nfsd: track last inode only in use_wgather case
[deliverable/linux.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);
582}
583
1da177e4
LT
584static inline void
585renew_client(struct nfs4_client *clp)
586{
587 /*
588 * Move client to the end to the LRU list.
589 */
590 dprintk("renewing client (clientid %08x/%08x)\n",
591 clp->cl_clientid.cl_boot,
592 clp->cl_clientid.cl_id);
593 list_move_tail(&clp->cl_lru, &client_lru);
594 clp->cl_time = get_seconds();
595}
596
597/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
598static int
599STALE_CLIENTID(clientid_t *clid)
600{
601 if (clid->cl_boot == boot_time)
602 return 0;
60adfc50
AA
603 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
604 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT
605 return 1;
606}
607
608/*
609 * XXX Should we use a slab cache ?
610 * This type of memory management is somewhat inefficient, but we use it
611 * anyway since SETCLIENTID is not a common operation.
612 */
35bba9a3 613static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT
614{
615 struct nfs4_client *clp;
616
35bba9a3
BF
617 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
618 if (clp == NULL)
619 return NULL;
620 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
621 if (clp->cl_name.data == NULL) {
622 kfree(clp);
623 return NULL;
1da177e4 624 }
35bba9a3
BF
625 memcpy(clp->cl_name.data, name.data, name.len);
626 clp->cl_name.len = name.len;
1da177e4
LT
627 return clp;
628}
629
1b1a9b31
BF
630static void
631shutdown_callback_client(struct nfs4_client *clp)
632{
c237dc03 633 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 634
1b1a9b31 635 if (clnt) {
404ec117
BF
636 /*
637 * Callback threads take a reference on the client, so there
638 * should be no outstanding callbacks at this point.
639 */
c237dc03 640 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF
641 rpc_shutdown_client(clnt);
642 }
c237dc03
BF
643 if (clp->cl_cb_conn.cb_cred) {
644 put_rpccred(clp->cl_cb_conn.cb_cred);
645 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 646 }
1b1a9b31
BF
647}
648
1da177e4
LT
649static inline void
650free_client(struct nfs4_client *clp)
651{
1b1a9b31 652 shutdown_callback_client(clp);
38eb76a5
AA
653 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
654 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT
655 if (clp->cl_cred.cr_group_info)
656 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 657 kfree(clp->cl_principal);
1da177e4
LT
658 kfree(clp->cl_name.data);
659 kfree(clp);
660}
661
662void
663put_nfs4_client(struct nfs4_client *clp)
664{
665 if (atomic_dec_and_test(&clp->cl_count))
666 free_client(clp);
667}
668
669static void
670expire_client(struct nfs4_client *clp)
671{
672 struct nfs4_stateowner *sop;
673 struct nfs4_delegation *dp;
1da177e4
LT
674 struct list_head reaplist;
675
676 dprintk("NFSD: expire_client cl_count %d\n",
677 atomic_read(&clp->cl_count));
678
1da177e4
LT
679 INIT_LIST_HEAD(&reaplist);
680 spin_lock(&recall_lock);
ea1da636
N
681 while (!list_empty(&clp->cl_delegations)) {
682 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT
683 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
684 dp->dl_flock);
ea1da636 685 list_del_init(&dp->dl_perclnt);
1da177e4
LT
686 list_move(&dp->dl_recall_lru, &reaplist);
687 }
688 spin_unlock(&recall_lock);
689 while (!list_empty(&reaplist)) {
690 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
691 list_del_init(&dp->dl_recall_lru);
692 unhash_delegation(dp);
693 }
694 list_del(&clp->cl_idhash);
695 list_del(&clp->cl_strhash);
696 list_del(&clp->cl_lru);
ea1da636
N
697 while (!list_empty(&clp->cl_openowners)) {
698 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 699 release_openowner(sop);
1da177e4 700 }
c4bf7868
ME
701 while (!list_empty(&clp->cl_sessions)) {
702 struct nfsd4_session *ses;
703 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
704 se_perclnt);
705 release_session(ses);
706 }
1da177e4
LT
707 put_nfs4_client(clp);
708}
709
35bba9a3
BF
710static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
711{
1da177e4
LT
712 struct nfs4_client *clp;
713
35bba9a3
BF
714 clp = alloc_client(name);
715 if (clp == NULL)
716 return NULL;
a55370a3 717 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 718 atomic_set(&clp->cl_count, 1);
c237dc03 719 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT
720 INIT_LIST_HEAD(&clp->cl_idhash);
721 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N
722 INIT_LIST_HEAD(&clp->cl_openowners);
723 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 724 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 725 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT
726 return clp;
727}
728
35bba9a3
BF
729static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
730{
731 memcpy(target->cl_verifier.data, source->data,
732 sizeof(target->cl_verifier.data));
1da177e4
LT
733}
734
35bba9a3
BF
735static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
736{
1da177e4
LT
737 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
738 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
739}
740
35bba9a3
BF
741static void copy_cred(struct svc_cred *target, struct svc_cred *source)
742{
1da177e4
LT
743 target->cr_uid = source->cr_uid;
744 target->cr_gid = source->cr_gid;
745 target->cr_group_info = source->cr_group_info;
746 get_group_info(target->cr_group_info);
747}
748
35bba9a3 749static int same_name(const char *n1, const char *n2)
599e0a22 750{
a55370a3 751 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT
752}
753
754static int
599e0a22
BF
755same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
756{
757 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT
758}
759
760static int
599e0a22
BF
761same_clid(clientid_t *cl1, clientid_t *cl2)
762{
763 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT
764}
765
766/* XXX what about NGROUP */
767static int
599e0a22
BF
768same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
769{
770 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT
771}
772
5ec7b46c
BF
773static void gen_clid(struct nfs4_client *clp)
774{
775 static u32 current_clientid = 1;
776
1da177e4
LT
777 clp->cl_clientid.cl_boot = boot_time;
778 clp->cl_clientid.cl_id = current_clientid++;
779}
780
deda2faa
BF
781static void gen_confirm(struct nfs4_client *clp)
782{
783 static u32 i;
784 u32 *p;
1da177e4 785
1da177e4 786 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF
787 *p++ = get_seconds();
788 *p++ = i++;
1da177e4
LT
789}
790
35bba9a3
BF
791static int check_name(struct xdr_netobj name)
792{
1da177e4
LT
793 if (name.len == 0)
794 return 0;
795 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 796 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT
797 return 0;
798 }
799 return 1;
800}
801
fd39ca9a 802static void
1da177e4
LT
803add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
804{
805 unsigned int idhashval;
806
807 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
808 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
809 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
810 list_add_tail(&clp->cl_lru, &client_lru);
811 clp->cl_time = get_seconds();
812}
813
fd39ca9a 814static void
1da177e4
LT
815move_to_confirmed(struct nfs4_client *clp)
816{
817 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
818 unsigned int strhashval;
819
820 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
821 list_del_init(&clp->cl_strhash);
f116629d 822 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 823 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT
824 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
825 renew_client(clp);
826}
827
828static struct nfs4_client *
829find_confirmed_client(clientid_t *clid)
830{
831 struct nfs4_client *clp;
832 unsigned int idhashval = clientid_hashval(clid->cl_id);
833
834 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 835 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
836 return clp;
837 }
838 return NULL;
839}
840
841static struct nfs4_client *
842find_unconfirmed_client(clientid_t *clid)
843{
844 struct nfs4_client *clp;
845 unsigned int idhashval = clientid_hashval(clid->cl_id);
846
847 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 848 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
849 return clp;
850 }
851 return NULL;
852}
853
a1bcecd2
AA
854/*
855 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
856 * parameter. Matching is based on the fact the at least one of the
857 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
858 *
859 * FIXME: we need to unify the clientid namespaces for nfsv4.x
860 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
861 * and SET_CLIENTID{,_CONFIRM}
862 */
863static inline int
864match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
865{
866 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
867 return use_exchange_id == has_exchange_flags;
868}
869
28ce6054 870static struct nfs4_client *
a1bcecd2
AA
871find_confirmed_client_by_str(const char *dname, unsigned int hashval,
872 bool use_exchange_id)
28ce6054
N
873{
874 struct nfs4_client *clp;
875
876 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
877 if (same_name(clp->cl_recdir, dname) &&
878 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
879 return clp;
880 }
881 return NULL;
882}
883
884static struct nfs4_client *
a1bcecd2
AA
885find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
886 bool use_exchange_id)
28ce6054
N
887{
888 struct nfs4_client *clp;
889
890 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
891 if (same_name(clp->cl_recdir, dname) &&
892 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
893 return clp;
894 }
895 return NULL;
896}
897
1da177e4
LT
898/* a helper function for parse_callback */
899static int
900parse_octet(unsigned int *lenp, char **addrp)
901{
902 unsigned int len = *lenp;
903 char *p = *addrp;
904 int n = -1;
905 char c;
906
907 for (;;) {
908 if (!len)
909 break;
910 len--;
911 c = *p++;
912 if (c == '.')
913 break;
914 if ((c < '0') || (c > '9')) {
915 n = -1;
916 break;
917 }
918 if (n < 0)
919 n = 0;
920 n = (n * 10) + (c - '0');
921 if (n > 255) {
922 n = -1;
923 break;
924 }
925 }
926 *lenp = len;
927 *addrp = p;
928 return n;
929}
930
931/* parse and set the setclientid ipv4 callback address */
fd39ca9a 932static int
1da177e4
LT
933parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
934{
935 int temp = 0;
936 u32 cbaddr = 0;
937 u16 cbport = 0;
938 u32 addrlen = addr_len;
939 char *addr = addr_val;
940 int i, shift;
941
942 /* ipaddress */
943 shift = 24;
944 for(i = 4; i > 0 ; i--) {
945 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
946 return 0;
947 }
948 cbaddr |= (temp << shift);
949 if (shift > 0)
950 shift -= 8;
951 }
952 *cbaddrp = cbaddr;
953
954 /* port */
955 shift = 8;
956 for(i = 2; i > 0 ; i--) {
957 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
958 return 0;
959 }
960 cbport |= (temp << shift);
961 if (shift > 0)
962 shift -= 8;
963 }
964 *cbportp = cbport;
965 return 1;
966}
967
fd39ca9a 968static void
1da177e4
LT
969gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
970{
c237dc03 971 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT
972
973 /* Currently, we only support tcp for the callback channel */
974 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
975 goto out_err;
976
977 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
978 &cb->cb_addr, &cb->cb_port)))
979 goto out_err;
980 cb->cb_prog = se->se_callback_prog;
981 cb->cb_ident = se->se_callback_ident;
1da177e4
LT
982 return;
983out_err:
849823c5 984 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT
985 "will not receive delegations\n",
986 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
987
1da177e4
LT
988 return;
989}
990
074fe897
AA
991void
992nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
993{
994 struct nfsd4_compoundres *resp = rqstp->rq_resp;
995
996 resp->cstate.statp = statp;
997}
998
999/*
1000 * Dereference the result pages.
1001 */
1002static void
1003nfsd4_release_respages(struct page **respages, short resused)
1004{
1005 int i;
1006
1007 dprintk("--> %s\n", __func__);
1008 for (i = 0; i < resused; i++) {
1009 if (!respages[i])
1010 continue;
1011 put_page(respages[i]);
1012 respages[i] = NULL;
1013 }
1014}
1015
1016static void
1017nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1018{
1019 int i;
1020
1021 for (i = 0; i < count; i++) {
1022 topages[i] = frompages[i];
1023 if (!topages[i])
1024 continue;
1025 get_page(topages[i]);
1026 }
1027}
1028
1029/*
1030 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1031 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1032 * length of the XDR response is less than se_fmaxresp_cached
1033 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1034 * of the reply (e.g. readdir).
1035 *
1036 * Store the base and length of the rq_req.head[0] page
1037 * of the NFSv4.1 data, just past the rpc header.
1038 */
1039void
1040nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1041{
1042 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1043 struct svc_rqst *rqstp = resp->rqstp;
1044 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1045 struct nfsd4_op *op = &args->ops[resp->opcnt];
1046 struct kvec *resv = &rqstp->rq_res.head[0];
1047
1048 dprintk("--> %s entry %p\n", __func__, entry);
1049
1050 /* Don't cache a failed OP_SEQUENCE. */
1051 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1052 return;
bf864a31 1053
074fe897 1054 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA
1055 entry->ce_opcnt = resp->opcnt;
1056 entry->ce_status = resp->cstate.status;
1057
1058 /*
1059 * Don't need a page to cache just the sequence operation - the slot
1060 * does this for us!
1061 */
1062
1063 if (nfsd4_not_cached(resp)) {
1064 entry->ce_resused = 0;
1065 entry->ce_rpchdrlen = 0;
1066 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1067 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1068 return;
1069 }
074fe897
AA
1070 entry->ce_resused = rqstp->rq_resused;
1071 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1072 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1073 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1074 entry->ce_resused);
074fe897
AA
1075 entry->ce_datav.iov_base = resp->cstate.statp;
1076 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1077 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA
1078 /* Current request rpc header length*/
1079 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1080 (char *)page_address(rqstp->rq_respages[0]);
1081}
1082
1083/*
1084 * We keep the rpc header, but take the nfs reply from the replycache.
1085 */
1086static int
1087nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1088 struct nfsd4_cache_entry *entry)
1089{
1090 struct svc_rqst *rqstp = resp->rqstp;
1091 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1092 int len;
1093
1094 /* Current request rpc header length*/
1095 len = (char *)resp->cstate.statp -
1096 (char *)page_address(rqstp->rq_respages[0]);
1097 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1098 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1099 entry->ce_datav.iov_len);
1100 return 0;
1101 }
1102 /* copy the cached reply nfsd data past the current rpc header */
1103 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1104 entry->ce_datav.iov_len);
1105 resv->iov_len = len + entry->ce_datav.iov_len;
1106 return 1;
1107}
1108
1109/*
1110 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1111 * cached page. Replace any futher replay pages from the cache.
1112 */
1113__be32
bf864a31
AA
1114nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1115 struct nfsd4_sequence *seq)
074fe897
AA
1116{
1117 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1118 __be32 status;
1119
1120 dprintk("--> %s entry %p\n", __func__, entry);
1121
bf864a31
AA
1122 /*
1123 * If this is just the sequence operation, we did not keep
1124 * a page in the cache entry because we can just use the
1125 * slot info stored in struct nfsd4_sequence that was checked
1126 * against the slot in nfsd4_sequence().
1127 *
1128 * This occurs when seq->cachethis is FALSE, or when the client
1129 * session inactivity timer fires and a solo sequence operation
1130 * is sent (lease renewal).
1131 */
1132 if (seq && nfsd4_not_cached(resp)) {
1133 seq->maxslots = resp->cstate.session->se_fnumslots;
1134 return nfs_ok;
1135 }
074fe897
AA
1136
1137 if (!nfsd41_copy_replay_data(resp, entry)) {
1138 /*
1139 * Not enough room to use the replay rpc header, send the
1140 * cached header. Release all the allocated result pages.
1141 */
1142 svc_free_res_pages(resp->rqstp);
1143 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1144 entry->ce_resused);
1145 } else {
1146 /* Release all but the first allocated result page */
1147
1148 resp->rqstp->rq_resused--;
1149 svc_free_res_pages(resp->rqstp);
1150
1151 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1152 &entry->ce_respages[1],
1153 entry->ce_resused - 1);
1154 }
1155
1156 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA
1157 resp->opcnt = entry->ce_opcnt;
1158 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA
1159 status = entry->ce_status;
1160
1161 return status;
1162}
1163
0733d213
AA
1164/*
1165 * Set the exchange_id flags returned by the server.
1166 */
1167static void
1168nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1169{
1170 /* pNFS is not supported */
1171 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1172
1173 /* Referrals are supported, Migration is not. */
1174 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1175
1176 /* set the wire flags to return to client. */
1177 clid->flags = new->cl_exchange_flags;
1178}
1179
069b6ad4
AA
1180__be32
1181nfsd4_exchange_id(struct svc_rqst *rqstp,
1182 struct nfsd4_compound_state *cstate,
1183 struct nfsd4_exchange_id *exid)
1184{
0733d213
AA
1185 struct nfs4_client *unconf, *conf, *new;
1186 int status;
1187 unsigned int strhashval;
1188 char dname[HEXDIR_LEN];
1189 nfs4_verifier verf = exid->verifier;
1190 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1191
1192 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1193 " ip_addr=%u flags %x, spa_how %d\n",
1194 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1195 ip_addr, exid->flags, exid->spa_how);
1196
1197 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1198 return nfserr_inval;
1199
1200 /* Currently only support SP4_NONE */
1201 switch (exid->spa_how) {
1202 case SP4_NONE:
1203 break;
1204 case SP4_SSV:
1205 return nfserr_encr_alg_unsupp;
1206 default:
1207 BUG(); /* checked by xdr code */
1208 case SP4_MACH_CRED:
1209 return nfserr_serverfault; /* no excuse :-/ */
1210 }
1211
1212 status = nfs4_make_rec_clidname(dname, &exid->clname);
1213
1214 if (status)
1215 goto error;
1216
1217 strhashval = clientstr_hashval(dname);
1218
1219 nfs4_lock_state();
1220 status = nfs_ok;
1221
a1bcecd2 1222 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1223 if (conf) {
1224 if (!same_verf(&verf, &conf->cl_verifier)) {
1225 /* 18.35.4 case 8 */
1226 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1227 status = nfserr_not_same;
1228 goto out;
1229 }
1230 /* Client reboot: destroy old state */
1231 expire_client(conf);
1232 goto out_new;
1233 }
1234 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1235 /* 18.35.4 case 9 */
1236 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1237 status = nfserr_perm;
1238 goto out;
1239 }
1240 expire_client(conf);
1241 goto out_new;
1242 }
1243 if (ip_addr != conf->cl_addr &&
1244 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1245 /* Client collision. 18.35.4 case 3 */
1246 status = nfserr_clid_inuse;
1247 goto out;
1248 }
1249 /*
1250 * Set bit when the owner id and verifier map to an already
1251 * confirmed client id (18.35.3).
1252 */
1253 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1254
1255 /*
1256 * Falling into 18.35.4 case 2, possible router replay.
1257 * Leave confirmed record intact and return same result.
1258 */
1259 copy_verf(conf, &verf);
1260 new = conf;
1261 goto out_copy;
1262 } else {
1263 /* 18.35.4 case 7 */
1264 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1265 status = nfserr_noent;
1266 goto out;
1267 }
1268 }
1269
a1bcecd2 1270 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1271 if (unconf) {
1272 /*
1273 * Possible retry or client restart. Per 18.35.4 case 4,
1274 * a new unconfirmed record should be generated regardless
1275 * of whether any properties have changed.
1276 */
1277 expire_client(unconf);
1278 }
1279
1280out_new:
1281 /* Normal case */
1282 new = create_client(exid->clname, dname);
1283 if (new == NULL) {
1284 status = nfserr_resource;
1285 goto out;
1286 }
1287
1288 copy_verf(new, &verf);
1289 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1290 new->cl_addr = ip_addr;
1291 gen_clid(new);
1292 gen_confirm(new);
1293 add_to_unconfirmed(new, strhashval);
1294out_copy:
1295 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1296 exid->clientid.cl_id = new->cl_clientid.cl_id;
1297
38eb76a5
AA
1298 new->cl_slot.sl_seqid = 0;
1299 exid->seqid = 1;
0733d213
AA
1300 nfsd4_set_ex_flags(new, exid);
1301
1302 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1303 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA
1304 status = nfs_ok;
1305
1306out:
1307 nfs4_unlock_state();
1308error:
1309 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1310 return status;
069b6ad4
AA
1311}
1312
b85d4c01
BH
1313static int
1314check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1315{
1316 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1317 slot->sl_seqid);
1318
1319 /* The slot is in use, and no response has been sent. */
1320 if (slot->sl_inuse) {
1321 if (seqid == slot->sl_seqid)
1322 return nfserr_jukebox;
1323 else
1324 return nfserr_seq_misordered;
1325 }
1326 /* Normal */
1327 if (likely(seqid == slot->sl_seqid + 1))
1328 return nfs_ok;
1329 /* Replay */
1330 if (seqid == slot->sl_seqid)
1331 return nfserr_replay_cache;
1332 /* Wraparound */
1333 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1334 return nfs_ok;
1335 /* Misordered replay or misordered new request */
1336 return nfserr_seq_misordered;
1337}
1338
069b6ad4
AA
1339__be32
1340nfsd4_create_session(struct svc_rqst *rqstp,
1341 struct nfsd4_compound_state *cstate,
1342 struct nfsd4_create_session *cr_ses)
1343{
ec6b5d7b 1344 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1345 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1346 struct nfs4_client *conf, *unconf;
38eb76a5 1347 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA
1348 int status = 0;
1349
1350 nfs4_lock_state();
1351 unconf = find_unconfirmed_client(&cr_ses->clientid);
1352 conf = find_confirmed_client(&cr_ses->clientid);
1353
1354 if (conf) {
38eb76a5
AA
1355 slot = &conf->cl_slot;
1356 status = check_slot_seqid(cr_ses->seqid, slot);
1357 if (status == nfserr_replay_cache) {
ec6b5d7b 1358 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA
1359 slot->sl_seqid);
1360 cstate->slot = slot;
1361 cstate->status = status;
1362 /* Return the cached reply status */
bf864a31 1363 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA
1364 goto out;
1365 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA
1366 status = nfserr_seq_misordered;
1367 dprintk("Sequence misordered!\n");
1368 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1369 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA
1370 goto out;
1371 }
38eb76a5 1372 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA
1373 } else if (unconf) {
1374 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1375 (ip_addr != unconf->cl_addr)) {
1376 status = nfserr_clid_inuse;
1377 goto out;
1378 }
1379
38eb76a5
AA
1380 slot = &unconf->cl_slot;
1381 status = check_slot_seqid(cr_ses->seqid, slot);
1382 if (status) {
1383 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA
1384 status = nfserr_seq_misordered;
1385 goto out;
1386 }
1387
38eb76a5 1388 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA
1389 move_to_confirmed(unconf);
1390
1391 /*
1392 * We do not support RDMA or persistent sessions
1393 */
1394 cr_ses->flags &= ~SESSION4_PERSIST;
1395 cr_ses->flags &= ~SESSION4_RDMA;
1396
1397 conf = unconf;
1398 } else {
1399 status = nfserr_stale_clientid;
1400 goto out;
1401 }
1402
1403 status = alloc_init_session(rqstp, conf, cr_ses);
1404 if (status)
1405 goto out;
1406
ec6b5d7b
AA
1407 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1408 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1409 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1410
38eb76a5
AA
1411 slot->sl_inuse = true;
1412 cstate->slot = slot;
bf864a31
AA
1413 /* Ensure a page is used for the cache */
1414 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA
1415out:
1416 nfs4_unlock_state();
1417 dprintk("%s returns %d\n", __func__, ntohl(status));
1418 return status;
069b6ad4
AA
1419}
1420
1421__be32
1422nfsd4_destroy_session(struct svc_rqst *r,
1423 struct nfsd4_compound_state *cstate,
1424 struct nfsd4_destroy_session *sessionid)
1425{
e10e0cfc
BH
1426 struct nfsd4_session *ses;
1427 u32 status = nfserr_badsession;
1428
1429 /* Notes:
1430 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1431 * - Should we return nfserr_back_chan_busy if waiting for
1432 * callbacks on to-be-destroyed session?
1433 * - Do we need to clear any callback info from previous session?
1434 */
1435
1436 dump_sessionid(__func__, &sessionid->sessionid);
1437 spin_lock(&sessionid_lock);
1438 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1439 if (!ses) {
1440 spin_unlock(&sessionid_lock);
1441 goto out;
1442 }
1443
1444 unhash_session(ses);
1445 spin_unlock(&sessionid_lock);
1446
1447 /* wait for callbacks */
1448 shutdown_callback_client(ses->se_client);
1449 nfsd4_put_session(ses);
1450 status = nfs_ok;
1451out:
1452 dprintk("%s returns %d\n", __func__, ntohl(status));
1453 return status;
069b6ad4
AA
1454}
1455
1456__be32
b85d4c01 1457nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA
1458 struct nfsd4_compound_state *cstate,
1459 struct nfsd4_sequence *seq)
1460{
f9bb94c4 1461 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH
1462 struct nfsd4_session *session;
1463 struct nfsd4_slot *slot;
1464 int status;
1465
f9bb94c4
AA
1466 if (resp->opcnt != 1)
1467 return nfserr_sequence_pos;
1468
b85d4c01
BH
1469 spin_lock(&sessionid_lock);
1470 status = nfserr_badsession;
1471 session = find_in_sessionid_hashtbl(&seq->sessionid);
1472 if (!session)
1473 goto out;
1474
1475 status = nfserr_badslot;
1476 if (seq->slotid >= session->se_fnumslots)
1477 goto out;
1478
1479 slot = &session->se_slots[seq->slotid];
1480 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1481
1482 status = check_slot_seqid(seq->seqid, slot);
1483 if (status == nfserr_replay_cache) {
1484 cstate->slot = slot;
1485 cstate->session = session;
da3846a2 1486 /* Return the cached reply status and set cstate->status
bf864a31
AA
1487 * for nfsd4_svc_encode_compoundres processing */
1488 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1489 cstate->status = nfserr_replay_cache;
b85d4c01
BH
1490 goto replay_cache;
1491 }
1492 if (status)
1493 goto out;
1494
1495 /* Success! bump slot seqid */
1496 slot->sl_inuse = true;
1497 slot->sl_seqid = seq->seqid;
bf864a31
AA
1498 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1499 /* Always set the cache entry cachethis for solo sequence */
1500 if (nfsd4_is_solo_sequence(resp))
1501 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH
1502
1503 cstate->slot = slot;
1504 cstate->session = session;
1505
1506replay_cache:
1507 /* Renew the clientid on success and on replay.
1508 * Hold a session reference until done processing the compound:
1509 * nfsd4_put_session called only if the cstate slot is set.
1510 */
1511 renew_client(session->se_client);
1512 nfsd4_get_session(session);
1513out:
1514 spin_unlock(&sessionid_lock);
1515 dprintk("%s: return %d\n", __func__, ntohl(status));
1516 return status;
069b6ad4
AA
1517}
1518
b37ad28b 1519__be32
b591480b
BF
1520nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1521 struct nfsd4_setclientid *setclid)
1da177e4 1522{
27459f09 1523 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT
1524 struct xdr_netobj clname = {
1525 .len = setclid->se_namelen,
1526 .data = setclid->se_name,
1527 };
1528 nfs4_verifier clverifier = setclid->se_verf;
1529 unsigned int strhashval;
28ce6054 1530 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1531 __be32 status;
68e76ad0 1532 char *princ;
a55370a3 1533 char dname[HEXDIR_LEN];
1da177e4 1534
1da177e4 1535 if (!check_name(clname))
73aea4ec 1536 return nfserr_inval;
1da177e4 1537
a55370a3
N
1538 status = nfs4_make_rec_clidname(dname, &clname);
1539 if (status)
73aea4ec 1540 return status;
a55370a3 1541
1da177e4
LT
1542 /*
1543 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1544 * We get here on a DRC miss.
1545 */
1546
a55370a3 1547 strhashval = clientstr_hashval(dname);
1da177e4 1548
1da177e4 1549 nfs4_lock_state();
a1bcecd2 1550 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1551 if (conf) {
a186e767 1552 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1553 status = nfserr_clid_inuse;
026722c2
BF
1554 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1555 dprintk("NFSD: setclientid: string in use by client"
1556 " at %pI4\n", &conf->cl_addr);
1da177e4
LT
1557 goto out;
1558 }
1da177e4 1559 }
a186e767
BF
1560 /*
1561 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1562 * has a description of SETCLIENTID request processing consisting
1563 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1564 */
a1bcecd2 1565 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT
1566 status = nfserr_resource;
1567 if (!conf) {
a186e767
BF
1568 /*
1569 * RFC 3530 14.2.33 CASE 4:
1570 * placed first, because it is the normal case
1da177e4
LT
1571 */
1572 if (unconf)
1573 expire_client(unconf);
a55370a3
N
1574 new = create_client(clname, dname);
1575 if (new == NULL)
1da177e4 1576 goto out;
1da177e4 1577 gen_clid(new);
599e0a22 1578 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1579 /*
a186e767
BF
1580 * RFC 3530 14.2.33 CASE 1:
1581 * probable callback update
1da177e4 1582 */
31f4a6c1
N
1583 if (unconf) {
1584 /* Note this is removing unconfirmed {*x***},
1585 * which is stronger than RFC recommended {vxc**}.
1586 * This has the advantage that there is at most
1587 * one {*x***} in either list at any time.
1588 */
1589 expire_client(unconf);
1da177e4 1590 }
a55370a3
N
1591 new = create_client(clname, dname);
1592 if (new == NULL)
1da177e4 1593 goto out;
1da177e4 1594 copy_clid(new, conf);
1da177e4
LT
1595 } else if (!unconf) {
1596 /*
a186e767
BF
1597 * RFC 3530 14.2.33 CASE 2:
1598 * probable client reboot; state will be removed if
1599 * confirmed.
1da177e4 1600 */
a55370a3
N
1601 new = create_client(clname, dname);
1602 if (new == NULL)
1da177e4 1603 goto out;
1da177e4 1604 gen_clid(new);
49ba8781 1605 } else {
a186e767
BF
1606 /*
1607 * RFC 3530 14.2.33 CASE 3:
1608 * probable client reboot; state will be removed if
1609 * confirmed.
1da177e4
LT
1610 */
1611 expire_client(unconf);
a55370a3
N
1612 new = create_client(clname, dname);
1613 if (new == NULL)
1da177e4 1614 goto out;
1da177e4 1615 gen_clid(new);
1da177e4 1616 }
c175b83c
BF
1617 copy_verf(new, &clverifier);
1618 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1619 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK
1620 princ = svc_gss_principal(rqstp);
1621 if (princ) {
1622 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1623 if (new->cl_principal == NULL) {
1624 free_client(new);
1625 goto out;
1626 }
1627 }
c175b83c
BF
1628 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1629 gen_confirm(new);
1630 gen_callback(new, setclid);
1631 add_to_unconfirmed(new, strhashval);
1da177e4
LT
1632 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1633 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1634 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1635 status = nfs_ok;
1636out:
1637 nfs4_unlock_state();
1638 return status;
1639}
1640
1641
1642/*
a186e767
BF
1643 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1644 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1645 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1646 */
b37ad28b 1647__be32
b591480b
BF
1648nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1649 struct nfsd4_compound_state *cstate,
1650 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1651{
27459f09 1652 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1653 struct nfs4_client *conf, *unconf;
1da177e4
LT
1654 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1655 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1656 __be32 status;
1da177e4
LT
1657
1658 if (STALE_CLIENTID(clid))
1659 return nfserr_stale_clientid;
1660 /*
1661 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1662 * We get here on a DRC miss.
1663 */
1664
1665 nfs4_lock_state();
21ab45a4
N
1666
1667 conf = find_confirmed_client(clid);
1668 unconf = find_unconfirmed_client(clid);
1669
1670 status = nfserr_clid_inuse;
27459f09 1671 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1672 goto out;
27459f09 1673 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N
1674 goto out;
1675
a186e767
BF
1676 /*
1677 * section 14.2.34 of RFC 3530 has a description of
1678 * SETCLIENTID_CONFIRM request processing consisting
1679 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1680 */
366e0c1d 1681 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF
1682 /*
1683 * RFC 3530 14.2.34 CASE 1:
1684 * callback update
1685 */
599e0a22 1686 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT
1687 status = nfserr_clid_inuse;
1688 else {
1a69c179
N
1689 /* XXX: We just turn off callbacks until we can handle
1690 * change request correctly. */
c237dc03 1691 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1692 expire_client(unconf);
1da177e4 1693 status = nfs_ok;
1a69c179 1694
1da177e4 1695 }
f3aba4e5 1696 } else if (conf && !unconf) {
a186e767
BF
1697 /*
1698 * RFC 3530 14.2.34 CASE 2:
1699 * probable retransmitted request; play it safe and
1700 * do nothing.
7c79f737 1701 */
599e0a22 1702 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1703 status = nfserr_clid_inuse;
21ab45a4 1704 else
1da177e4 1705 status = nfs_ok;
7c79f737 1706 } else if (!conf && unconf
599e0a22 1707 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF
1708 /*
1709 * RFC 3530 14.2.34 CASE 3:
1710 * Normal case; new or rebooted client:
7c79f737 1711 */
599e0a22 1712 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT
1713 status = nfserr_clid_inuse;
1714 } else {
1a69c179
N
1715 unsigned int hash =
1716 clientstr_hashval(unconf->cl_recdir);
1717 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1718 hash, false);
1a69c179 1719 if (conf) {
c7b9a459 1720 nfsd4_remove_clid_dir(conf);
1a69c179
N
1721 expire_client(conf);
1722 }
1da177e4 1723 move_to_confirmed(unconf);
21ab45a4 1724 conf = unconf;
46f8a64b 1725 nfsd4_probe_callback(conf);
1a69c179 1726 status = nfs_ok;
1da177e4 1727 }
599e0a22
BF
1728 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1729 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1730 &confirm)))) {
a186e767
BF
1731 /*
1732 * RFC 3530 14.2.34 CASE 4:
1733 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1734 */
1da177e4 1735 status = nfserr_stale_clientid;
7c79f737 1736 } else {
08e8987c
N
1737 /* check that we have hit one of the cases...*/
1738 status = nfserr_clid_inuse;
1739 }
1da177e4 1740out:
1da177e4
LT
1741 nfs4_unlock_state();
1742 return status;
1743}
1744
1da177e4
LT
1745/* OPEN Share state helper functions */
1746static inline struct nfs4_file *
1747alloc_init_file(struct inode *ino)
1748{
1749 struct nfs4_file *fp;
1750 unsigned int hashval = file_hashval(ino);
1751
e60d4398
N
1752 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1753 if (fp) {
8b671b80 1754 atomic_set(&fp->fi_ref, 1);
1da177e4 1755 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N
1756 INIT_LIST_HEAD(&fp->fi_stateids);
1757 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1758 spin_lock(&recall_lock);
1da177e4 1759 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1760 spin_unlock(&recall_lock);
1da177e4
LT
1761 fp->fi_inode = igrab(ino);
1762 fp->fi_id = current_fileid++;
47f9940c 1763 fp->fi_had_conflict = false;
1da177e4
LT
1764 return fp;
1765 }
1766 return NULL;
1767}
1768
e60d4398 1769static void
e18b890b 1770nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1771{
e60d4398
N
1772 if (*slab == NULL)
1773 return;
1a1d92c1 1774 kmem_cache_destroy(*slab);
e60d4398 1775 *slab = NULL;
1da177e4
LT
1776}
1777
e8ff2a84 1778void
1da177e4
LT
1779nfsd4_free_slabs(void)
1780{
e60d4398
N
1781 nfsd4_free_slab(&stateowner_slab);
1782 nfsd4_free_slab(&file_slab);
5ac049ac 1783 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1784 nfsd4_free_slab(&deleg_slab);
e60d4398 1785}
1da177e4 1786
e60d4398
N
1787static int
1788nfsd4_init_slabs(void)
1789{
1790 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1791 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N
1792 if (stateowner_slab == NULL)
1793 goto out_nomem;
1794 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1795 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N
1796 if (file_slab == NULL)
1797 goto out_nomem;
5ac049ac 1798 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1799 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N
1800 if (stateid_slab == NULL)
1801 goto out_nomem;
5b2d21c1 1802 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1803 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N
1804 if (deleg_slab == NULL)
1805 goto out_nomem;
e60d4398
N
1806 return 0;
1807out_nomem:
1808 nfsd4_free_slabs();
1809 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1810 return -ENOMEM;
1da177e4
LT
1811}
1812
1813void
1814nfs4_free_stateowner(struct kref *kref)
1815{
1816 struct nfs4_stateowner *sop =
1817 container_of(kref, struct nfs4_stateowner, so_ref);
1818 kfree(sop->so_owner.data);
1819 kmem_cache_free(stateowner_slab, sop);
1820}
1821
1822static inline struct nfs4_stateowner *
1823alloc_stateowner(struct xdr_netobj *owner)
1824{
1825 struct nfs4_stateowner *sop;
1826
1827 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1828 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1829 memcpy(sop->so_owner.data, owner->data, owner->len);
1830 sop->so_owner.len = owner->len;
1831 kref_init(&sop->so_ref);
1832 return sop;
1833 }
1834 kmem_cache_free(stateowner_slab, sop);
1835 }
1836 return NULL;
1837}
1838
1839static struct nfs4_stateowner *
1840alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1841 struct nfs4_stateowner *sop;
1842 struct nfs4_replay *rp;
1843 unsigned int idhashval;
1844
1845 if (!(sop = alloc_stateowner(&open->op_owner)))
1846 return NULL;
1847 idhashval = ownerid_hashval(current_ownerid);
1848 INIT_LIST_HEAD(&sop->so_idhash);
1849 INIT_LIST_HEAD(&sop->so_strhash);
1850 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
1851 INIT_LIST_HEAD(&sop->so_stateids);
1852 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT
1853 INIT_LIST_HEAD(&sop->so_close_lru);
1854 sop->so_time = 0;
1855 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1856 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1857 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT
1858 sop->so_is_open_owner = 1;
1859 sop->so_id = current_ownerid++;
1860 sop->so_client = clp;
1861 sop->so_seqid = open->op_seqid;
1862 sop->so_confirmed = 0;
1863 rp = &sop->so_replay;
de1ae286 1864 rp->rp_status = nfserr_serverfault;
1da177e4
LT
1865 rp->rp_buflen = 0;
1866 rp->rp_buf = rp->rp_ibuf;
1867 return sop;
1868}
1869
1da177e4
LT
1870static inline void
1871init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1872 struct nfs4_stateowner *sop = open->op_stateowner;
1873 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1874
1875 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N
1876 INIT_LIST_HEAD(&stp->st_perstateowner);
1877 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT
1878 INIT_LIST_HEAD(&stp->st_perfile);
1879 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1880 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1881 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1882 stp->st_stateowner = sop;
13cd2184 1883 get_nfs4_file(fp);
1da177e4 1884 stp->st_file = fp;
78155ed7 1885 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
1886 stp->st_stateid.si_stateownerid = sop->so_id;
1887 stp->st_stateid.si_fileid = fp->fi_id;
1888 stp->st_stateid.si_generation = 0;
1889 stp->st_access_bmap = 0;
1890 stp->st_deny_bmap = 0;
84459a11
AA
1891 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1892 &stp->st_access_bmap);
1da177e4 1893 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1894 stp->st_openstp = NULL;
1da177e4
LT
1895}
1896
fd39ca9a 1897static void
1da177e4
LT
1898move_to_close_lru(struct nfs4_stateowner *sop)
1899{
1900 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1901
358dd55a 1902 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT
1903 sop->so_time = get_seconds();
1904}
1905
1da177e4 1906static int
599e0a22
BF
1907same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1908 clientid_t *clid)
1909{
1910 return (sop->so_owner.len == owner->len) &&
1911 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1912 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT
1913}
1914
1915static struct nfs4_stateowner *
1916find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1917{
1918 struct nfs4_stateowner *so = NULL;
1919
1920 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1921 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT
1922 return so;
1923 }
1924 return NULL;
1925}
1926
1927/* search file_hashtbl[] for file */
1928static struct nfs4_file *
1929find_file(struct inode *ino)
1930{
1931 unsigned int hashval = file_hashval(ino);
1932 struct nfs4_file *fp;
1933
8b671b80 1934 spin_lock(&recall_lock);
1da177e4 1935 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N
1936 if (fp->fi_inode == ino) {
1937 get_nfs4_file(fp);
8b671b80 1938 spin_unlock(&recall_lock);
1da177e4 1939 return fp;
13cd2184 1940 }
1da177e4 1941 }
8b671b80 1942 spin_unlock(&recall_lock);
1da177e4
LT
1943 return NULL;
1944}
1945
d87a8ade 1946static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1947{
d87a8ade 1948 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1949 return 0;
d87a8ade
AA
1950 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1951 return 0;
1952 x &= ~NFS4_SHARE_ACCESS_MASK;
1953 if (minorversion && x) {
1954 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1955 return 0;
1956 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1957 return 0;
1958 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1959 }
1960 if (x)
8838dc43
BF
1961 return 0;
1962 return 1;
ba5a6a19
BF
1963}
1964
8838dc43 1965static inline int deny_valid(u32 x)
ba5a6a19 1966{
8838dc43
BF
1967 /* Note: unlike access bits, deny bits may be zero. */
1968 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1969}
1da177e4 1970
4f83aa30
BF
1971/*
1972 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1973 * st_{access,deny}_bmap field of the stateid, in order to track not
1974 * only what share bits are currently in force, but also what
1975 * combinations of share bits previous opens have used. This allows us
1976 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1977 * return an error if the client attempt to downgrade to a combination
1978 * of share bits not explicable by closing some of its previous opens.
1979 *
1980 * XXX: This enforcement is actually incomplete, since we don't keep
1981 * track of access/deny bit combinations; so, e.g., we allow:
1982 *
1983 * OPEN allow read, deny write
1984 * OPEN allow both, deny none
1985 * DOWNGRADE allow read, deny none
1986 *
1987 * which we should reject.
1988 */
fd39ca9a 1989static void
1da177e4
LT
1990set_access(unsigned int *access, unsigned long bmap) {
1991 int i;
1992
1993 *access = 0;
1994 for (i = 1; i < 4; i++) {
1995 if (test_bit(i, &bmap))
1996 *access |= i;
1997 }
1998}
1999
fd39ca9a 2000static void
1da177e4
LT
2001set_deny(unsigned int *deny, unsigned long bmap) {
2002 int i;
2003
2004 *deny = 0;
2005 for (i = 0; i < 4; i++) {
2006 if (test_bit(i, &bmap))
2007 *deny |= i ;
2008 }
2009}
2010
2011static int
2012test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2013 unsigned int access, deny;
2014
2015 set_access(&access, stp->st_access_bmap);
2016 set_deny(&deny, stp->st_deny_bmap);
2017 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2018 return 0;
2019 return 1;
2020}
2021
2022/*
2023 * Called to check deny when READ with all zero stateid or
2024 * WRITE with all zero or all one stateid
2025 */
b37ad28b 2026static __be32
1da177e4
LT
2027nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2028{
2029 struct inode *ino = current_fh->fh_dentry->d_inode;
2030 struct nfs4_file *fp;
2031 struct nfs4_stateid *stp;
b37ad28b 2032 __be32 ret;
1da177e4
LT
2033
2034 dprintk("NFSD: nfs4_share_conflict\n");
2035
2036 fp = find_file(ino);
13cd2184
N
2037 if (!fp)
2038 return nfs_ok;
b700949b 2039 ret = nfserr_locked;
1da177e4 2040 /* Search for conflicting share reservations */
13cd2184
N
2041 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2042 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2043 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2044 goto out;
1da177e4 2045 }
13cd2184
N
2046 ret = nfs_ok;
2047out:
2048 put_nfs4_file(fp);
2049 return ret;
1da177e4
LT
2050}
2051
2052static inline void
2053nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2054{
2055 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2056 drop_file_write_access(filp);
1da177e4
LT
2057 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2058 }
2059}
2060
1da177e4
LT
2061/*
2062 * Spawn a thread to perform a recall on the delegation represented
2063 * by the lease (file_lock)
2064 *
2065 * Called from break_lease() with lock_kernel() held.
2066 * Note: we assume break_lease will only call this *once* for any given
2067 * lease.
2068 */
2069static
2070void nfsd_break_deleg_cb(struct file_lock *fl)
2071{
63e4863f 2072 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT
2073
2074 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2075 if (!dp)
2076 return;
2077
2078 /* We're assuming the state code never drops its reference
2079 * without first removing the lease. Since we're in this lease
2080 * callback (and since the lease code is serialized by the kernel
2081 * lock) we know the server hasn't removed the lease yet, we know
2082 * it's safe to take a reference: */
2083 atomic_inc(&dp->dl_count);
cfdcad4d 2084 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT
2085
2086 spin_lock(&recall_lock);
2087 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2088 spin_unlock(&recall_lock);
2089
2090 /* only place dl_time is set. protected by lock_kernel*/
2091 dp->dl_time = get_seconds();
2092
0272e1fd
BF
2093 /*
2094 * We don't want the locks code to timeout the lease for us;
2095 * we'll remove it ourself if the delegation isn't returned
2096 * in time.
2097 */
2098 fl->fl_break_time = 0;
1da177e4 2099
63e4863f
BF
2100 dp->dl_file->fi_had_conflict = true;
2101 nfsd4_cb_recall(dp);
1da177e4
LT
2102}
2103
2104/*
2105 * The file_lock is being reapd.
2106 *
2107 * Called by locks_free_lock() with lock_kernel() held.
2108 */
2109static
2110void nfsd_release_deleg_cb(struct file_lock *fl)
2111{
2112 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2113
2114 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2115
2116 if (!(fl->fl_flags & FL_LEASE) || !dp)
2117 return;
2118 dp->dl_flock = NULL;
2119}
2120
2121/*
2122 * Set the delegation file_lock back pointer.
2123 *
a9933cea 2124 * Called from setlease() with lock_kernel() held.
1da177e4
LT
2125 */
2126static
2127void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2128{
2129 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2130
2131 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2132 if (!dp)
2133 return;
2134 dp->dl_flock = new;
2135}
2136
2137/*
a9933cea 2138 * Called from setlease() with lock_kernel() held
1da177e4
LT
2139 */
2140static
2141int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2142{
2143 struct nfs4_delegation *onlistd =
2144 (struct nfs4_delegation *)onlist->fl_owner;
2145 struct nfs4_delegation *tryd =
2146 (struct nfs4_delegation *)try->fl_owner;
2147
2148 if (onlist->fl_lmops != try->fl_lmops)
2149 return 0;
2150
2151 return onlistd->dl_client == tryd->dl_client;
2152}
2153
2154
2155static
2156int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2157{
2158 if (arg & F_UNLCK)
2159 return lease_modify(onlist, arg);
2160 else
2161 return -EAGAIN;
2162}
2163
fd39ca9a 2164static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT
2165 .fl_break = nfsd_break_deleg_cb,
2166 .fl_release_private = nfsd_release_deleg_cb,
2167 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2168 .fl_mylease = nfsd_same_client_deleg_cb,
2169 .fl_change = nfsd_change_deleg_cb,
2170};
2171
2172
b37ad28b 2173__be32
6668958f
AA
2174nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2175 struct nfsd4_open *open)
1da177e4 2176{
1da177e4
LT
2177 clientid_t *clientid = &open->op_clientid;
2178 struct nfs4_client *clp = NULL;
2179 unsigned int strhashval;
2180 struct nfs4_stateowner *sop = NULL;
2181
1da177e4 2182 if (!check_name(open->op_owner))
0f442aa2 2183 return nfserr_inval;
1da177e4
LT
2184
2185 if (STALE_CLIENTID(&open->op_clientid))
2186 return nfserr_stale_clientid;
2187
2188 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2189 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF
2190 open->op_stateowner = sop;
2191 if (!sop) {
2192 /* Make sure the client's lease hasn't expired. */
1da177e4
LT
2193 clp = find_confirmed_client(clientid);
2194 if (clp == NULL)
0f442aa2
BF
2195 return nfserr_expired;
2196 goto renew;
1da177e4 2197 }
6668958f
AA
2198 /* When sessions are used, skip open sequenceid processing */
2199 if (nfsd4_has_session(cstate))
2200 goto renew;
0f442aa2
BF
2201 if (!sop->so_confirmed) {
2202 /* Replace unconfirmed owners without checking for replay. */
2203 clp = sop->so_client;
f044ff83 2204 release_openowner(sop);
0f442aa2
BF
2205 open->op_stateowner = NULL;
2206 goto renew;
2207 }
2208 if (open->op_seqid == sop->so_seqid - 1) {
2209 if (sop->so_replay.rp_buflen)
a90b061c 2210 return nfserr_replay_me;
0f442aa2
BF
2211 /* The original OPEN failed so spectacularly
2212 * that we don't even have replay data saved!
2213 * Therefore, we have no choice but to continue
2214 * processing this OPEN; presumably, we'll
2215 * fail again for the same reason.
2216 */
2217 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2218 goto renew;
2219 }
2220 if (open->op_seqid != sop->so_seqid)
2221 return nfserr_bad_seqid;
1da177e4 2222renew:
0f442aa2
BF
2223 if (open->op_stateowner == NULL) {
2224 sop = alloc_init_open_stateowner(strhashval, clp, open);
2225 if (sop == NULL)
2226 return nfserr_resource;
2227 open->op_stateowner = sop;
2228 }
2229 list_del_init(&sop->so_close_lru);
1da177e4 2230 renew_client(sop->so_client);
0f442aa2 2231 return nfs_ok;
1da177e4
LT
2232}
2233
b37ad28b 2234static inline __be32
4a6e43e6
N
2235nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2236{
2237 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2238 return nfserr_openmode;
2239 else
2240 return nfs_ok;
2241}
2242
52f4fb43
N
2243static struct nfs4_delegation *
2244find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2245{
2246 struct nfs4_delegation *dp;
2247
ea1da636 2248 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N
2249 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2250 return dp;
2251 }
2252 return NULL;
2253}
2254
b37ad28b 2255static __be32
567d9829
N
2256nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2257 struct nfs4_delegation **dp)
2258{
2259 int flags;
b37ad28b 2260 __be32 status = nfserr_bad_stateid;
567d9829
N
2261
2262 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2263 if (*dp == NULL)
c44c5eeb 2264 goto out;
567d9829
N
2265 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2266 RD_STATE : WR_STATE;
2267 status = nfs4_check_delegmode(*dp, flags);
2268 if (status)
2269 *dp = NULL;
c44c5eeb
N
2270out:
2271 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2272 return nfs_ok;
2273 if (status)
2274 return status;
2275 open->op_stateowner->so_confirmed = 1;
2276 return nfs_ok;
567d9829
N
2277}
2278
b37ad28b 2279static __be32
1da177e4
LT
2280nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2281{
2282 struct nfs4_stateid *local;
b37ad28b 2283 __be32 status = nfserr_share_denied;
1da177e4
LT
2284 struct nfs4_stateowner *sop = open->op_stateowner;
2285
8beefa24 2286 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT
2287 /* ignore lock owners */
2288 if (local->st_stateowner->so_is_open_owner == 0)
2289 continue;
2290 /* remember if we have seen this open owner */
2291 if (local->st_stateowner == sop)
2292 *stpp = local;
2293 /* check for conflicting share reservations */
2294 if (!test_share(local, open))
2295 goto out;
2296 }
2297 status = 0;
2298out:
2299 return status;
2300}
2301
5ac049ac
N
2302static inline struct nfs4_stateid *
2303nfs4_alloc_stateid(void)
2304{
2305 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2306}
2307
b37ad28b 2308static __be32
1da177e4 2309nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2310 struct nfs4_delegation *dp,
1da177e4
LT
2311 struct svc_fh *cur_fh, int flags)
2312{
2313 struct nfs4_stateid *stp;
1da177e4 2314
5ac049ac 2315 stp = nfs4_alloc_stateid();
1da177e4
LT
2316 if (stp == NULL)
2317 return nfserr_resource;
2318
567d9829
N
2319 if (dp) {
2320 get_file(dp->dl_vfs_file);
2321 stp->st_vfs_file = dp->dl_vfs_file;
2322 } else {
b37ad28b 2323 __be32 status;
567d9829
N
2324 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2325 &stp->st_vfs_file);
2326 if (status) {
2327 if (status == nfserr_dropit)
2328 status = nfserr_jukebox;
5ac049ac 2329 kmem_cache_free(stateid_slab, stp);
567d9829
N
2330 return status;
2331 }
1da177e4 2332 }
1da177e4
LT
2333 *stpp = stp;
2334 return 0;
2335}
2336
b37ad28b 2337static inline __be32
1da177e4
LT
2338nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2339 struct nfsd4_open *open)
2340{
2341 struct iattr iattr = {
2342 .ia_valid = ATTR_SIZE,
2343 .ia_size = 0,
2344 };
2345 if (!open->op_truncate)
2346 return 0;
2347 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2348 return nfserr_inval;
1da177e4
LT
2349 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2350}
2351
b37ad28b 2352static __be32
1da177e4
LT
2353nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2354{
2355 struct file *filp = stp->st_vfs_file;
7eaa36e2 2356 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2357 unsigned int share_access, new_writer;
b37ad28b 2358 __be32 status;
1da177e4
LT
2359
2360 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF
2361 new_writer = (~share_access) & open->op_share_access
2362 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2363
6c26d08f 2364 if (new_writer) {
b37ad28b
AV
2365 int err = get_write_access(inode);
2366 if (err)
2367 return nfserrno(err);
e518f056
BH
2368 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2369 if (err)
2370 return nfserrno(err);
2371 file_take_write(filp);
6c26d08f 2372 }
1da177e4
LT
2373 status = nfsd4_truncate(rqstp, cur_fh, open);
2374 if (status) {
6c26d08f
BF
2375 if (new_writer)
2376 put_write_access(inode);
1da177e4
LT
2377 return status;
2378 }
2379 /* remember the open */
6c26d08f 2380 filp->f_mode |= open->op_share_access;
b55e0ba1
BF
2381 __set_bit(open->op_share_access, &stp->st_access_bmap);
2382 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT
2383
2384 return nfs_ok;
2385}
2386
2387
1da177e4 2388static void
37515177 2389nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2390{
37515177
N
2391 open->op_stateowner->so_confirmed = 1;
2392 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT
2393}
2394
2395/*
2396 * Attempt to hand out a delegation.
2397 */
2398static void
2399nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2400{
2401 struct nfs4_delegation *dp;
2402 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2403 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT
2404 struct file_lock fl, *flp = &fl;
2405 int status, flag = 0;
2406
2407 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N
2408 open->op_recall = 0;
2409 switch (open->op_claim_type) {
2410 case NFS4_OPEN_CLAIM_PREVIOUS:
2411 if (!atomic_read(&cb->cb_set))
2412 open->op_recall = 1;
2413 flag = open->op_delegate_type;
2414 if (flag == NFS4_OPEN_DELEGATE_NONE)
2415 goto out;
2416 break;
2417 case NFS4_OPEN_CLAIM_NULL:
2418 /* Let's not give out any delegations till everyone's
2419 * had the chance to reclaim theirs.... */
af558e33 2420 if (locks_in_grace())
7b190fec
N
2421 goto out;
2422 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2423 goto out;
2424 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2425 flag = NFS4_OPEN_DELEGATE_WRITE;
2426 else
2427 flag = NFS4_OPEN_DELEGATE_READ;
2428 break;
2429 default:
2430 goto out;
2431 }
1da177e4
LT
2432
2433 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2434 if (dp == NULL) {
2435 flag = NFS4_OPEN_DELEGATE_NONE;
2436 goto out;
2437 }
2438 locks_init_lock(&fl);
2439 fl.fl_lmops = &nfsd_lease_mng_ops;
2440 fl.fl_flags = FL_LEASE;
9167f501 2441 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT
2442 fl.fl_end = OFFSET_MAX;
2443 fl.fl_owner = (fl_owner_t)dp;
2444 fl.fl_file = stp->st_vfs_file;
2445 fl.fl_pid = current->tgid;
2446
a9933cea 2447 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT
2448 * the lock_manager callbacks fl_mylease and fl_change are used
2449 */
9167f501 2450 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2451 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2452 unhash_delegation(dp);
1da177e4
LT
2453 flag = NFS4_OPEN_DELEGATE_NONE;
2454 goto out;
2455 }
2456
2457 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2458
2459 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2460 dp->dl_stateid.si_boot,
2461 dp->dl_stateid.si_stateownerid,
2462 dp->dl_stateid.si_fileid,
2463 dp->dl_stateid.si_generation);
2464out:
7b190fec
N
2465 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2466 && flag == NFS4_OPEN_DELEGATE_NONE
2467 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2468 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT
2469 open->op_delegate_type = flag;
2470}
2471
2472/*
2473 * called with nfs4_lock_state() held.
2474 */
b37ad28b 2475__be32
1da177e4
LT
2476nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2477{
6668958f 2478 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT
2479 struct nfs4_file *fp = NULL;
2480 struct inode *ino = current_fh->fh_dentry->d_inode;
2481 struct nfs4_stateid *stp = NULL;
567d9829 2482 struct nfs4_delegation *dp = NULL;
b37ad28b 2483 __be32 status;
1da177e4
LT
2484
2485 status = nfserr_inval;
d87a8ade 2486 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2487 || !deny_valid(open->op_share_deny))
1da177e4
LT
2488 goto out;
2489 /*
2490 * Lookup file; if found, lookup stateid and check open request,
2491 * and check for delegations in the process of being recalled.
2492 * If not found, create the nfs4_file struct
2493 */
2494 fp = find_file(ino);
2495 if (fp) {
2496 if ((status = nfs4_check_open(fp, open, &stp)))
2497 goto out;
c44c5eeb
N
2498 status = nfs4_check_deleg(fp, open, &dp);
2499 if (status)
2500 goto out;
1da177e4 2501 } else {
c44c5eeb
N
2502 status = nfserr_bad_stateid;
2503 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2504 goto out;
1da177e4
LT
2505 status = nfserr_resource;
2506 fp = alloc_init_file(ino);
2507 if (fp == NULL)
2508 goto out;
2509 }
2510
2511 /*
2512 * OPEN the file, or upgrade an existing OPEN.
2513 * If truncate fails, the OPEN fails.
2514 */
2515 if (stp) {
2516 /* Stateid was found, this is an OPEN upgrade */
2517 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2518 if (status)
2519 goto out;
444c2c07 2520 update_stateid(&stp->st_stateid);
1da177e4
LT
2521 } else {
2522 /* Stateid was not found, this is a new OPEN */
2523 int flags = 0;
9ecb6a08 2524 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2525 flags |= NFSD_MAY_READ;
1da177e4 2526 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2527 flags |= NFSD_MAY_WRITE;
567d9829
N
2528 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2529 if (status)
1da177e4
LT
2530 goto out;
2531 init_stateid(stp, fp, open);
2532 status = nfsd4_truncate(rqstp, current_fh, open);
2533 if (status) {
2283963f 2534 release_open_stateid(stp);
1da177e4
LT
2535 goto out;
2536 }
6668958f
AA
2537 if (nfsd4_has_session(&resp->cstate))
2538 update_stateid(&stp->st_stateid);
1da177e4
LT
2539 }
2540 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2541
6668958f
AA
2542 if (nfsd4_has_session(&resp->cstate))
2543 open->op_stateowner->so_confirmed = 1;
2544
1da177e4
LT
2545 /*
2546 * Attempt to hand out a delegation. No error return, because the
2547 * OPEN succeeds even if we fail.
2548 */
2549 nfs4_open_delegation(current_fh, open, stp);
2550
2551 status = nfs_ok;
2552
2553 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2554 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2555 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2556out:
13cd2184
N
2557 if (fp)
2558 put_nfs4_file(fp);
37515177
N
2559 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2560 nfs4_set_claim_prev(open);
1da177e4
LT
2561 /*
2562 * To finish the open response, we just need to set the rflags.
2563 */
2564 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA
2565 if (!open->op_stateowner->so_confirmed &&
2566 !nfsd4_has_session(&resp->cstate))
1da177e4
LT
2567 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2568
2569 return status;
2570}
2571
b37ad28b 2572__be32
b591480b
BF
2573nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2574 clientid_t *clid)
1da177e4
LT
2575{
2576 struct nfs4_client *clp;
b37ad28b 2577 __be32 status;
1da177e4
LT
2578
2579 nfs4_lock_state();
2580 dprintk("process_renew(%08x/%08x): starting\n",
2581 clid->cl_boot, clid->cl_id);
2582 status = nfserr_stale_clientid;
2583 if (STALE_CLIENTID(clid))
2584 goto out;
2585 clp = find_confirmed_client(clid);
2586 status = nfserr_expired;
2587 if (clp == NULL) {
2588 /* We assume the client took too long to RENEW. */
2589 dprintk("nfsd4_renew: clientid not found!\n");
2590 goto out;
2591 }
2592 renew_client(clp);
2593 status = nfserr_cb_path_down;
ea1da636 2594 if (!list_empty(&clp->cl_delegations)
c237dc03 2595 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT
2596 goto out;
2597 status = nfs_ok;
2598out:
2599 nfs4_unlock_state();
2600 return status;
2601}
2602
af558e33
BF
2603struct lock_manager nfsd4_manager = {
2604};
2605
a76b4319 2606static void
af558e33 2607nfsd4_end_grace(void)
a76b4319
N
2608{
2609 dprintk("NFSD: end of grace period\n");
c7b9a459 2610 nfsd4_recdir_purge_old();
af558e33 2611 locks_end_grace(&nfsd4_manager);
a76b4319
N
2612}
2613
fd39ca9a 2614static time_t
1da177e4
LT
2615nfs4_laundromat(void)
2616{
2617 struct nfs4_client *clp;
2618 struct nfs4_stateowner *sop;
2619 struct nfs4_delegation *dp;
2620 struct list_head *pos, *next, reaplist;
2621 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2622 time_t t, clientid_val = NFSD_LEASE_TIME;
2623 time_t u, test_val = NFSD_LEASE_TIME;
2624
2625 nfs4_lock_state();
2626
2627 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF
2628 if (locks_in_grace())
2629 nfsd4_end_grace();
1da177e4
LT
2630 list_for_each_safe(pos, next, &client_lru) {
2631 clp = list_entry(pos, struct nfs4_client, cl_lru);
2632 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2633 t = clp->cl_time - cutoff;
2634 if (clientid_val > t)
2635 clientid_val = t;
2636 break;
2637 }
2638 dprintk("NFSD: purging unused client (clientid %08x)\n",
2639 clp->cl_clientid.cl_id);
c7b9a459 2640 nfsd4_remove_clid_dir(clp);
1da177e4
LT
2641 expire_client(clp);
2642 }
2643 INIT_LIST_HEAD(&reaplist);
2644 spin_lock(&recall_lock);
2645 list_for_each_safe(pos, next, &del_recall_lru) {
2646 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2647 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2648 u = dp->dl_time - cutoff;
2649 if (test_val > u)
2650 test_val = u;
2651 break;
2652 }
2653 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2654 dp, dp->dl_flock);
2655 list_move(&dp->dl_recall_lru, &reaplist);
2656 }
2657 spin_unlock(&recall_lock);
2658 list_for_each_safe(pos, next, &reaplist) {
2659 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2660 list_del_init(&dp->dl_recall_lru);
2661 unhash_delegation(dp);
2662 }
2663 test_val = NFSD_LEASE_TIME;
2664 list_for_each_safe(pos, next, &close_lru) {
2665 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2666 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2667 u = sop->so_time - cutoff;
2668 if (test_val > u)
2669 test_val = u;
2670 break;
2671 }
2672 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2673 sop->so_id);
f044ff83 2674 release_openowner(sop);
1da177e4
LT
2675 }
2676 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2677 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2678 nfs4_unlock_state();
2679 return clientid_val;
2680}
2681
a254b246
HH
2682static struct workqueue_struct *laundry_wq;
2683static void laundromat_main(struct work_struct *);
2684static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2685
2686static void
c4028958 2687laundromat_main(struct work_struct *not_used)
1da177e4
LT
2688{
2689 time_t t;
2690
2691 t = nfs4_laundromat();
2692 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2693 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT
2694}
2695
fd39ca9a 2696static struct nfs4_stateowner *
f8816512
N
2697search_close_lru(u32 st_id, int flags)
2698{
1da177e4
LT
2699 struct nfs4_stateowner *local = NULL;
2700
1da177e4
LT
2701 if (flags & CLOSE_STATE) {
2702 list_for_each_entry(local, &close_lru, so_close_lru) {
2703 if (local->so_id == st_id)
2704 return local;
2705 }
2706 }
2707 return NULL;
2708}
2709
2710static inline int
2711nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2712{
7eaa36e2 2713 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT
2714}
2715
2716static int
2717STALE_STATEID(stateid_t *stateid)
2718{
78155ed7
BN
2719 if (time_after((unsigned long)boot_time,
2720 (unsigned long)stateid->si_boot)) {
2721 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2722 stateid->si_boot, stateid->si_stateownerid,
2723 stateid->si_fileid, stateid->si_generation);
2724 return 1;
2725 }
2726 return 0;
2727}
2728
2729static int
2730EXPIRED_STATEID(stateid_t *stateid)
2731{
2732 if (time_before((unsigned long)boot_time,
2733 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2734 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN
2735 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2736 stateid->si_boot, stateid->si_stateownerid,
2737 stateid->si_fileid, stateid->si_generation);
2738 return 1;
2739 }
2740 return 0;
2741}
2742
2743static __be32
2744stateid_error_map(stateid_t *stateid)
2745{
2746 if (STALE_STATEID(stateid))
2747 return nfserr_stale_stateid;
2748 if (EXPIRED_STATEID(stateid))
2749 return nfserr_expired;
2750
2751 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2752 stateid->si_boot, stateid->si_stateownerid,
2753 stateid->si_fileid, stateid->si_generation);
2754 return nfserr_bad_stateid;
1da177e4
LT
2755}
2756
2757static inline int
2758access_permit_read(unsigned long access_bmap)
2759{
2760 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2761 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2762 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2763}
2764
2765static inline int
2766access_permit_write(unsigned long access_bmap)
2767{
2768 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2769 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2770}
2771
2772static
b37ad28b 2773__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2774{
b37ad28b 2775 __be32 status = nfserr_openmode;
1da177e4
LT
2776
2777 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2778 goto out;
2779 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2780 goto out;
2781 status = nfs_ok;
2782out:
2783 return status;
2784}
2785
b37ad28b 2786static inline __be32
1da177e4
LT
2787check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2788{
203a8c8e 2789 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2790 return nfs_ok;
af558e33 2791 else if (locks_in_grace()) {
1da177e4
LT
2792 /* Answer in remaining cases depends on existance of
2793 * conflicting state; so we must wait out the grace period. */
2794 return nfserr_grace;
2795 } else if (flags & WR_STATE)
2796 return nfs4_share_conflict(current_fh,
2797 NFS4_SHARE_DENY_WRITE);
2798 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2799 return nfs4_share_conflict(current_fh,
2800 NFS4_SHARE_DENY_READ);
2801}
2802
2803/*
2804 * Allow READ/WRITE during grace period on recovered state only for files
2805 * that are not able to provide mandatory locking.
2806 */
2807static inline int
18f82731 2808grace_disallows_io(struct inode *inode)
1da177e4 2809{
203a8c8e 2810 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT
2811}
2812
6668958f 2813static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2814{
6668958f
AA
2815 /*
2816 * When sessions are used the stateid generation number is ignored
2817 * when it is zero.
2818 */
2819 if ((flags & HAS_SESSION) && in->si_generation == 0)
2820 goto out;
2821
0836f587
BF
2822 /* If the client sends us a stateid from the future, it's buggy: */
2823 if (in->si_generation > ref->si_generation)
2824 return nfserr_bad_stateid;
2825 /*
2826 * The following, however, can happen. For example, if the
2827 * client sends an open and some IO at the same time, the open
2828 * may bump si_generation while the IO is still in flight.
2829 * Thanks to hard links and renames, the client never knows what
2830 * file an open will affect. So it could avoid that situation
2831 * only by serializing all opens and IO from the same open
2832 * owner. To recover from the old_stateid error, the client
2833 * will just have to retry the IO:
2834 */
2835 if (in->si_generation < ref->si_generation)
2836 return nfserr_old_stateid;
6668958f 2837out:
0836f587
BF
2838 return nfs_ok;
2839}
2840
3e633079
BF
2841static int is_delegation_stateid(stateid_t *stateid)
2842{
2843 return stateid->si_fileid == 0;
2844}
2845
1da177e4
LT
2846/*
2847* Checks for stateid operations
2848*/
b37ad28b 2849__be32
dd453dfd
BH
2850nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2851 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT
2852{
2853 struct nfs4_stateid *stp = NULL;
2854 struct nfs4_delegation *dp = NULL;
dd453dfd 2855 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2856 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2857 __be32 status;
1da177e4 2858
1da177e4
LT
2859 if (filpp)
2860 *filpp = NULL;
2861
18f82731 2862 if (grace_disallows_io(ino))
1da177e4
LT
2863 return nfserr_grace;
2864
6668958f
AA
2865 if (nfsd4_has_session(cstate))
2866 flags |= HAS_SESSION;
2867
1da177e4
LT
2868 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2869 return check_special_stateids(current_fh, stateid, flags);
2870
1da177e4
LT
2871 status = nfserr_stale_stateid;
2872 if (STALE_STATEID(stateid))
2873 goto out;
2874
1da177e4 2875 status = nfserr_bad_stateid;
3e633079 2876 if (is_delegation_stateid(stateid)) {
a4455be0 2877 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN
2878 if (!dp) {
2879 status = stateid_error_map(stateid);
1da177e4 2880 goto out;
78155ed7 2881 }
6668958f
AA
2882 status = check_stateid_generation(stateid, &dp->dl_stateid,
2883 flags);
0c2a498f
BF
2884 if (status)
2885 goto out;
dc9bf700
BF
2886 status = nfs4_check_delegmode(dp, flags);
2887 if (status)
2888 goto out;
2889 renew_client(dp->dl_client);
dc9bf700
BF
2890 if (filpp)
2891 *filpp = dp->dl_vfs_file;
1da177e4 2892 } else { /* open or lock stateid */
a4455be0 2893 stp = find_stateid(stateid, flags);
78155ed7
BN
2894 if (!stp) {
2895 status = stateid_error_map(stateid);
1da177e4 2896 goto out;
78155ed7 2897 }
6150ef0d 2898 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT
2899 goto out;
2900 if (!stp->st_stateowner->so_confirmed)
2901 goto out;
6668958f
AA
2902 status = check_stateid_generation(stateid, &stp->st_stateid,
2903 flags);
0c2a498f
BF
2904 if (status)
2905 goto out;
a4455be0
BF
2906 status = nfs4_check_openmode(stp, flags);
2907 if (status)
1da177e4
LT
2908 goto out;
2909 renew_client(stp->st_stateowner->so_client);
2910 if (filpp)
2911 *filpp = stp->st_vfs_file;
1da177e4
LT
2912 }
2913 status = nfs_ok;
2914out:
2915 return status;
2916}
2917
4c4cd222
N
2918static inline int
2919setlkflg (int type)
2920{
2921 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2922 RD_STATE : WR_STATE;
2923}
1da177e4
LT
2924
2925/*
2926 * Checks for sequence id mutating operations.
2927 */
b37ad28b 2928static __be32
dd453dfd
BH
2929nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2930 stateid_t *stateid, int flags,
2931 struct nfs4_stateowner **sopp,
2932 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2933{
1da177e4
LT
2934 struct nfs4_stateid *stp;
2935 struct nfs4_stateowner *sop;
dd453dfd 2936 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2937 __be32 status;
1da177e4
LT
2938
2939 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2940 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2941 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2942 stateid->si_generation);
3a4f98bb 2943
1da177e4
LT
2944 *stpp = NULL;
2945 *sopp = NULL;
2946
1da177e4 2947 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2948 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2949 return nfserr_bad_stateid;
1da177e4
LT
2950 }
2951
1da177e4 2952 if (STALE_STATEID(stateid))
3a4f98bb 2953 return nfserr_stale_stateid;
6668958f
AA
2954
2955 if (nfsd4_has_session(cstate))
2956 flags |= HAS_SESSION;
2957
1da177e4
LT
2958 /*
2959 * We return BAD_STATEID if filehandle doesn't match stateid,
2960 * the confirmed flag is incorrecly set, or the generation
2961 * number is incorrect.
1da177e4 2962 */
f8816512
N
2963 stp = find_stateid(stateid, flags);
2964 if (stp == NULL) {
2965 /*
2966 * Also, we should make sure this isn't just the result of
2967 * a replayed close:
2968 */
2969 sop = search_close_lru(stateid->si_stateownerid, flags);
2970 if (sop == NULL)
78155ed7 2971 return stateid_error_map(stateid);
f8816512
N
2972 *sopp = sop;
2973 goto check_replay;
2974 }
1da177e4 2975
39325bd0
BF
2976 *stpp = stp;
2977 *sopp = sop = stp->st_stateowner;
2978
4c4cd222 2979 if (lock) {
4c4cd222 2980 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2981 struct nfs4_client *clp = sop->so_client;
4c4cd222 2982 int lkflg = 0;
b37ad28b 2983 __be32 status;
4c4cd222
N
2984
2985 lkflg = setlkflg(lock->lk_type);
2986
2987 if (lock->lk_is_new) {
599e0a22
BF
2988 if (!sop->so_is_open_owner)
2989 return nfserr_bad_stateid;
60adfc50
AA
2990 if (!(flags & HAS_SESSION) &&
2991 !same_clid(&clp->cl_clientid, lockclid))
2992 return nfserr_bad_stateid;
599e0a22
BF
2993 /* stp is the open stateid */
2994 status = nfs4_check_openmode(stp, lkflg);
2995 if (status)
2996 return status;
2997 } else {
2998 /* stp is the lock stateid */
2999 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3000 if (status)
3001 return status;
4c4cd222 3002 }
1da177e4
LT
3003 }
3004
f3362737 3005 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3006 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3007 return nfserr_bad_stateid;
1da177e4
LT
3008 }
3009
1da177e4
LT
3010 /*
3011 * We now validate the seqid and stateid generation numbers.
3012 * For the moment, we ignore the possibility of
3013 * generation number wraparound.
3014 */
6668958f 3015 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT
3016 goto check_replay;
3017
3a4f98bb 3018 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3019 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N
3020 " unconfirmed stateowner!\n");
3021 return nfserr_bad_stateid;
1da177e4 3022 }
3a4f98bb 3023 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3024 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N
3025 " confirmed yet!\n");
3026 return nfserr_bad_stateid;
1da177e4 3027 }
6668958f 3028 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF
3029 if (status)
3030 return status;
52fd004e 3031 renew_client(sop->so_client);
3a4f98bb 3032 return nfs_ok;
1da177e4 3033
1da177e4 3034check_replay:
bd9aac52 3035 if (seqid == sop->so_seqid - 1) {
849823c5 3036 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3037 /* indicate replay to calling function */
a90b061c 3038 return nfserr_replay_me;
1da177e4 3039 }
2fdada03 3040 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N
3041 sop->so_seqid, seqid);
3042 *sopp = NULL;
3043 return nfserr_bad_seqid;
1da177e4
LT
3044}
3045
b37ad28b 3046__be32
ca364317 3047nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3048 struct nfsd4_open_confirm *oc)
1da177e4 3049{
b37ad28b 3050 __be32 status;
1da177e4
LT
3051 struct nfs4_stateowner *sop;
3052 struct nfs4_stateid *stp;
3053
3054 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF
3055 (int)cstate->current_fh.fh_dentry->d_name.len,
3056 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3057
ca364317 3058 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF
3059 if (status)
3060 return status;
1da177e4
LT
3061
3062 nfs4_lock_state();
3063
dd453dfd 3064 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3065 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3066 CONFIRM | OPEN_STATE,
1da177e4
LT
3067 &oc->oc_stateowner, &stp, NULL)))
3068 goto out;
3069
3070 sop = oc->oc_stateowner;
3071 sop->so_confirmed = 1;
3072 update_stateid(&stp->st_stateid);
3073 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3074 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3075 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3076 stp->st_stateid.si_boot,
3077 stp->st_stateid.si_stateownerid,
3078 stp->st_stateid.si_fileid,
3079 stp->st_stateid.si_generation);
c7b9a459
N
3080
3081 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3082out:
f2327d9a 3083 if (oc->oc_stateowner) {
1da177e4 3084 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3085 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3086 }
1da177e4
LT
3087 nfs4_unlock_state();
3088 return status;
3089}
3090
3091
3092/*
3093 * unset all bits in union bitmap (bmap) that
3094 * do not exist in share (from successful OPEN_DOWNGRADE)
3095 */
3096static void
3097reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3098{
3099 int i;
3100 for (i = 1; i < 4; i++) {
3101 if ((i & access) != i)
3102 __clear_bit(i, bmap);
3103 }
3104}
3105
3106static void
3107reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3108{
3109 int i;
3110 for (i = 0; i < 4; i++) {
3111 if ((i & deny) != i)
3112 __clear_bit(i, bmap);
3113 }
3114}
3115
b37ad28b 3116__be32
ca364317
BF
3117nfsd4_open_downgrade(struct svc_rqst *rqstp,
3118 struct nfsd4_compound_state *cstate,
a4f1706a 3119 struct nfsd4_open_downgrade *od)
1da177e4 3120{
b37ad28b 3121 __be32 status;
1da177e4
LT
3122 struct nfs4_stateid *stp;
3123 unsigned int share_access;
3124
3125 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF
3126 (int)cstate->current_fh.fh_dentry->d_name.len,
3127 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3128
d87a8ade 3129 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3130 || !deny_valid(od->od_share_deny))
1da177e4
LT
3131 return nfserr_inval;
3132
3133 nfs4_lock_state();
dd453dfd 3134 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3135 od->od_seqid,
1da177e4 3136 &od->od_stateid,
f3362737 3137 OPEN_STATE,
1da177e4
LT
3138 &od->od_stateowner, &stp, NULL)))
3139 goto out;
3140
3141 status = nfserr_inval;
3142 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3143 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3144 stp->st_access_bmap, od->od_share_access);
3145 goto out;
3146 }
3147 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3148 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3149 stp->st_deny_bmap, od->od_share_deny);
3150 goto out;
3151 }
3152 set_access(&share_access, stp->st_access_bmap);
3153 nfs4_file_downgrade(stp->st_vfs_file,
3154 share_access & ~od->od_share_access);
3155
3156 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3157 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3158
3159 update_stateid(&stp->st_stateid);
3160 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3161 status = nfs_ok;
3162out:
f2327d9a 3163 if (od->od_stateowner) {
1da177e4 3164 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3165 cstate->replay_owner = od->od_stateowner;
f2327d9a 3166 }
1da177e4
LT
3167 nfs4_unlock_state();
3168 return status;
3169}
3170
3171/*
3172 * nfs4_unlock_state() called after encode
3173 */
b37ad28b 3174__be32
ca364317 3175nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3176 struct nfsd4_close *close)
1da177e4 3177{
b37ad28b 3178 __be32 status;
1da177e4
LT
3179 struct nfs4_stateid *stp;
3180
3181 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF
3182 (int)cstate->current_fh.fh_dentry->d_name.len,
3183 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT
3184
3185 nfs4_lock_state();
3186 /* check close_lru for replay */
dd453dfd 3187 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3188 close->cl_seqid,
1da177e4 3189 &close->cl_stateid,
f3362737 3190 OPEN_STATE | CLOSE_STATE,
1da177e4
LT
3191 &close->cl_stateowner, &stp, NULL)))
3192 goto out;
1da177e4
LT
3193 status = nfs_ok;
3194 update_stateid(&stp->st_stateid);
3195 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3196
04ef5954 3197 /* release_stateid() calls nfsd_close() if needed */
2283963f 3198 release_open_stateid(stp);
04ef5954
BF
3199
3200 /* place unused nfs4_stateowners on so_close_lru list to be
3201 * released by the laundromat service after the lease period
3202 * to enable us to handle CLOSE replay
3203 */
3204 if (list_empty(&close->cl_stateowner->so_stateids))
3205 move_to_close_lru(close->cl_stateowner);
1da177e4 3206out:
f2327d9a 3207 if (close->cl_stateowner) {
1da177e4 3208 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3209 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3210 }
1da177e4
LT
3211 nfs4_unlock_state();
3212 return status;
3213}
3214
b37ad28b 3215__be32
ca364317
BF
3216nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3217 struct nfsd4_delegreturn *dr)
1da177e4 3218{
203a8c8e
BF
3219 struct nfs4_delegation *dp;
3220 stateid_t *stateid = &dr->dr_stateid;
3221 struct inode *inode;
b37ad28b 3222 __be32 status;
6668958f 3223 int flags = 0;
1da177e4 3224
ca364317 3225 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF
3226 return status;
3227 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3228
6668958f
AA
3229 if (nfsd4_has_session(cstate))
3230 flags |= HAS_SESSION;
1da177e4 3231 nfs4_lock_state();
203a8c8e
BF
3232 status = nfserr_bad_stateid;
3233 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3234 goto out;
3235 status = nfserr_stale_stateid;
3236 if (STALE_STATEID(stateid))
3237 goto out;
7e0f7cf5 3238 status = nfserr_bad_stateid;
203a8c8e
BF
3239 if (!is_delegation_stateid(stateid))
3240 goto out;
203a8c8e 3241 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN
3242 if (!dp) {
3243 status = stateid_error_map(stateid);
203a8c8e 3244 goto out;
78155ed7 3245 }
6668958f 3246 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF
3247 if (status)
3248 goto out;
3249 renew_client(dp->dl_client);
3250
3251 unhash_delegation(dp);
1da177e4 3252out:
203a8c8e
BF
3253 nfs4_unlock_state();
3254
1da177e4
LT
3255 return status;
3256}
3257
3258
3259/*
3260 * Lock owner state (byte-range locks)
3261 */
3262#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3263#define LOCK_HASH_BITS 8
3264#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3265#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3266
87df4de8
BH
3267static inline u64
3268end_offset(u64 start, u64 len)
3269{
3270 u64 end;
3271
3272 end = start + len;
3273 return end >= start ? end: NFS4_MAX_UINT64;
3274}
3275
3276/* last octet in a range */
3277static inline u64
3278last_byte_offset(u64 start, u64 len)
3279{
3280 u64 end;
3281
3282 BUG_ON(!len);
3283 end = start + len;
3284 return end > start ? end - 1: NFS4_MAX_UINT64;
3285}
3286
1da177e4
LT
3287#define lockownerid_hashval(id) \
3288 ((id) & LOCK_HASH_MASK)
3289
3290static inline unsigned int
3291lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3292 struct xdr_netobj *ownername)
3293{
3294 return (file_hashval(inode) + cl_id
3295 + opaque_hashval(ownername->data, ownername->len))
3296 & LOCK_HASH_MASK;
3297}
3298
3299static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3300static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3301static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3302
fd39ca9a 3303static struct nfs4_stateid *
1da177e4
LT
3304find_stateid(stateid_t *stid, int flags)
3305{
9346eff0 3306 struct nfs4_stateid *local;
1da177e4
LT
3307 u32 st_id = stid->si_stateownerid;
3308 u32 f_id = stid->si_fileid;
3309 unsigned int hashval;
3310
3311 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3312 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3313 hashval = stateid_hashval(st_id, f_id);
3314 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3315 if ((local->st_stateid.si_stateownerid == st_id) &&
3316 (local->st_stateid.si_fileid == f_id))
3317 return local;
3318 }
3319 }
9346eff0
KK
3320
3321 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3322 hashval = stateid_hashval(st_id, f_id);
3323 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3324 if ((local->st_stateid.si_stateownerid == st_id) &&
3325 (local->st_stateid.si_fileid == f_id))
3326 return local;
3327 }
849823c5 3328 }
1da177e4
LT
3329 return NULL;
3330}
3331
3332static struct nfs4_delegation *
3333find_delegation_stateid(struct inode *ino, stateid_t *stid)
3334{
13cd2184
N
3335 struct nfs4_file *fp;
3336 struct nfs4_delegation *dl;
1da177e4
LT
3337
3338 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3339 stid->si_boot, stid->si_stateownerid,
3340 stid->si_fileid, stid->si_generation);
3341
1da177e4 3342 fp = find_file(ino);
13cd2184
N
3343 if (!fp)
3344 return NULL;
3345 dl = find_delegation_file(fp, stid);
3346 put_nfs4_file(fp);
3347 return dl;
1da177e4
LT
3348}
3349
3350/*
3351 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3352 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3353 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3354 * locking, this prevents us from being completely protocol-compliant. The
3355 * real solution to this problem is to start using unsigned file offsets in
3356 * the VFS, but this is a very deep change!
3357 */
3358static inline void
3359nfs4_transform_lock_offset(struct file_lock *lock)
3360{
3361 if (lock->fl_start < 0)
3362 lock->fl_start = OFFSET_MAX;
3363 if (lock->fl_end < 0)
3364 lock->fl_end = OFFSET_MAX;
3365}
3366
d5b9026a
N
3367/* Hack!: For now, we're defining this just so we can use a pointer to it
3368 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3369static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3370};
1da177e4
LT
3371
3372static inline void
3373nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3374{
d5b9026a
N
3375 struct nfs4_stateowner *sop;
3376 unsigned int hval;
1da177e4 3377
d5b9026a
N
3378 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3379 sop = (struct nfs4_stateowner *) fl->fl_owner;
3380 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT
3381 kref_get(&sop->so_ref);
3382 deny->ld_sop = sop;
3383 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N
3384 } else {
3385 deny->ld_sop = NULL;
3386 deny->ld_clientid.cl_boot = 0;
3387 deny->ld_clientid.cl_id = 0;
1da177e4
LT
3388 }
3389 deny->ld_start = fl->fl_start;
87df4de8
BH
3390 deny->ld_length = NFS4_MAX_UINT64;
3391 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT
3392 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3393 deny->ld_type = NFS4_READ_LT;
3394 if (fl->fl_type != F_RDLCK)
3395 deny->ld_type = NFS4_WRITE_LT;
3396}
3397
1da177e4
LT
3398static struct nfs4_stateowner *
3399find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3400 struct xdr_netobj *owner)
3401{
3402 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3403 struct nfs4_stateowner *op;
3404
3405 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3406 if (same_owner_str(op, owner, clid))
1da177e4
LT
3407 return op;
3408 }
3409 return NULL;
3410}
3411
3412/*
3413 * Alloc a lock owner structure.
3414 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3415 * occured.
3416 *
3417 * strhashval = lock_ownerstr_hashval
1da177e4
LT
3418 */
3419
3420static struct nfs4_stateowner *
3421alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3422 struct nfs4_stateowner *sop;
3423 struct nfs4_replay *rp;
3424 unsigned int idhashval;
3425
3426 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3427 return NULL;
3428 idhashval = lockownerid_hashval(current_ownerid);
3429 INIT_LIST_HEAD(&sop->so_idhash);
3430 INIT_LIST_HEAD(&sop->so_strhash);
3431 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
3432 INIT_LIST_HEAD(&sop->so_stateids);
3433 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT
3434 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3435 sop->so_time = 0;
3436 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3437 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3438 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT
3439 sop->so_is_open_owner = 0;
3440 sop->so_id = current_ownerid++;
3441 sop->so_client = clp;
b59e3c0e
NB
3442 /* It is the openowner seqid that will be incremented in encode in the
3443 * case of new lockowners; so increment the lock seqid manually: */
3444 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT
3445 sop->so_confirmed = 1;
3446 rp = &sop->so_replay;
de1ae286 3447 rp->rp_status = nfserr_serverfault;
1da177e4
LT
3448 rp->rp_buflen = 0;
3449 rp->rp_buf = rp->rp_ibuf;
3450 return sop;
3451}
3452
fd39ca9a 3453static struct nfs4_stateid *
1da177e4
LT
3454alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3455{
3456 struct nfs4_stateid *stp;
3457 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3458
5ac049ac
N
3459 stp = nfs4_alloc_stateid();
3460 if (stp == NULL)
1da177e4
LT
3461 goto out;
3462 INIT_LIST_HEAD(&stp->st_hash);
3463 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N
3464 INIT_LIST_HEAD(&stp->st_perstateowner);
3465 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3466 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3467 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3468 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3469 stp->st_stateowner = sop;
13cd2184 3470 get_nfs4_file(fp);
1da177e4 3471 stp->st_file = fp;
78155ed7 3472 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
3473 stp->st_stateid.si_stateownerid = sop->so_id;
3474 stp->st_stateid.si_fileid = fp->fi_id;
3475 stp->st_stateid.si_generation = 0;
3476 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3477 stp->st_access_bmap = open_stp->st_access_bmap;
3478 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3479 stp->st_openstp = open_stp;
1da177e4
LT
3480
3481out:
3482 return stp;
3483}
3484
fd39ca9a 3485static int
1da177e4
LT
3486check_lock_length(u64 offset, u64 length)
3487{
87df4de8 3488 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT
3489 LOFF_OVERFLOW(offset, length)));
3490}
3491
3492/*
3493 * LOCK operation
3494 */
b37ad28b 3495__be32
ca364317 3496nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3497 struct nfsd4_lock *lock)
1da177e4 3498{
3e9e3dbe 3499 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3500 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT
3501 struct nfs4_stateid *lock_stp;
3502 struct file *filp;
3503 struct file_lock file_lock;
8dc7c311 3504 struct file_lock conflock;
b37ad28b 3505 __be32 status = 0;
1da177e4 3506 unsigned int strhashval;
150b3934 3507 unsigned int cmd;
b8dd7b9a 3508 int err;
1da177e4
LT
3509
3510 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3511 (long long) lock->lk_offset,
3512 (long long) lock->lk_length);
3513
1da177e4
LT
3514 if (check_lock_length(lock->lk_offset, lock->lk_length))
3515 return nfserr_inval;
3516
ca364317 3517 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3518 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA
3519 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3520 return status;
3521 }
3522
1da177e4
LT
3523 nfs4_lock_state();
3524
3525 if (lock->lk_is_new) {
893f8770
N
3526 /*
3527 * Client indicates that this is a new lockowner.
3528 * Use open owner and open stateid to create lock owner and
3529 * lock stateid.
3530 */
1da177e4
LT
3531 struct nfs4_stateid *open_stp = NULL;
3532 struct nfs4_file *fp;
3533
3534 status = nfserr_stale_clientid;
60adfc50
AA
3535 if (!nfsd4_has_session(cstate) &&
3536 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3537 goto out;
1da177e4 3538
1da177e4 3539 /* validate and update open stateid and open seqid */
dd453dfd 3540 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3541 lock->lk_new_open_seqid,
3542 &lock->lk_new_open_stateid,
f3362737 3543 OPEN_STATE,
3a65588a 3544 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3545 lock);
37515177 3546 if (status)
1da177e4 3547 goto out;
3a65588a 3548 open_sop = lock->lk_replay_owner;
1da177e4
LT
3549 /* create lockowner and lock stateid */
3550 fp = open_stp->st_file;
3551 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3552 open_sop->so_client->cl_clientid.cl_id,
3553 &lock->v.new.owner);
3e9e3dbe
N
3554 /* XXX: Do we need to check for duplicate stateowners on
3555 * the same file, or should they just be allowed (and
3556 * create new stateids)? */
1da177e4 3557 status = nfserr_resource;
b59e3c0e
NB
3558 lock_sop = alloc_init_lock_stateowner(strhashval,
3559 open_sop->so_client, open_stp, lock);
3560 if (lock_sop == NULL)
1da177e4 3561 goto out;
b59e3c0e 3562 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3563 if (lock_stp == NULL)
1da177e4 3564 goto out;
1da177e4
LT
3565 } else {
3566 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3567 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3568 lock->lk_old_lock_seqid,
3569 &lock->lk_old_lock_stateid,
f3362737 3570 LOCK_STATE,
3a65588a 3571 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT
3572 if (status)
3573 goto out;
3a65588a 3574 lock_sop = lock->lk_replay_owner;
1da177e4 3575 }
3a65588a 3576 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT
3577 filp = lock_stp->st_vfs_file;
3578
0dd395dc 3579 status = nfserr_grace;
af558e33 3580 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N
3581 goto out;
3582 status = nfserr_no_grace;
af558e33 3583 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N
3584 goto out;
3585
1da177e4
LT
3586 locks_init_lock(&file_lock);
3587 switch (lock->lk_type) {
3588 case NFS4_READ_LT:
3589 case NFS4_READW_LT:
3590 file_lock.fl_type = F_RDLCK;
150b3934 3591 cmd = F_SETLK;
1da177e4
LT
3592 break;
3593 case NFS4_WRITE_LT:
3594 case NFS4_WRITEW_LT:
3595 file_lock.fl_type = F_WRLCK;
150b3934 3596 cmd = F_SETLK;
1da177e4
LT
3597 break;
3598 default:
3599 status = nfserr_inval;
3600 goto out;
3601 }
b59e3c0e 3602 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT
3603 file_lock.fl_pid = current->tgid;
3604 file_lock.fl_file = filp;
3605 file_lock.fl_flags = FL_POSIX;
d5b9026a 3606 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3607
3608 file_lock.fl_start = lock->lk_offset;
87df4de8 3609 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT
3610 nfs4_transform_lock_offset(&file_lock);
3611
3612 /*
3613 * Try to lock the file in the VFS.
3614 * Note: locks.c uses the BKL to protect the inode's lock list.
3615 */
3616
fd85b817 3617 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3618 switch (-err) {
1da177e4
LT
3619 case 0: /* success! */
3620 update_stateid(&lock_stp->st_stateid);
3621 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3622 sizeof(stateid_t));
b8dd7b9a 3623 status = 0;
eb76b3fd
AA
3624 break;
3625 case (EAGAIN): /* conflock holds conflicting lock */
3626 status = nfserr_denied;
3627 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3628 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3629 break;
1da177e4
LT
3630 case (EDEADLK):
3631 status = nfserr_deadlock;
eb76b3fd 3632 break;
1da177e4 3633 default:
fd85b817 3634 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA
3635 status = nfserr_resource;
3636 break;
1da177e4 3637 }
1da177e4 3638out:
8a280510 3639 if (status && lock->lk_is_new && lock_sop)
f044ff83 3640 release_lockowner(lock_sop);
3a65588a
BF
3641 if (lock->lk_replay_owner) {
3642 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3643 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3644 }
1da177e4
LT
3645 nfs4_unlock_state();
3646 return status;
3647}
3648
55ef1274
BF
3649/*
3650 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3651 * so we do a temporary open here just to get an open file to pass to
3652 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3653 * inode operation.)
3654 */
3655static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3656{
3657 struct file *file;
3658 int err;
3659
3660 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3661 if (err)
3662 return err;
3663 err = vfs_test_lock(file, lock);
3664 nfsd_close(file);
3665 return err;
3666}
3667
1da177e4
LT
3668/*
3669 * LOCKT operation
3670 */
b37ad28b 3671__be32
ca364317
BF
3672nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3673 struct nfsd4_lockt *lockt)
1da177e4
LT
3674{
3675 struct inode *inode;
1da177e4 3676 struct file_lock file_lock;
fd85b817 3677 int error;
b37ad28b 3678 __be32 status;
1da177e4 3679
af558e33 3680 if (locks_in_grace())
1da177e4
LT
3681 return nfserr_grace;
3682
3683 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3684 return nfserr_inval;
3685
3686 lockt->lt_stateowner = NULL;
3687 nfs4_lock_state();
3688
3689 status = nfserr_stale_clientid;
60adfc50 3690 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3691 goto out;
1da177e4 3692
ca364317 3693 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3694 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT
3695 if (status == nfserr_symlink)
3696 status = nfserr_inval;
3697 goto out;
3698 }
3699
ca364317 3700 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
3701 locks_init_lock(&file_lock);
3702 switch (lockt->lt_type) {
3703 case NFS4_READ_LT:
3704 case NFS4_READW_LT:
3705 file_lock.fl_type = F_RDLCK;
3706 break;
3707 case NFS4_WRITE_LT:
3708 case NFS4_WRITEW_LT:
3709 file_lock.fl_type = F_WRLCK;
3710 break;
3711 default:
2fdada03 3712 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT
3713 status = nfserr_inval;
3714 goto out;
3715 }
3716
3717 lockt->lt_stateowner = find_lockstateowner_str(inode,
3718 &lockt->lt_clientid, &lockt->lt_owner);
3719 if (lockt->lt_stateowner)
3720 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3721 file_lock.fl_pid = current->tgid;
3722 file_lock.fl_flags = FL_POSIX;
3723
3724 file_lock.fl_start = lockt->lt_offset;
87df4de8 3725 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT
3726
3727 nfs4_transform_lock_offset(&file_lock);
3728
1da177e4 3729 status = nfs_ok;
55ef1274 3730 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME
3731 if (error) {
3732 status = nfserrno(error);
3733 goto out;
3734 }
9d6a8c5c 3735 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3736 status = nfserr_denied;
9d6a8c5c 3737 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT
3738 }
3739out:
3740 nfs4_unlock_state();
3741 return status;
3742}
3743
b37ad28b 3744__be32
ca364317 3745nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3746 struct nfsd4_locku *locku)
1da177e4
LT
3747{
3748 struct nfs4_stateid *stp;
3749 struct file *filp = NULL;
3750 struct file_lock file_lock;
b37ad28b 3751 __be32 status;
b8dd7b9a 3752 int err;
1da177e4
LT
3753
3754 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3755 (long long) locku->lu_offset,
3756 (long long) locku->lu_length);
3757
3758 if (check_lock_length(locku->lu_offset, locku->lu_length))
3759 return nfserr_inval;
3760
3761 nfs4_lock_state();
3762
dd453dfd 3763 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3764 locku->lu_seqid,
3765 &locku->lu_stateid,
f3362737 3766 LOCK_STATE,
1da177e4
LT
3767 &locku->lu_stateowner, &stp, NULL)))
3768 goto out;
3769
3770 filp = stp->st_vfs_file;
3771 BUG_ON(!filp);
3772 locks_init_lock(&file_lock);
3773 file_lock.fl_type = F_UNLCK;
3774 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3775 file_lock.fl_pid = current->tgid;
3776 file_lock.fl_file = filp;
3777 file_lock.fl_flags = FL_POSIX;
d5b9026a 3778 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3779 file_lock.fl_start = locku->lu_offset;
3780
87df4de8 3781 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT
3782 nfs4_transform_lock_offset(&file_lock);
3783
3784 /*
3785 * Try to unlock the file in the VFS.
3786 */
fd85b817 3787 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3788 if (err) {
fd85b817 3789 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT
3790 goto out_nfserr;
3791 }
3792 /*
3793 * OK, unlock succeeded; the only thing left to do is update the stateid.
3794 */
3795 update_stateid(&stp->st_stateid);
3796 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3797
3798out:
f2327d9a 3799 if (locku->lu_stateowner) {
1da177e4 3800 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3801 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3802 }
1da177e4
LT
3803 nfs4_unlock_state();
3804 return status;
3805
3806out_nfserr:
b8dd7b9a 3807 status = nfserrno(err);
1da177e4
LT
3808 goto out;
3809}
3810
3811/*
3812 * returns
3813 * 1: locks held by lockowner
3814 * 0: no locks held by lockowner
3815 */
3816static int
3817check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3818{
3819 struct file_lock **flpp;
7eaa36e2 3820 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT
3821 int status = 0;
3822
3823 lock_kernel();
3824 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3825 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT
3826 status = 1;
3827 goto out;
796dadfd 3828 }
1da177e4
LT
3829 }
3830out:
3831 unlock_kernel();
3832 return status;
3833}
3834
b37ad28b 3835__be32
b591480b
BF
3836nfsd4_release_lockowner(struct svc_rqst *rqstp,
3837 struct nfsd4_compound_state *cstate,
3838 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT
3839{
3840 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N
3841 struct nfs4_stateowner *sop;
3842 struct nfs4_stateid *stp;
1da177e4 3843 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N
3844 struct list_head matches;
3845 int i;
b37ad28b 3846 __be32 status;
1da177e4
LT
3847
3848 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3849 clid->cl_boot, clid->cl_id);
3850
3851 /* XXX check for lease expiration */
3852
3853 status = nfserr_stale_clientid;
849823c5 3854 if (STALE_CLIENTID(clid))
1da177e4 3855 return status;
1da177e4
LT
3856
3857 nfs4_lock_state();
3858
3e9e3dbe
N
3859 status = nfserr_locks_held;
3860 /* XXX: we're doing a linear search through all the lockowners.
3861 * Yipes! For now we'll just hope clients aren't really using
3862 * release_lockowner much, but eventually we have to fix these
3863 * data structures. */
3864 INIT_LIST_HEAD(&matches);
3865 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3866 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3867 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N
3868 continue;
3869 list_for_each_entry(stp, &sop->so_stateids,
3870 st_perstateowner) {
3871 if (check_for_locks(stp->st_vfs_file, sop))
3872 goto out;
3873 /* Note: so_perclient unused for lockowners,
3874 * so it's OK to fool with here. */
3875 list_add(&sop->so_perclient, &matches);
3876 }
1da177e4 3877 }
3e9e3dbe
N
3878 }
3879 /* Clients probably won't expect us to return with some (but not all)
3880 * of the lockowner state released; so don't release any until all
3881 * have been checked. */
3882 status = nfs_ok;
0fa822e4
N
3883 while (!list_empty(&matches)) {
3884 sop = list_entry(matches.next, struct nfs4_stateowner,
3885 so_perclient);
3886 /* unhash_stateowner deletes so_perclient only
3887 * for openowners. */
3888 list_del(&sop->so_perclient);
f044ff83 3889 release_lockowner(sop);
1da177e4
LT
3890 }
3891out:
3892 nfs4_unlock_state();
3893 return status;
3894}
3895
3896static inline struct nfs4_client_reclaim *
a55370a3 3897alloc_reclaim(void)
1da177e4 3898{
a55370a3 3899 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT
3900}
3901
c7b9a459 3902int
a1bcecd2 3903nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N
3904{
3905 unsigned int strhashval = clientstr_hashval(name);
3906 struct nfs4_client *clp;
3907
a1bcecd2 3908 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N
3909 return clp ? 1 : 0;
3910}
3911
1da177e4
LT
3912/*
3913 * failure => all reset bets are off, nfserr_no_grace...
3914 */
190e4fbf
N
3915int
3916nfs4_client_to_reclaim(const char *name)
1da177e4
LT
3917{
3918 unsigned int strhashval;
3919 struct nfs4_client_reclaim *crp = NULL;
3920
a55370a3
N
3921 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3922 crp = alloc_reclaim();
1da177e4
LT
3923 if (!crp)
3924 return 0;
a55370a3 3925 strhashval = clientstr_hashval(name);
1da177e4
LT
3926 INIT_LIST_HEAD(&crp->cr_strhash);
3927 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3928 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT
3929 reclaim_str_hashtbl_size++;
3930 return 1;
3931}
3932
3933static void
3934nfs4_release_reclaim(void)
3935{
3936 struct nfs4_client_reclaim *crp = NULL;
3937 int i;
3938
1da177e4
LT
3939 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3940 while (!list_empty(&reclaim_str_hashtbl[i])) {
3941 crp = list_entry(reclaim_str_hashtbl[i].next,
3942 struct nfs4_client_reclaim, cr_strhash);
3943 list_del(&crp->cr_strhash);
1da177e4
LT
3944 kfree(crp);
3945 reclaim_str_hashtbl_size--;
3946 }
3947 }
3948 BUG_ON(reclaim_str_hashtbl_size);
3949}
3950
3951/*
3952 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3953static struct nfs4_client_reclaim *
1da177e4
LT
3954nfs4_find_reclaim_client(clientid_t *clid)
3955{
3956 unsigned int strhashval;
3957 struct nfs4_client *clp;
3958 struct nfs4_client_reclaim *crp = NULL;
3959
3960
3961 /* find clientid in conf_id_hashtbl */
3962 clp = find_confirmed_client(clid);
3963 if (clp == NULL)
3964 return NULL;
3965
a55370a3
N
3966 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3967 clp->cl_name.len, clp->cl_name.data,
3968 clp->cl_recdir);
1da177e4
LT
3969
3970 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3971 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3972 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3973 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT
3974 return crp;
3975 }
3976 }
3977 return NULL;
3978}
3979
3980/*
3981* Called from OPEN. Look for clientid in reclaim list.
3982*/
b37ad28b 3983__be32
1da177e4
LT
3984nfs4_check_open_reclaim(clientid_t *clid)
3985{
dfc83565 3986 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT
3987}
3988
ac4d8ff2 3989/* initialization to perform at module load time: */
1da177e4 3990
e8ff2a84 3991int
ac4d8ff2 3992nfs4_state_init(void)
1da177e4 3993{
e8ff2a84 3994 int i, status;
1da177e4 3995
e8ff2a84
BF
3996 status = nfsd4_init_slabs();
3997 if (status)
3998 return status;
1da177e4
LT
3999 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4000 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4001 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4002 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4003 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4004 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4005 }
5282fd72
ME
4006 for (i = 0; i < SESSION_HASH_SIZE; i++)
4007 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT
4008 for (i = 0; i < FILE_HASH_SIZE; i++) {
4009 INIT_LIST_HEAD(&file_hashtbl[i]);
4010 }
4011 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4012 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4013 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4014 }
4015 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4016 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4017 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4018 }
4019 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4020 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4021 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4022 }
1da177e4 4023 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT
4024 INIT_LIST_HEAD(&close_lru);
4025 INIT_LIST_HEAD(&client_lru);
4026 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4027 reclaim_str_hashtbl_size = 0;
e8ff2a84 4028 return 0;
ac4d8ff2
N
4029}
4030
190e4fbf
N
4031static void
4032nfsd4_load_reboot_recovery_data(void)
4033{
4034 int status;
4035
0964a3d3
N
4036 nfs4_lock_state();
4037 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4038 status = nfsd4_recdir_load();
0964a3d3 4039 nfs4_unlock_state();
190e4fbf
N
4040 if (status)
4041 printk("NFSD: Failure reading reboot recovery data\n");
4042}
4043
9a8db97e
ME
4044unsigned long
4045get_nfs4_grace_period(void)
4046{
4047 return max(user_lease_time, lease_time) * HZ;
4048}
4049
c2f1a551
MS
4050/*
4051 * Since the lifetime of a delegation isn't limited to that of an open, a
4052 * client may quite reasonably hang on to a delegation as long as it has
4053 * the inode cached. This becomes an obvious problem the first time a
4054 * client's inode cache approaches the size of the server's total memory.
4055 *
4056 * For now we avoid this problem by imposing a hard limit on the number
4057 * of delegations, which varies according to the server's memory size.
4058 */
4059static void
4060set_max_delegations(void)
4061{
4062 /*
4063 * Allow at most 4 delegations per megabyte of RAM. Quick
4064 * estimates suggest that in the worst case (where every delegation
4065 * is for a different inode), a delegation could take about 1.5K,
4066 * giving a worst case usage of about 6% of memory.
4067 */
4068 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4069}
4070
ac4d8ff2
N
4071/* initialization to perform when the nfsd service is started: */
4072
4073static void
4074__nfs4_state_start(void)
4075{
9a8db97e 4076 unsigned long grace_time;
ac4d8ff2 4077
1da177e4 4078 boot_time = get_seconds();
af558e33 4079 grace_time = get_nfs4_grace_period();
d99a05ad 4080 lease_time = user_lease_time;
af558e33 4081 locks_start_grace(&nfsd4_manager);
9a8db97e
ME
4082 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4083 grace_time/HZ);
58da282b 4084 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4085 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4086 set_max_delegations();
1da177e4
LT
4087}
4088
e8ff2a84 4089void
76a3550e 4090nfs4_state_start(void)
1da177e4 4091{
1da177e4 4092 if (nfs4_init)
e8ff2a84 4093 return;
190e4fbf 4094 nfsd4_load_reboot_recovery_data();
76a3550e 4095 __nfs4_state_start();
1da177e4 4096 nfs4_init = 1;
e8ff2a84 4097 return;
1da177e4
LT
4098}
4099
1da177e4
LT
4100time_t
4101nfs4_lease_time(void)
4102{
4103 return lease_time;
4104}
4105
4106static void
4107__nfs4_state_shutdown(void)
4108{
4109 int i;
4110 struct nfs4_client *clp = NULL;
4111 struct nfs4_delegation *dp = NULL;
1da177e4
LT
4112 struct list_head *pos, *next, reaplist;
4113
1da177e4
LT
4114 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4115 while (!list_empty(&conf_id_hashtbl[i])) {
4116 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4117 expire_client(clp);
4118 }
4119 while (!list_empty(&unconf_str_hashtbl[i])) {
4120 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4121 expire_client(clp);
4122 }
4123 }
4124 INIT_LIST_HEAD(&reaplist);
4125 spin_lock(&recall_lock);
4126 list_for_each_safe(pos, next, &del_recall_lru) {
4127 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4128 list_move(&dp->dl_recall_lru, &reaplist);
4129 }
4130 spin_unlock(&recall_lock);
4131 list_for_each_safe(pos, next, &reaplist) {
4132 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4133 list_del_init(&dp->dl_recall_lru);
4134 unhash_delegation(dp);
4135 }
4136
190e4fbf 4137 nfsd4_shutdown_recdir();
1da177e4 4138 nfs4_init = 0;
1da177e4
LT
4139}
4140
4141void
4142nfs4_state_shutdown(void)
4143{
5e8d5c29
N
4144 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4145 destroy_workqueue(laundry_wq);
2c5e7615 4146 locks_end_grace(&nfsd4_manager);
1da177e4
LT
4147 nfs4_lock_state();
4148 nfs4_release_reclaim();
4149 __nfs4_state_shutdown();
1da177e4
LT
4150 nfs4_unlock_state();
4151}
4152
3dd98a3b
JL
4153/*
4154 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4155 * accessed when nfsd is starting.
4156 */
0964a3d3
N
4157static void
4158nfs4_set_recdir(char *recdir)
4159{
0964a3d3 4160 strcpy(user_recovery_dirname, recdir);
0964a3d3
N
4161}
4162
4163/*
4164 * Change the NFSv4 recovery directory to recdir.
4165 */
4166int
4167nfs4_reset_recoverydir(char *recdir)
4168{
4169 int status;
a63bb996 4170 struct path path;
0964a3d3 4171
a63bb996 4172 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N
4173 if (status)
4174 return status;
4175 status = -ENOTDIR;
a63bb996 4176 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N
4177 nfs4_set_recdir(recdir);
4178 status = 0;
4179 }
a63bb996 4180 path_put(&path);
0964a3d3
N
4181 return status;
4182}
4183
3dd98a3b
JL
4184char *
4185nfs4_recoverydir(void)
4186{
4187 return user_recovery_dirname;
4188}
4189
1da177e4
LT
4190/*
4191 * Called when leasetime is changed.
4192 *
d99a05ad
N
4193 * The only way the protocol gives us to handle on-the-fly lease changes is to
4194 * simulate a reboot. Instead of doing that, we just wait till the next time
4195 * we start to register any changes in lease time. If the administrator
4196 * really wants to change the lease time *now*, they can go ahead and bring
4197 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL
4198 *
4199 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4200 * when nfsd is starting
1da177e4
LT
4201 */
4202void
4203nfs4_reset_lease(time_t leasetime)
4204{
d99a05ad 4205 user_lease_time = leasetime;
1da177e4 4206}
This page took 1.002872 seconds and 5 git commands to generate.