projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: Backchannel: cleanup nfs4.0 callback encode routines
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
1da177e4
LT		 58
59#define NFSDDBG_FACILITY NFSDDBG_PROC
60
61/* Globals */
62static time_t lease_time = 90; /* default lease time */
d99a05ad 63static time_t user_lease_time = 90;
fd39ca9a 64static time_t boot_time;
1da177e4
LT		 65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N		 69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
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 }
0733d213
AA		 1250 /*
1251 * Set bit when the owner id and verifier map to an already
1252 * confirmed client id (18.35.3).
1253 */
1254 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1255
1256 /*
1257 * Falling into 18.35.4 case 2, possible router replay.
1258 * Leave confirmed record intact and return same result.
1259 */
1260 copy_verf(conf, &verf);
1261 new = conf;
1262 goto out_copy;
6ddbbbfe
MS		 1263 }
1264
1265 /* 18.35.4 case 7 */
1266 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1267 status = nfserr_noent;
1268 goto out;
0733d213
AA		 1269 }
1270
a1bcecd2 1271 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1272 if (unconf) {
1273 /*
1274 * Possible retry or client restart. Per 18.35.4 case 4,
1275 * a new unconfirmed record should be generated regardless
1276 * of whether any properties have changed.
1277 */
1278 expire_client(unconf);
1279 }
1280
1281out_new:
1282 /* Normal case */
1283 new = create_client(exid->clname, dname);
1284 if (new == NULL) {
1285 status = nfserr_resource;
1286 goto out;
1287 }
1288
1289 copy_verf(new, &verf);
1290 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1291 new->cl_addr = ip_addr;
1292 gen_clid(new);
1293 gen_confirm(new);
1294 add_to_unconfirmed(new, strhashval);
1295out_copy:
1296 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1297 exid->clientid.cl_id = new->cl_clientid.cl_id;
1298
38eb76a5
AA		 1299 new->cl_slot.sl_seqid = 0;
1300 exid->seqid = 1;
0733d213
AA		 1301 nfsd4_set_ex_flags(new, exid);
1302
1303 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1304 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1305 status = nfs_ok;
1306
1307out:
1308 nfs4_unlock_state();
1309error:
1310 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1311 return status;
069b6ad4
AA		 1312}
1313
b85d4c01
BH		 1314static int
1315check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1316{
1317 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1318 slot->sl_seqid);
1319
1320 /* The slot is in use, and no response has been sent. */
1321 if (slot->sl_inuse) {
1322 if (seqid == slot->sl_seqid)
1323 return nfserr_jukebox;
1324 else
1325 return nfserr_seq_misordered;
1326 }
1327 /* Normal */
1328 if (likely(seqid == slot->sl_seqid + 1))
1329 return nfs_ok;
1330 /* Replay */
1331 if (seqid == slot->sl_seqid)
1332 return nfserr_replay_cache;
1333 /* Wraparound */
1334 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1335 return nfs_ok;
1336 /* Misordered replay or misordered new request */
1337 return nfserr_seq_misordered;
1338}
1339
069b6ad4
AA		 1340__be32
1341nfsd4_create_session(struct svc_rqst *rqstp,
1342 struct nfsd4_compound_state *cstate,
1343 struct nfsd4_create_session *cr_ses)
1344{
ec6b5d7b 1345 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1346 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1347 struct nfs4_client *conf, *unconf;
38eb76a5 1348 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1349 int status = 0;
1350
1351 nfs4_lock_state();
1352 unconf = find_unconfirmed_client(&cr_ses->clientid);
1353 conf = find_confirmed_client(&cr_ses->clientid);
1354
1355 if (conf) {
38eb76a5
AA		 1356 slot = &conf->cl_slot;
1357 status = check_slot_seqid(cr_ses->seqid, slot);
1358 if (status == nfserr_replay_cache) {
ec6b5d7b 1359 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1360 slot->sl_seqid);
1361 cstate->slot = slot;
1362 cstate->status = status;
1363 /* Return the cached reply status */
bf864a31 1364 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1365 goto out;
1366 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1367 status = nfserr_seq_misordered;
1368 dprintk("Sequence misordered!\n");
1369 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1370 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1371 goto out;
1372 }
38eb76a5 1373 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1374 } else if (unconf) {
1375 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1376 (ip_addr != unconf->cl_addr)) {
1377 status = nfserr_clid_inuse;
1378 goto out;
1379 }
1380
38eb76a5
AA		 1381 slot = &unconf->cl_slot;
1382 status = check_slot_seqid(cr_ses->seqid, slot);
1383 if (status) {
1384 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1385 status = nfserr_seq_misordered;
1386 goto out;
1387 }
1388
38eb76a5 1389 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1390 move_to_confirmed(unconf);
1391
1392 /*
1393 * We do not support RDMA or persistent sessions
1394 */
1395 cr_ses->flags &= ~SESSION4_PERSIST;
1396 cr_ses->flags &= ~SESSION4_RDMA;
1397
1398 conf = unconf;
1399 } else {
1400 status = nfserr_stale_clientid;
1401 goto out;
1402 }
1403
1404 status = alloc_init_session(rqstp, conf, cr_ses);
1405 if (status)
1406 goto out;
1407
ec6b5d7b
AA		 1408 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1409 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1410 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1411
38eb76a5
AA		 1412 slot->sl_inuse = true;
1413 cstate->slot = slot;
bf864a31
AA		 1414 /* Ensure a page is used for the cache */
1415 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1416out:
1417 nfs4_unlock_state();
1418 dprintk("%s returns %d\n", __func__, ntohl(status));
1419 return status;
069b6ad4
AA		 1420}
1421
1422__be32
1423nfsd4_destroy_session(struct svc_rqst *r,
1424 struct nfsd4_compound_state *cstate,
1425 struct nfsd4_destroy_session *sessionid)
1426{
e10e0cfc
BH		 1427 struct nfsd4_session *ses;
1428 u32 status = nfserr_badsession;
1429
1430 /* Notes:
1431 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1432 * - Should we return nfserr_back_chan_busy if waiting for
1433 * callbacks on to-be-destroyed session?
1434 * - Do we need to clear any callback info from previous session?
1435 */
1436
1437 dump_sessionid(__func__, &sessionid->sessionid);
1438 spin_lock(&sessionid_lock);
1439 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1440 if (!ses) {
1441 spin_unlock(&sessionid_lock);
1442 goto out;
1443 }
1444
1445 unhash_session(ses);
1446 spin_unlock(&sessionid_lock);
1447
1448 /* wait for callbacks */
1449 shutdown_callback_client(ses->se_client);
1450 nfsd4_put_session(ses);
1451 status = nfs_ok;
1452out:
1453 dprintk("%s returns %d\n", __func__, ntohl(status));
1454 return status;
069b6ad4
AA		 1455}
1456
1457__be32
b85d4c01 1458nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1459 struct nfsd4_compound_state *cstate,
1460 struct nfsd4_sequence *seq)
1461{
f9bb94c4 1462 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1463 struct nfsd4_session *session;
1464 struct nfsd4_slot *slot;
1465 int status;
1466
f9bb94c4
AA		 1467 if (resp->opcnt != 1)
1468 return nfserr_sequence_pos;
1469
b85d4c01
BH		 1470 spin_lock(&sessionid_lock);
1471 status = nfserr_badsession;
1472 session = find_in_sessionid_hashtbl(&seq->sessionid);
1473 if (!session)
1474 goto out;
1475
1476 status = nfserr_badslot;
6c18ba9f 1477 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1478 goto out;
1479
1480 slot = &session->se_slots[seq->slotid];
1481 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1482
1483 status = check_slot_seqid(seq->seqid, slot);
1484 if (status == nfserr_replay_cache) {
1485 cstate->slot = slot;
1486 cstate->session = session;
da3846a2 1487 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1488 * for nfsd4_svc_encode_compoundres processing */
1489 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1490 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1491 goto replay_cache;
1492 }
1493 if (status)
1494 goto out;
1495
1496 /* Success! bump slot seqid */
1497 slot->sl_inuse = true;
1498 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1499 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1500 /* Always set the cache entry cachethis for solo sequence */
1501 if (nfsd4_is_solo_sequence(resp))
1502 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1503
1504 cstate->slot = slot;
1505 cstate->session = session;
1506
1507replay_cache:
1508 /* Renew the clientid on success and on replay.
1509 * Hold a session reference until done processing the compound:
1510 * nfsd4_put_session called only if the cstate slot is set.
1511 */
1512 renew_client(session->se_client);
1513 nfsd4_get_session(session);
1514out:
1515 spin_unlock(&sessionid_lock);
1516 dprintk("%s: return %d\n", __func__, ntohl(status));
1517 return status;
069b6ad4
AA		 1518}
1519
b37ad28b 1520__be32
b591480b
BF		 1521nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1522 struct nfsd4_setclientid *setclid)
1da177e4 1523{
27459f09 1524 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1525 struct xdr_netobj clname = {
1526 .len = setclid->se_namelen,
1527 .data = setclid->se_name,
1528 };
1529 nfs4_verifier clverifier = setclid->se_verf;
1530 unsigned int strhashval;
28ce6054 1531 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1532 __be32 status;
68e76ad0 1533 char *princ;
a55370a3 1534 char dname[HEXDIR_LEN];
1da177e4 1535
1da177e4 1536 if (!check_name(clname))
73aea4ec 1537 return nfserr_inval;
1da177e4 1538
a55370a3
N		 1539 status = nfs4_make_rec_clidname(dname, &clname);
1540 if (status)
73aea4ec 1541 return status;
a55370a3 1542
1da177e4
LT		 1543 /*
1544 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1545 * We get here on a DRC miss.
1546 */
1547
a55370a3 1548 strhashval = clientstr_hashval(dname);
1da177e4 1549
1da177e4 1550 nfs4_lock_state();
a1bcecd2 1551 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1552 if (conf) {
a186e767 1553 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1554 status = nfserr_clid_inuse;
026722c2
BF		 1555 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1556 dprintk("NFSD: setclientid: string in use by client"
1557 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1558 goto out;
1559 }
1da177e4 1560 }
a186e767
BF		 1561 /*
1562 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1563 * has a description of SETCLIENTID request processing consisting
1564 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1565 */
a1bcecd2 1566 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1567 status = nfserr_resource;
1568 if (!conf) {
a186e767
BF		 1569 /*
1570 * RFC 3530 14.2.33 CASE 4:
1571 * placed first, because it is the normal case
1da177e4
LT		 1572 */
1573 if (unconf)
1574 expire_client(unconf);
a55370a3
N		 1575 new = create_client(clname, dname);
1576 if (new == NULL)
1da177e4 1577 goto out;
1da177e4 1578 gen_clid(new);
599e0a22 1579 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1580 /*
a186e767
BF		 1581 * RFC 3530 14.2.33 CASE 1:
1582 * probable callback update
1da177e4 1583 */
31f4a6c1
N		 1584 if (unconf) {
1585 /* Note this is removing unconfirmed {*x***},
1586 * which is stronger than RFC recommended {vxc**}.
1587 * This has the advantage that there is at most
1588 * one {*x***} in either list at any time.
1589 */
1590 expire_client(unconf);
1da177e4 1591 }
a55370a3
N		 1592 new = create_client(clname, dname);
1593 if (new == NULL)
1da177e4 1594 goto out;
1da177e4 1595 copy_clid(new, conf);
1da177e4
LT		 1596 } else if (!unconf) {
1597 /*
a186e767
BF		 1598 * RFC 3530 14.2.33 CASE 2:
1599 * probable client reboot; state will be removed if
1600 * confirmed.
1da177e4 1601 */
a55370a3
N		 1602 new = create_client(clname, dname);
1603 if (new == NULL)
1da177e4 1604 goto out;
1da177e4 1605 gen_clid(new);
49ba8781 1606 } else {
a186e767
BF		 1607 /*
1608 * RFC 3530 14.2.33 CASE 3:
1609 * probable client reboot; state will be removed if
1610 * confirmed.
1da177e4
LT		 1611 */
1612 expire_client(unconf);
a55370a3
N		 1613 new = create_client(clname, dname);
1614 if (new == NULL)
1da177e4 1615 goto out;
1da177e4 1616 gen_clid(new);
1da177e4 1617 }
c175b83c
BF		 1618 copy_verf(new, &clverifier);
1619 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1620 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1621 princ = svc_gss_principal(rqstp);
1622 if (princ) {
1623 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1624 if (new->cl_principal == NULL) {
1625 free_client(new);
1626 goto out;
1627 }
1628 }
c175b83c
BF		 1629 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1630 gen_confirm(new);
1631 gen_callback(new, setclid);
1632 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1633 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1634 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1635 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1636 status = nfs_ok;
1637out:
1638 nfs4_unlock_state();
1639 return status;
1640}
1641
1642
1643/*
a186e767
BF		 1644 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1645 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1646 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1647 */
b37ad28b 1648__be32
b591480b
BF		 1649nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1650 struct nfsd4_compound_state *cstate,
1651 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1652{
27459f09 1653 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1654 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1655 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1656 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1657 __be32 status;
1da177e4
LT		 1658
1659 if (STALE_CLIENTID(clid))
1660 return nfserr_stale_clientid;
1661 /*
1662 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1663 * We get here on a DRC miss.
1664 */
1665
1666 nfs4_lock_state();
21ab45a4
N		 1667
1668 conf = find_confirmed_client(clid);
1669 unconf = find_unconfirmed_client(clid);
1670
1671 status = nfserr_clid_inuse;
27459f09 1672 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1673 goto out;
27459f09 1674 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1675 goto out;
1676
a186e767
BF		 1677 /*
1678 * section 14.2.34 of RFC 3530 has a description of
1679 * SETCLIENTID_CONFIRM request processing consisting
1680 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1681 */
366e0c1d 1682 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1683 /*
1684 * RFC 3530 14.2.34 CASE 1:
1685 * callback update
1686 */
599e0a22 1687 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1688 status = nfserr_clid_inuse;
1689 else {
1a69c179
N		 1690 /* XXX: We just turn off callbacks until we can handle
1691 * change request correctly. */
c237dc03 1692 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1693 expire_client(unconf);
1da177e4 1694 status = nfs_ok;
1a69c179 1695
1da177e4 1696 }
f3aba4e5 1697 } else if (conf && !unconf) {
a186e767
BF		 1698 /*
1699 * RFC 3530 14.2.34 CASE 2:
1700 * probable retransmitted request; play it safe and
1701 * do nothing.
7c79f737 1702 */
599e0a22 1703 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1704 status = nfserr_clid_inuse;
21ab45a4 1705 else
1da177e4 1706 status = nfs_ok;
7c79f737 1707 } else if (!conf && unconf
599e0a22 1708 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1709 /*
1710 * RFC 3530 14.2.34 CASE 3:
1711 * Normal case; new or rebooted client:
7c79f737 1712 */
599e0a22 1713 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1714 status = nfserr_clid_inuse;
1715 } else {
1a69c179
N		 1716 unsigned int hash =
1717 clientstr_hashval(unconf->cl_recdir);
1718 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1719 hash, false);
1a69c179 1720 if (conf) {
c7b9a459 1721 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1722 expire_client(conf);
1723 }
1da177e4 1724 move_to_confirmed(unconf);
21ab45a4 1725 conf = unconf;
46f8a64b 1726 nfsd4_probe_callback(conf);
1a69c179 1727 status = nfs_ok;
1da177e4 1728 }
599e0a22
BF		 1729 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1730 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1731 &confirm)))) {
a186e767
BF		 1732 /*
1733 * RFC 3530 14.2.34 CASE 4:
1734 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1735 */
1da177e4 1736 status = nfserr_stale_clientid;
7c79f737 1737 } else {
08e8987c
N		 1738 /* check that we have hit one of the cases...*/
1739 status = nfserr_clid_inuse;
1740 }
1da177e4 1741out:
1da177e4
LT		 1742 nfs4_unlock_state();
1743 return status;
1744}
1745
1da177e4
LT		 1746/* OPEN Share state helper functions */
1747static inline struct nfs4_file *
1748alloc_init_file(struct inode *ino)
1749{
1750 struct nfs4_file *fp;
1751 unsigned int hashval = file_hashval(ino);
1752
e60d4398
N		 1753 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1754 if (fp) {
8b671b80 1755 atomic_set(&fp->fi_ref, 1);
1da177e4 1756 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1757 INIT_LIST_HEAD(&fp->fi_stateids);
1758 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1759 spin_lock(&recall_lock);
1da177e4 1760 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1761 spin_unlock(&recall_lock);
1da177e4
LT		 1762 fp->fi_inode = igrab(ino);
1763 fp->fi_id = current_fileid++;
47f9940c 1764 fp->fi_had_conflict = false;
1da177e4
LT		 1765 return fp;
1766 }
1767 return NULL;
1768}
1769
e60d4398 1770static void
e18b890b 1771nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1772{
e60d4398
N		 1773 if (*slab == NULL)
1774 return;
1a1d92c1 1775 kmem_cache_destroy(*slab);
e60d4398 1776 *slab = NULL;
1da177e4
LT		 1777}
1778
e8ff2a84 1779void
1da177e4
LT		 1780nfsd4_free_slabs(void)
1781{
e60d4398
N		 1782 nfsd4_free_slab(&stateowner_slab);
1783 nfsd4_free_slab(&file_slab);
5ac049ac 1784 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1785 nfsd4_free_slab(&deleg_slab);
e60d4398 1786}
1da177e4 1787
e60d4398
N		 1788static int
1789nfsd4_init_slabs(void)
1790{
1791 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1792 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1793 if (stateowner_slab == NULL)
1794 goto out_nomem;
1795 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1796 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1797 if (file_slab == NULL)
1798 goto out_nomem;
5ac049ac 1799 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1800 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1801 if (stateid_slab == NULL)
1802 goto out_nomem;
5b2d21c1 1803 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1804 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1805 if (deleg_slab == NULL)
1806 goto out_nomem;
e60d4398
N		 1807 return 0;
1808out_nomem:
1809 nfsd4_free_slabs();
1810 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1811 return -ENOMEM;
1da177e4
LT		 1812}
1813
1814void
1815nfs4_free_stateowner(struct kref *kref)
1816{
1817 struct nfs4_stateowner *sop =
1818 container_of(kref, struct nfs4_stateowner, so_ref);
1819 kfree(sop->so_owner.data);
1820 kmem_cache_free(stateowner_slab, sop);
1821}
1822
1823static inline struct nfs4_stateowner *
1824alloc_stateowner(struct xdr_netobj *owner)
1825{
1826 struct nfs4_stateowner *sop;
1827
1828 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1829 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1830 memcpy(sop->so_owner.data, owner->data, owner->len);
1831 sop->so_owner.len = owner->len;
1832 kref_init(&sop->so_ref);
1833 return sop;
1834 }
1835 kmem_cache_free(stateowner_slab, sop);
1836 }
1837 return NULL;
1838}
1839
1840static struct nfs4_stateowner *
1841alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1842 struct nfs4_stateowner *sop;
1843 struct nfs4_replay *rp;
1844 unsigned int idhashval;
1845
1846 if (!(sop = alloc_stateowner(&open->op_owner)))
1847 return NULL;
1848 idhashval = ownerid_hashval(current_ownerid);
1849 INIT_LIST_HEAD(&sop->so_idhash);
1850 INIT_LIST_HEAD(&sop->so_strhash);
1851 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1852 INIT_LIST_HEAD(&sop->so_stateids);
1853 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1854 INIT_LIST_HEAD(&sop->so_close_lru);
1855 sop->so_time = 0;
1856 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1857 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1858 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1859 sop->so_is_open_owner = 1;
1860 sop->so_id = current_ownerid++;
1861 sop->so_client = clp;
1862 sop->so_seqid = open->op_seqid;
1863 sop->so_confirmed = 0;
1864 rp = &sop->so_replay;
de1ae286 1865 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1866 rp->rp_buflen = 0;
1867 rp->rp_buf = rp->rp_ibuf;
1868 return sop;
1869}
1870
1da177e4
LT		 1871static inline void
1872init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1873 struct nfs4_stateowner *sop = open->op_stateowner;
1874 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1875
1876 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1877 INIT_LIST_HEAD(&stp->st_perstateowner);
1878 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1879 INIT_LIST_HEAD(&stp->st_perfile);
1880 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1881 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1882 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1883 stp->st_stateowner = sop;
13cd2184 1884 get_nfs4_file(fp);
1da177e4 1885 stp->st_file = fp;
78155ed7 1886 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1887 stp->st_stateid.si_stateownerid = sop->so_id;
1888 stp->st_stateid.si_fileid = fp->fi_id;
1889 stp->st_stateid.si_generation = 0;
1890 stp->st_access_bmap = 0;
1891 stp->st_deny_bmap = 0;
84459a11
AA		 1892 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1893 &stp->st_access_bmap);
1da177e4 1894 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1895 stp->st_openstp = NULL;
1da177e4
LT		 1896}
1897
fd39ca9a 1898static void
1da177e4
LT		 1899move_to_close_lru(struct nfs4_stateowner *sop)
1900{
1901 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1902
358dd55a 1903 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1904 sop->so_time = get_seconds();
1905}
1906
1da177e4 1907static int
599e0a22
BF		 1908same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1909 clientid_t *clid)
1910{
1911 return (sop->so_owner.len == owner->len) &&
1912 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1913 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1914}
1915
1916static struct nfs4_stateowner *
1917find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1918{
1919 struct nfs4_stateowner *so = NULL;
1920
1921 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1922 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1923 return so;
1924 }
1925 return NULL;
1926}
1927
1928/* search file_hashtbl[] for file */
1929static struct nfs4_file *
1930find_file(struct inode *ino)
1931{
1932 unsigned int hashval = file_hashval(ino);
1933 struct nfs4_file *fp;
1934
8b671b80 1935 spin_lock(&recall_lock);
1da177e4 1936 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1937 if (fp->fi_inode == ino) {
1938 get_nfs4_file(fp);
8b671b80 1939 spin_unlock(&recall_lock);
1da177e4 1940 return fp;
13cd2184 1941 }
1da177e4 1942 }
8b671b80 1943 spin_unlock(&recall_lock);
1da177e4
LT		 1944 return NULL;
1945}
1946
d87a8ade 1947static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1948{
d87a8ade 1949 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1950 return 0;
d87a8ade
AA		 1951 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1952 return 0;
1953 x &= ~NFS4_SHARE_ACCESS_MASK;
1954 if (minorversion && x) {
1955 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1956 return 0;
1957 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1958 return 0;
1959 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1960 }
1961 if (x)
8838dc43
BF		 1962 return 0;
1963 return 1;
ba5a6a19
BF		 1964}
1965
8838dc43 1966static inline int deny_valid(u32 x)
ba5a6a19 1967{
8838dc43
BF		 1968 /* Note: unlike access bits, deny bits may be zero. */
1969 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1970}
1da177e4 1971
4f83aa30
BF		 1972/*
1973 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1974 * st_{access,deny}_bmap field of the stateid, in order to track not
1975 * only what share bits are currently in force, but also what
1976 * combinations of share bits previous opens have used. This allows us
1977 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1978 * return an error if the client attempt to downgrade to a combination
1979 * of share bits not explicable by closing some of its previous opens.
1980 *
1981 * XXX: This enforcement is actually incomplete, since we don't keep
1982 * track of access/deny bit combinations; so, e.g., we allow:
1983 *
1984 * OPEN allow read, deny write
1985 * OPEN allow both, deny none
1986 * DOWNGRADE allow read, deny none
1987 *
1988 * which we should reject.
1989 */
fd39ca9a 1990static void
1da177e4
LT		 1991set_access(unsigned int *access, unsigned long bmap) {
1992 int i;
1993
1994 *access = 0;
1995 for (i = 1; i < 4; i++) {
1996 if (test_bit(i, &bmap))
1997 *access |= i;
1998 }
1999}
2000
fd39ca9a 2001static void
1da177e4
LT		 2002set_deny(unsigned int *deny, unsigned long bmap) {
2003 int i;
2004
2005 *deny = 0;
2006 for (i = 0; i < 4; i++) {
2007 if (test_bit(i, &bmap))
2008 *deny |= i ;
2009 }
2010}
2011
2012static int
2013test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2014 unsigned int access, deny;
2015
2016 set_access(&access, stp->st_access_bmap);
2017 set_deny(&deny, stp->st_deny_bmap);
2018 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2019 return 0;
2020 return 1;
2021}
2022
2023/*
2024 * Called to check deny when READ with all zero stateid or
2025 * WRITE with all zero or all one stateid
2026 */
b37ad28b 2027static __be32
1da177e4
LT		 2028nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2029{
2030 struct inode *ino = current_fh->fh_dentry->d_inode;
2031 struct nfs4_file *fp;
2032 struct nfs4_stateid *stp;
b37ad28b 2033 __be32 ret;
1da177e4
LT		 2034
2035 dprintk("NFSD: nfs4_share_conflict\n");
2036
2037 fp = find_file(ino);
13cd2184
N		 2038 if (!fp)
2039 return nfs_ok;
b700949b 2040 ret = nfserr_locked;
1da177e4 2041 /* Search for conflicting share reservations */
13cd2184
N		 2042 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2043 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2044 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2045 goto out;
1da177e4 2046 }
13cd2184
N		 2047 ret = nfs_ok;
2048out:
2049 put_nfs4_file(fp);
2050 return ret;
1da177e4
LT		 2051}
2052
2053static inline void
2054nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2055{
2056 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2057 drop_file_write_access(filp);
1da177e4
LT		 2058 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2059 }
2060}
2061
1da177e4
LT		 2062/*
2063 * Spawn a thread to perform a recall on the delegation represented
2064 * by the lease (file_lock)
2065 *
2066 * Called from break_lease() with lock_kernel() held.
2067 * Note: we assume break_lease will only call this *once* for any given
2068 * lease.
2069 */
2070static
2071void nfsd_break_deleg_cb(struct file_lock *fl)
2072{
63e4863f 2073 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2074
2075 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2076 if (!dp)
2077 return;
2078
2079 /* We're assuming the state code never drops its reference
2080 * without first removing the lease. Since we're in this lease
2081 * callback (and since the lease code is serialized by the kernel
2082 * lock) we know the server hasn't removed the lease yet, we know
2083 * it's safe to take a reference: */
2084 atomic_inc(&dp->dl_count);
cfdcad4d 2085 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2086
2087 spin_lock(&recall_lock);
2088 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2089 spin_unlock(&recall_lock);
2090
2091 /* only place dl_time is set. protected by lock_kernel*/
2092 dp->dl_time = get_seconds();
2093
0272e1fd
BF		 2094 /*
2095 * We don't want the locks code to timeout the lease for us;
2096 * we'll remove it ourself if the delegation isn't returned
2097 * in time.
2098 */
2099 fl->fl_break_time = 0;
1da177e4 2100
63e4863f
BF		 2101 dp->dl_file->fi_had_conflict = true;
2102 nfsd4_cb_recall(dp);
1da177e4
LT		 2103}
2104
2105/*
2106 * The file_lock is being reapd.
2107 *
2108 * Called by locks_free_lock() with lock_kernel() held.
2109 */
2110static
2111void nfsd_release_deleg_cb(struct file_lock *fl)
2112{
2113 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2114
2115 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2116
2117 if (!(fl->fl_flags & FL_LEASE) || !dp)
2118 return;
2119 dp->dl_flock = NULL;
2120}
2121
2122/*
2123 * Set the delegation file_lock back pointer.
2124 *
a9933cea 2125 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2126 */
2127static
2128void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2129{
2130 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2131
2132 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2133 if (!dp)
2134 return;
2135 dp->dl_flock = new;
2136}
2137
2138/*
a9933cea 2139 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2140 */
2141static
2142int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2143{
2144 struct nfs4_delegation *onlistd =
2145 (struct nfs4_delegation *)onlist->fl_owner;
2146 struct nfs4_delegation *tryd =
2147 (struct nfs4_delegation *)try->fl_owner;
2148
2149 if (onlist->fl_lmops != try->fl_lmops)
2150 return 0;
2151
2152 return onlistd->dl_client == tryd->dl_client;
2153}
2154
2155
2156static
2157int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2158{
2159 if (arg & F_UNLCK)
2160 return lease_modify(onlist, arg);
2161 else
2162 return -EAGAIN;
2163}
2164
fd39ca9a 2165static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2166 .fl_break = nfsd_break_deleg_cb,
2167 .fl_release_private = nfsd_release_deleg_cb,
2168 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2169 .fl_mylease = nfsd_same_client_deleg_cb,
2170 .fl_change = nfsd_change_deleg_cb,
2171};
2172
2173
b37ad28b 2174__be32
6668958f
AA		 2175nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2176 struct nfsd4_open *open)
1da177e4 2177{
1da177e4
LT		 2178 clientid_t *clientid = &open->op_clientid;
2179 struct nfs4_client *clp = NULL;
2180 unsigned int strhashval;
2181 struct nfs4_stateowner *sop = NULL;
2182
1da177e4 2183 if (!check_name(open->op_owner))
0f442aa2 2184 return nfserr_inval;
1da177e4
LT		 2185
2186 if (STALE_CLIENTID(&open->op_clientid))
2187 return nfserr_stale_clientid;
2188
2189 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2190 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2191 open->op_stateowner = sop;
2192 if (!sop) {
2193 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2194 clp = find_confirmed_client(clientid);
2195 if (clp == NULL)
0f442aa2
BF		 2196 return nfserr_expired;
2197 goto renew;
1da177e4 2198 }
6668958f
AA		 2199 /* When sessions are used, skip open sequenceid processing */
2200 if (nfsd4_has_session(cstate))
2201 goto renew;
0f442aa2
BF		 2202 if (!sop->so_confirmed) {
2203 /* Replace unconfirmed owners without checking for replay. */
2204 clp = sop->so_client;
f044ff83 2205 release_openowner(sop);
0f442aa2
BF		 2206 open->op_stateowner = NULL;
2207 goto renew;
2208 }
2209 if (open->op_seqid == sop->so_seqid - 1) {
2210 if (sop->so_replay.rp_buflen)
a90b061c 2211 return nfserr_replay_me;
0f442aa2
BF		 2212 /* The original OPEN failed so spectacularly
2213 * that we don't even have replay data saved!
2214 * Therefore, we have no choice but to continue
2215 * processing this OPEN; presumably, we'll
2216 * fail again for the same reason.
2217 */
2218 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2219 goto renew;
2220 }
2221 if (open->op_seqid != sop->so_seqid)
2222 return nfserr_bad_seqid;
1da177e4 2223renew:
0f442aa2
BF		 2224 if (open->op_stateowner == NULL) {
2225 sop = alloc_init_open_stateowner(strhashval, clp, open);
2226 if (sop == NULL)
2227 return nfserr_resource;
2228 open->op_stateowner = sop;
2229 }
2230 list_del_init(&sop->so_close_lru);
1da177e4 2231 renew_client(sop->so_client);
0f442aa2 2232 return nfs_ok;
1da177e4
LT		 2233}
2234
b37ad28b 2235static inline __be32
4a6e43e6
N		 2236nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2237{
2238 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2239 return nfserr_openmode;
2240 else
2241 return nfs_ok;
2242}
2243
52f4fb43
N		 2244static struct nfs4_delegation *
2245find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2246{
2247 struct nfs4_delegation *dp;
2248
ea1da636 2249 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2250 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2251 return dp;
2252 }
2253 return NULL;
2254}
2255
b37ad28b 2256static __be32
567d9829
N		 2257nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2258 struct nfs4_delegation **dp)
2259{
2260 int flags;
b37ad28b 2261 __be32 status = nfserr_bad_stateid;
567d9829
N		 2262
2263 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2264 if (*dp == NULL)
c44c5eeb 2265 goto out;
567d9829
N		 2266 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2267 RD_STATE : WR_STATE;
2268 status = nfs4_check_delegmode(*dp, flags);
2269 if (status)
2270 *dp = NULL;
c44c5eeb
N		 2271out:
2272 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2273 return nfs_ok;
2274 if (status)
2275 return status;
2276 open->op_stateowner->so_confirmed = 1;
2277 return nfs_ok;
567d9829
N		 2278}
2279
b37ad28b 2280static __be32
1da177e4
LT		 2281nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2282{
2283 struct nfs4_stateid *local;
b37ad28b 2284 __be32 status = nfserr_share_denied;
1da177e4
LT		 2285 struct nfs4_stateowner *sop = open->op_stateowner;
2286
8beefa24 2287 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2288 /* ignore lock owners */
2289 if (local->st_stateowner->so_is_open_owner == 0)
2290 continue;
2291 /* remember if we have seen this open owner */
2292 if (local->st_stateowner == sop)
2293 *stpp = local;
2294 /* check for conflicting share reservations */
2295 if (!test_share(local, open))
2296 goto out;
2297 }
2298 status = 0;
2299out:
2300 return status;
2301}
2302
5ac049ac
N		 2303static inline struct nfs4_stateid *
2304nfs4_alloc_stateid(void)
2305{
2306 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2307}
2308
b37ad28b 2309static __be32
1da177e4 2310nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2311 struct nfs4_delegation *dp,
1da177e4
LT		 2312 struct svc_fh *cur_fh, int flags)
2313{
2314 struct nfs4_stateid *stp;
1da177e4 2315
5ac049ac 2316 stp = nfs4_alloc_stateid();
1da177e4
LT		 2317 if (stp == NULL)
2318 return nfserr_resource;
2319
567d9829
N		 2320 if (dp) {
2321 get_file(dp->dl_vfs_file);
2322 stp->st_vfs_file = dp->dl_vfs_file;
2323 } else {
b37ad28b 2324 __be32 status;
567d9829
N		 2325 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2326 &stp->st_vfs_file);
2327 if (status) {
2328 if (status == nfserr_dropit)
2329 status = nfserr_jukebox;
5ac049ac 2330 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2331 return status;
2332 }
1da177e4 2333 }
1da177e4
LT		 2334 *stpp = stp;
2335 return 0;
2336}
2337
b37ad28b 2338static inline __be32
1da177e4
LT		 2339nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2340 struct nfsd4_open *open)
2341{
2342 struct iattr iattr = {
2343 .ia_valid = ATTR_SIZE,
2344 .ia_size = 0,
2345 };
2346 if (!open->op_truncate)
2347 return 0;
2348 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2349 return nfserr_inval;
1da177e4
LT		 2350 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2351}
2352
b37ad28b 2353static __be32
1da177e4
LT		 2354nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2355{
2356 struct file *filp = stp->st_vfs_file;
7eaa36e2 2357 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2358 unsigned int share_access, new_writer;
b37ad28b 2359 __be32 status;
1da177e4
LT		 2360
2361 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2362 new_writer = (~share_access) & open->op_share_access
2363 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2364
6c26d08f 2365 if (new_writer) {
b37ad28b
AV		 2366 int err = get_write_access(inode);
2367 if (err)
2368 return nfserrno(err);
e518f056
BH		 2369 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2370 if (err)
2371 return nfserrno(err);
2372 file_take_write(filp);
6c26d08f 2373 }
1da177e4
LT		 2374 status = nfsd4_truncate(rqstp, cur_fh, open);
2375 if (status) {
6c26d08f
BF		 2376 if (new_writer)
2377 put_write_access(inode);
1da177e4
LT		 2378 return status;
2379 }
2380 /* remember the open */
6c26d08f 2381 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2382 __set_bit(open->op_share_access, &stp->st_access_bmap);
2383 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2384
2385 return nfs_ok;
2386}
2387
2388
1da177e4 2389static void
37515177 2390nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2391{
37515177
N		 2392 open->op_stateowner->so_confirmed = 1;
2393 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2394}
2395
2396/*
2397 * Attempt to hand out a delegation.
2398 */
2399static void
2400nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2401{
2402 struct nfs4_delegation *dp;
2403 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2404 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2405 struct file_lock fl, *flp = &fl;
2406 int status, flag = 0;
2407
2408 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2409 open->op_recall = 0;
2410 switch (open->op_claim_type) {
2411 case NFS4_OPEN_CLAIM_PREVIOUS:
2412 if (!atomic_read(&cb->cb_set))
2413 open->op_recall = 1;
2414 flag = open->op_delegate_type;
2415 if (flag == NFS4_OPEN_DELEGATE_NONE)
2416 goto out;
2417 break;
2418 case NFS4_OPEN_CLAIM_NULL:
2419 /* Let's not give out any delegations till everyone's
2420 * had the chance to reclaim theirs.... */
af558e33 2421 if (locks_in_grace())
7b190fec
N		 2422 goto out;
2423 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2424 goto out;
2425 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2426 flag = NFS4_OPEN_DELEGATE_WRITE;
2427 else
2428 flag = NFS4_OPEN_DELEGATE_READ;
2429 break;
2430 default:
2431 goto out;
2432 }
1da177e4
LT		 2433
2434 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2435 if (dp == NULL) {
2436 flag = NFS4_OPEN_DELEGATE_NONE;
2437 goto out;
2438 }
2439 locks_init_lock(&fl);
2440 fl.fl_lmops = &nfsd_lease_mng_ops;
2441 fl.fl_flags = FL_LEASE;
9167f501 2442 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2443 fl.fl_end = OFFSET_MAX;
2444 fl.fl_owner = (fl_owner_t)dp;
2445 fl.fl_file = stp->st_vfs_file;
2446 fl.fl_pid = current->tgid;
2447
a9933cea 2448 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2449 * the lock_manager callbacks fl_mylease and fl_change are used
2450 */
9167f501 2451 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2452 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2453 unhash_delegation(dp);
1da177e4
LT		 2454 flag = NFS4_OPEN_DELEGATE_NONE;
2455 goto out;
2456 }
2457
2458 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2459
2460 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2461 dp->dl_stateid.si_boot,
2462 dp->dl_stateid.si_stateownerid,
2463 dp->dl_stateid.si_fileid,
2464 dp->dl_stateid.si_generation);
2465out:
7b190fec
N		 2466 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2467 && flag == NFS4_OPEN_DELEGATE_NONE
2468 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2469 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2470 open->op_delegate_type = flag;
2471}
2472
2473/*
2474 * called with nfs4_lock_state() held.
2475 */
b37ad28b 2476__be32
1da177e4
LT		 2477nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2478{
6668958f 2479 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2480 struct nfs4_file *fp = NULL;
2481 struct inode *ino = current_fh->fh_dentry->d_inode;
2482 struct nfs4_stateid *stp = NULL;
567d9829 2483 struct nfs4_delegation *dp = NULL;
b37ad28b 2484 __be32 status;
1da177e4
LT		 2485
2486 status = nfserr_inval;
d87a8ade 2487 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2488 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2489 goto out;
2490 /*
2491 * Lookup file; if found, lookup stateid and check open request,
2492 * and check for delegations in the process of being recalled.
2493 * If not found, create the nfs4_file struct
2494 */
2495 fp = find_file(ino);
2496 if (fp) {
2497 if ((status = nfs4_check_open(fp, open, &stp)))
2498 goto out;
c44c5eeb
N		 2499 status = nfs4_check_deleg(fp, open, &dp);
2500 if (status)
2501 goto out;
1da177e4 2502 } else {
c44c5eeb
N		 2503 status = nfserr_bad_stateid;
2504 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2505 goto out;
1da177e4
LT		 2506 status = nfserr_resource;
2507 fp = alloc_init_file(ino);
2508 if (fp == NULL)
2509 goto out;
2510 }
2511
2512 /*
2513 * OPEN the file, or upgrade an existing OPEN.
2514 * If truncate fails, the OPEN fails.
2515 */
2516 if (stp) {
2517 /* Stateid was found, this is an OPEN upgrade */
2518 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2519 if (status)
2520 goto out;
444c2c07 2521 update_stateid(&stp->st_stateid);
1da177e4
LT		 2522 } else {
2523 /* Stateid was not found, this is a new OPEN */
2524 int flags = 0;
9ecb6a08 2525 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2526 flags |= NFSD_MAY_READ;
1da177e4 2527 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2528 flags |= NFSD_MAY_WRITE;
567d9829
N		 2529 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2530 if (status)
1da177e4
LT		 2531 goto out;
2532 init_stateid(stp, fp, open);
2533 status = nfsd4_truncate(rqstp, current_fh, open);
2534 if (status) {
2283963f 2535 release_open_stateid(stp);
1da177e4
LT		 2536 goto out;
2537 }
6668958f
AA		 2538 if (nfsd4_has_session(&resp->cstate))
2539 update_stateid(&stp->st_stateid);
1da177e4
LT		 2540 }
2541 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2542
6668958f
AA		 2543 if (nfsd4_has_session(&resp->cstate))
2544 open->op_stateowner->so_confirmed = 1;
2545
1da177e4
LT		 2546 /*
2547 * Attempt to hand out a delegation. No error return, because the
2548 * OPEN succeeds even if we fail.
2549 */
2550 nfs4_open_delegation(current_fh, open, stp);
2551
2552 status = nfs_ok;
2553
2554 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2555 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2556 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2557out:
13cd2184
N		 2558 if (fp)
2559 put_nfs4_file(fp);
37515177
N		 2560 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2561 nfs4_set_claim_prev(open);
1da177e4
LT		 2562 /*
2563 * To finish the open response, we just need to set the rflags.
2564 */
2565 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2566 if (!open->op_stateowner->so_confirmed &&
2567 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2568 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2569
2570 return status;
2571}
2572
b37ad28b 2573__be32
b591480b
BF		 2574nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2575 clientid_t *clid)
1da177e4
LT		 2576{
2577 struct nfs4_client *clp;
b37ad28b 2578 __be32 status;
1da177e4
LT		 2579
2580 nfs4_lock_state();
2581 dprintk("process_renew(%08x/%08x): starting\n",
2582 clid->cl_boot, clid->cl_id);
2583 status = nfserr_stale_clientid;
2584 if (STALE_CLIENTID(clid))
2585 goto out;
2586 clp = find_confirmed_client(clid);
2587 status = nfserr_expired;
2588 if (clp == NULL) {
2589 /* We assume the client took too long to RENEW. */
2590 dprintk("nfsd4_renew: clientid not found!\n");
2591 goto out;
2592 }
2593 renew_client(clp);
2594 status = nfserr_cb_path_down;
ea1da636 2595 if (!list_empty(&clp->cl_delegations)
c237dc03 2596 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2597 goto out;
2598 status = nfs_ok;
2599out:
2600 nfs4_unlock_state();
2601 return status;
2602}
2603
af558e33
BF		 2604struct lock_manager nfsd4_manager = {
2605};
2606
a76b4319 2607static void
af558e33 2608nfsd4_end_grace(void)
a76b4319
N		 2609{
2610 dprintk("NFSD: end of grace period\n");
c7b9a459 2611 nfsd4_recdir_purge_old();
af558e33 2612 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2613}
2614
fd39ca9a 2615static time_t
1da177e4
LT		 2616nfs4_laundromat(void)
2617{
2618 struct nfs4_client *clp;
2619 struct nfs4_stateowner *sop;
2620 struct nfs4_delegation *dp;
2621 struct list_head *pos, *next, reaplist;
2622 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2623 time_t t, clientid_val = NFSD_LEASE_TIME;
2624 time_t u, test_val = NFSD_LEASE_TIME;
2625
2626 nfs4_lock_state();
2627
2628 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2629 if (locks_in_grace())
2630 nfsd4_end_grace();
1da177e4
LT		 2631 list_for_each_safe(pos, next, &client_lru) {
2632 clp = list_entry(pos, struct nfs4_client, cl_lru);
2633 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2634 t = clp->cl_time - cutoff;
2635 if (clientid_val > t)
2636 clientid_val = t;
2637 break;
2638 }
2639 dprintk("NFSD: purging unused client (clientid %08x)\n",
2640 clp->cl_clientid.cl_id);
c7b9a459 2641 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2642 expire_client(clp);
2643 }
2644 INIT_LIST_HEAD(&reaplist);
2645 spin_lock(&recall_lock);
2646 list_for_each_safe(pos, next, &del_recall_lru) {
2647 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2648 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2649 u = dp->dl_time - cutoff;
2650 if (test_val > u)
2651 test_val = u;
2652 break;
2653 }
2654 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2655 dp, dp->dl_flock);
2656 list_move(&dp->dl_recall_lru, &reaplist);
2657 }
2658 spin_unlock(&recall_lock);
2659 list_for_each_safe(pos, next, &reaplist) {
2660 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2661 list_del_init(&dp->dl_recall_lru);
2662 unhash_delegation(dp);
2663 }
2664 test_val = NFSD_LEASE_TIME;
2665 list_for_each_safe(pos, next, &close_lru) {
2666 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2667 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2668 u = sop->so_time - cutoff;
2669 if (test_val > u)
2670 test_val = u;
2671 break;
2672 }
2673 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2674 sop->so_id);
f044ff83 2675 release_openowner(sop);
1da177e4
LT		 2676 }
2677 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2678 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2679 nfs4_unlock_state();
2680 return clientid_val;
2681}
2682
a254b246
HH		 2683static struct workqueue_struct *laundry_wq;
2684static void laundromat_main(struct work_struct *);
2685static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2686
2687static void
c4028958 2688laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2689{
2690 time_t t;
2691
2692 t = nfs4_laundromat();
2693 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2694 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2695}
2696
fd39ca9a 2697static struct nfs4_stateowner *
f8816512
N		 2698search_close_lru(u32 st_id, int flags)
2699{
1da177e4
LT		 2700 struct nfs4_stateowner *local = NULL;
2701
1da177e4
LT		 2702 if (flags & CLOSE_STATE) {
2703 list_for_each_entry(local, &close_lru, so_close_lru) {
2704 if (local->so_id == st_id)
2705 return local;
2706 }
2707 }
2708 return NULL;
2709}
2710
2711static inline int
2712nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2713{
7eaa36e2 2714 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2715}
2716
2717static int
2718STALE_STATEID(stateid_t *stateid)
2719{
78155ed7
BN		 2720 if (time_after((unsigned long)boot_time,
2721 (unsigned long)stateid->si_boot)) {
2722 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2723 stateid->si_boot, stateid->si_stateownerid,
2724 stateid->si_fileid, stateid->si_generation);
2725 return 1;
2726 }
2727 return 0;
2728}
2729
2730static int
2731EXPIRED_STATEID(stateid_t *stateid)
2732{
2733 if (time_before((unsigned long)boot_time,
2734 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2735 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2736 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2737 stateid->si_boot, stateid->si_stateownerid,
2738 stateid->si_fileid, stateid->si_generation);
2739 return 1;
2740 }
2741 return 0;
2742}
2743
2744static __be32
2745stateid_error_map(stateid_t *stateid)
2746{
2747 if (STALE_STATEID(stateid))
2748 return nfserr_stale_stateid;
2749 if (EXPIRED_STATEID(stateid))
2750 return nfserr_expired;
2751
2752 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2753 stateid->si_boot, stateid->si_stateownerid,
2754 stateid->si_fileid, stateid->si_generation);
2755 return nfserr_bad_stateid;
1da177e4
LT		 2756}
2757
2758static inline int
2759access_permit_read(unsigned long access_bmap)
2760{
2761 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2762 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2763 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2764}
2765
2766static inline int
2767access_permit_write(unsigned long access_bmap)
2768{
2769 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2770 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2771}
2772
2773static
b37ad28b 2774__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2775{
b37ad28b 2776 __be32 status = nfserr_openmode;
1da177e4
LT		 2777
2778 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2779 goto out;
2780 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2781 goto out;
2782 status = nfs_ok;
2783out:
2784 return status;
2785}
2786
b37ad28b 2787static inline __be32
1da177e4
LT		 2788check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2789{
203a8c8e 2790 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2791 return nfs_ok;
af558e33 2792 else if (locks_in_grace()) {
1da177e4
LT		 2793 /* Answer in remaining cases depends on existance of
2794 * conflicting state; so we must wait out the grace period. */
2795 return nfserr_grace;
2796 } else if (flags & WR_STATE)
2797 return nfs4_share_conflict(current_fh,
2798 NFS4_SHARE_DENY_WRITE);
2799 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2800 return nfs4_share_conflict(current_fh,
2801 NFS4_SHARE_DENY_READ);
2802}
2803
2804/*
2805 * Allow READ/WRITE during grace period on recovered state only for files
2806 * that are not able to provide mandatory locking.
2807 */
2808static inline int
18f82731 2809grace_disallows_io(struct inode *inode)
1da177e4 2810{
203a8c8e 2811 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2812}
2813
6668958f 2814static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2815{
6668958f
AA		 2816 /*
2817 * When sessions are used the stateid generation number is ignored
2818 * when it is zero.
2819 */
2820 if ((flags & HAS_SESSION) && in->si_generation == 0)
2821 goto out;
2822
0836f587
BF		 2823 /* If the client sends us a stateid from the future, it's buggy: */
2824 if (in->si_generation > ref->si_generation)
2825 return nfserr_bad_stateid;
2826 /*
2827 * The following, however, can happen. For example, if the
2828 * client sends an open and some IO at the same time, the open
2829 * may bump si_generation while the IO is still in flight.
2830 * Thanks to hard links and renames, the client never knows what
2831 * file an open will affect. So it could avoid that situation
2832 * only by serializing all opens and IO from the same open
2833 * owner. To recover from the old_stateid error, the client
2834 * will just have to retry the IO:
2835 */
2836 if (in->si_generation < ref->si_generation)
2837 return nfserr_old_stateid;
6668958f 2838out:
0836f587
BF		 2839 return nfs_ok;
2840}
2841
3e633079
BF		 2842static int is_delegation_stateid(stateid_t *stateid)
2843{
2844 return stateid->si_fileid == 0;
2845}
2846
1da177e4
LT		 2847/*
2848* Checks for stateid operations
2849*/
b37ad28b 2850__be32
dd453dfd
BH		 2851nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2852 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2853{
2854 struct nfs4_stateid *stp = NULL;
2855 struct nfs4_delegation *dp = NULL;
dd453dfd 2856 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2857 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2858 __be32 status;
1da177e4 2859
1da177e4
LT		 2860 if (filpp)
2861 *filpp = NULL;
2862
18f82731 2863 if (grace_disallows_io(ino))
1da177e4
LT		 2864 return nfserr_grace;
2865
6668958f
AA		 2866 if (nfsd4_has_session(cstate))
2867 flags |= HAS_SESSION;
2868
1da177e4
LT		 2869 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2870 return check_special_stateids(current_fh, stateid, flags);
2871
1da177e4
LT		 2872 status = nfserr_stale_stateid;
2873 if (STALE_STATEID(stateid))
2874 goto out;
2875
1da177e4 2876 status = nfserr_bad_stateid;
3e633079 2877 if (is_delegation_stateid(stateid)) {
a4455be0 2878 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2879 if (!dp) {
2880 status = stateid_error_map(stateid);
1da177e4 2881 goto out;
78155ed7 2882 }
6668958f
AA		 2883 status = check_stateid_generation(stateid, &dp->dl_stateid,
2884 flags);
0c2a498f
BF		 2885 if (status)
2886 goto out;
dc9bf700
BF		 2887 status = nfs4_check_delegmode(dp, flags);
2888 if (status)
2889 goto out;
2890 renew_client(dp->dl_client);
dc9bf700
BF		 2891 if (filpp)
2892 *filpp = dp->dl_vfs_file;
1da177e4 2893 } else { /* open or lock stateid */
a4455be0 2894 stp = find_stateid(stateid, flags);
78155ed7
BN		 2895 if (!stp) {
2896 status = stateid_error_map(stateid);
1da177e4 2897 goto out;
78155ed7 2898 }
6150ef0d 2899 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2900 goto out;
2901 if (!stp->st_stateowner->so_confirmed)
2902 goto out;
6668958f
AA		 2903 status = check_stateid_generation(stateid, &stp->st_stateid,
2904 flags);
0c2a498f
BF		 2905 if (status)
2906 goto out;
a4455be0
BF		 2907 status = nfs4_check_openmode(stp, flags);
2908 if (status)
1da177e4
LT		 2909 goto out;
2910 renew_client(stp->st_stateowner->so_client);
2911 if (filpp)
2912 *filpp = stp->st_vfs_file;
1da177e4
LT		 2913 }
2914 status = nfs_ok;
2915out:
2916 return status;
2917}
2918
4c4cd222
N		 2919static inline int
2920setlkflg (int type)
2921{
2922 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2923 RD_STATE : WR_STATE;
2924}
1da177e4
LT		 2925
2926/*
2927 * Checks for sequence id mutating operations.
2928 */
b37ad28b 2929static __be32
dd453dfd
BH		 2930nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2931 stateid_t *stateid, int flags,
2932 struct nfs4_stateowner **sopp,
2933 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2934{
1da177e4
LT		 2935 struct nfs4_stateid *stp;
2936 struct nfs4_stateowner *sop;
dd453dfd 2937 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2938 __be32 status;
1da177e4
LT		 2939
2940 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2941 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2942 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2943 stateid->si_generation);
3a4f98bb 2944
1da177e4
LT		 2945 *stpp = NULL;
2946 *sopp = NULL;
2947
1da177e4 2948 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2949 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2950 return nfserr_bad_stateid;
1da177e4
LT		 2951 }
2952
1da177e4 2953 if (STALE_STATEID(stateid))
3a4f98bb 2954 return nfserr_stale_stateid;
6668958f
AA		 2955
2956 if (nfsd4_has_session(cstate))
2957 flags |= HAS_SESSION;
2958
1da177e4
LT		 2959 /*
2960 * We return BAD_STATEID if filehandle doesn't match stateid,
2961 * the confirmed flag is incorrecly set, or the generation
2962 * number is incorrect.
1da177e4 2963 */
f8816512
N		 2964 stp = find_stateid(stateid, flags);
2965 if (stp == NULL) {
2966 /*
2967 * Also, we should make sure this isn't just the result of
2968 * a replayed close:
2969 */
2970 sop = search_close_lru(stateid->si_stateownerid, flags);
2971 if (sop == NULL)
78155ed7 2972 return stateid_error_map(stateid);
f8816512
N		 2973 *sopp = sop;
2974 goto check_replay;
2975 }
1da177e4 2976
39325bd0
BF		 2977 *stpp = stp;
2978 *sopp = sop = stp->st_stateowner;
2979
4c4cd222 2980 if (lock) {
4c4cd222 2981 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2982 struct nfs4_client *clp = sop->so_client;
4c4cd222 2983 int lkflg = 0;
b37ad28b 2984 __be32 status;
4c4cd222
N		 2985
2986 lkflg = setlkflg(lock->lk_type);
2987
2988 if (lock->lk_is_new) {
599e0a22
BF		 2989 if (!sop->so_is_open_owner)
2990 return nfserr_bad_stateid;
60adfc50
AA		 2991 if (!(flags & HAS_SESSION) &&
2992 !same_clid(&clp->cl_clientid, lockclid))
2993 return nfserr_bad_stateid;
599e0a22
BF		 2994 /* stp is the open stateid */
2995 status = nfs4_check_openmode(stp, lkflg);
2996 if (status)
2997 return status;
2998 } else {
2999 /* stp is the lock stateid */
3000 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3001 if (status)
3002 return status;
4c4cd222 3003 }
1da177e4
LT		 3004 }
3005
f3362737 3006 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3007 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3008 return nfserr_bad_stateid;
1da177e4
LT		 3009 }
3010
1da177e4
LT		 3011 /*
3012 * We now validate the seqid and stateid generation numbers.
3013 * For the moment, we ignore the possibility of
3014 * generation number wraparound.
3015 */
6668958f 3016 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3017 goto check_replay;
3018
3a4f98bb 3019 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3020 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3021 " unconfirmed stateowner!\n");
3022 return nfserr_bad_stateid;
1da177e4 3023 }
3a4f98bb 3024 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3025 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3026 " confirmed yet!\n");
3027 return nfserr_bad_stateid;
1da177e4 3028 }
6668958f 3029 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3030 if (status)
3031 return status;
52fd004e 3032 renew_client(sop->so_client);
3a4f98bb 3033 return nfs_ok;
1da177e4 3034
1da177e4 3035check_replay:
bd9aac52 3036 if (seqid == sop->so_seqid - 1) {
849823c5 3037 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3038 /* indicate replay to calling function */
a90b061c 3039 return nfserr_replay_me;
1da177e4 3040 }
2fdada03 3041 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3042 sop->so_seqid, seqid);
3043 *sopp = NULL;
3044 return nfserr_bad_seqid;
1da177e4
LT		 3045}
3046
b37ad28b 3047__be32
ca364317 3048nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3049 struct nfsd4_open_confirm *oc)
1da177e4 3050{
b37ad28b 3051 __be32 status;
1da177e4
LT		 3052 struct nfs4_stateowner *sop;
3053 struct nfs4_stateid *stp;
3054
3055 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3056 (int)cstate->current_fh.fh_dentry->d_name.len,
3057 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3058
ca364317 3059 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3060 if (status)
3061 return status;
1da177e4
LT		 3062
3063 nfs4_lock_state();
3064
dd453dfd 3065 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3066 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3067 CONFIRM | OPEN_STATE,
1da177e4
LT		 3068 &oc->oc_stateowner, &stp, NULL)))
3069 goto out;
3070
3071 sop = oc->oc_stateowner;
3072 sop->so_confirmed = 1;
3073 update_stateid(&stp->st_stateid);
3074 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3075 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3076 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3077 stp->st_stateid.si_boot,
3078 stp->st_stateid.si_stateownerid,
3079 stp->st_stateid.si_fileid,
3080 stp->st_stateid.si_generation);
c7b9a459
N		 3081
3082 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3083out:
f2327d9a 3084 if (oc->oc_stateowner) {
1da177e4 3085 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3086 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3087 }
1da177e4
LT		 3088 nfs4_unlock_state();
3089 return status;
3090}
3091
3092
3093/*
3094 * unset all bits in union bitmap (bmap) that
3095 * do not exist in share (from successful OPEN_DOWNGRADE)
3096 */
3097static void
3098reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3099{
3100 int i;
3101 for (i = 1; i < 4; i++) {
3102 if ((i & access) != i)
3103 __clear_bit(i, bmap);
3104 }
3105}
3106
3107static void
3108reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3109{
3110 int i;
3111 for (i = 0; i < 4; i++) {
3112 if ((i & deny) != i)
3113 __clear_bit(i, bmap);
3114 }
3115}
3116
b37ad28b 3117__be32
ca364317
BF		 3118nfsd4_open_downgrade(struct svc_rqst *rqstp,
3119 struct nfsd4_compound_state *cstate,
a4f1706a 3120 struct nfsd4_open_downgrade *od)
1da177e4 3121{
b37ad28b 3122 __be32 status;
1da177e4
LT		 3123 struct nfs4_stateid *stp;
3124 unsigned int share_access;
3125
3126 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3127 (int)cstate->current_fh.fh_dentry->d_name.len,
3128 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3129
d87a8ade 3130 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3131 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3132 return nfserr_inval;
3133
3134 nfs4_lock_state();
dd453dfd 3135 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3136 od->od_seqid,
1da177e4 3137 &od->od_stateid,
f3362737 3138 OPEN_STATE,
1da177e4
LT		 3139 &od->od_stateowner, &stp, NULL)))
3140 goto out;
3141
3142 status = nfserr_inval;
3143 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3144 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3145 stp->st_access_bmap, od->od_share_access);
3146 goto out;
3147 }
3148 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3149 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3150 stp->st_deny_bmap, od->od_share_deny);
3151 goto out;
3152 }
3153 set_access(&share_access, stp->st_access_bmap);
3154 nfs4_file_downgrade(stp->st_vfs_file,
3155 share_access & ~od->od_share_access);
3156
3157 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3158 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3159
3160 update_stateid(&stp->st_stateid);
3161 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3162 status = nfs_ok;
3163out:
f2327d9a 3164 if (od->od_stateowner) {
1da177e4 3165 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3166 cstate->replay_owner = od->od_stateowner;
f2327d9a 3167 }
1da177e4
LT		 3168 nfs4_unlock_state();
3169 return status;
3170}
3171
3172/*
3173 * nfs4_unlock_state() called after encode
3174 */
b37ad28b 3175__be32
ca364317 3176nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3177 struct nfsd4_close *close)
1da177e4 3178{
b37ad28b 3179 __be32 status;
1da177e4
LT		 3180 struct nfs4_stateid *stp;
3181
3182 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3183 (int)cstate->current_fh.fh_dentry->d_name.len,
3184 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3185
3186 nfs4_lock_state();
3187 /* check close_lru for replay */
dd453dfd 3188 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3189 close->cl_seqid,
1da177e4 3190 &close->cl_stateid,
f3362737 3191 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3192 &close->cl_stateowner, &stp, NULL)))
3193 goto out;
1da177e4
LT		 3194 status = nfs_ok;
3195 update_stateid(&stp->st_stateid);
3196 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3197
04ef5954 3198 /* release_stateid() calls nfsd_close() if needed */
2283963f 3199 release_open_stateid(stp);
04ef5954
BF		 3200
3201 /* place unused nfs4_stateowners on so_close_lru list to be
3202 * released by the laundromat service after the lease period
3203 * to enable us to handle CLOSE replay
3204 */
3205 if (list_empty(&close->cl_stateowner->so_stateids))
3206 move_to_close_lru(close->cl_stateowner);
1da177e4 3207out:
f2327d9a 3208 if (close->cl_stateowner) {
1da177e4 3209 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3210 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3211 }
1da177e4
LT		 3212 nfs4_unlock_state();
3213 return status;
3214}
3215
b37ad28b 3216__be32
ca364317
BF		 3217nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3218 struct nfsd4_delegreturn *dr)
1da177e4 3219{
203a8c8e
BF		 3220 struct nfs4_delegation *dp;
3221 stateid_t *stateid = &dr->dr_stateid;
3222 struct inode *inode;
b37ad28b 3223 __be32 status;
6668958f 3224 int flags = 0;
1da177e4 3225
ca364317 3226 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3227 return status;
3228 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3229
6668958f
AA		 3230 if (nfsd4_has_session(cstate))
3231 flags |= HAS_SESSION;
1da177e4 3232 nfs4_lock_state();
203a8c8e
BF		 3233 status = nfserr_bad_stateid;
3234 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3235 goto out;
3236 status = nfserr_stale_stateid;
3237 if (STALE_STATEID(stateid))
3238 goto out;
7e0f7cf5 3239 status = nfserr_bad_stateid;
203a8c8e
BF		 3240 if (!is_delegation_stateid(stateid))
3241 goto out;
203a8c8e 3242 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3243 if (!dp) {
3244 status = stateid_error_map(stateid);
203a8c8e 3245 goto out;
78155ed7 3246 }
6668958f 3247 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3248 if (status)
3249 goto out;
3250 renew_client(dp->dl_client);
3251
3252 unhash_delegation(dp);
1da177e4 3253out:
203a8c8e
BF		 3254 nfs4_unlock_state();
3255
1da177e4
LT		 3256 return status;
3257}
3258
3259
3260/*
3261 * Lock owner state (byte-range locks)
3262 */
3263#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3264#define LOCK_HASH_BITS 8
3265#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3266#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3267
87df4de8
BH		 3268static inline u64
3269end_offset(u64 start, u64 len)
3270{
3271 u64 end;
3272
3273 end = start + len;
3274 return end >= start ? end: NFS4_MAX_UINT64;
3275}
3276
3277/* last octet in a range */
3278static inline u64
3279last_byte_offset(u64 start, u64 len)
3280{
3281 u64 end;
3282
3283 BUG_ON(!len);
3284 end = start + len;
3285 return end > start ? end - 1: NFS4_MAX_UINT64;
3286}
3287
1da177e4
LT		 3288#define lockownerid_hashval(id) \
3289 ((id) & LOCK_HASH_MASK)
3290
3291static inline unsigned int
3292lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3293 struct xdr_netobj *ownername)
3294{
3295 return (file_hashval(inode) + cl_id
3296 + opaque_hashval(ownername->data, ownername->len))
3297 & LOCK_HASH_MASK;
3298}
3299
3300static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3301static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3302static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3303
fd39ca9a 3304static struct nfs4_stateid *
1da177e4
LT		 3305find_stateid(stateid_t *stid, int flags)
3306{
9346eff0 3307 struct nfs4_stateid *local;
1da177e4
LT		 3308 u32 st_id = stid->si_stateownerid;
3309 u32 f_id = stid->si_fileid;
3310 unsigned int hashval;
3311
3312 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3313 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3314 hashval = stateid_hashval(st_id, f_id);
3315 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3316 if ((local->st_stateid.si_stateownerid == st_id) &&
3317 (local->st_stateid.si_fileid == f_id))
3318 return local;
3319 }
3320 }
9346eff0
KK		 3321
3322 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3323 hashval = stateid_hashval(st_id, f_id);
3324 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3325 if ((local->st_stateid.si_stateownerid == st_id) &&
3326 (local->st_stateid.si_fileid == f_id))
3327 return local;
3328 }
849823c5 3329 }
1da177e4
LT		 3330 return NULL;
3331}
3332
3333static struct nfs4_delegation *
3334find_delegation_stateid(struct inode *ino, stateid_t *stid)
3335{
13cd2184
N		 3336 struct nfs4_file *fp;
3337 struct nfs4_delegation *dl;
1da177e4
LT		 3338
3339 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3340 stid->si_boot, stid->si_stateownerid,
3341 stid->si_fileid, stid->si_generation);
3342
1da177e4 3343 fp = find_file(ino);
13cd2184
N		 3344 if (!fp)
3345 return NULL;
3346 dl = find_delegation_file(fp, stid);
3347 put_nfs4_file(fp);
3348 return dl;
1da177e4
LT		 3349}
3350
3351/*
3352 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3353 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3354 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3355 * locking, this prevents us from being completely protocol-compliant. The
3356 * real solution to this problem is to start using unsigned file offsets in
3357 * the VFS, but this is a very deep change!
3358 */
3359static inline void
3360nfs4_transform_lock_offset(struct file_lock *lock)
3361{
3362 if (lock->fl_start < 0)
3363 lock->fl_start = OFFSET_MAX;
3364 if (lock->fl_end < 0)
3365 lock->fl_end = OFFSET_MAX;
3366}
3367
d5b9026a
N		 3368/* Hack!: For now, we're defining this just so we can use a pointer to it
3369 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3370static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3371};
1da177e4
LT		 3372
3373static inline void
3374nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3375{
d5b9026a
N		 3376 struct nfs4_stateowner *sop;
3377 unsigned int hval;
1da177e4 3378
d5b9026a
N		 3379 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3380 sop = (struct nfs4_stateowner *) fl->fl_owner;
3381 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3382 kref_get(&sop->so_ref);
3383 deny->ld_sop = sop;
3384 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3385 } else {
3386 deny->ld_sop = NULL;
3387 deny->ld_clientid.cl_boot = 0;
3388 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3389 }
3390 deny->ld_start = fl->fl_start;
87df4de8
BH		 3391 deny->ld_length = NFS4_MAX_UINT64;
3392 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3393 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3394 deny->ld_type = NFS4_READ_LT;
3395 if (fl->fl_type != F_RDLCK)
3396 deny->ld_type = NFS4_WRITE_LT;
3397}
3398
1da177e4
LT		 3399static struct nfs4_stateowner *
3400find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3401 struct xdr_netobj *owner)
3402{
3403 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3404 struct nfs4_stateowner *op;
3405
3406 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3407 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3408 return op;
3409 }
3410 return NULL;
3411}
3412
3413/*
3414 * Alloc a lock owner structure.
3415 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3416 * occured.
3417 *
3418 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3419 */
3420
3421static struct nfs4_stateowner *
3422alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3423 struct nfs4_stateowner *sop;
3424 struct nfs4_replay *rp;
3425 unsigned int idhashval;
3426
3427 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3428 return NULL;
3429 idhashval = lockownerid_hashval(current_ownerid);
3430 INIT_LIST_HEAD(&sop->so_idhash);
3431 INIT_LIST_HEAD(&sop->so_strhash);
3432 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3433 INIT_LIST_HEAD(&sop->so_stateids);
3434 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3435 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3436 sop->so_time = 0;
3437 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3438 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3439 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3440 sop->so_is_open_owner = 0;
3441 sop->so_id = current_ownerid++;
3442 sop->so_client = clp;
b59e3c0e
NB		 3443 /* It is the openowner seqid that will be incremented in encode in the
3444 * case of new lockowners; so increment the lock seqid manually: */
3445 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3446 sop->so_confirmed = 1;
3447 rp = &sop->so_replay;
de1ae286 3448 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3449 rp->rp_buflen = 0;
3450 rp->rp_buf = rp->rp_ibuf;
3451 return sop;
3452}
3453
fd39ca9a 3454static struct nfs4_stateid *
1da177e4
LT		 3455alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3456{
3457 struct nfs4_stateid *stp;
3458 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3459
5ac049ac
N		 3460 stp = nfs4_alloc_stateid();
3461 if (stp == NULL)
1da177e4
LT		 3462 goto out;
3463 INIT_LIST_HEAD(&stp->st_hash);
3464 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3465 INIT_LIST_HEAD(&stp->st_perstateowner);
3466 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3467 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3468 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3469 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3470 stp->st_stateowner = sop;
13cd2184 3471 get_nfs4_file(fp);
1da177e4 3472 stp->st_file = fp;
78155ed7 3473 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3474 stp->st_stateid.si_stateownerid = sop->so_id;
3475 stp->st_stateid.si_fileid = fp->fi_id;
3476 stp->st_stateid.si_generation = 0;
3477 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3478 stp->st_access_bmap = open_stp->st_access_bmap;
3479 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3480 stp->st_openstp = open_stp;
1da177e4
LT		 3481
3482out:
3483 return stp;
3484}
3485
fd39ca9a 3486static int
1da177e4
LT		 3487check_lock_length(u64 offset, u64 length)
3488{
87df4de8 3489 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3490 LOFF_OVERFLOW(offset, length)));
3491}
3492
3493/*
3494 * LOCK operation
3495 */
b37ad28b 3496__be32
ca364317 3497nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3498 struct nfsd4_lock *lock)
1da177e4 3499{
3e9e3dbe 3500 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3501 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3502 struct nfs4_stateid *lock_stp;
3503 struct file *filp;
3504 struct file_lock file_lock;
8dc7c311 3505 struct file_lock conflock;
b37ad28b 3506 __be32 status = 0;
1da177e4 3507 unsigned int strhashval;
150b3934 3508 unsigned int cmd;
b8dd7b9a 3509 int err;
1da177e4
LT		 3510
3511 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3512 (long long) lock->lk_offset,
3513 (long long) lock->lk_length);
3514
1da177e4
LT		 3515 if (check_lock_length(lock->lk_offset, lock->lk_length))
3516 return nfserr_inval;
3517
ca364317 3518 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3519 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3520 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3521 return status;
3522 }
3523
1da177e4
LT		 3524 nfs4_lock_state();
3525
3526 if (lock->lk_is_new) {
893f8770
N		 3527 /*
3528 * Client indicates that this is a new lockowner.
3529 * Use open owner and open stateid to create lock owner and
3530 * lock stateid.
3531 */
1da177e4
LT		 3532 struct nfs4_stateid *open_stp = NULL;
3533 struct nfs4_file *fp;
3534
3535 status = nfserr_stale_clientid;
60adfc50
AA		 3536 if (!nfsd4_has_session(cstate) &&
3537 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3538 goto out;
1da177e4 3539
1da177e4 3540 /* validate and update open stateid and open seqid */
dd453dfd 3541 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3542 lock->lk_new_open_seqid,
3543 &lock->lk_new_open_stateid,
f3362737 3544 OPEN_STATE,
3a65588a 3545 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3546 lock);
37515177 3547 if (status)
1da177e4 3548 goto out;
3a65588a 3549 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3550 /* create lockowner and lock stateid */
3551 fp = open_stp->st_file;
3552 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3553 open_sop->so_client->cl_clientid.cl_id,
3554 &lock->v.new.owner);
3e9e3dbe
N		 3555 /* XXX: Do we need to check for duplicate stateowners on
3556 * the same file, or should they just be allowed (and
3557 * create new stateids)? */
1da177e4 3558 status = nfserr_resource;
b59e3c0e
NB		 3559 lock_sop = alloc_init_lock_stateowner(strhashval,
3560 open_sop->so_client, open_stp, lock);
3561 if (lock_sop == NULL)
1da177e4 3562 goto out;
b59e3c0e 3563 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3564 if (lock_stp == NULL)
1da177e4 3565 goto out;
1da177e4
LT		 3566 } else {
3567 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3568 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3569 lock->lk_old_lock_seqid,
3570 &lock->lk_old_lock_stateid,
f3362737 3571 LOCK_STATE,
3a65588a 3572 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3573 if (status)
3574 goto out;
3a65588a 3575 lock_sop = lock->lk_replay_owner;
1da177e4 3576 }
3a65588a 3577 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3578 filp = lock_stp->st_vfs_file;
3579
0dd395dc 3580 status = nfserr_grace;
af558e33 3581 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3582 goto out;
3583 status = nfserr_no_grace;
af558e33 3584 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3585 goto out;
3586
1da177e4
LT		 3587 locks_init_lock(&file_lock);
3588 switch (lock->lk_type) {
3589 case NFS4_READ_LT:
3590 case NFS4_READW_LT:
3591 file_lock.fl_type = F_RDLCK;
150b3934 3592 cmd = F_SETLK;
1da177e4
LT		 3593 break;
3594 case NFS4_WRITE_LT:
3595 case NFS4_WRITEW_LT:
3596 file_lock.fl_type = F_WRLCK;
150b3934 3597 cmd = F_SETLK;
1da177e4
LT		 3598 break;
3599 default:
3600 status = nfserr_inval;
3601 goto out;
3602 }
b59e3c0e 3603 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3604 file_lock.fl_pid = current->tgid;
3605 file_lock.fl_file = filp;
3606 file_lock.fl_flags = FL_POSIX;
d5b9026a 3607 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3608
3609 file_lock.fl_start = lock->lk_offset;
87df4de8 3610 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3611 nfs4_transform_lock_offset(&file_lock);
3612
3613 /*
3614 * Try to lock the file in the VFS.
3615 * Note: locks.c uses the BKL to protect the inode's lock list.
3616 */
3617
fd85b817 3618 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3619 switch (-err) {
1da177e4
LT		 3620 case 0: /* success! */
3621 update_stateid(&lock_stp->st_stateid);
3622 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3623 sizeof(stateid_t));
b8dd7b9a 3624 status = 0;
eb76b3fd
AA		 3625 break;
3626 case (EAGAIN): /* conflock holds conflicting lock */
3627 status = nfserr_denied;
3628 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3629 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3630 break;
1da177e4
LT		 3631 case (EDEADLK):
3632 status = nfserr_deadlock;
eb76b3fd 3633 break;
1da177e4 3634 default:
fd85b817 3635 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3636 status = nfserr_resource;
3637 break;
1da177e4 3638 }
1da177e4 3639out:
8a280510 3640 if (status && lock->lk_is_new && lock_sop)
f044ff83 3641 release_lockowner(lock_sop);
3a65588a
BF		 3642 if (lock->lk_replay_owner) {
3643 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3644 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3645 }
1da177e4
LT		 3646 nfs4_unlock_state();
3647 return status;
3648}
3649
55ef1274
BF		 3650/*
3651 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3652 * so we do a temporary open here just to get an open file to pass to
3653 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3654 * inode operation.)
3655 */
3656static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3657{
3658 struct file *file;
3659 int err;
3660
3661 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3662 if (err)
3663 return err;
3664 err = vfs_test_lock(file, lock);
3665 nfsd_close(file);
3666 return err;
3667}
3668
1da177e4
LT		 3669/*
3670 * LOCKT operation
3671 */
b37ad28b 3672__be32
ca364317
BF		 3673nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3674 struct nfsd4_lockt *lockt)
1da177e4
LT		 3675{
3676 struct inode *inode;
1da177e4 3677 struct file_lock file_lock;
fd85b817 3678 int error;
b37ad28b 3679 __be32 status;
1da177e4 3680
af558e33 3681 if (locks_in_grace())
1da177e4
LT		 3682 return nfserr_grace;
3683
3684 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3685 return nfserr_inval;
3686
3687 lockt->lt_stateowner = NULL;
3688 nfs4_lock_state();
3689
3690 status = nfserr_stale_clientid;
60adfc50 3691 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3692 goto out;
1da177e4 3693
ca364317 3694 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3695 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3696 if (status == nfserr_symlink)
3697 status = nfserr_inval;
3698 goto out;
3699 }
3700
ca364317 3701 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3702 locks_init_lock(&file_lock);
3703 switch (lockt->lt_type) {
3704 case NFS4_READ_LT:
3705 case NFS4_READW_LT:
3706 file_lock.fl_type = F_RDLCK;
3707 break;
3708 case NFS4_WRITE_LT:
3709 case NFS4_WRITEW_LT:
3710 file_lock.fl_type = F_WRLCK;
3711 break;
3712 default:
2fdada03 3713 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3714 status = nfserr_inval;
3715 goto out;
3716 }
3717
3718 lockt->lt_stateowner = find_lockstateowner_str(inode,
3719 &lockt->lt_clientid, &lockt->lt_owner);
3720 if (lockt->lt_stateowner)
3721 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3722 file_lock.fl_pid = current->tgid;
3723 file_lock.fl_flags = FL_POSIX;
3724
3725 file_lock.fl_start = lockt->lt_offset;
87df4de8 3726 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3727
3728 nfs4_transform_lock_offset(&file_lock);
3729
1da177e4 3730 status = nfs_ok;
55ef1274 3731 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3732 if (error) {
3733 status = nfserrno(error);
3734 goto out;
3735 }
9d6a8c5c 3736 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3737 status = nfserr_denied;
9d6a8c5c 3738 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3739 }
3740out:
3741 nfs4_unlock_state();
3742 return status;
3743}
3744
b37ad28b 3745__be32
ca364317 3746nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3747 struct nfsd4_locku *locku)
1da177e4
LT		 3748{
3749 struct nfs4_stateid *stp;
3750 struct file *filp = NULL;
3751 struct file_lock file_lock;
b37ad28b 3752 __be32 status;
b8dd7b9a 3753 int err;
1da177e4
LT		 3754
3755 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3756 (long long) locku->lu_offset,
3757 (long long) locku->lu_length);
3758
3759 if (check_lock_length(locku->lu_offset, locku->lu_length))
3760 return nfserr_inval;
3761
3762 nfs4_lock_state();
3763
dd453dfd 3764 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3765 locku->lu_seqid,
3766 &locku->lu_stateid,
f3362737 3767 LOCK_STATE,
1da177e4
LT		 3768 &locku->lu_stateowner, &stp, NULL)))
3769 goto out;
3770
3771 filp = stp->st_vfs_file;
3772 BUG_ON(!filp);
3773 locks_init_lock(&file_lock);
3774 file_lock.fl_type = F_UNLCK;
3775 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3776 file_lock.fl_pid = current->tgid;
3777 file_lock.fl_file = filp;
3778 file_lock.fl_flags = FL_POSIX;
d5b9026a 3779 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3780 file_lock.fl_start = locku->lu_offset;
3781
87df4de8 3782 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3783 nfs4_transform_lock_offset(&file_lock);
3784
3785 /*
3786 * Try to unlock the file in the VFS.
3787 */
fd85b817 3788 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3789 if (err) {
fd85b817 3790 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3791 goto out_nfserr;
3792 }
3793 /*
3794 * OK, unlock succeeded; the only thing left to do is update the stateid.
3795 */
3796 update_stateid(&stp->st_stateid);
3797 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3798
3799out:
f2327d9a 3800 if (locku->lu_stateowner) {
1da177e4 3801 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3802 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3803 }
1da177e4
LT		 3804 nfs4_unlock_state();
3805 return status;
3806
3807out_nfserr:
b8dd7b9a 3808 status = nfserrno(err);
1da177e4
LT		 3809 goto out;
3810}
3811
3812/*
3813 * returns
3814 * 1: locks held by lockowner
3815 * 0: no locks held by lockowner
3816 */
3817static int
3818check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3819{
3820 struct file_lock **flpp;
7eaa36e2 3821 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3822 int status = 0;
3823
3824 lock_kernel();
3825 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3826 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3827 status = 1;
3828 goto out;
796dadfd 3829 }
1da177e4
LT		 3830 }
3831out:
3832 unlock_kernel();
3833 return status;
3834}
3835
b37ad28b 3836__be32
b591480b
BF		 3837nfsd4_release_lockowner(struct svc_rqst *rqstp,
3838 struct nfsd4_compound_state *cstate,
3839 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3840{
3841 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3842 struct nfs4_stateowner *sop;
3843 struct nfs4_stateid *stp;
1da177e4 3844 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3845 struct list_head matches;
3846 int i;
b37ad28b 3847 __be32 status;
1da177e4
LT		 3848
3849 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3850 clid->cl_boot, clid->cl_id);
3851
3852 /* XXX check for lease expiration */
3853
3854 status = nfserr_stale_clientid;
849823c5 3855 if (STALE_CLIENTID(clid))
1da177e4 3856 return status;
1da177e4
LT		 3857
3858 nfs4_lock_state();
3859
3e9e3dbe
N		 3860 status = nfserr_locks_held;
3861 /* XXX: we're doing a linear search through all the lockowners.
3862 * Yipes! For now we'll just hope clients aren't really using
3863 * release_lockowner much, but eventually we have to fix these
3864 * data structures. */
3865 INIT_LIST_HEAD(&matches);
3866 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3867 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3868 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3869 continue;
3870 list_for_each_entry(stp, &sop->so_stateids,
3871 st_perstateowner) {
3872 if (check_for_locks(stp->st_vfs_file, sop))
3873 goto out;
3874 /* Note: so_perclient unused for lockowners,
3875 * so it's OK to fool with here. */
3876 list_add(&sop->so_perclient, &matches);
3877 }
1da177e4 3878 }
3e9e3dbe
N		 3879 }
3880 /* Clients probably won't expect us to return with some (but not all)
3881 * of the lockowner state released; so don't release any until all
3882 * have been checked. */
3883 status = nfs_ok;
0fa822e4
N		 3884 while (!list_empty(&matches)) {
3885 sop = list_entry(matches.next, struct nfs4_stateowner,
3886 so_perclient);
3887 /* unhash_stateowner deletes so_perclient only
3888 * for openowners. */
3889 list_del(&sop->so_perclient);
f044ff83 3890 release_lockowner(sop);
1da177e4
LT		 3891 }
3892out:
3893 nfs4_unlock_state();
3894 return status;
3895}
3896
3897static inline struct nfs4_client_reclaim *
a55370a3 3898alloc_reclaim(void)
1da177e4 3899{
a55370a3 3900 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3901}
3902
c7b9a459 3903int
a1bcecd2 3904nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3905{
3906 unsigned int strhashval = clientstr_hashval(name);
3907 struct nfs4_client *clp;
3908
a1bcecd2 3909 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3910 return clp ? 1 : 0;
3911}
3912
1da177e4
LT		 3913/*
3914 * failure => all reset bets are off, nfserr_no_grace...
3915 */
190e4fbf
N		 3916int
3917nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3918{
3919 unsigned int strhashval;
3920 struct nfs4_client_reclaim *crp = NULL;
3921
a55370a3
N		 3922 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3923 crp = alloc_reclaim();
1da177e4
LT		 3924 if (!crp)
3925 return 0;
a55370a3 3926 strhashval = clientstr_hashval(name);
1da177e4
LT		 3927 INIT_LIST_HEAD(&crp->cr_strhash);
3928 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3929 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3930 reclaim_str_hashtbl_size++;
3931 return 1;
3932}
3933
3934static void
3935nfs4_release_reclaim(void)
3936{
3937 struct nfs4_client_reclaim *crp = NULL;
3938 int i;
3939
1da177e4
LT		 3940 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3941 while (!list_empty(&reclaim_str_hashtbl[i])) {
3942 crp = list_entry(reclaim_str_hashtbl[i].next,
3943 struct nfs4_client_reclaim, cr_strhash);
3944 list_del(&crp->cr_strhash);
1da177e4
LT		 3945 kfree(crp);
3946 reclaim_str_hashtbl_size--;
3947 }
3948 }
3949 BUG_ON(reclaim_str_hashtbl_size);
3950}
3951
3952/*
3953 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3954static struct nfs4_client_reclaim *
1da177e4
LT		 3955nfs4_find_reclaim_client(clientid_t *clid)
3956{
3957 unsigned int strhashval;
3958 struct nfs4_client *clp;
3959 struct nfs4_client_reclaim *crp = NULL;
3960
3961
3962 /* find clientid in conf_id_hashtbl */
3963 clp = find_confirmed_client(clid);
3964 if (clp == NULL)
3965 return NULL;
3966
a55370a3
N		 3967 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3968 clp->cl_name.len, clp->cl_name.data,
3969 clp->cl_recdir);
1da177e4
LT		 3970
3971 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3972 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3973 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3974 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3975 return crp;
3976 }
3977 }
3978 return NULL;
3979}
3980
3981/*
3982* Called from OPEN. Look for clientid in reclaim list.
3983*/
b37ad28b 3984__be32
1da177e4
LT		 3985nfs4_check_open_reclaim(clientid_t *clid)
3986{
dfc83565 3987 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3988}
3989
ac4d8ff2 3990/* initialization to perform at module load time: */
1da177e4 3991
e8ff2a84 3992int
ac4d8ff2 3993nfs4_state_init(void)
1da177e4 3994{
e8ff2a84 3995 int i, status;
1da177e4 3996
e8ff2a84
BF		 3997 status = nfsd4_init_slabs();
3998 if (status)
3999 return status;
1da177e4
LT		 4000 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4001 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4002 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4003 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4004 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4005 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4006 }
5282fd72
ME		 4007 for (i = 0; i < SESSION_HASH_SIZE; i++)
4008 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4009 for (i = 0; i < FILE_HASH_SIZE; i++) {
4010 INIT_LIST_HEAD(&file_hashtbl[i]);
4011 }
4012 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4013 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4014 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4015 }
4016 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4017 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4018 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4019 }
4020 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4021 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4022 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4023 }
1da177e4 4024 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4025 INIT_LIST_HEAD(&close_lru);
4026 INIT_LIST_HEAD(&client_lru);
4027 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4028 reclaim_str_hashtbl_size = 0;
e8ff2a84 4029 return 0;
ac4d8ff2
N		 4030}
4031
190e4fbf
N		 4032static void
4033nfsd4_load_reboot_recovery_data(void)
4034{
4035 int status;
4036
0964a3d3
N		 4037 nfs4_lock_state();
4038 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4039 status = nfsd4_recdir_load();
0964a3d3 4040 nfs4_unlock_state();
190e4fbf
N		 4041 if (status)
4042 printk("NFSD: Failure reading reboot recovery data\n");
4043}
4044
9a8db97e
ME		 4045unsigned long
4046get_nfs4_grace_period(void)
4047{
4048 return max(user_lease_time, lease_time) * HZ;
4049}
4050
c2f1a551
MS		 4051/*
4052 * Since the lifetime of a delegation isn't limited to that of an open, a
4053 * client may quite reasonably hang on to a delegation as long as it has
4054 * the inode cached. This becomes an obvious problem the first time a
4055 * client's inode cache approaches the size of the server's total memory.
4056 *
4057 * For now we avoid this problem by imposing a hard limit on the number
4058 * of delegations, which varies according to the server's memory size.
4059 */
4060static void
4061set_max_delegations(void)
4062{
4063 /*
4064 * Allow at most 4 delegations per megabyte of RAM. Quick
4065 * estimates suggest that in the worst case (where every delegation
4066 * is for a different inode), a delegation could take about 1.5K,
4067 * giving a worst case usage of about 6% of memory.
4068 */
4069 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4070}
4071
ac4d8ff2
N		 4072/* initialization to perform when the nfsd service is started: */
4073
4074static void
4075__nfs4_state_start(void)
4076{
9a8db97e 4077 unsigned long grace_time;
ac4d8ff2 4078
1da177e4 4079 boot_time = get_seconds();
af558e33 4080 grace_time = get_nfs4_grace_period();
d99a05ad 4081 lease_time = user_lease_time;
af558e33 4082 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4083 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4084 grace_time/HZ);
58da282b 4085 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4086 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4087 set_max_delegations();
1da177e4
LT		 4088}
4089
e8ff2a84 4090void
76a3550e 4091nfs4_state_start(void)
1da177e4 4092{
1da177e4 4093 if (nfs4_init)
e8ff2a84 4094 return;
190e4fbf 4095 nfsd4_load_reboot_recovery_data();
76a3550e 4096 __nfs4_state_start();
1da177e4 4097 nfs4_init = 1;
e8ff2a84 4098 return;
1da177e4
LT		 4099}
4100
1da177e4
LT		 4101time_t
4102nfs4_lease_time(void)
4103{
4104 return lease_time;
4105}
4106
4107static void
4108__nfs4_state_shutdown(void)
4109{
4110 int i;
4111 struct nfs4_client *clp = NULL;
4112 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4113 struct list_head *pos, *next, reaplist;
4114
1da177e4
LT		 4115 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4116 while (!list_empty(&conf_id_hashtbl[i])) {
4117 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4118 expire_client(clp);
4119 }
4120 while (!list_empty(&unconf_str_hashtbl[i])) {
4121 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4122 expire_client(clp);
4123 }
4124 }
4125 INIT_LIST_HEAD(&reaplist);
4126 spin_lock(&recall_lock);
4127 list_for_each_safe(pos, next, &del_recall_lru) {
4128 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4129 list_move(&dp->dl_recall_lru, &reaplist);
4130 }
4131 spin_unlock(&recall_lock);
4132 list_for_each_safe(pos, next, &reaplist) {
4133 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4134 list_del_init(&dp->dl_recall_lru);
4135 unhash_delegation(dp);
4136 }
4137
190e4fbf 4138 nfsd4_shutdown_recdir();
1da177e4 4139 nfs4_init = 0;
1da177e4
LT		 4140}
4141
4142void
4143nfs4_state_shutdown(void)
4144{
5e8d5c29
N		 4145 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4146 destroy_workqueue(laundry_wq);
2c5e7615 4147 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4148 nfs4_lock_state();
4149 nfs4_release_reclaim();
4150 __nfs4_state_shutdown();
1da177e4
LT		 4151 nfs4_unlock_state();
4152}
4153
3dd98a3b
JL		 4154/*
4155 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4156 * accessed when nfsd is starting.
4157 */
0964a3d3
N		 4158static void
4159nfs4_set_recdir(char *recdir)
4160{
0964a3d3 4161 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4162}
4163
4164/*
4165 * Change the NFSv4 recovery directory to recdir.
4166 */
4167int
4168nfs4_reset_recoverydir(char *recdir)
4169{
4170 int status;
a63bb996 4171 struct path path;
0964a3d3 4172
a63bb996 4173 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4174 if (status)
4175 return status;
4176 status = -ENOTDIR;
a63bb996 4177 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4178 nfs4_set_recdir(recdir);
4179 status = 0;
4180 }
a63bb996 4181 path_put(&path);
0964a3d3
N		 4182 return status;
4183}
4184
3dd98a3b
JL		 4185char *
4186nfs4_recoverydir(void)
4187{
4188 return user_recovery_dirname;
4189}
4190
1da177e4
LT		 4191/*
4192 * Called when leasetime is changed.
4193 *
d99a05ad
N		 4194 * The only way the protocol gives us to handle on-the-fly lease changes is to
4195 * simulate a reboot. Instead of doing that, we just wait till the next time
4196 * we start to register any changes in lease time. If the administrator
4197 * really wants to change the lease time *now*, they can go ahead and bring
4198 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4199 *
4200 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4201 * when nfsd is starting
1da177e4
LT		 4202 */
4203void
4204nfs4_reset_lease(time_t leasetime)
4205{
d99a05ad 4206 user_lease_time = leasetime;
1da177e4 4207}
Linux 6.x block layer and io_uring tree(s)