projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: track recall retries in nfs4_delegation
[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
428 spin_lock(&nfsd_serv->sv_lock);
429 if (np + nfsd_serv->sv_drc_pages_used > nfsd_serv->sv_drc_max_pages)
430 np = nfsd_serv->sv_drc_max_pages - nfsd_serv->sv_drc_pages_used;
431 nfsd_serv->sv_drc_pages_used += np;
432 spin_unlock(&nfsd_serv->sv_lock);
433
434 if (np <= 0) {
435 status = nfserr_resource;
436 fchan->maxreqs = 0;
437 } else
438 fchan->maxreqs = np / NFSD_PAGES_PER_SLOT;
439
440 return status;
441}
442
443/*
444 * fchan holds the client values on input, and the server values on output
445 */
446static int init_forechannel_attrs(struct svc_rqst *rqstp,
447 struct nfsd4_session *session,
448 struct nfsd4_channel_attrs *fchan)
449{
450 int status = 0;
451 __u32 maxcount = svc_max_payload(rqstp);
452
453 /* headerpadsz set to zero in encode routine */
454
455 /* Use the client's max request and max response size if possible */
456 if (fchan->maxreq_sz > maxcount)
457 fchan->maxreq_sz = maxcount;
458 session->se_fmaxreq_sz = fchan->maxreq_sz;
459
460 if (fchan->maxresp_sz > maxcount)
461 fchan->maxresp_sz = maxcount;
462 session->se_fmaxresp_sz = fchan->maxresp_sz;
463
464 /* Set the max response cached size our default which is
465 * a multiple of PAGE_SIZE and small */
466 session->se_fmaxresp_cached = NFSD_PAGES_PER_SLOT * PAGE_SIZE;
467 fchan->maxresp_cached = session->se_fmaxresp_cached;
468
469 /* Use the client's maxops if possible */
470 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
471 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
472 session->se_fmaxops = fchan->maxops;
473
474 /* try to use the client requested number of slots */
475 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
476 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
477
478 /* FIXME: Error means no more DRC pages so the server should
479 * recover pages from existing sessions. For now fail session
480 * creation.
481 */
482 status = set_forechannel_maxreqs(fchan);
483
484 session->se_fnumslots = fchan->maxreqs;
485 return status;
486}
487
488static int
489alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
490 struct nfsd4_create_session *cses)
491{
492 struct nfsd4_session *new, tmp;
493 int idx, status = nfserr_resource, slotsize;
494
495 memset(&tmp, 0, sizeof(tmp));
496
497 /* FIXME: For now, we just accept the client back channel attributes. */
498 status = init_forechannel_attrs(rqstp, &tmp, &cses->fore_channel);
499 if (status)
500 goto out;
501
502 /* allocate struct nfsd4_session and slot table in one piece */
503 slotsize = tmp.se_fnumslots * sizeof(struct nfsd4_slot);
504 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
505 if (!new)
506 goto out;
507
508 memcpy(new, &tmp, sizeof(*new));
509
510 new->se_client = clp;
511 gen_sessionid(new);
512 idx = hash_sessionid(&new->se_sessionid);
513 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
514 NFS4_MAX_SESSIONID_LEN);
515
516 new->se_flags = cses->flags;
517 kref_init(&new->se_ref);
518 spin_lock(&sessionid_lock);
519 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
520 list_add(&new->se_perclnt, &clp->cl_sessions);
521 spin_unlock(&sessionid_lock);
522
523 status = nfs_ok;
524out:
525 return status;
526}
527
5282fd72
ME		 528/* caller must hold sessionid_lock */
529static struct nfsd4_session *
530find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
531{
532 struct nfsd4_session *elem;
533 int idx;
534
535 dump_sessionid(__func__, sessionid);
536 idx = hash_sessionid(sessionid);
537 dprintk("%s: idx is %d\n", __func__, idx);
538 /* Search in the appropriate list */
539 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
540 dump_sessionid("list traversal", &elem->se_sessionid);
541 if (!memcmp(elem->se_sessionid.data, sessionid->data,
542 NFS4_MAX_SESSIONID_LEN)) {
543 return elem;
544 }
545 }
546
547 dprintk("%s: session not found\n", __func__);
548 return NULL;
549}
550
551/* caller must hold sessionid_lock */
7116ed6b 552static void
5282fd72 553unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 554{
555 list_del(&ses->se_hash);
556 list_del(&ses->se_perclnt);
5282fd72
ME		 557}
558
559static void
560release_session(struct nfsd4_session *ses)
561{
562 spin_lock(&sessionid_lock);
563 unhash_session(ses);
564 spin_unlock(&sessionid_lock);
7116ed6b
AA		 565 nfsd4_put_session(ses);
566}
567
14778a13
AA		 568static void nfsd4_release_respages(struct page **respages, short resused);
569
7116ed6b
AA		 570void
571free_session(struct kref *kref)
572{
573 struct nfsd4_session *ses;
14778a13 574 int i;
7116ed6b
AA		 575
576 ses = container_of(kref, struct nfsd4_session, se_ref);
14778a13
AA		 577 for (i = 0; i < ses->se_fnumslots; i++) {
578 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
579 nfsd4_release_respages(e->ce_respages, e->ce_resused);
580 }
7116ed6b
AA		 581 kfree(ses->se_slots);
582 kfree(ses);
583}
584
1da177e4
LT		 585static inline void
586renew_client(struct nfs4_client *clp)
587{
588 /*
589 * Move client to the end to the LRU list.
590 */
591 dprintk("renewing client (clientid %08x/%08x)\n",
592 clp->cl_clientid.cl_boot,
593 clp->cl_clientid.cl_id);
594 list_move_tail(&clp->cl_lru, &client_lru);
595 clp->cl_time = get_seconds();
596}
597
598/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
599static int
600STALE_CLIENTID(clientid_t *clid)
601{
602 if (clid->cl_boot == boot_time)
603 return 0;
60adfc50
AA		 604 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
605 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 606 return 1;
607}
608
609/*
610 * XXX Should we use a slab cache ?
611 * This type of memory management is somewhat inefficient, but we use it
612 * anyway since SETCLIENTID is not a common operation.
613 */
35bba9a3 614static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 615{
616 struct nfs4_client *clp;
617
35bba9a3
BF		 618 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
619 if (clp == NULL)
620 return NULL;
621 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
622 if (clp->cl_name.data == NULL) {
623 kfree(clp);
624 return NULL;
1da177e4 625 }
35bba9a3
BF		 626 memcpy(clp->cl_name.data, name.data, name.len);
627 clp->cl_name.len = name.len;
1da177e4
LT		 628 return clp;
629}
630
1b1a9b31
BF		 631static void
632shutdown_callback_client(struct nfs4_client *clp)
633{
c237dc03 634 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 635
1b1a9b31 636 if (clnt) {
404ec117
BF		 637 /*
638 * Callback threads take a reference on the client, so there
639 * should be no outstanding callbacks at this point.
640 */
c237dc03 641 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 642 rpc_shutdown_client(clnt);
643 }
c237dc03
BF		 644 if (clp->cl_cb_conn.cb_cred) {
645 put_rpccred(clp->cl_cb_conn.cb_cred);
646 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 647 }
1b1a9b31
BF		 648}
649
1da177e4
LT		 650static inline void
651free_client(struct nfs4_client *clp)
652{
1b1a9b31 653 shutdown_callback_client(clp);
38eb76a5
AA		 654 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
655 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 656 if (clp->cl_cred.cr_group_info)
657 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 658 kfree(clp->cl_principal);
1da177e4
LT		 659 kfree(clp->cl_name.data);
660 kfree(clp);
661}
662
663void
664put_nfs4_client(struct nfs4_client *clp)
665{
666 if (atomic_dec_and_test(&clp->cl_count))
667 free_client(clp);
668}
669
670static void
671expire_client(struct nfs4_client *clp)
672{
673 struct nfs4_stateowner *sop;
674 struct nfs4_delegation *dp;
1da177e4
LT		 675 struct list_head reaplist;
676
677 dprintk("NFSD: expire_client cl_count %d\n",
678 atomic_read(&clp->cl_count));
679
1da177e4
LT		 680 INIT_LIST_HEAD(&reaplist);
681 spin_lock(&recall_lock);
ea1da636
N		 682 while (!list_empty(&clp->cl_delegations)) {
683 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 684 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
685 dp->dl_flock);
ea1da636 686 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 687 list_move(&dp->dl_recall_lru, &reaplist);
688 }
689 spin_unlock(&recall_lock);
690 while (!list_empty(&reaplist)) {
691 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
692 list_del_init(&dp->dl_recall_lru);
693 unhash_delegation(dp);
694 }
695 list_del(&clp->cl_idhash);
696 list_del(&clp->cl_strhash);
697 list_del(&clp->cl_lru);
ea1da636
N		 698 while (!list_empty(&clp->cl_openowners)) {
699 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 700 release_openowner(sop);
1da177e4 701 }
c4bf7868
ME		 702 while (!list_empty(&clp->cl_sessions)) {
703 struct nfsd4_session *ses;
704 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
705 se_perclnt);
706 release_session(ses);
707 }
1da177e4
LT		 708 put_nfs4_client(clp);
709}
710
35bba9a3
BF		 711static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
712{
1da177e4
LT		 713 struct nfs4_client *clp;
714
35bba9a3
BF		 715 clp = alloc_client(name);
716 if (clp == NULL)
717 return NULL;
a55370a3 718 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 719 atomic_set(&clp->cl_count, 1);
c237dc03 720 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 721 INIT_LIST_HEAD(&clp->cl_idhash);
722 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 723 INIT_LIST_HEAD(&clp->cl_openowners);
724 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 725 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 726 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 727 return clp;
728}
729
35bba9a3
BF		 730static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
731{
732 memcpy(target->cl_verifier.data, source->data,
733 sizeof(target->cl_verifier.data));
1da177e4
LT		 734}
735
35bba9a3
BF		 736static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
737{
1da177e4
LT		 738 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
739 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
740}
741
35bba9a3
BF		 742static void copy_cred(struct svc_cred *target, struct svc_cred *source)
743{
1da177e4
LT		 744 target->cr_uid = source->cr_uid;
745 target->cr_gid = source->cr_gid;
746 target->cr_group_info = source->cr_group_info;
747 get_group_info(target->cr_group_info);
748}
749
35bba9a3 750static int same_name(const char *n1, const char *n2)
599e0a22 751{
a55370a3 752 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 753}
754
755static int
599e0a22
BF		 756same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
757{
758 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 759}
760
761static int
599e0a22
BF		 762same_clid(clientid_t *cl1, clientid_t *cl2)
763{
764 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 765}
766
767/* XXX what about NGROUP */
768static int
599e0a22
BF		 769same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
770{
771 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 772}
773
5ec7b46c
BF		 774static void gen_clid(struct nfs4_client *clp)
775{
776 static u32 current_clientid = 1;
777
1da177e4
LT		 778 clp->cl_clientid.cl_boot = boot_time;
779 clp->cl_clientid.cl_id = current_clientid++;
780}
781
deda2faa
BF		 782static void gen_confirm(struct nfs4_client *clp)
783{
784 static u32 i;
785 u32 *p;
1da177e4 786
1da177e4 787 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 788 *p++ = get_seconds();
789 *p++ = i++;
1da177e4
LT		 790}
791
35bba9a3
BF		 792static int check_name(struct xdr_netobj name)
793{
1da177e4
LT		 794 if (name.len == 0)
795 return 0;
796 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 797 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 798 return 0;
799 }
800 return 1;
801}
802
fd39ca9a 803static void
1da177e4
LT		 804add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
805{
806 unsigned int idhashval;
807
808 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
809 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
810 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
811 list_add_tail(&clp->cl_lru, &client_lru);
812 clp->cl_time = get_seconds();
813}
814
fd39ca9a 815static void
1da177e4
LT		 816move_to_confirmed(struct nfs4_client *clp)
817{
818 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
819 unsigned int strhashval;
820
821 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
822 list_del_init(&clp->cl_strhash);
f116629d 823 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 824 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 825 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
826 renew_client(clp);
827}
828
829static struct nfs4_client *
830find_confirmed_client(clientid_t *clid)
831{
832 struct nfs4_client *clp;
833 unsigned int idhashval = clientid_hashval(clid->cl_id);
834
835 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 836 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 837 return clp;
838 }
839 return NULL;
840}
841
842static struct nfs4_client *
843find_unconfirmed_client(clientid_t *clid)
844{
845 struct nfs4_client *clp;
846 unsigned int idhashval = clientid_hashval(clid->cl_id);
847
848 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 849 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 850 return clp;
851 }
852 return NULL;
853}
854
a1bcecd2
AA		 855/*
856 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
857 * parameter. Matching is based on the fact the at least one of the
858 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
859 *
860 * FIXME: we need to unify the clientid namespaces for nfsv4.x
861 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
862 * and SET_CLIENTID{,_CONFIRM}
863 */
864static inline int
865match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
866{
867 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
868 return use_exchange_id == has_exchange_flags;
869}
870
28ce6054 871static struct nfs4_client *
a1bcecd2
AA		 872find_confirmed_client_by_str(const char *dname, unsigned int hashval,
873 bool use_exchange_id)
28ce6054
N		 874{
875 struct nfs4_client *clp;
876
877 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 878 if (same_name(clp->cl_recdir, dname) &&
879 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 880 return clp;
881 }
882 return NULL;
883}
884
885static struct nfs4_client *
a1bcecd2
AA		 886find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
887 bool use_exchange_id)
28ce6054
N		 888{
889 struct nfs4_client *clp;
890
891 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 892 if (same_name(clp->cl_recdir, dname) &&
893 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 894 return clp;
895 }
896 return NULL;
897}
898
1da177e4
LT		 899/* a helper function for parse_callback */
900static int
901parse_octet(unsigned int *lenp, char **addrp)
902{
903 unsigned int len = *lenp;
904 char *p = *addrp;
905 int n = -1;
906 char c;
907
908 for (;;) {
909 if (!len)
910 break;
911 len--;
912 c = *p++;
913 if (c == '.')
914 break;
915 if ((c < '0') || (c > '9')) {
916 n = -1;
917 break;
918 }
919 if (n < 0)
920 n = 0;
921 n = (n * 10) + (c - '0');
922 if (n > 255) {
923 n = -1;
924 break;
925 }
926 }
927 *lenp = len;
928 *addrp = p;
929 return n;
930}
931
932/* parse and set the setclientid ipv4 callback address */
fd39ca9a 933static int
1da177e4
LT		 934parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
935{
936 int temp = 0;
937 u32 cbaddr = 0;
938 u16 cbport = 0;
939 u32 addrlen = addr_len;
940 char *addr = addr_val;
941 int i, shift;
942
943 /* ipaddress */
944 shift = 24;
945 for(i = 4; i > 0 ; i--) {
946 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
947 return 0;
948 }
949 cbaddr |= (temp << shift);
950 if (shift > 0)
951 shift -= 8;
952 }
953 *cbaddrp = cbaddr;
954
955 /* port */
956 shift = 8;
957 for(i = 2; i > 0 ; i--) {
958 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
959 return 0;
960 }
961 cbport |= (temp << shift);
962 if (shift > 0)
963 shift -= 8;
964 }
965 *cbportp = cbport;
966 return 1;
967}
968
fd39ca9a 969static void
1da177e4
LT		 970gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
971{
c237dc03 972 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT		 973
974 /* Currently, we only support tcp for the callback channel */
975 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
976 goto out_err;
977
978 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
979 &cb->cb_addr, &cb->cb_port)))
980 goto out_err;
981 cb->cb_prog = se->se_callback_prog;
982 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 983 return;
984out_err:
849823c5 985 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 986 "will not receive delegations\n",
987 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
988
1da177e4
LT		 989 return;
990}
991
074fe897
AA		 992void
993nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
994{
995 struct nfsd4_compoundres *resp = rqstp->rq_resp;
996
997 resp->cstate.statp = statp;
998}
999
1000/*
1001 * Dereference the result pages.
1002 */
1003static void
1004nfsd4_release_respages(struct page **respages, short resused)
1005{
1006 int i;
1007
1008 dprintk("--> %s\n", __func__);
1009 for (i = 0; i < resused; i++) {
1010 if (!respages[i])
1011 continue;
1012 put_page(respages[i]);
1013 respages[i] = NULL;
1014 }
1015}
1016
1017static void
1018nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1019{
1020 int i;
1021
1022 for (i = 0; i < count; i++) {
1023 topages[i] = frompages[i];
1024 if (!topages[i])
1025 continue;
1026 get_page(topages[i]);
1027 }
1028}
1029
1030/*
1031 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1032 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1033 * length of the XDR response is less than se_fmaxresp_cached
1034 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1035 * of the reply (e.g. readdir).
1036 *
1037 * Store the base and length of the rq_req.head[0] page
1038 * of the NFSv4.1 data, just past the rpc header.
1039 */
1040void
1041nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1042{
1043 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1044 struct svc_rqst *rqstp = resp->rqstp;
1045 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1046 struct nfsd4_op *op = &args->ops[resp->opcnt];
1047 struct kvec *resv = &rqstp->rq_res.head[0];
1048
1049 dprintk("--> %s entry %p\n", __func__, entry);
1050
1051 /* Don't cache a failed OP_SEQUENCE. */
1052 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1053 return;
bf864a31 1054
074fe897 1055 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1056 entry->ce_opcnt = resp->opcnt;
1057 entry->ce_status = resp->cstate.status;
1058
1059 /*
1060 * Don't need a page to cache just the sequence operation - the slot
1061 * does this for us!
1062 */
1063
1064 if (nfsd4_not_cached(resp)) {
1065 entry->ce_resused = 0;
1066 entry->ce_rpchdrlen = 0;
1067 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1068 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1069 return;
1070 }
074fe897
AA		 1071 entry->ce_resused = rqstp->rq_resused;
1072 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1073 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1074 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1075 entry->ce_resused);
074fe897
AA		 1076 entry->ce_datav.iov_base = resp->cstate.statp;
1077 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1078 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1079 /* Current request rpc header length*/
1080 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1081 (char *)page_address(rqstp->rq_respages[0]);
1082}
1083
1084/*
1085 * We keep the rpc header, but take the nfs reply from the replycache.
1086 */
1087static int
1088nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1089 struct nfsd4_cache_entry *entry)
1090{
1091 struct svc_rqst *rqstp = resp->rqstp;
1092 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1093 int len;
1094
1095 /* Current request rpc header length*/
1096 len = (char *)resp->cstate.statp -
1097 (char *)page_address(rqstp->rq_respages[0]);
1098 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1099 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1100 entry->ce_datav.iov_len);
1101 return 0;
1102 }
1103 /* copy the cached reply nfsd data past the current rpc header */
1104 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1105 entry->ce_datav.iov_len);
1106 resv->iov_len = len + entry->ce_datav.iov_len;
1107 return 1;
1108}
1109
1110/*
1111 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1112 * cached page. Replace any futher replay pages from the cache.
1113 */
1114__be32
bf864a31
AA		 1115nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1116 struct nfsd4_sequence *seq)
074fe897
AA		 1117{
1118 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1119 __be32 status;
1120
1121 dprintk("--> %s entry %p\n", __func__, entry);
1122
bf864a31
AA		 1123 /*
1124 * If this is just the sequence operation, we did not keep
1125 * a page in the cache entry because we can just use the
1126 * slot info stored in struct nfsd4_sequence that was checked
1127 * against the slot in nfsd4_sequence().
1128 *
1129 * This occurs when seq->cachethis is FALSE, or when the client
1130 * session inactivity timer fires and a solo sequence operation
1131 * is sent (lease renewal).
1132 */
1133 if (seq && nfsd4_not_cached(resp)) {
1134 seq->maxslots = resp->cstate.session->se_fnumslots;
1135 return nfs_ok;
1136 }
074fe897
AA		 1137
1138 if (!nfsd41_copy_replay_data(resp, entry)) {
1139 /*
1140 * Not enough room to use the replay rpc header, send the
1141 * cached header. Release all the allocated result pages.
1142 */
1143 svc_free_res_pages(resp->rqstp);
1144 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1145 entry->ce_resused);
1146 } else {
1147 /* Release all but the first allocated result page */
1148
1149 resp->rqstp->rq_resused--;
1150 svc_free_res_pages(resp->rqstp);
1151
1152 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1153 &entry->ce_respages[1],
1154 entry->ce_resused - 1);
1155 }
1156
1157 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1158 resp->opcnt = entry->ce_opcnt;
1159 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1160 status = entry->ce_status;
1161
1162 return status;
1163}
1164
0733d213
AA		 1165/*
1166 * Set the exchange_id flags returned by the server.
1167 */
1168static void
1169nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1170{
1171 /* pNFS is not supported */
1172 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1173
1174 /* Referrals are supported, Migration is not. */
1175 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1176
1177 /* set the wire flags to return to client. */
1178 clid->flags = new->cl_exchange_flags;
1179}
1180
069b6ad4
AA		 1181__be32
1182nfsd4_exchange_id(struct svc_rqst *rqstp,
1183 struct nfsd4_compound_state *cstate,
1184 struct nfsd4_exchange_id *exid)
1185{
0733d213
AA		 1186 struct nfs4_client *unconf, *conf, *new;
1187 int status;
1188 unsigned int strhashval;
1189 char dname[HEXDIR_LEN];
1190 nfs4_verifier verf = exid->verifier;
1191 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1192
1193 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1194 " ip_addr=%u flags %x, spa_how %d\n",
1195 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1196 ip_addr, exid->flags, exid->spa_how);
1197
1198 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1199 return nfserr_inval;
1200
1201 /* Currently only support SP4_NONE */
1202 switch (exid->spa_how) {
1203 case SP4_NONE:
1204 break;
1205 case SP4_SSV:
1206 return nfserr_encr_alg_unsupp;
1207 default:
1208 BUG(); /* checked by xdr code */
1209 case SP4_MACH_CRED:
1210 return nfserr_serverfault; /* no excuse :-/ */
1211 }
1212
1213 status = nfs4_make_rec_clidname(dname, &exid->clname);
1214
1215 if (status)
1216 goto error;
1217
1218 strhashval = clientstr_hashval(dname);
1219
1220 nfs4_lock_state();
1221 status = nfs_ok;
1222
a1bcecd2 1223 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1224 if (conf) {
1225 if (!same_verf(&verf, &conf->cl_verifier)) {
1226 /* 18.35.4 case 8 */
1227 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1228 status = nfserr_not_same;
1229 goto out;
1230 }
1231 /* Client reboot: destroy old state */
1232 expire_client(conf);
1233 goto out_new;
1234 }
1235 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1236 /* 18.35.4 case 9 */
1237 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1238 status = nfserr_perm;
1239 goto out;
1240 }
1241 expire_client(conf);
1242 goto out_new;
1243 }
1244 if (ip_addr != conf->cl_addr &&
1245 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1246 /* Client collision. 18.35.4 case 3 */
1247 status = nfserr_clid_inuse;
1248 goto out;
1249 }
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;
1263 } else {
1264 /* 18.35.4 case 7 */
1265 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1266 status = nfserr_noent;
1267 goto out;
1268 }
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;
1477 if (seq->slotid >= session->se_fnumslots)
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
2062/*
2063 * Recall a delegation
2064 */
2065static int
2066do_recall(void *__dp)
2067{
2068 struct nfs4_delegation *dp = __dp;
2069
47f9940c 2070 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2071 nfsd4_cb_recall(dp);
2072 return 0;
2073}
2074
2075/*
2076 * Spawn a thread to perform a recall on the delegation represented
2077 * by the lease (file_lock)
2078 *
2079 * Called from break_lease() with lock_kernel() held.
2080 * Note: we assume break_lease will only call this *once* for any given
2081 * lease.
2082 */
2083static
2084void nfsd_break_deleg_cb(struct file_lock *fl)
2085{
2086 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2087 struct task_struct *t;
2088
2089 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2090 if (!dp)
2091 return;
2092
2093 /* We're assuming the state code never drops its reference
2094 * without first removing the lease. Since we're in this lease
2095 * callback (and since the lease code is serialized by the kernel
2096 * lock) we know the server hasn't removed the lease yet, we know
2097 * it's safe to take a reference: */
2098 atomic_inc(&dp->dl_count);
cfdcad4d 2099 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2100
2101 spin_lock(&recall_lock);
2102 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2103 spin_unlock(&recall_lock);
2104
2105 /* only place dl_time is set. protected by lock_kernel*/
2106 dp->dl_time = get_seconds();
2107
0272e1fd
BF		 2108 /*
2109 * We don't want the locks code to timeout the lease for us;
2110 * we'll remove it ourself if the delegation isn't returned
2111 * in time.
2112 */
2113 fl->fl_break_time = 0;
1da177e4
LT		 2114
2115 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2116 if (IS_ERR(t)) {
2117 struct nfs4_client *clp = dp->dl_client;
2118
2119 printk(KERN_INFO "NFSD: Callback thread failed for "
2120 "for client (clientid %08x/%08x)\n",
2121 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2122 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2123 nfs4_put_delegation(dp);
2124 }
2125}
2126
2127/*
2128 * The file_lock is being reapd.
2129 *
2130 * Called by locks_free_lock() with lock_kernel() held.
2131 */
2132static
2133void nfsd_release_deleg_cb(struct file_lock *fl)
2134{
2135 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2136
2137 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2138
2139 if (!(fl->fl_flags & FL_LEASE) || !dp)
2140 return;
2141 dp->dl_flock = NULL;
2142}
2143
2144/*
2145 * Set the delegation file_lock back pointer.
2146 *
a9933cea 2147 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2148 */
2149static
2150void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2151{
2152 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2153
2154 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2155 if (!dp)
2156 return;
2157 dp->dl_flock = new;
2158}
2159
2160/*
a9933cea 2161 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2162 */
2163static
2164int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2165{
2166 struct nfs4_delegation *onlistd =
2167 (struct nfs4_delegation *)onlist->fl_owner;
2168 struct nfs4_delegation *tryd =
2169 (struct nfs4_delegation *)try->fl_owner;
2170
2171 if (onlist->fl_lmops != try->fl_lmops)
2172 return 0;
2173
2174 return onlistd->dl_client == tryd->dl_client;
2175}
2176
2177
2178static
2179int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2180{
2181 if (arg & F_UNLCK)
2182 return lease_modify(onlist, arg);
2183 else
2184 return -EAGAIN;
2185}
2186
fd39ca9a 2187static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2188 .fl_break = nfsd_break_deleg_cb,
2189 .fl_release_private = nfsd_release_deleg_cb,
2190 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2191 .fl_mylease = nfsd_same_client_deleg_cb,
2192 .fl_change = nfsd_change_deleg_cb,
2193};
2194
2195
b37ad28b 2196__be32
6668958f
AA		 2197nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2198 struct nfsd4_open *open)
1da177e4 2199{
1da177e4
LT		 2200 clientid_t *clientid = &open->op_clientid;
2201 struct nfs4_client *clp = NULL;
2202 unsigned int strhashval;
2203 struct nfs4_stateowner *sop = NULL;
2204
1da177e4 2205 if (!check_name(open->op_owner))
0f442aa2 2206 return nfserr_inval;
1da177e4
LT		 2207
2208 if (STALE_CLIENTID(&open->op_clientid))
2209 return nfserr_stale_clientid;
2210
2211 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2212 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2213 open->op_stateowner = sop;
2214 if (!sop) {
2215 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2216 clp = find_confirmed_client(clientid);
2217 if (clp == NULL)
0f442aa2
BF		 2218 return nfserr_expired;
2219 goto renew;
1da177e4 2220 }
6668958f
AA		 2221 /* When sessions are used, skip open sequenceid processing */
2222 if (nfsd4_has_session(cstate))
2223 goto renew;
0f442aa2
BF		 2224 if (!sop->so_confirmed) {
2225 /* Replace unconfirmed owners without checking for replay. */
2226 clp = sop->so_client;
f044ff83 2227 release_openowner(sop);
0f442aa2
BF		 2228 open->op_stateowner = NULL;
2229 goto renew;
2230 }
2231 if (open->op_seqid == sop->so_seqid - 1) {
2232 if (sop->so_replay.rp_buflen)
a90b061c 2233 return nfserr_replay_me;
0f442aa2
BF		 2234 /* The original OPEN failed so spectacularly
2235 * that we don't even have replay data saved!
2236 * Therefore, we have no choice but to continue
2237 * processing this OPEN; presumably, we'll
2238 * fail again for the same reason.
2239 */
2240 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2241 goto renew;
2242 }
2243 if (open->op_seqid != sop->so_seqid)
2244 return nfserr_bad_seqid;
1da177e4 2245renew:
0f442aa2
BF		 2246 if (open->op_stateowner == NULL) {
2247 sop = alloc_init_open_stateowner(strhashval, clp, open);
2248 if (sop == NULL)
2249 return nfserr_resource;
2250 open->op_stateowner = sop;
2251 }
2252 list_del_init(&sop->so_close_lru);
1da177e4 2253 renew_client(sop->so_client);
0f442aa2 2254 return nfs_ok;
1da177e4
LT		 2255}
2256
b37ad28b 2257static inline __be32
4a6e43e6
N		 2258nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2259{
2260 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2261 return nfserr_openmode;
2262 else
2263 return nfs_ok;
2264}
2265
52f4fb43
N		 2266static struct nfs4_delegation *
2267find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2268{
2269 struct nfs4_delegation *dp;
2270
ea1da636 2271 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2272 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2273 return dp;
2274 }
2275 return NULL;
2276}
2277
b37ad28b 2278static __be32
567d9829
N		 2279nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2280 struct nfs4_delegation **dp)
2281{
2282 int flags;
b37ad28b 2283 __be32 status = nfserr_bad_stateid;
567d9829
N		 2284
2285 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2286 if (*dp == NULL)
c44c5eeb 2287 goto out;
567d9829
N		 2288 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2289 RD_STATE : WR_STATE;
2290 status = nfs4_check_delegmode(*dp, flags);
2291 if (status)
2292 *dp = NULL;
c44c5eeb
N		 2293out:
2294 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2295 return nfs_ok;
2296 if (status)
2297 return status;
2298 open->op_stateowner->so_confirmed = 1;
2299 return nfs_ok;
567d9829
N		 2300}
2301
b37ad28b 2302static __be32
1da177e4
LT		 2303nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2304{
2305 struct nfs4_stateid *local;
b37ad28b 2306 __be32 status = nfserr_share_denied;
1da177e4
LT		 2307 struct nfs4_stateowner *sop = open->op_stateowner;
2308
8beefa24 2309 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2310 /* ignore lock owners */
2311 if (local->st_stateowner->so_is_open_owner == 0)
2312 continue;
2313 /* remember if we have seen this open owner */
2314 if (local->st_stateowner == sop)
2315 *stpp = local;
2316 /* check for conflicting share reservations */
2317 if (!test_share(local, open))
2318 goto out;
2319 }
2320 status = 0;
2321out:
2322 return status;
2323}
2324
5ac049ac
N		 2325static inline struct nfs4_stateid *
2326nfs4_alloc_stateid(void)
2327{
2328 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2329}
2330
b37ad28b 2331static __be32
1da177e4 2332nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2333 struct nfs4_delegation *dp,
1da177e4
LT		 2334 struct svc_fh *cur_fh, int flags)
2335{
2336 struct nfs4_stateid *stp;
1da177e4 2337
5ac049ac 2338 stp = nfs4_alloc_stateid();
1da177e4
LT		 2339 if (stp == NULL)
2340 return nfserr_resource;
2341
567d9829
N		 2342 if (dp) {
2343 get_file(dp->dl_vfs_file);
2344 stp->st_vfs_file = dp->dl_vfs_file;
2345 } else {
b37ad28b 2346 __be32 status;
567d9829
N		 2347 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2348 &stp->st_vfs_file);
2349 if (status) {
2350 if (status == nfserr_dropit)
2351 status = nfserr_jukebox;
5ac049ac 2352 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2353 return status;
2354 }
1da177e4 2355 }
1da177e4
LT		 2356 *stpp = stp;
2357 return 0;
2358}
2359
b37ad28b 2360static inline __be32
1da177e4
LT		 2361nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2362 struct nfsd4_open *open)
2363{
2364 struct iattr iattr = {
2365 .ia_valid = ATTR_SIZE,
2366 .ia_size = 0,
2367 };
2368 if (!open->op_truncate)
2369 return 0;
2370 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2371 return nfserr_inval;
1da177e4
LT		 2372 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2373}
2374
b37ad28b 2375static __be32
1da177e4
LT		 2376nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2377{
2378 struct file *filp = stp->st_vfs_file;
7eaa36e2 2379 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2380 unsigned int share_access, new_writer;
b37ad28b 2381 __be32 status;
1da177e4
LT		 2382
2383 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2384 new_writer = (~share_access) & open->op_share_access
2385 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2386
6c26d08f 2387 if (new_writer) {
b37ad28b
AV		 2388 int err = get_write_access(inode);
2389 if (err)
2390 return nfserrno(err);
e518f056
BH		 2391 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2392 if (err)
2393 return nfserrno(err);
2394 file_take_write(filp);
6c26d08f 2395 }
1da177e4
LT		 2396 status = nfsd4_truncate(rqstp, cur_fh, open);
2397 if (status) {
6c26d08f
BF		 2398 if (new_writer)
2399 put_write_access(inode);
1da177e4
LT		 2400 return status;
2401 }
2402 /* remember the open */
6c26d08f 2403 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2404 __set_bit(open->op_share_access, &stp->st_access_bmap);
2405 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2406
2407 return nfs_ok;
2408}
2409
2410
1da177e4 2411static void
37515177 2412nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2413{
37515177
N		 2414 open->op_stateowner->so_confirmed = 1;
2415 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2416}
2417
2418/*
2419 * Attempt to hand out a delegation.
2420 */
2421static void
2422nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2423{
2424 struct nfs4_delegation *dp;
2425 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2426 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2427 struct file_lock fl, *flp = &fl;
2428 int status, flag = 0;
2429
2430 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2431 open->op_recall = 0;
2432 switch (open->op_claim_type) {
2433 case NFS4_OPEN_CLAIM_PREVIOUS:
2434 if (!atomic_read(&cb->cb_set))
2435 open->op_recall = 1;
2436 flag = open->op_delegate_type;
2437 if (flag == NFS4_OPEN_DELEGATE_NONE)
2438 goto out;
2439 break;
2440 case NFS4_OPEN_CLAIM_NULL:
2441 /* Let's not give out any delegations till everyone's
2442 * had the chance to reclaim theirs.... */
af558e33 2443 if (locks_in_grace())
7b190fec
N		 2444 goto out;
2445 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2446 goto out;
2447 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2448 flag = NFS4_OPEN_DELEGATE_WRITE;
2449 else
2450 flag = NFS4_OPEN_DELEGATE_READ;
2451 break;
2452 default:
2453 goto out;
2454 }
1da177e4
LT		 2455
2456 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2457 if (dp == NULL) {
2458 flag = NFS4_OPEN_DELEGATE_NONE;
2459 goto out;
2460 }
2461 locks_init_lock(&fl);
2462 fl.fl_lmops = &nfsd_lease_mng_ops;
2463 fl.fl_flags = FL_LEASE;
9167f501 2464 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2465 fl.fl_end = OFFSET_MAX;
2466 fl.fl_owner = (fl_owner_t)dp;
2467 fl.fl_file = stp->st_vfs_file;
2468 fl.fl_pid = current->tgid;
2469
a9933cea 2470 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2471 * the lock_manager callbacks fl_mylease and fl_change are used
2472 */
9167f501 2473 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2474 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2475 unhash_delegation(dp);
1da177e4
LT		 2476 flag = NFS4_OPEN_DELEGATE_NONE;
2477 goto out;
2478 }
2479
2480 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2481
2482 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2483 dp->dl_stateid.si_boot,
2484 dp->dl_stateid.si_stateownerid,
2485 dp->dl_stateid.si_fileid,
2486 dp->dl_stateid.si_generation);
2487out:
7b190fec
N		 2488 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2489 && flag == NFS4_OPEN_DELEGATE_NONE
2490 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2491 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2492 open->op_delegate_type = flag;
2493}
2494
2495/*
2496 * called with nfs4_lock_state() held.
2497 */
b37ad28b 2498__be32
1da177e4
LT		 2499nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2500{
6668958f 2501 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2502 struct nfs4_file *fp = NULL;
2503 struct inode *ino = current_fh->fh_dentry->d_inode;
2504 struct nfs4_stateid *stp = NULL;
567d9829 2505 struct nfs4_delegation *dp = NULL;
b37ad28b 2506 __be32 status;
1da177e4
LT		 2507
2508 status = nfserr_inval;
d87a8ade 2509 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2510 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2511 goto out;
2512 /*
2513 * Lookup file; if found, lookup stateid and check open request,
2514 * and check for delegations in the process of being recalled.
2515 * If not found, create the nfs4_file struct
2516 */
2517 fp = find_file(ino);
2518 if (fp) {
2519 if ((status = nfs4_check_open(fp, open, &stp)))
2520 goto out;
c44c5eeb
N		 2521 status = nfs4_check_deleg(fp, open, &dp);
2522 if (status)
2523 goto out;
1da177e4 2524 } else {
c44c5eeb
N		 2525 status = nfserr_bad_stateid;
2526 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2527 goto out;
1da177e4
LT		 2528 status = nfserr_resource;
2529 fp = alloc_init_file(ino);
2530 if (fp == NULL)
2531 goto out;
2532 }
2533
2534 /*
2535 * OPEN the file, or upgrade an existing OPEN.
2536 * If truncate fails, the OPEN fails.
2537 */
2538 if (stp) {
2539 /* Stateid was found, this is an OPEN upgrade */
2540 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2541 if (status)
2542 goto out;
444c2c07 2543 update_stateid(&stp->st_stateid);
1da177e4
LT		 2544 } else {
2545 /* Stateid was not found, this is a new OPEN */
2546 int flags = 0;
9ecb6a08 2547 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2548 flags |= NFSD_MAY_READ;
1da177e4 2549 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2550 flags |= NFSD_MAY_WRITE;
567d9829
N		 2551 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2552 if (status)
1da177e4
LT		 2553 goto out;
2554 init_stateid(stp, fp, open);
2555 status = nfsd4_truncate(rqstp, current_fh, open);
2556 if (status) {
2283963f 2557 release_open_stateid(stp);
1da177e4
LT		 2558 goto out;
2559 }
6668958f
AA		 2560 if (nfsd4_has_session(&resp->cstate))
2561 update_stateid(&stp->st_stateid);
1da177e4
LT		 2562 }
2563 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2564
6668958f
AA		 2565 if (nfsd4_has_session(&resp->cstate))
2566 open->op_stateowner->so_confirmed = 1;
2567
1da177e4
LT		 2568 /*
2569 * Attempt to hand out a delegation. No error return, because the
2570 * OPEN succeeds even if we fail.
2571 */
2572 nfs4_open_delegation(current_fh, open, stp);
2573
2574 status = nfs_ok;
2575
2576 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2577 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2578 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2579out:
13cd2184
N		 2580 if (fp)
2581 put_nfs4_file(fp);
37515177
N		 2582 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2583 nfs4_set_claim_prev(open);
1da177e4
LT		 2584 /*
2585 * To finish the open response, we just need to set the rflags.
2586 */
2587 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2588 if (!open->op_stateowner->so_confirmed &&
2589 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2590 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2591
2592 return status;
2593}
2594
b37ad28b 2595__be32
b591480b
BF		 2596nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2597 clientid_t *clid)
1da177e4
LT		 2598{
2599 struct nfs4_client *clp;
b37ad28b 2600 __be32 status;
1da177e4
LT		 2601
2602 nfs4_lock_state();
2603 dprintk("process_renew(%08x/%08x): starting\n",
2604 clid->cl_boot, clid->cl_id);
2605 status = nfserr_stale_clientid;
2606 if (STALE_CLIENTID(clid))
2607 goto out;
2608 clp = find_confirmed_client(clid);
2609 status = nfserr_expired;
2610 if (clp == NULL) {
2611 /* We assume the client took too long to RENEW. */
2612 dprintk("nfsd4_renew: clientid not found!\n");
2613 goto out;
2614 }
2615 renew_client(clp);
2616 status = nfserr_cb_path_down;
ea1da636 2617 if (!list_empty(&clp->cl_delegations)
c237dc03 2618 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2619 goto out;
2620 status = nfs_ok;
2621out:
2622 nfs4_unlock_state();
2623 return status;
2624}
2625
af558e33
BF		 2626struct lock_manager nfsd4_manager = {
2627};
2628
a76b4319 2629static void
af558e33 2630nfsd4_end_grace(void)
a76b4319
N		 2631{
2632 dprintk("NFSD: end of grace period\n");
c7b9a459 2633 nfsd4_recdir_purge_old();
af558e33 2634 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2635}
2636
fd39ca9a 2637static time_t
1da177e4
LT		 2638nfs4_laundromat(void)
2639{
2640 struct nfs4_client *clp;
2641 struct nfs4_stateowner *sop;
2642 struct nfs4_delegation *dp;
2643 struct list_head *pos, *next, reaplist;
2644 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2645 time_t t, clientid_val = NFSD_LEASE_TIME;
2646 time_t u, test_val = NFSD_LEASE_TIME;
2647
2648 nfs4_lock_state();
2649
2650 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2651 if (locks_in_grace())
2652 nfsd4_end_grace();
1da177e4
LT		 2653 list_for_each_safe(pos, next, &client_lru) {
2654 clp = list_entry(pos, struct nfs4_client, cl_lru);
2655 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2656 t = clp->cl_time - cutoff;
2657 if (clientid_val > t)
2658 clientid_val = t;
2659 break;
2660 }
2661 dprintk("NFSD: purging unused client (clientid %08x)\n",
2662 clp->cl_clientid.cl_id);
c7b9a459 2663 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2664 expire_client(clp);
2665 }
2666 INIT_LIST_HEAD(&reaplist);
2667 spin_lock(&recall_lock);
2668 list_for_each_safe(pos, next, &del_recall_lru) {
2669 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2670 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2671 u = dp->dl_time - cutoff;
2672 if (test_val > u)
2673 test_val = u;
2674 break;
2675 }
2676 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2677 dp, dp->dl_flock);
2678 list_move(&dp->dl_recall_lru, &reaplist);
2679 }
2680 spin_unlock(&recall_lock);
2681 list_for_each_safe(pos, next, &reaplist) {
2682 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2683 list_del_init(&dp->dl_recall_lru);
2684 unhash_delegation(dp);
2685 }
2686 test_val = NFSD_LEASE_TIME;
2687 list_for_each_safe(pos, next, &close_lru) {
2688 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2689 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2690 u = sop->so_time - cutoff;
2691 if (test_val > u)
2692 test_val = u;
2693 break;
2694 }
2695 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2696 sop->so_id);
f044ff83 2697 release_openowner(sop);
1da177e4
LT		 2698 }
2699 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2700 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2701 nfs4_unlock_state();
2702 return clientid_val;
2703}
2704
a254b246
HH		 2705static struct workqueue_struct *laundry_wq;
2706static void laundromat_main(struct work_struct *);
2707static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2708
2709static void
c4028958 2710laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2711{
2712 time_t t;
2713
2714 t = nfs4_laundromat();
2715 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2716 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2717}
2718
fd39ca9a 2719static struct nfs4_stateowner *
f8816512
N		 2720search_close_lru(u32 st_id, int flags)
2721{
1da177e4
LT		 2722 struct nfs4_stateowner *local = NULL;
2723
1da177e4
LT		 2724 if (flags & CLOSE_STATE) {
2725 list_for_each_entry(local, &close_lru, so_close_lru) {
2726 if (local->so_id == st_id)
2727 return local;
2728 }
2729 }
2730 return NULL;
2731}
2732
2733static inline int
2734nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2735{
7eaa36e2 2736 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2737}
2738
2739static int
2740STALE_STATEID(stateid_t *stateid)
2741{
78155ed7
BN		 2742 if (time_after((unsigned long)boot_time,
2743 (unsigned long)stateid->si_boot)) {
2744 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2745 stateid->si_boot, stateid->si_stateownerid,
2746 stateid->si_fileid, stateid->si_generation);
2747 return 1;
2748 }
2749 return 0;
2750}
2751
2752static int
2753EXPIRED_STATEID(stateid_t *stateid)
2754{
2755 if (time_before((unsigned long)boot_time,
2756 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2757 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2758 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2759 stateid->si_boot, stateid->si_stateownerid,
2760 stateid->si_fileid, stateid->si_generation);
2761 return 1;
2762 }
2763 return 0;
2764}
2765
2766static __be32
2767stateid_error_map(stateid_t *stateid)
2768{
2769 if (STALE_STATEID(stateid))
2770 return nfserr_stale_stateid;
2771 if (EXPIRED_STATEID(stateid))
2772 return nfserr_expired;
2773
2774 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2775 stateid->si_boot, stateid->si_stateownerid,
2776 stateid->si_fileid, stateid->si_generation);
2777 return nfserr_bad_stateid;
1da177e4
LT		 2778}
2779
2780static inline int
2781access_permit_read(unsigned long access_bmap)
2782{
2783 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2784 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2785 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2786}
2787
2788static inline int
2789access_permit_write(unsigned long access_bmap)
2790{
2791 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2792 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2793}
2794
2795static
b37ad28b 2796__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2797{
b37ad28b 2798 __be32 status = nfserr_openmode;
1da177e4
LT		 2799
2800 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2801 goto out;
2802 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2803 goto out;
2804 status = nfs_ok;
2805out:
2806 return status;
2807}
2808
b37ad28b 2809static inline __be32
1da177e4
LT		 2810check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2811{
203a8c8e 2812 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2813 return nfs_ok;
af558e33 2814 else if (locks_in_grace()) {
1da177e4
LT		 2815 /* Answer in remaining cases depends on existance of
2816 * conflicting state; so we must wait out the grace period. */
2817 return nfserr_grace;
2818 } else if (flags & WR_STATE)
2819 return nfs4_share_conflict(current_fh,
2820 NFS4_SHARE_DENY_WRITE);
2821 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2822 return nfs4_share_conflict(current_fh,
2823 NFS4_SHARE_DENY_READ);
2824}
2825
2826/*
2827 * Allow READ/WRITE during grace period on recovered state only for files
2828 * that are not able to provide mandatory locking.
2829 */
2830static inline int
18f82731 2831grace_disallows_io(struct inode *inode)
1da177e4 2832{
203a8c8e 2833 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2834}
2835
6668958f 2836static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2837{
6668958f
AA		 2838 /*
2839 * When sessions are used the stateid generation number is ignored
2840 * when it is zero.
2841 */
2842 if ((flags & HAS_SESSION) && in->si_generation == 0)
2843 goto out;
2844
0836f587
BF		 2845 /* If the client sends us a stateid from the future, it's buggy: */
2846 if (in->si_generation > ref->si_generation)
2847 return nfserr_bad_stateid;
2848 /*
2849 * The following, however, can happen. For example, if the
2850 * client sends an open and some IO at the same time, the open
2851 * may bump si_generation while the IO is still in flight.
2852 * Thanks to hard links and renames, the client never knows what
2853 * file an open will affect. So it could avoid that situation
2854 * only by serializing all opens and IO from the same open
2855 * owner. To recover from the old_stateid error, the client
2856 * will just have to retry the IO:
2857 */
2858 if (in->si_generation < ref->si_generation)
2859 return nfserr_old_stateid;
6668958f 2860out:
0836f587
BF		 2861 return nfs_ok;
2862}
2863
3e633079
BF		 2864static int is_delegation_stateid(stateid_t *stateid)
2865{
2866 return stateid->si_fileid == 0;
2867}
2868
1da177e4
LT		 2869/*
2870* Checks for stateid operations
2871*/
b37ad28b 2872__be32
dd453dfd
BH		 2873nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2874 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2875{
2876 struct nfs4_stateid *stp = NULL;
2877 struct nfs4_delegation *dp = NULL;
dd453dfd 2878 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2879 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2880 __be32 status;
1da177e4 2881
1da177e4
LT		 2882 if (filpp)
2883 *filpp = NULL;
2884
18f82731 2885 if (grace_disallows_io(ino))
1da177e4
LT		 2886 return nfserr_grace;
2887
6668958f
AA		 2888 if (nfsd4_has_session(cstate))
2889 flags |= HAS_SESSION;
2890
1da177e4
LT		 2891 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2892 return check_special_stateids(current_fh, stateid, flags);
2893
1da177e4
LT		 2894 status = nfserr_stale_stateid;
2895 if (STALE_STATEID(stateid))
2896 goto out;
2897
1da177e4 2898 status = nfserr_bad_stateid;
3e633079 2899 if (is_delegation_stateid(stateid)) {
a4455be0 2900 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2901 if (!dp) {
2902 status = stateid_error_map(stateid);
1da177e4 2903 goto out;
78155ed7 2904 }
6668958f
AA		 2905 status = check_stateid_generation(stateid, &dp->dl_stateid,
2906 flags);
0c2a498f
BF		 2907 if (status)
2908 goto out;
dc9bf700
BF		 2909 status = nfs4_check_delegmode(dp, flags);
2910 if (status)
2911 goto out;
2912 renew_client(dp->dl_client);
dc9bf700
BF		 2913 if (filpp)
2914 *filpp = dp->dl_vfs_file;
1da177e4 2915 } else { /* open or lock stateid */
a4455be0 2916 stp = find_stateid(stateid, flags);
78155ed7
BN		 2917 if (!stp) {
2918 status = stateid_error_map(stateid);
1da177e4 2919 goto out;
78155ed7 2920 }
6150ef0d 2921 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2922 goto out;
2923 if (!stp->st_stateowner->so_confirmed)
2924 goto out;
6668958f
AA		 2925 status = check_stateid_generation(stateid, &stp->st_stateid,
2926 flags);
0c2a498f
BF		 2927 if (status)
2928 goto out;
a4455be0
BF		 2929 status = nfs4_check_openmode(stp, flags);
2930 if (status)
1da177e4
LT		 2931 goto out;
2932 renew_client(stp->st_stateowner->so_client);
2933 if (filpp)
2934 *filpp = stp->st_vfs_file;
1da177e4
LT		 2935 }
2936 status = nfs_ok;
2937out:
2938 return status;
2939}
2940
4c4cd222
N		 2941static inline int
2942setlkflg (int type)
2943{
2944 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2945 RD_STATE : WR_STATE;
2946}
1da177e4
LT		 2947
2948/*
2949 * Checks for sequence id mutating operations.
2950 */
b37ad28b 2951static __be32
dd453dfd
BH		 2952nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2953 stateid_t *stateid, int flags,
2954 struct nfs4_stateowner **sopp,
2955 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2956{
1da177e4
LT		 2957 struct nfs4_stateid *stp;
2958 struct nfs4_stateowner *sop;
dd453dfd 2959 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2960 __be32 status;
1da177e4
LT		 2961
2962 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2963 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2964 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2965 stateid->si_generation);
3a4f98bb 2966
1da177e4
LT		 2967 *stpp = NULL;
2968 *sopp = NULL;
2969
1da177e4 2970 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2971 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2972 return nfserr_bad_stateid;
1da177e4
LT		 2973 }
2974
1da177e4 2975 if (STALE_STATEID(stateid))
3a4f98bb 2976 return nfserr_stale_stateid;
6668958f
AA		 2977
2978 if (nfsd4_has_session(cstate))
2979 flags |= HAS_SESSION;
2980
1da177e4
LT		 2981 /*
2982 * We return BAD_STATEID if filehandle doesn't match stateid,
2983 * the confirmed flag is incorrecly set, or the generation
2984 * number is incorrect.
1da177e4 2985 */
f8816512
N		 2986 stp = find_stateid(stateid, flags);
2987 if (stp == NULL) {
2988 /*
2989 * Also, we should make sure this isn't just the result of
2990 * a replayed close:
2991 */
2992 sop = search_close_lru(stateid->si_stateownerid, flags);
2993 if (sop == NULL)
78155ed7 2994 return stateid_error_map(stateid);
f8816512
N		 2995 *sopp = sop;
2996 goto check_replay;
2997 }
1da177e4 2998
39325bd0
BF		 2999 *stpp = stp;
3000 *sopp = sop = stp->st_stateowner;
3001
4c4cd222 3002 if (lock) {
4c4cd222 3003 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 3004 struct nfs4_client *clp = sop->so_client;
4c4cd222 3005 int lkflg = 0;
b37ad28b 3006 __be32 status;
4c4cd222
N		 3007
3008 lkflg = setlkflg(lock->lk_type);
3009
3010 if (lock->lk_is_new) {
599e0a22
BF		 3011 if (!sop->so_is_open_owner)
3012 return nfserr_bad_stateid;
60adfc50
AA		 3013 if (!(flags & HAS_SESSION) &&
3014 !same_clid(&clp->cl_clientid, lockclid))
3015 return nfserr_bad_stateid;
599e0a22
BF		 3016 /* stp is the open stateid */
3017 status = nfs4_check_openmode(stp, lkflg);
3018 if (status)
3019 return status;
3020 } else {
3021 /* stp is the lock stateid */
3022 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3023 if (status)
3024 return status;
4c4cd222 3025 }
1da177e4
LT		 3026 }
3027
f3362737 3028 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3029 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3030 return nfserr_bad_stateid;
1da177e4
LT		 3031 }
3032
1da177e4
LT		 3033 /*
3034 * We now validate the seqid and stateid generation numbers.
3035 * For the moment, we ignore the possibility of
3036 * generation number wraparound.
3037 */
6668958f 3038 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3039 goto check_replay;
3040
3a4f98bb 3041 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3042 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3043 " unconfirmed stateowner!\n");
3044 return nfserr_bad_stateid;
1da177e4 3045 }
3a4f98bb 3046 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3047 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3048 " confirmed yet!\n");
3049 return nfserr_bad_stateid;
1da177e4 3050 }
6668958f 3051 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3052 if (status)
3053 return status;
52fd004e 3054 renew_client(sop->so_client);
3a4f98bb 3055 return nfs_ok;
1da177e4 3056
1da177e4 3057check_replay:
bd9aac52 3058 if (seqid == sop->so_seqid - 1) {
849823c5 3059 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3060 /* indicate replay to calling function */
a90b061c 3061 return nfserr_replay_me;
1da177e4 3062 }
2fdada03 3063 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3064 sop->so_seqid, seqid);
3065 *sopp = NULL;
3066 return nfserr_bad_seqid;
1da177e4
LT		 3067}
3068
b37ad28b 3069__be32
ca364317 3070nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3071 struct nfsd4_open_confirm *oc)
1da177e4 3072{
b37ad28b 3073 __be32 status;
1da177e4
LT		 3074 struct nfs4_stateowner *sop;
3075 struct nfs4_stateid *stp;
3076
3077 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3078 (int)cstate->current_fh.fh_dentry->d_name.len,
3079 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3080
ca364317 3081 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3082 if (status)
3083 return status;
1da177e4
LT		 3084
3085 nfs4_lock_state();
3086
dd453dfd 3087 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3088 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3089 CONFIRM | OPEN_STATE,
1da177e4
LT		 3090 &oc->oc_stateowner, &stp, NULL)))
3091 goto out;
3092
3093 sop = oc->oc_stateowner;
3094 sop->so_confirmed = 1;
3095 update_stateid(&stp->st_stateid);
3096 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3097 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3098 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3099 stp->st_stateid.si_boot,
3100 stp->st_stateid.si_stateownerid,
3101 stp->st_stateid.si_fileid,
3102 stp->st_stateid.si_generation);
c7b9a459
N		 3103
3104 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3105out:
f2327d9a 3106 if (oc->oc_stateowner) {
1da177e4 3107 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3108 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3109 }
1da177e4
LT		 3110 nfs4_unlock_state();
3111 return status;
3112}
3113
3114
3115/*
3116 * unset all bits in union bitmap (bmap) that
3117 * do not exist in share (from successful OPEN_DOWNGRADE)
3118 */
3119static void
3120reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3121{
3122 int i;
3123 for (i = 1; i < 4; i++) {
3124 if ((i & access) != i)
3125 __clear_bit(i, bmap);
3126 }
3127}
3128
3129static void
3130reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3131{
3132 int i;
3133 for (i = 0; i < 4; i++) {
3134 if ((i & deny) != i)
3135 __clear_bit(i, bmap);
3136 }
3137}
3138
b37ad28b 3139__be32
ca364317
BF		 3140nfsd4_open_downgrade(struct svc_rqst *rqstp,
3141 struct nfsd4_compound_state *cstate,
a4f1706a 3142 struct nfsd4_open_downgrade *od)
1da177e4 3143{
b37ad28b 3144 __be32 status;
1da177e4
LT		 3145 struct nfs4_stateid *stp;
3146 unsigned int share_access;
3147
3148 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3149 (int)cstate->current_fh.fh_dentry->d_name.len,
3150 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3151
d87a8ade 3152 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3153 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3154 return nfserr_inval;
3155
3156 nfs4_lock_state();
dd453dfd 3157 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3158 od->od_seqid,
1da177e4 3159 &od->od_stateid,
f3362737 3160 OPEN_STATE,
1da177e4
LT		 3161 &od->od_stateowner, &stp, NULL)))
3162 goto out;
3163
3164 status = nfserr_inval;
3165 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3166 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3167 stp->st_access_bmap, od->od_share_access);
3168 goto out;
3169 }
3170 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3171 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3172 stp->st_deny_bmap, od->od_share_deny);
3173 goto out;
3174 }
3175 set_access(&share_access, stp->st_access_bmap);
3176 nfs4_file_downgrade(stp->st_vfs_file,
3177 share_access & ~od->od_share_access);
3178
3179 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3180 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3181
3182 update_stateid(&stp->st_stateid);
3183 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3184 status = nfs_ok;
3185out:
f2327d9a 3186 if (od->od_stateowner) {
1da177e4 3187 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3188 cstate->replay_owner = od->od_stateowner;
f2327d9a 3189 }
1da177e4
LT		 3190 nfs4_unlock_state();
3191 return status;
3192}
3193
3194/*
3195 * nfs4_unlock_state() called after encode
3196 */
b37ad28b 3197__be32
ca364317 3198nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3199 struct nfsd4_close *close)
1da177e4 3200{
b37ad28b 3201 __be32 status;
1da177e4
LT		 3202 struct nfs4_stateid *stp;
3203
3204 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3205 (int)cstate->current_fh.fh_dentry->d_name.len,
3206 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3207
3208 nfs4_lock_state();
3209 /* check close_lru for replay */
dd453dfd 3210 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3211 close->cl_seqid,
1da177e4 3212 &close->cl_stateid,
f3362737 3213 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3214 &close->cl_stateowner, &stp, NULL)))
3215 goto out;
1da177e4
LT		 3216 status = nfs_ok;
3217 update_stateid(&stp->st_stateid);
3218 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3219
04ef5954 3220 /* release_stateid() calls nfsd_close() if needed */
2283963f 3221 release_open_stateid(stp);
04ef5954
BF		 3222
3223 /* place unused nfs4_stateowners on so_close_lru list to be
3224 * released by the laundromat service after the lease period
3225 * to enable us to handle CLOSE replay
3226 */
3227 if (list_empty(&close->cl_stateowner->so_stateids))
3228 move_to_close_lru(close->cl_stateowner);
1da177e4 3229out:
f2327d9a 3230 if (close->cl_stateowner) {
1da177e4 3231 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3232 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3233 }
1da177e4
LT		 3234 nfs4_unlock_state();
3235 return status;
3236}
3237
b37ad28b 3238__be32
ca364317
BF		 3239nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3240 struct nfsd4_delegreturn *dr)
1da177e4 3241{
203a8c8e
BF		 3242 struct nfs4_delegation *dp;
3243 stateid_t *stateid = &dr->dr_stateid;
3244 struct inode *inode;
b37ad28b 3245 __be32 status;
6668958f 3246 int flags = 0;
1da177e4 3247
ca364317 3248 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3249 return status;
3250 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3251
6668958f
AA		 3252 if (nfsd4_has_session(cstate))
3253 flags |= HAS_SESSION;
1da177e4 3254 nfs4_lock_state();
203a8c8e
BF		 3255 status = nfserr_bad_stateid;
3256 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3257 goto out;
3258 status = nfserr_stale_stateid;
3259 if (STALE_STATEID(stateid))
3260 goto out;
7e0f7cf5 3261 status = nfserr_bad_stateid;
203a8c8e
BF		 3262 if (!is_delegation_stateid(stateid))
3263 goto out;
203a8c8e 3264 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3265 if (!dp) {
3266 status = stateid_error_map(stateid);
203a8c8e 3267 goto out;
78155ed7 3268 }
6668958f 3269 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3270 if (status)
3271 goto out;
3272 renew_client(dp->dl_client);
3273
3274 unhash_delegation(dp);
1da177e4 3275out:
203a8c8e
BF		 3276 nfs4_unlock_state();
3277
1da177e4
LT		 3278 return status;
3279}
3280
3281
3282/*
3283 * Lock owner state (byte-range locks)
3284 */
3285#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3286#define LOCK_HASH_BITS 8
3287#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3288#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3289
87df4de8
BH		 3290static inline u64
3291end_offset(u64 start, u64 len)
3292{
3293 u64 end;
3294
3295 end = start + len;
3296 return end >= start ? end: NFS4_MAX_UINT64;
3297}
3298
3299/* last octet in a range */
3300static inline u64
3301last_byte_offset(u64 start, u64 len)
3302{
3303 u64 end;
3304
3305 BUG_ON(!len);
3306 end = start + len;
3307 return end > start ? end - 1: NFS4_MAX_UINT64;
3308}
3309
1da177e4
LT		 3310#define lockownerid_hashval(id) \
3311 ((id) & LOCK_HASH_MASK)
3312
3313static inline unsigned int
3314lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3315 struct xdr_netobj *ownername)
3316{
3317 return (file_hashval(inode) + cl_id
3318 + opaque_hashval(ownername->data, ownername->len))
3319 & LOCK_HASH_MASK;
3320}
3321
3322static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3323static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3324static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3325
fd39ca9a 3326static struct nfs4_stateid *
1da177e4
LT		 3327find_stateid(stateid_t *stid, int flags)
3328{
9346eff0 3329 struct nfs4_stateid *local;
1da177e4
LT		 3330 u32 st_id = stid->si_stateownerid;
3331 u32 f_id = stid->si_fileid;
3332 unsigned int hashval;
3333
3334 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3335 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3336 hashval = stateid_hashval(st_id, f_id);
3337 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3338 if ((local->st_stateid.si_stateownerid == st_id) &&
3339 (local->st_stateid.si_fileid == f_id))
3340 return local;
3341 }
3342 }
9346eff0
KK		 3343
3344 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3345 hashval = stateid_hashval(st_id, f_id);
3346 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3347 if ((local->st_stateid.si_stateownerid == st_id) &&
3348 (local->st_stateid.si_fileid == f_id))
3349 return local;
3350 }
849823c5 3351 }
1da177e4
LT		 3352 return NULL;
3353}
3354
3355static struct nfs4_delegation *
3356find_delegation_stateid(struct inode *ino, stateid_t *stid)
3357{
13cd2184
N		 3358 struct nfs4_file *fp;
3359 struct nfs4_delegation *dl;
1da177e4
LT		 3360
3361 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3362 stid->si_boot, stid->si_stateownerid,
3363 stid->si_fileid, stid->si_generation);
3364
1da177e4 3365 fp = find_file(ino);
13cd2184
N		 3366 if (!fp)
3367 return NULL;
3368 dl = find_delegation_file(fp, stid);
3369 put_nfs4_file(fp);
3370 return dl;
1da177e4
LT		 3371}
3372
3373/*
3374 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3375 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3376 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3377 * locking, this prevents us from being completely protocol-compliant. The
3378 * real solution to this problem is to start using unsigned file offsets in
3379 * the VFS, but this is a very deep change!
3380 */
3381static inline void
3382nfs4_transform_lock_offset(struct file_lock *lock)
3383{
3384 if (lock->fl_start < 0)
3385 lock->fl_start = OFFSET_MAX;
3386 if (lock->fl_end < 0)
3387 lock->fl_end = OFFSET_MAX;
3388}
3389
d5b9026a
N		 3390/* Hack!: For now, we're defining this just so we can use a pointer to it
3391 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3392static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3393};
1da177e4
LT		 3394
3395static inline void
3396nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3397{
d5b9026a
N		 3398 struct nfs4_stateowner *sop;
3399 unsigned int hval;
1da177e4 3400
d5b9026a
N		 3401 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3402 sop = (struct nfs4_stateowner *) fl->fl_owner;
3403 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3404 kref_get(&sop->so_ref);
3405 deny->ld_sop = sop;
3406 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3407 } else {
3408 deny->ld_sop = NULL;
3409 deny->ld_clientid.cl_boot = 0;
3410 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3411 }
3412 deny->ld_start = fl->fl_start;
87df4de8
BH		 3413 deny->ld_length = NFS4_MAX_UINT64;
3414 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3415 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3416 deny->ld_type = NFS4_READ_LT;
3417 if (fl->fl_type != F_RDLCK)
3418 deny->ld_type = NFS4_WRITE_LT;
3419}
3420
1da177e4
LT		 3421static struct nfs4_stateowner *
3422find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3423 struct xdr_netobj *owner)
3424{
3425 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3426 struct nfs4_stateowner *op;
3427
3428 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3429 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3430 return op;
3431 }
3432 return NULL;
3433}
3434
3435/*
3436 * Alloc a lock owner structure.
3437 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3438 * occured.
3439 *
3440 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3441 */
3442
3443static struct nfs4_stateowner *
3444alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3445 struct nfs4_stateowner *sop;
3446 struct nfs4_replay *rp;
3447 unsigned int idhashval;
3448
3449 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3450 return NULL;
3451 idhashval = lockownerid_hashval(current_ownerid);
3452 INIT_LIST_HEAD(&sop->so_idhash);
3453 INIT_LIST_HEAD(&sop->so_strhash);
3454 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3455 INIT_LIST_HEAD(&sop->so_stateids);
3456 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3457 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3458 sop->so_time = 0;
3459 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3460 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3461 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3462 sop->so_is_open_owner = 0;
3463 sop->so_id = current_ownerid++;
3464 sop->so_client = clp;
b59e3c0e
NB		 3465 /* It is the openowner seqid that will be incremented in encode in the
3466 * case of new lockowners; so increment the lock seqid manually: */
3467 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3468 sop->so_confirmed = 1;
3469 rp = &sop->so_replay;
de1ae286 3470 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3471 rp->rp_buflen = 0;
3472 rp->rp_buf = rp->rp_ibuf;
3473 return sop;
3474}
3475
fd39ca9a 3476static struct nfs4_stateid *
1da177e4
LT		 3477alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3478{
3479 struct nfs4_stateid *stp;
3480 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3481
5ac049ac
N		 3482 stp = nfs4_alloc_stateid();
3483 if (stp == NULL)
1da177e4
LT		 3484 goto out;
3485 INIT_LIST_HEAD(&stp->st_hash);
3486 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3487 INIT_LIST_HEAD(&stp->st_perstateowner);
3488 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3489 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3490 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3491 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3492 stp->st_stateowner = sop;
13cd2184 3493 get_nfs4_file(fp);
1da177e4 3494 stp->st_file = fp;
78155ed7 3495 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3496 stp->st_stateid.si_stateownerid = sop->so_id;
3497 stp->st_stateid.si_fileid = fp->fi_id;
3498 stp->st_stateid.si_generation = 0;
3499 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3500 stp->st_access_bmap = open_stp->st_access_bmap;
3501 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3502 stp->st_openstp = open_stp;
1da177e4
LT		 3503
3504out:
3505 return stp;
3506}
3507
fd39ca9a 3508static int
1da177e4
LT		 3509check_lock_length(u64 offset, u64 length)
3510{
87df4de8 3511 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3512 LOFF_OVERFLOW(offset, length)));
3513}
3514
3515/*
3516 * LOCK operation
3517 */
b37ad28b 3518__be32
ca364317 3519nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3520 struct nfsd4_lock *lock)
1da177e4 3521{
3e9e3dbe 3522 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3523 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3524 struct nfs4_stateid *lock_stp;
3525 struct file *filp;
3526 struct file_lock file_lock;
8dc7c311 3527 struct file_lock conflock;
b37ad28b 3528 __be32 status = 0;
1da177e4 3529 unsigned int strhashval;
150b3934 3530 unsigned int cmd;
b8dd7b9a 3531 int err;
1da177e4
LT		 3532
3533 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3534 (long long) lock->lk_offset,
3535 (long long) lock->lk_length);
3536
1da177e4
LT		 3537 if (check_lock_length(lock->lk_offset, lock->lk_length))
3538 return nfserr_inval;
3539
ca364317 3540 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3541 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3542 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3543 return status;
3544 }
3545
1da177e4
LT		 3546 nfs4_lock_state();
3547
3548 if (lock->lk_is_new) {
893f8770
N		 3549 /*
3550 * Client indicates that this is a new lockowner.
3551 * Use open owner and open stateid to create lock owner and
3552 * lock stateid.
3553 */
1da177e4
LT		 3554 struct nfs4_stateid *open_stp = NULL;
3555 struct nfs4_file *fp;
3556
3557 status = nfserr_stale_clientid;
60adfc50
AA		 3558 if (!nfsd4_has_session(cstate) &&
3559 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3560 goto out;
1da177e4 3561
1da177e4 3562 /* validate and update open stateid and open seqid */
dd453dfd 3563 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3564 lock->lk_new_open_seqid,
3565 &lock->lk_new_open_stateid,
f3362737 3566 OPEN_STATE,
3a65588a 3567 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3568 lock);
37515177 3569 if (status)
1da177e4 3570 goto out;
3a65588a 3571 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3572 /* create lockowner and lock stateid */
3573 fp = open_stp->st_file;
3574 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3575 open_sop->so_client->cl_clientid.cl_id,
3576 &lock->v.new.owner);
3e9e3dbe
N		 3577 /* XXX: Do we need to check for duplicate stateowners on
3578 * the same file, or should they just be allowed (and
3579 * create new stateids)? */
1da177e4 3580 status = nfserr_resource;
b59e3c0e
NB		 3581 lock_sop = alloc_init_lock_stateowner(strhashval,
3582 open_sop->so_client, open_stp, lock);
3583 if (lock_sop == NULL)
1da177e4 3584 goto out;
b59e3c0e 3585 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3586 if (lock_stp == NULL)
1da177e4 3587 goto out;
1da177e4
LT		 3588 } else {
3589 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3590 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3591 lock->lk_old_lock_seqid,
3592 &lock->lk_old_lock_stateid,
f3362737 3593 LOCK_STATE,
3a65588a 3594 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3595 if (status)
3596 goto out;
3a65588a 3597 lock_sop = lock->lk_replay_owner;
1da177e4 3598 }
3a65588a 3599 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3600 filp = lock_stp->st_vfs_file;
3601
0dd395dc 3602 status = nfserr_grace;
af558e33 3603 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3604 goto out;
3605 status = nfserr_no_grace;
af558e33 3606 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3607 goto out;
3608
1da177e4
LT		 3609 locks_init_lock(&file_lock);
3610 switch (lock->lk_type) {
3611 case NFS4_READ_LT:
3612 case NFS4_READW_LT:
3613 file_lock.fl_type = F_RDLCK;
150b3934 3614 cmd = F_SETLK;
1da177e4
LT		 3615 break;
3616 case NFS4_WRITE_LT:
3617 case NFS4_WRITEW_LT:
3618 file_lock.fl_type = F_WRLCK;
150b3934 3619 cmd = F_SETLK;
1da177e4
LT		 3620 break;
3621 default:
3622 status = nfserr_inval;
3623 goto out;
3624 }
b59e3c0e 3625 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3626 file_lock.fl_pid = current->tgid;
3627 file_lock.fl_file = filp;
3628 file_lock.fl_flags = FL_POSIX;
d5b9026a 3629 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3630
3631 file_lock.fl_start = lock->lk_offset;
87df4de8 3632 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3633 nfs4_transform_lock_offset(&file_lock);
3634
3635 /*
3636 * Try to lock the file in the VFS.
3637 * Note: locks.c uses the BKL to protect the inode's lock list.
3638 */
3639
fd85b817 3640 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3641 switch (-err) {
1da177e4
LT		 3642 case 0: /* success! */
3643 update_stateid(&lock_stp->st_stateid);
3644 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3645 sizeof(stateid_t));
b8dd7b9a 3646 status = 0;
eb76b3fd
AA		 3647 break;
3648 case (EAGAIN): /* conflock holds conflicting lock */
3649 status = nfserr_denied;
3650 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3651 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3652 break;
1da177e4
LT		 3653 case (EDEADLK):
3654 status = nfserr_deadlock;
eb76b3fd 3655 break;
1da177e4 3656 default:
fd85b817 3657 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3658 status = nfserr_resource;
3659 break;
1da177e4 3660 }
1da177e4 3661out:
8a280510 3662 if (status && lock->lk_is_new && lock_sop)
f044ff83 3663 release_lockowner(lock_sop);
3a65588a
BF		 3664 if (lock->lk_replay_owner) {
3665 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3666 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3667 }
1da177e4
LT		 3668 nfs4_unlock_state();
3669 return status;
3670}
3671
55ef1274
BF		 3672/*
3673 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3674 * so we do a temporary open here just to get an open file to pass to
3675 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3676 * inode operation.)
3677 */
3678static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3679{
3680 struct file *file;
3681 int err;
3682
3683 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3684 if (err)
3685 return err;
3686 err = vfs_test_lock(file, lock);
3687 nfsd_close(file);
3688 return err;
3689}
3690
1da177e4
LT		 3691/*
3692 * LOCKT operation
3693 */
b37ad28b 3694__be32
ca364317
BF		 3695nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3696 struct nfsd4_lockt *lockt)
1da177e4
LT		 3697{
3698 struct inode *inode;
1da177e4 3699 struct file_lock file_lock;
fd85b817 3700 int error;
b37ad28b 3701 __be32 status;
1da177e4 3702
af558e33 3703 if (locks_in_grace())
1da177e4
LT		 3704 return nfserr_grace;
3705
3706 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3707 return nfserr_inval;
3708
3709 lockt->lt_stateowner = NULL;
3710 nfs4_lock_state();
3711
3712 status = nfserr_stale_clientid;
60adfc50 3713 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3714 goto out;
1da177e4 3715
ca364317 3716 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3717 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3718 if (status == nfserr_symlink)
3719 status = nfserr_inval;
3720 goto out;
3721 }
3722
ca364317 3723 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3724 locks_init_lock(&file_lock);
3725 switch (lockt->lt_type) {
3726 case NFS4_READ_LT:
3727 case NFS4_READW_LT:
3728 file_lock.fl_type = F_RDLCK;
3729 break;
3730 case NFS4_WRITE_LT:
3731 case NFS4_WRITEW_LT:
3732 file_lock.fl_type = F_WRLCK;
3733 break;
3734 default:
2fdada03 3735 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3736 status = nfserr_inval;
3737 goto out;
3738 }
3739
3740 lockt->lt_stateowner = find_lockstateowner_str(inode,
3741 &lockt->lt_clientid, &lockt->lt_owner);
3742 if (lockt->lt_stateowner)
3743 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3744 file_lock.fl_pid = current->tgid;
3745 file_lock.fl_flags = FL_POSIX;
3746
3747 file_lock.fl_start = lockt->lt_offset;
87df4de8 3748 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3749
3750 nfs4_transform_lock_offset(&file_lock);
3751
1da177e4 3752 status = nfs_ok;
55ef1274 3753 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3754 if (error) {
3755 status = nfserrno(error);
3756 goto out;
3757 }
9d6a8c5c 3758 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3759 status = nfserr_denied;
9d6a8c5c 3760 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3761 }
3762out:
3763 nfs4_unlock_state();
3764 return status;
3765}
3766
b37ad28b 3767__be32
ca364317 3768nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3769 struct nfsd4_locku *locku)
1da177e4
LT		 3770{
3771 struct nfs4_stateid *stp;
3772 struct file *filp = NULL;
3773 struct file_lock file_lock;
b37ad28b 3774 __be32 status;
b8dd7b9a 3775 int err;
1da177e4
LT		 3776
3777 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3778 (long long) locku->lu_offset,
3779 (long long) locku->lu_length);
3780
3781 if (check_lock_length(locku->lu_offset, locku->lu_length))
3782 return nfserr_inval;
3783
3784 nfs4_lock_state();
3785
dd453dfd 3786 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3787 locku->lu_seqid,
3788 &locku->lu_stateid,
f3362737 3789 LOCK_STATE,
1da177e4
LT		 3790 &locku->lu_stateowner, &stp, NULL)))
3791 goto out;
3792
3793 filp = stp->st_vfs_file;
3794 BUG_ON(!filp);
3795 locks_init_lock(&file_lock);
3796 file_lock.fl_type = F_UNLCK;
3797 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3798 file_lock.fl_pid = current->tgid;
3799 file_lock.fl_file = filp;
3800 file_lock.fl_flags = FL_POSIX;
d5b9026a 3801 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3802 file_lock.fl_start = locku->lu_offset;
3803
87df4de8 3804 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3805 nfs4_transform_lock_offset(&file_lock);
3806
3807 /*
3808 * Try to unlock the file in the VFS.
3809 */
fd85b817 3810 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3811 if (err) {
fd85b817 3812 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3813 goto out_nfserr;
3814 }
3815 /*
3816 * OK, unlock succeeded; the only thing left to do is update the stateid.
3817 */
3818 update_stateid(&stp->st_stateid);
3819 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3820
3821out:
f2327d9a 3822 if (locku->lu_stateowner) {
1da177e4 3823 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3824 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3825 }
1da177e4
LT		 3826 nfs4_unlock_state();
3827 return status;
3828
3829out_nfserr:
b8dd7b9a 3830 status = nfserrno(err);
1da177e4
LT		 3831 goto out;
3832}
3833
3834/*
3835 * returns
3836 * 1: locks held by lockowner
3837 * 0: no locks held by lockowner
3838 */
3839static int
3840check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3841{
3842 struct file_lock **flpp;
7eaa36e2 3843 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3844 int status = 0;
3845
3846 lock_kernel();
3847 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3848 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3849 status = 1;
3850 goto out;
796dadfd 3851 }
1da177e4
LT		 3852 }
3853out:
3854 unlock_kernel();
3855 return status;
3856}
3857
b37ad28b 3858__be32
b591480b
BF		 3859nfsd4_release_lockowner(struct svc_rqst *rqstp,
3860 struct nfsd4_compound_state *cstate,
3861 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3862{
3863 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3864 struct nfs4_stateowner *sop;
3865 struct nfs4_stateid *stp;
1da177e4 3866 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3867 struct list_head matches;
3868 int i;
b37ad28b 3869 __be32 status;
1da177e4
LT		 3870
3871 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3872 clid->cl_boot, clid->cl_id);
3873
3874 /* XXX check for lease expiration */
3875
3876 status = nfserr_stale_clientid;
849823c5 3877 if (STALE_CLIENTID(clid))
1da177e4 3878 return status;
1da177e4
LT		 3879
3880 nfs4_lock_state();
3881
3e9e3dbe
N		 3882 status = nfserr_locks_held;
3883 /* XXX: we're doing a linear search through all the lockowners.
3884 * Yipes! For now we'll just hope clients aren't really using
3885 * release_lockowner much, but eventually we have to fix these
3886 * data structures. */
3887 INIT_LIST_HEAD(&matches);
3888 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3889 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3890 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3891 continue;
3892 list_for_each_entry(stp, &sop->so_stateids,
3893 st_perstateowner) {
3894 if (check_for_locks(stp->st_vfs_file, sop))
3895 goto out;
3896 /* Note: so_perclient unused for lockowners,
3897 * so it's OK to fool with here. */
3898 list_add(&sop->so_perclient, &matches);
3899 }
1da177e4 3900 }
3e9e3dbe
N		 3901 }
3902 /* Clients probably won't expect us to return with some (but not all)
3903 * of the lockowner state released; so don't release any until all
3904 * have been checked. */
3905 status = nfs_ok;
0fa822e4
N		 3906 while (!list_empty(&matches)) {
3907 sop = list_entry(matches.next, struct nfs4_stateowner,
3908 so_perclient);
3909 /* unhash_stateowner deletes so_perclient only
3910 * for openowners. */
3911 list_del(&sop->so_perclient);
f044ff83 3912 release_lockowner(sop);
1da177e4
LT		 3913 }
3914out:
3915 nfs4_unlock_state();
3916 return status;
3917}
3918
3919static inline struct nfs4_client_reclaim *
a55370a3 3920alloc_reclaim(void)
1da177e4 3921{
a55370a3 3922 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3923}
3924
c7b9a459 3925int
a1bcecd2 3926nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3927{
3928 unsigned int strhashval = clientstr_hashval(name);
3929 struct nfs4_client *clp;
3930
a1bcecd2 3931 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3932 return clp ? 1 : 0;
3933}
3934
1da177e4
LT		 3935/*
3936 * failure => all reset bets are off, nfserr_no_grace...
3937 */
190e4fbf
N		 3938int
3939nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3940{
3941 unsigned int strhashval;
3942 struct nfs4_client_reclaim *crp = NULL;
3943
a55370a3
N		 3944 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3945 crp = alloc_reclaim();
1da177e4
LT		 3946 if (!crp)
3947 return 0;
a55370a3 3948 strhashval = clientstr_hashval(name);
1da177e4
LT		 3949 INIT_LIST_HEAD(&crp->cr_strhash);
3950 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3951 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3952 reclaim_str_hashtbl_size++;
3953 return 1;
3954}
3955
3956static void
3957nfs4_release_reclaim(void)
3958{
3959 struct nfs4_client_reclaim *crp = NULL;
3960 int i;
3961
1da177e4
LT		 3962 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3963 while (!list_empty(&reclaim_str_hashtbl[i])) {
3964 crp = list_entry(reclaim_str_hashtbl[i].next,
3965 struct nfs4_client_reclaim, cr_strhash);
3966 list_del(&crp->cr_strhash);
1da177e4
LT		 3967 kfree(crp);
3968 reclaim_str_hashtbl_size--;
3969 }
3970 }
3971 BUG_ON(reclaim_str_hashtbl_size);
3972}
3973
3974/*
3975 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3976static struct nfs4_client_reclaim *
1da177e4
LT		 3977nfs4_find_reclaim_client(clientid_t *clid)
3978{
3979 unsigned int strhashval;
3980 struct nfs4_client *clp;
3981 struct nfs4_client_reclaim *crp = NULL;
3982
3983
3984 /* find clientid in conf_id_hashtbl */
3985 clp = find_confirmed_client(clid);
3986 if (clp == NULL)
3987 return NULL;
3988
a55370a3
N		 3989 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3990 clp->cl_name.len, clp->cl_name.data,
3991 clp->cl_recdir);
1da177e4
LT		 3992
3993 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3994 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3995 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3996 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3997 return crp;
3998 }
3999 }
4000 return NULL;
4001}
4002
4003/*
4004* Called from OPEN. Look for clientid in reclaim list.
4005*/
b37ad28b 4006__be32
1da177e4
LT		 4007nfs4_check_open_reclaim(clientid_t *clid)
4008{
dfc83565 4009 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 4010}
4011
ac4d8ff2 4012/* initialization to perform at module load time: */
1da177e4 4013
e8ff2a84 4014int
ac4d8ff2 4015nfs4_state_init(void)
1da177e4 4016{
e8ff2a84 4017 int i, status;
1da177e4 4018
e8ff2a84
BF		 4019 status = nfsd4_init_slabs();
4020 if (status)
4021 return status;
1da177e4
LT		 4022 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4023 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4024 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4025 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4026 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4027 }
5282fd72
ME		 4028 for (i = 0; i < SESSION_HASH_SIZE; i++)
4029 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4030 for (i = 0; i < FILE_HASH_SIZE; i++) {
4031 INIT_LIST_HEAD(&file_hashtbl[i]);
4032 }
4033 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4034 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4035 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4036 }
4037 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4038 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4039 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4040 }
4041 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4042 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4043 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4044 }
1da177e4 4045 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4046 INIT_LIST_HEAD(&close_lru);
4047 INIT_LIST_HEAD(&client_lru);
4048 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 4049 for (i = 0; i < CLIENT_HASH_SIZE; i++)
4050 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4051 reclaim_str_hashtbl_size = 0;
e8ff2a84 4052 return 0;
ac4d8ff2
N		 4053}
4054
190e4fbf
N		 4055static void
4056nfsd4_load_reboot_recovery_data(void)
4057{
4058 int status;
4059
0964a3d3
N		 4060 nfs4_lock_state();
4061 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4062 status = nfsd4_recdir_load();
0964a3d3 4063 nfs4_unlock_state();
190e4fbf
N		 4064 if (status)
4065 printk("NFSD: Failure reading reboot recovery data\n");
4066}
4067
9a8db97e
ME		 4068unsigned long
4069get_nfs4_grace_period(void)
4070{
4071 return max(user_lease_time, lease_time) * HZ;
4072}
4073
c2f1a551
MS		 4074/*
4075 * Since the lifetime of a delegation isn't limited to that of an open, a
4076 * client may quite reasonably hang on to a delegation as long as it has
4077 * the inode cached. This becomes an obvious problem the first time a
4078 * client's inode cache approaches the size of the server's total memory.
4079 *
4080 * For now we avoid this problem by imposing a hard limit on the number
4081 * of delegations, which varies according to the server's memory size.
4082 */
4083static void
4084set_max_delegations(void)
4085{
4086 /*
4087 * Allow at most 4 delegations per megabyte of RAM. Quick
4088 * estimates suggest that in the worst case (where every delegation
4089 * is for a different inode), a delegation could take about 1.5K,
4090 * giving a worst case usage of about 6% of memory.
4091 */
4092 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4093}
4094
ac4d8ff2
N		 4095/* initialization to perform when the nfsd service is started: */
4096
4097static void
4098__nfs4_state_start(void)
4099{
9a8db97e 4100 unsigned long grace_time;
ac4d8ff2 4101
1da177e4 4102 boot_time = get_seconds();
af558e33 4103 grace_time = get_nfs4_grace_period();
d99a05ad 4104 lease_time = user_lease_time;
af558e33 4105 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4106 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4107 grace_time/HZ);
58da282b 4108 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4109 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4110 set_max_delegations();
1da177e4
LT		 4111}
4112
e8ff2a84 4113void
76a3550e 4114nfs4_state_start(void)
1da177e4 4115{
1da177e4 4116 if (nfs4_init)
e8ff2a84 4117 return;
190e4fbf 4118 nfsd4_load_reboot_recovery_data();
76a3550e 4119 __nfs4_state_start();
1da177e4 4120 nfs4_init = 1;
e8ff2a84 4121 return;
1da177e4
LT		 4122}
4123
1da177e4
LT		 4124time_t
4125nfs4_lease_time(void)
4126{
4127 return lease_time;
4128}
4129
4130static void
4131__nfs4_state_shutdown(void)
4132{
4133 int i;
4134 struct nfs4_client *clp = NULL;
4135 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4136 struct list_head *pos, *next, reaplist;
4137
1da177e4
LT		 4138 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4139 while (!list_empty(&conf_id_hashtbl[i])) {
4140 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4141 expire_client(clp);
4142 }
4143 while (!list_empty(&unconf_str_hashtbl[i])) {
4144 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4145 expire_client(clp);
4146 }
4147 }
4148 INIT_LIST_HEAD(&reaplist);
4149 spin_lock(&recall_lock);
4150 list_for_each_safe(pos, next, &del_recall_lru) {
4151 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4152 list_move(&dp->dl_recall_lru, &reaplist);
4153 }
4154 spin_unlock(&recall_lock);
4155 list_for_each_safe(pos, next, &reaplist) {
4156 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4157 list_del_init(&dp->dl_recall_lru);
4158 unhash_delegation(dp);
4159 }
4160
190e4fbf 4161 nfsd4_shutdown_recdir();
1da177e4 4162 nfs4_init = 0;
1da177e4
LT		 4163}
4164
4165void
4166nfs4_state_shutdown(void)
4167{
5e8d5c29
N		 4168 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4169 destroy_workqueue(laundry_wq);
2c5e7615 4170 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4171 nfs4_lock_state();
4172 nfs4_release_reclaim();
4173 __nfs4_state_shutdown();
1da177e4
LT		 4174 nfs4_unlock_state();
4175}
4176
3dd98a3b
JL		 4177/*
4178 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4179 * accessed when nfsd is starting.
4180 */
0964a3d3
N		 4181static void
4182nfs4_set_recdir(char *recdir)
4183{
0964a3d3 4184 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4185}
4186
4187/*
4188 * Change the NFSv4 recovery directory to recdir.
4189 */
4190int
4191nfs4_reset_recoverydir(char *recdir)
4192{
4193 int status;
a63bb996 4194 struct path path;
0964a3d3 4195
a63bb996 4196 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4197 if (status)
4198 return status;
4199 status = -ENOTDIR;
a63bb996 4200 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4201 nfs4_set_recdir(recdir);
4202 status = 0;
4203 }
a63bb996 4204 path_put(&path);
0964a3d3
N		 4205 return status;
4206}
4207
3dd98a3b
JL		 4208char *
4209nfs4_recoverydir(void)
4210{
4211 return user_recovery_dirname;
4212}
4213
1da177e4
LT		 4214/*
4215 * Called when leasetime is changed.
4216 *
d99a05ad
N		 4217 * The only way the protocol gives us to handle on-the-fly lease changes is to
4218 * simulate a reboot. Instead of doing that, we just wait till the next time
4219 * we start to register any changes in lease time. If the administrator
4220 * really wants to change the lease time *now*, they can go ahead and bring
4221 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4222 *
4223 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4224 * when nfsd is starting
1da177e4
LT		 4225 */
4226void
4227nfs4_reset_lease(time_t leasetime)
4228{
d99a05ad 4229 user_lease_time = leasetime;
1da177e4 4230}
Linux 6.x block layer and io_uring tree(s)