(reclaim_str_hashtbl[i].next,
3949 struct nfs4_client_reclaim, cr_strhash);
3950 list_del(&crp->cr_strhash);
1da177e4
LT		 3951 kfree(crp);
3952 reclaim_str_hashtbl_size--;
3953 }
3954 }
3955 BUG_ON(reclaim_str_hashtbl_size);
3956}
3957
3958/*
3959 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3960static struct nfs4_client_reclaim *
1da177e4
LT		 3961nfs4_find_reclaim_client(clientid_t *clid)
3962{
3963 unsigned int strhashval;
3964 struct nfs4_client *clp;
3965 struct nfs4_client_reclaim *crp = NULL;
3966
3967
3968 /* find clientid in conf_id_hashtbl */
3969 clp = find_confirmed_client(clid);
3970 if (clp == NULL)
3971 return NULL;
3972
a55370a3
N		 3973 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3974 clp->cl_name.len, clp->cl_name.data,
3975 clp->cl_recdir);
1da177e4
LT		 3976
3977 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3978 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3979 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3980 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3981 return crp;
3982 }
3983 }
3984 return NULL;
3985}
3986
3987/*
3988* Called from OPEN. Look for clientid in reclaim list.
3989*/
b37ad28b 3990__be32
1da177e4
LT		 3991nfs4_check_open_reclaim(clientid_t *clid)
3992{
dfc83565 3993 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3994}
3995
ac4d8ff2 3996/* initialization to perform at module load time: */
1da177e4 3997
e8ff2a84 3998int
ac4d8ff2 3999nfs4_state_init(void)
1da177e4 4000{
e8ff2a84 4001 int i, status;
1da177e4 4002
e8ff2a84
BF		 4003 status = nfsd4_init_slabs();
4004 if (status)
4005 return status;
1da177e4
LT		 4006 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4007 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4008 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4009 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4010 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4011 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4012 }
5282fd72
ME		 4013 for (i = 0; i < SESSION_HASH_SIZE; i++)
4014 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4015 for (i = 0; i < FILE_HASH_SIZE; i++) {
4016 INIT_LIST_HEAD(&file_hashtbl[i]);
4017 }
4018 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4019 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4020 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4021 }
4022 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4023 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4024 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4025 }
4026 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4027 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4028 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4029 }
1da177e4 4030 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4031 INIT_LIST_HEAD(&close_lru);
4032 INIT_LIST_HEAD(&client_lru);
4033 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4034 reclaim_str_hashtbl_size = 0;
e8ff2a84 4035 return 0;
ac4d8ff2
N		 4036}
4037
190e4fbf
N		 4038static void
4039nfsd4_load_reboot_recovery_data(void)
4040{
4041 int status;
4042
0964a3d3
N		 4043 nfs4_lock_state();
4044 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4045 status = nfsd4_recdir_load();
0964a3d3 4046 nfs4_unlock_state();
190e4fbf
N		 4047 if (status)
4048 printk("NFSD: Failure reading reboot recovery data\n");
4049}
4050
9a8db97e
ME		 4051unsigned long
4052get_nfs4_grace_period(void)
4053{
4054 return max(user_lease_time, lease_time) * HZ;
4055}
4056
c2f1a551
MS		 4057/*
4058 * Since the lifetime of a delegation isn't limited to that of an open, a
4059 * client may quite reasonably hang on to a delegation as long as it has
4060 * the inode cached. This becomes an obvious problem the first time a
4061 * client's inode cache approaches the size of the server's total memory.
4062 *
4063 * For now we avoid this problem by imposing a hard limit on the number
4064 * of delegations, which varies according to the server's memory size.
4065 */
4066static void
4067set_max_delegations(void)
4068{
4069 /*
4070 * Allow at most 4 delegations per megabyte of RAM. Quick
4071 * estimates suggest that in the worst case (where every delegation
4072 * is for a different inode), a delegation could take about 1.5K,
4073 * giving a worst case usage of about 6% of memory.
4074 */
4075 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4076}
4077
ac4d8ff2
N		 4078/* initialization to perform when the nfsd service is started: */
4079
4080static void
4081__nfs4_state_start(void)
4082{
9a8db97e 4083 unsigned long grace_time;
ac4d8ff2 4084
1da177e4 4085 boot_time = get_seconds();
af558e33 4086 grace_time = get_nfs4_grace_period();
d99a05ad 4087 lease_time = user_lease_time;
af558e33 4088 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4089 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4090 grace_time/HZ);
58da282b 4091 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4092 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4093 set_max_delegations();
1da177e4
LT		 4094}
4095
e8ff2a84 4096void
76a3550e 4097nfs4_state_start(void)
1da177e4 4098{
1da177e4 4099 if (nfs4_init)
e8ff2a84 4100 return;
190e4fbf 4101 nfsd4_load_reboot_recovery_data();
76a3550e 4102 __nfs4_state_start();
1da177e4 4103 nfs4_init = 1;
e8ff2a84 4104 return;
1da177e4
LT		 4105}
4106
1da177e4
LT		 4107time_t
4108nfs4_lease_time(void)
4109{
4110 return lease_time;
4111}
4112
4113static void
4114__nfs4_state_shutdown(void)
4115{
4116 int i;
4117 struct nfs4_client *clp = NULL;
4118 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4119 struct list_head *pos, *next, reaplist;
4120
1da177e4
LT		 4121 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4122 while (!list_empty(&conf_id_hashtbl[i])) {
4123 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4124 expire_client(clp);
4125 }
4126 while (!list_empty(&unconf_str_hashtbl[i])) {
4127 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4128 expire_client(clp);
4129 }
4130 }
4131 INIT_LIST_HEAD(&reaplist);
4132 spin_lock(&recall_lock);
4133 list_for_each_safe(pos, next, &del_recall_lru) {
4134 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4135 list_move(&dp->dl_recall_lru, &reaplist);
4136 }
4137 spin_unlock(&recall_lock);
4138 list_for_each_safe(pos, next, &reaplist) {
4139 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4140 list_del_init(&dp->dl_recall_lru);
4141 unhash_delegation(dp);
4142 }
4143
190e4fbf 4144 nfsd4_shutdown_recdir();
1da177e4 4145 nfs4_init = 0;
1da177e4
LT		 4146}
4147
4148void
4149nfs4_state_shutdown(void)
4150{
5e8d5c29
N		 4151 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4152 destroy_workqueue(laundry_wq);
2c5e7615 4153 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4154 nfs4_lock_state();
4155 nfs4_release_reclaim();
4156 __nfs4_state_shutdown();
1da177e4
LT		 4157 nfs4_unlock_state();
4158}
4159
3dd98a3b
JL		 4160/*
4161 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4162 * accessed when nfsd is starting.
4163 */
0964a3d3
N		 4164static void
4165nfs4_set_recdir(char *recdir)
4166{
0964a3d3 4167 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4168}
4169
4170/*
4171 * Change the NFSv4 recovery directory to recdir.
4172 */
4173int
4174nfs4_reset_recoverydir(char *recdir)
4175{
4176 int status;
a63bb996 4177 struct path path;
0964a3d3 4178
a63bb996 4179 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4180 if (status)
4181 return status;
4182 status = -ENOTDIR;
a63bb996 4183 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4184 nfs4_set_recdir(recdir);
4185 status = 0;
4186 }
a63bb996 4187 path_put(&path);
0964a3d3
N		 4188 return status;
4189}
4190
3dd98a3b
JL		 4191char *
4192nfs4_recoverydir(void)
4193{
4194 return user_recovery_dirname;
4195}
4196
1da177e4
LT		 4197/*
4198 * Called when leasetime is changed.
4199 *
d99a05ad
N		 4200 * The only way the protocol gives us to handle on-the-fly lease changes is to
4201 * simulate a reboot. Instead of doing that, we just wait till the next time
4202 * we start to register any changes in lease time. If the administrator
4203 * really wants to change the lease time *now*, they can go ahead and bring
4204 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4205 *
4206 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4207 * when nfsd is starting
1da177e4
LT		 4208 */
4209void
4210nfs4_reset_lease(time_t leasetime)
4211{
d99a05ad 4212 user_lease_time = leasetime;
1da177e4 4213}
Clone of davem's net-next-2.6 tree