projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: replace callback thread by asynchronous rpc
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
1da177e4
LT		 58
59#define NFSDDBG_FACILITY NFSDDBG_PROC
60
61/* Globals */
62static time_t lease_time = 90; /* default lease time */
d99a05ad 63static time_t user_lease_time = 90;
fd39ca9a 64static time_t boot_time;
1da177e4
LT		 65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N		 69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 71static u64 current_sessionid = 1;
fd39ca9a
N		 72
73#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
74#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 75
1da177e4 76/* forward declarations */
fd39ca9a 77static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 78static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 79static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80static void nfs4_set_recdir(char *recdir);
1da177e4 81
8b671b80
BF		 82/* Locking: */
83
84/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 85static DEFINE_MUTEX(client_mutex);
1da177e4 86
8b671b80
BF		 87/*
88 * Currently used for the del_recall_lru and file hash table. In an
89 * effort to decrease the scope of the client_mutex, this spinlock may
90 * eventually cover more:
91 */
92static DEFINE_SPINLOCK(recall_lock);
93
e18b890b
CL		 94static struct kmem_cache *stateowner_slab = NULL;
95static struct kmem_cache *file_slab = NULL;
96static struct kmem_cache *stateid_slab = NULL;
97static struct kmem_cache *deleg_slab = NULL;
e60d4398 98
1da177e4
LT		 99void
100nfs4_lock_state(void)
101{
353ab6e9 102 mutex_lock(&client_mutex);
1da177e4
LT		 103}
104
105void
106nfs4_unlock_state(void)
107{
353ab6e9 108 mutex_unlock(&client_mutex);
1da177e4
LT		 109}
110
111static inline u32
112opaque_hashval(const void *ptr, int nbytes)
113{
114 unsigned char *cptr = (unsigned char *) ptr;
115
116 u32 x = 0;
117 while (nbytes--) {
118 x *= 37;
119 x += *cptr++;
120 }
121 return x;
122}
123
1da177e4
LT		 124static struct list_head del_recall_lru;
125
13cd2184
N		 126static inline void
127put_nfs4_file(struct nfs4_file *fi)
128{
8b671b80
BF		 129 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
130 list_del(&fi->fi_hash);
131 spin_unlock(&recall_lock);
132 iput(fi->fi_inode);
133 kmem_cache_free(file_slab, fi);
134 }
13cd2184
N		 135}
136
137static inline void
138get_nfs4_file(struct nfs4_file *fi)
139{
8b671b80 140 atomic_inc(&fi->fi_ref);
13cd2184
N		 141}
142
ef0f3390 143static int num_delegations;
c2f1a551 144unsigned int max_delegations;
ef0f3390
N		 145
146/*
147 * Open owner state (share locks)
148 */
149
150/* hash tables for nfs4_stateowner */
151#define OWNER_HASH_BITS 8
152#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
153#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
154
155#define ownerid_hashval(id) \
156 ((id) & OWNER_HASH_MASK)
157#define ownerstr_hashval(clientid, ownername) \
158 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
159
160static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
161static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
162
163/* hash table for nfs4_file */
164#define FILE_HASH_BITS 8
165#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
166#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
167/* hash table for (open)nfs4_stateid */
168#define STATEID_HASH_BITS 10
169#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
170#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
171
172#define file_hashval(x) \
173 hash_ptr(x, FILE_HASH_BITS)
174#define stateid_hashval(owner_id, file_id) \
175 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
176
177static struct list_head file_hashtbl[FILE_HASH_SIZE];
178static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
179
1da177e4
LT		 180static struct nfs4_delegation *
181alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
182{
183 struct nfs4_delegation *dp;
184 struct nfs4_file *fp = stp->st_file;
185 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
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;
206 dp->dl_recall.cbr_dp = NULL;
207 dp->dl_recall.cbr_ident = cb->cb_ident;
208 dp->dl_recall.cbr_trunc = 0;
78155ed7 209 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 210 dp->dl_stateid.si_stateownerid = current_delegid++;
211 dp->dl_stateid.si_fileid = 0;
212 dp->dl_stateid.si_generation = 0;
6c02eaa1 213 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 214 dp->dl_time = 0;
215 atomic_set(&dp->dl_count, 1);
ea1da636
N		 216 list_add(&dp->dl_perfile, &fp->fi_delegations);
217 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 218 return dp;
219}
220
221void
222nfs4_put_delegation(struct nfs4_delegation *dp)
223{
224 if (atomic_dec_and_test(&dp->dl_count)) {
225 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 226 put_nfs4_file(dp->dl_file);
5b2d21c1 227 kmem_cache_free(deleg_slab, dp);
ef0f3390 228 num_delegations--;
1da177e4
LT		 229 }
230}
231
232/* Remove the associated file_lock first, then remove the delegation.
233 * lease_modify() is called to remove the FS_LEASE file_lock from
234 * the i_flock list, eventually calling nfsd's lock_manager
235 * fl_release_callback.
236 */
237static void
238nfs4_close_delegation(struct nfs4_delegation *dp)
239{
240 struct file *filp = dp->dl_vfs_file;
241
242 dprintk("NFSD: close_delegation dp %p\n",dp);
243 dp->dl_vfs_file = NULL;
244 /* The following nfsd_close may not actually close the file,
245 * but we want to remove the lease in any case. */
c907132d 246 if (dp->dl_flock)
a9933cea 247 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 248 nfsd_close(filp);
1da177e4
LT		 249}
250
251/* Called under the state lock. */
252static void
253unhash_delegation(struct nfs4_delegation *dp)
254{
ea1da636
N		 255 list_del_init(&dp->dl_perfile);
256 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 257 spin_lock(&recall_lock);
258 list_del_init(&dp->dl_recall_lru);
259 spin_unlock(&recall_lock);
260 nfs4_close_delegation(dp);
261 nfs4_put_delegation(dp);
262}
263
264/*
265 * SETCLIENTID state
266 */
267
268/* Hash tables for nfs4_clientid state */
269#define CLIENT_HASH_BITS 4
270#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
271#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
272
273#define clientid_hashval(id) \
274 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 275#define clientstr_hashval(name) \
276 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 277/*
278 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
279 * used in reboot/reset lease grace period processing
280 *
281 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
282 * setclientid_confirmed info.
283 *
284 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
285 * setclientid info.
286 *
287 * client_lru holds client queue ordered by nfs4_client.cl_time
288 * for lease renewal.
289 *
290 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
291 * for last close replay.
292 */
293static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
294static int reclaim_str_hashtbl_size = 0;
295static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
299static struct list_head client_lru;
300static struct list_head close_lru;
301
2283963f
BF		 302static void unhash_generic_stateid(struct nfs4_stateid *stp)
303{
304 list_del(&stp->st_hash);
305 list_del(&stp->st_perfile);
306 list_del(&stp->st_perstateowner);
307}
308
309static void free_generic_stateid(struct nfs4_stateid *stp)
310{
311 put_nfs4_file(stp->st_file);
312 kmem_cache_free(stateid_slab, stp);
313}
314
315static void release_lock_stateid(struct nfs4_stateid *stp)
316{
317 unhash_generic_stateid(stp);
318 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
319 free_generic_stateid(stp);
320}
321
f044ff83
BF		 322static void unhash_lockowner(struct nfs4_stateowner *sop)
323{
324 struct nfs4_stateid *stp;
325
326 list_del(&sop->so_idhash);
327 list_del(&sop->so_strhash);
328 list_del(&sop->so_perstateid);
329 while (!list_empty(&sop->so_stateids)) {
330 stp = list_first_entry(&sop->so_stateids,
331 struct nfs4_stateid, st_perstateowner);
332 release_lock_stateid(stp);
333 }
334}
335
336static void release_lockowner(struct nfs4_stateowner *sop)
337{
338 unhash_lockowner(sop);
339 nfs4_put_stateowner(sop);
340}
341
f1d110ca
BF		 342static void
343release_stateid_lockowners(struct nfs4_stateid *open_stp)
344{
345 struct nfs4_stateowner *lock_sop;
346
347 while (!list_empty(&open_stp->st_lockowners)) {
348 lock_sop = list_entry(open_stp->st_lockowners.next,
349 struct nfs4_stateowner, so_perstateid);
350 /* list_del(&open_stp->st_lockowners); */
351 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 352 release_lockowner(lock_sop);
f1d110ca
BF		 353 }
354}
355
2283963f
BF		 356static void release_open_stateid(struct nfs4_stateid *stp)
357{
358 unhash_generic_stateid(stp);
359 release_stateid_lockowners(stp);
360 nfsd_close(stp->st_vfs_file);
361 free_generic_stateid(stp);
362}
363
f044ff83 364static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 365{
366 struct nfs4_stateid *stp;
367
368 list_del(&sop->so_idhash);
369 list_del(&sop->so_strhash);
f044ff83
BF		 370 list_del(&sop->so_perclient);
371 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 372 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 373 stp = list_first_entry(&sop->so_stateids,
374 struct nfs4_stateid, st_perstateowner);
375 release_open_stateid(stp);
f1d110ca
BF		 376 }
377}
378
f044ff83 379static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 380{
f044ff83 381 unhash_openowner(sop);
f1d110ca
BF		 382 list_del(&sop->so_close_lru);
383 nfs4_put_stateowner(sop);
384}
385
5282fd72
ME		 386static DEFINE_SPINLOCK(sessionid_lock);
387#define SESSION_HASH_SIZE 512
388static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
389
390static inline int
391hash_sessionid(struct nfs4_sessionid *sessionid)
392{
393 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
394
395 return sid->sequence % SESSION_HASH_SIZE;
396}
397
398static inline void
399dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
400{
401 u32 *ptr = (u32 *)(&sessionid->data[0]);
402 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
403}
404
ec6b5d7b
AA		 405static void
406gen_sessionid(struct nfsd4_session *ses)
407{
408 struct nfs4_client *clp = ses->se_client;
409 struct nfsd4_sessionid *sid;
410
411 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
412 sid->clientid = clp->cl_clientid;
413 sid->sequence = current_sessionid++;
414 sid->reserved = 0;
415}
416
417/*
418 * Give the client the number of slots it requests bound by
419 * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
420 *
421 * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
422 * should (up to a point) re-negotiate active sessions and reduce their
423 * slot usage to make rooom for new connections. For now we just fail the
424 * create session.
425 */
426static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
427{
428 int status = 0, np = fchan->maxreqs * NFSD_PAGES_PER_SLOT;
429
430 spin_lock(&nfsd_serv->sv_lock);
431 if (np + nfsd_serv->sv_drc_pages_used > nfsd_serv->sv_drc_max_pages)
432 np = nfsd_serv->sv_drc_max_pages - nfsd_serv->sv_drc_pages_used;
433 nfsd_serv->sv_drc_pages_used += np;
434 spin_unlock(&nfsd_serv->sv_lock);
435
436 if (np <= 0) {
437 status = nfserr_resource;
438 fchan->maxreqs = 0;
439 } else
440 fchan->maxreqs = np / NFSD_PAGES_PER_SLOT;
441
442 return status;
443}
444
445/*
446 * fchan holds the client values on input, and the server values on output
447 */
448static int init_forechannel_attrs(struct svc_rqst *rqstp,
449 struct nfsd4_session *session,
450 struct nfsd4_channel_attrs *fchan)
451{
452 int status = 0;
453 __u32 maxcount = svc_max_payload(rqstp);
454
455 /* headerpadsz set to zero in encode routine */
456
457 /* Use the client's max request and max response size if possible */
458 if (fchan->maxreq_sz > maxcount)
459 fchan->maxreq_sz = maxcount;
460 session->se_fmaxreq_sz = fchan->maxreq_sz;
461
462 if (fchan->maxresp_sz > maxcount)
463 fchan->maxresp_sz = maxcount;
464 session->se_fmaxresp_sz = fchan->maxresp_sz;
465
466 /* Set the max response cached size our default which is
467 * a multiple of PAGE_SIZE and small */
468 session->se_fmaxresp_cached = NFSD_PAGES_PER_SLOT * PAGE_SIZE;
469 fchan->maxresp_cached = session->se_fmaxresp_cached;
470
471 /* Use the client's maxops if possible */
472 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
473 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
474 session->se_fmaxops = fchan->maxops;
475
476 /* try to use the client requested number of slots */
477 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
478 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
479
480 /* FIXME: Error means no more DRC pages so the server should
481 * recover pages from existing sessions. For now fail session
482 * creation.
483 */
484 status = set_forechannel_maxreqs(fchan);
485
486 session->se_fnumslots = fchan->maxreqs;
487 return status;
488}
489
490static int
491alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
492 struct nfsd4_create_session *cses)
493{
494 struct nfsd4_session *new, tmp;
495 int idx, status = nfserr_resource, slotsize;
496
497 memset(&tmp, 0, sizeof(tmp));
498
499 /* FIXME: For now, we just accept the client back channel attributes. */
500 status = init_forechannel_attrs(rqstp, &tmp, &cses->fore_channel);
501 if (status)
502 goto out;
503
504 /* allocate struct nfsd4_session and slot table in one piece */
505 slotsize = tmp.se_fnumslots * sizeof(struct nfsd4_slot);
506 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
507 if (!new)
508 goto out;
509
510 memcpy(new, &tmp, sizeof(*new));
511
512 new->se_client = clp;
513 gen_sessionid(new);
514 idx = hash_sessionid(&new->se_sessionid);
515 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
516 NFS4_MAX_SESSIONID_LEN);
517
518 new->se_flags = cses->flags;
519 kref_init(&new->se_ref);
520 spin_lock(&sessionid_lock);
521 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
522 list_add(&new->se_perclnt, &clp->cl_sessions);
523 spin_unlock(&sessionid_lock);
524
525 status = nfs_ok;
526out:
527 return status;
528}
529
5282fd72
ME		 530/* caller must hold sessionid_lock */
531static struct nfsd4_session *
532find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
533{
534 struct nfsd4_session *elem;
535 int idx;
536
537 dump_sessionid(__func__, sessionid);
538 idx = hash_sessionid(sessionid);
539 dprintk("%s: idx is %d\n", __func__, idx);
540 /* Search in the appropriate list */
541 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
542 dump_sessionid("list traversal", &elem->se_sessionid);
543 if (!memcmp(elem->se_sessionid.data, sessionid->data,
544 NFS4_MAX_SESSIONID_LEN)) {
545 return elem;
546 }
547 }
548
549 dprintk("%s: session not found\n", __func__);
550 return NULL;
551}
552
553/* caller must hold sessionid_lock */
7116ed6b 554static void
5282fd72 555unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 556{
557 list_del(&ses->se_hash);
558 list_del(&ses->se_perclnt);
5282fd72
ME		 559}
560
561static void
562release_session(struct nfsd4_session *ses)
563{
564 spin_lock(&sessionid_lock);
565 unhash_session(ses);
566 spin_unlock(&sessionid_lock);
7116ed6b
AA		 567 nfsd4_put_session(ses);
568}
569
14778a13
AA		 570static void nfsd4_release_respages(struct page **respages, short resused);
571
7116ed6b
AA		 572void
573free_session(struct kref *kref)
574{
575 struct nfsd4_session *ses;
14778a13 576 int i;
7116ed6b
AA		 577
578 ses = container_of(kref, struct nfsd4_session, se_ref);
14778a13
AA		 579 for (i = 0; i < ses->se_fnumslots; i++) {
580 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
581 nfsd4_release_respages(e->ce_respages, e->ce_resused);
582 }
7116ed6b
AA		 583 kfree(ses->se_slots);
584 kfree(ses);
585}
586
1da177e4
LT		 587static inline void
588renew_client(struct nfs4_client *clp)
589{
590 /*
591 * Move client to the end to the LRU list.
592 */
593 dprintk("renewing client (clientid %08x/%08x)\n",
594 clp->cl_clientid.cl_boot,
595 clp->cl_clientid.cl_id);
596 list_move_tail(&clp->cl_lru, &client_lru);
597 clp->cl_time = get_seconds();
598}
599
600/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
601static int
602STALE_CLIENTID(clientid_t *clid)
603{
604 if (clid->cl_boot == boot_time)
605 return 0;
60adfc50
AA		 606 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
607 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 608 return 1;
609}
610
611/*
612 * XXX Should we use a slab cache ?
613 * This type of memory management is somewhat inefficient, but we use it
614 * anyway since SETCLIENTID is not a common operation.
615 */
35bba9a3 616static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 617{
618 struct nfs4_client *clp;
619
35bba9a3
BF		 620 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
621 if (clp == NULL)
622 return NULL;
623 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
624 if (clp->cl_name.data == NULL) {
625 kfree(clp);
626 return NULL;
1da177e4 627 }
35bba9a3
BF		 628 memcpy(clp->cl_name.data, name.data, name.len);
629 clp->cl_name.len = name.len;
1da177e4
LT		 630 return clp;
631}
632
1b1a9b31
BF		 633static void
634shutdown_callback_client(struct nfs4_client *clp)
635{
636 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
637
1b1a9b31 638 if (clnt) {
404ec117
BF		 639 /*
640 * Callback threads take a reference on the client, so there
641 * should be no outstanding callbacks at this point.
642 */
1b1a9b31
BF		 643 clp->cl_callback.cb_client = NULL;
644 rpc_shutdown_client(clnt);
645 }
3cef9ab2
BF		 646 if (clp->cl_callback.cb_cred) {
647 put_rpccred(clp->cl_callback.cb_cred);
648 clp->cl_callback.cb_cred = NULL;
649 }
1b1a9b31
BF		 650}
651
1da177e4
LT		 652static inline void
653free_client(struct nfs4_client *clp)
654{
1b1a9b31 655 shutdown_callback_client(clp);
38eb76a5
AA		 656 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
657 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 658 if (clp->cl_cred.cr_group_info)
659 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 660 kfree(clp->cl_principal);
1da177e4
LT		 661 kfree(clp->cl_name.data);
662 kfree(clp);
663}
664
665void
666put_nfs4_client(struct nfs4_client *clp)
667{
668 if (atomic_dec_and_test(&clp->cl_count))
669 free_client(clp);
670}
671
672static void
673expire_client(struct nfs4_client *clp)
674{
675 struct nfs4_stateowner *sop;
676 struct nfs4_delegation *dp;
1da177e4
LT		 677 struct list_head reaplist;
678
679 dprintk("NFSD: expire_client cl_count %d\n",
680 atomic_read(&clp->cl_count));
681
1da177e4
LT		 682 INIT_LIST_HEAD(&reaplist);
683 spin_lock(&recall_lock);
ea1da636
N		 684 while (!list_empty(&clp->cl_delegations)) {
685 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 686 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
687 dp->dl_flock);
ea1da636 688 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 689 list_move(&dp->dl_recall_lru, &reaplist);
690 }
691 spin_unlock(&recall_lock);
692 while (!list_empty(&reaplist)) {
693 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
694 list_del_init(&dp->dl_recall_lru);
695 unhash_delegation(dp);
696 }
697 list_del(&clp->cl_idhash);
698 list_del(&clp->cl_strhash);
699 list_del(&clp->cl_lru);
ea1da636
N		 700 while (!list_empty(&clp->cl_openowners)) {
701 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 702 release_openowner(sop);
1da177e4 703 }
c4bf7868
ME		 704 while (!list_empty(&clp->cl_sessions)) {
705 struct nfsd4_session *ses;
706 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
707 se_perclnt);
708 release_session(ses);
709 }
1da177e4
LT		 710 put_nfs4_client(clp);
711}
712
35bba9a3
BF		 713static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
714{
1da177e4
LT		 715 struct nfs4_client *clp;
716
35bba9a3
BF		 717 clp = alloc_client(name);
718 if (clp == NULL)
719 return NULL;
a55370a3 720 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT		 721 atomic_set(&clp->cl_count, 1);
722 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT		 723 INIT_LIST_HEAD(&clp->cl_idhash);
724 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 725 INIT_LIST_HEAD(&clp->cl_openowners);
726 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 727 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 728 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 729 return clp;
730}
731
35bba9a3
BF		 732static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
733{
734 memcpy(target->cl_verifier.data, source->data,
735 sizeof(target->cl_verifier.data));
1da177e4
LT		 736}
737
35bba9a3
BF		 738static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
739{
1da177e4
LT		 740 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
741 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
742}
743
35bba9a3
BF		 744static void copy_cred(struct svc_cred *target, struct svc_cred *source)
745{
1da177e4
LT		 746 target->cr_uid = source->cr_uid;
747 target->cr_gid = source->cr_gid;
748 target->cr_group_info = source->cr_group_info;
749 get_group_info(target->cr_group_info);
750}
751
35bba9a3 752static int same_name(const char *n1, const char *n2)
599e0a22 753{
a55370a3 754 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 755}
756
757static int
599e0a22
BF		 758same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
759{
760 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 761}
762
763static int
599e0a22
BF		 764same_clid(clientid_t *cl1, clientid_t *cl2)
765{
766 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 767}
768
769/* XXX what about NGROUP */
770static int
599e0a22
BF		 771same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
772{
773 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 774}
775
5ec7b46c
BF		 776static void gen_clid(struct nfs4_client *clp)
777{
778 static u32 current_clientid = 1;
779
1da177e4
LT		 780 clp->cl_clientid.cl_boot = boot_time;
781 clp->cl_clientid.cl_id = current_clientid++;
782}
783
deda2faa
BF		 784static void gen_confirm(struct nfs4_client *clp)
785{
786 static u32 i;
787 u32 *p;
1da177e4 788
1da177e4 789 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 790 *p++ = get_seconds();
791 *p++ = i++;
1da177e4
LT		 792}
793
35bba9a3
BF		 794static int check_name(struct xdr_netobj name)
795{
1da177e4
LT		 796 if (name.len == 0)
797 return 0;
798 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 799 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 800 return 0;
801 }
802 return 1;
803}
804
fd39ca9a 805static void
1da177e4
LT		 806add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
807{
808 unsigned int idhashval;
809
810 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
811 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
812 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
813 list_add_tail(&clp->cl_lru, &client_lru);
814 clp->cl_time = get_seconds();
815}
816
fd39ca9a 817static void
1da177e4
LT		 818move_to_confirmed(struct nfs4_client *clp)
819{
820 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
821 unsigned int strhashval;
822
823 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
824 list_del_init(&clp->cl_strhash);
f116629d 825 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 826 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 827 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
828 renew_client(clp);
829}
830
831static struct nfs4_client *
832find_confirmed_client(clientid_t *clid)
833{
834 struct nfs4_client *clp;
835 unsigned int idhashval = clientid_hashval(clid->cl_id);
836
837 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 838 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 839 return clp;
840 }
841 return NULL;
842}
843
844static struct nfs4_client *
845find_unconfirmed_client(clientid_t *clid)
846{
847 struct nfs4_client *clp;
848 unsigned int idhashval = clientid_hashval(clid->cl_id);
849
850 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 851 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 852 return clp;
853 }
854 return NULL;
855}
856
a1bcecd2
AA		 857/*
858 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
859 * parameter. Matching is based on the fact the at least one of the
860 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
861 *
862 * FIXME: we need to unify the clientid namespaces for nfsv4.x
863 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
864 * and SET_CLIENTID{,_CONFIRM}
865 */
866static inline int
867match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
868{
869 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
870 return use_exchange_id == has_exchange_flags;
871}
872
28ce6054 873static struct nfs4_client *
a1bcecd2
AA		 874find_confirmed_client_by_str(const char *dname, unsigned int hashval,
875 bool use_exchange_id)
28ce6054
N		 876{
877 struct nfs4_client *clp;
878
879 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 880 if (same_name(clp->cl_recdir, dname) &&
881 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 882 return clp;
883 }
884 return NULL;
885}
886
887static struct nfs4_client *
a1bcecd2
AA		 888find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
889 bool use_exchange_id)
28ce6054
N		 890{
891 struct nfs4_client *clp;
892
893 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 894 if (same_name(clp->cl_recdir, dname) &&
895 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 896 return clp;
897 }
898 return NULL;
899}
900
1da177e4
LT		 901/* a helper function for parse_callback */
902static int
903parse_octet(unsigned int *lenp, char **addrp)
904{
905 unsigned int len = *lenp;
906 char *p = *addrp;
907 int n = -1;
908 char c;
909
910 for (;;) {
911 if (!len)
912 break;
913 len--;
914 c = *p++;
915 if (c == '.')
916 break;
917 if ((c < '0') || (c > '9')) {
918 n = -1;
919 break;
920 }
921 if (n < 0)
922 n = 0;
923 n = (n * 10) + (c - '0');
924 if (n > 255) {
925 n = -1;
926 break;
927 }
928 }
929 *lenp = len;
930 *addrp = p;
931 return n;
932}
933
934/* parse and set the setclientid ipv4 callback address */
fd39ca9a 935static int
1da177e4
LT		 936parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
937{
938 int temp = 0;
939 u32 cbaddr = 0;
940 u16 cbport = 0;
941 u32 addrlen = addr_len;
942 char *addr = addr_val;
943 int i, shift;
944
945 /* ipaddress */
946 shift = 24;
947 for(i = 4; i > 0 ; i--) {
948 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
949 return 0;
950 }
951 cbaddr |= (temp << shift);
952 if (shift > 0)
953 shift -= 8;
954 }
955 *cbaddrp = cbaddr;
956
957 /* port */
958 shift = 8;
959 for(i = 2; i > 0 ; i--) {
960 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
961 return 0;
962 }
963 cbport |= (temp << shift);
964 if (shift > 0)
965 shift -= 8;
966 }
967 *cbportp = cbport;
968 return 1;
969}
970
fd39ca9a 971static void
1da177e4
LT		 972gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
973{
974 struct nfs4_callback *cb = &clp->cl_callback;
975
976 /* Currently, we only support tcp for the callback channel */
977 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
978 goto out_err;
979
980 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
981 &cb->cb_addr, &cb->cb_port)))
982 goto out_err;
983 cb->cb_prog = se->se_callback_prog;
984 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 985 return;
986out_err:
849823c5 987 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 988 "will not receive delegations\n",
989 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
990
1da177e4
LT		 991 return;
992}
993
074fe897
AA		 994void
995nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
996{
997 struct nfsd4_compoundres *resp = rqstp->rq_resp;
998
999 resp->cstate.statp = statp;
1000}
1001
1002/*
1003 * Dereference the result pages.
1004 */
1005static void
1006nfsd4_release_respages(struct page **respages, short resused)
1007{
1008 int i;
1009
1010 dprintk("--> %s\n", __func__);
1011 for (i = 0; i < resused; i++) {
1012 if (!respages[i])
1013 continue;
1014 put_page(respages[i]);
1015 respages[i] = NULL;
1016 }
1017}
1018
1019static void
1020nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1021{
1022 int i;
1023
1024 for (i = 0; i < count; i++) {
1025 topages[i] = frompages[i];
1026 if (!topages[i])
1027 continue;
1028 get_page(topages[i]);
1029 }
1030}
1031
1032/*
1033 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1034 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1035 * length of the XDR response is less than se_fmaxresp_cached
1036 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1037 * of the reply (e.g. readdir).
1038 *
1039 * Store the base and length of the rq_req.head[0] page
1040 * of the NFSv4.1 data, just past the rpc header.
1041 */
1042void
1043nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1044{
1045 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1046 struct svc_rqst *rqstp = resp->rqstp;
1047 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1048 struct nfsd4_op *op = &args->ops[resp->opcnt];
1049 struct kvec *resv = &rqstp->rq_res.head[0];
1050
1051 dprintk("--> %s entry %p\n", __func__, entry);
1052
1053 /* Don't cache a failed OP_SEQUENCE. */
1054 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1055 return;
bf864a31 1056
074fe897 1057 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1058 entry->ce_opcnt = resp->opcnt;
1059 entry->ce_status = resp->cstate.status;
1060
1061 /*
1062 * Don't need a page to cache just the sequence operation - the slot
1063 * does this for us!
1064 */
1065
1066 if (nfsd4_not_cached(resp)) {
1067 entry->ce_resused = 0;
1068 entry->ce_rpchdrlen = 0;
1069 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1070 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1071 return;
1072 }
074fe897
AA		 1073 entry->ce_resused = rqstp->rq_resused;
1074 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1075 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1076 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1077 entry->ce_resused);
074fe897
AA		 1078 entry->ce_datav.iov_base = resp->cstate.statp;
1079 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1080 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1081 /* Current request rpc header length*/
1082 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1083 (char *)page_address(rqstp->rq_respages[0]);
1084}
1085
1086/*
1087 * We keep the rpc header, but take the nfs reply from the replycache.
1088 */
1089static int
1090nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1091 struct nfsd4_cache_entry *entry)
1092{
1093 struct svc_rqst *rqstp = resp->rqstp;
1094 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1095 int len;
1096
1097 /* Current request rpc header length*/
1098 len = (char *)resp->cstate.statp -
1099 (char *)page_address(rqstp->rq_respages[0]);
1100 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1101 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1102 entry->ce_datav.iov_len);
1103 return 0;
1104 }
1105 /* copy the cached reply nfsd data past the current rpc header */
1106 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1107 entry->ce_datav.iov_len);
1108 resv->iov_len = len + entry->ce_datav.iov_len;
1109 return 1;
1110}
1111
1112/*
1113 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1114 * cached page. Replace any futher replay pages from the cache.
1115 */
1116__be32
bf864a31
AA		 1117nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1118 struct nfsd4_sequence *seq)
074fe897
AA		 1119{
1120 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1121 __be32 status;
1122
1123 dprintk("--> %s entry %p\n", __func__, entry);
1124
bf864a31
AA		 1125 /*
1126 * If this is just the sequence operation, we did not keep
1127 * a page in the cache entry because we can just use the
1128 * slot info stored in struct nfsd4_sequence that was checked
1129 * against the slot in nfsd4_sequence().
1130 *
1131 * This occurs when seq->cachethis is FALSE, or when the client
1132 * session inactivity timer fires and a solo sequence operation
1133 * is sent (lease renewal).
1134 */
1135 if (seq && nfsd4_not_cached(resp)) {
1136 seq->maxslots = resp->cstate.session->se_fnumslots;
1137 return nfs_ok;
1138 }
074fe897
AA		 1139
1140 if (!nfsd41_copy_replay_data(resp, entry)) {
1141 /*
1142 * Not enough room to use the replay rpc header, send the
1143 * cached header. Release all the allocated result pages.
1144 */
1145 svc_free_res_pages(resp->rqstp);
1146 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1147 entry->ce_resused);
1148 } else {
1149 /* Release all but the first allocated result page */
1150
1151 resp->rqstp->rq_resused--;
1152 svc_free_res_pages(resp->rqstp);
1153
1154 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1155 &entry->ce_respages[1],
1156 entry->ce_resused - 1);
1157 }
1158
1159 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1160 resp->opcnt = entry->ce_opcnt;
1161 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1162 status = entry->ce_status;
1163
1164 return status;
1165}
1166
0733d213
AA		 1167/*
1168 * Set the exchange_id flags returned by the server.
1169 */
1170static void
1171nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1172{
1173 /* pNFS is not supported */
1174 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1175
1176 /* Referrals are supported, Migration is not. */
1177 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1178
1179 /* set the wire flags to return to client. */
1180 clid->flags = new->cl_exchange_flags;
1181}
1182
069b6ad4
AA		 1183__be32
1184nfsd4_exchange_id(struct svc_rqst *rqstp,
1185 struct nfsd4_compound_state *cstate,
1186 struct nfsd4_exchange_id *exid)
1187{
0733d213
AA		 1188 struct nfs4_client *unconf, *conf, *new;
1189 int status;
1190 unsigned int strhashval;
1191 char dname[HEXDIR_LEN];
1192 nfs4_verifier verf = exid->verifier;
1193 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1194
1195 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1196 " ip_addr=%u flags %x, spa_how %d\n",
1197 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1198 ip_addr, exid->flags, exid->spa_how);
1199
1200 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1201 return nfserr_inval;
1202
1203 /* Currently only support SP4_NONE */
1204 switch (exid->spa_how) {
1205 case SP4_NONE:
1206 break;
1207 case SP4_SSV:
1208 return nfserr_encr_alg_unsupp;
1209 default:
1210 BUG(); /* checked by xdr code */
1211 case SP4_MACH_CRED:
1212 return nfserr_serverfault; /* no excuse :-/ */
1213 }
1214
1215 status = nfs4_make_rec_clidname(dname, &exid->clname);
1216
1217 if (status)
1218 goto error;
1219
1220 strhashval = clientstr_hashval(dname);
1221
1222 nfs4_lock_state();
1223 status = nfs_ok;
1224
a1bcecd2 1225 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1226 if (conf) {
1227 if (!same_verf(&verf, &conf->cl_verifier)) {
1228 /* 18.35.4 case 8 */
1229 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1230 status = nfserr_not_same;
1231 goto out;
1232 }
1233 /* Client reboot: destroy old state */
1234 expire_client(conf);
1235 goto out_new;
1236 }
1237 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1238 /* 18.35.4 case 9 */
1239 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1240 status = nfserr_perm;
1241 goto out;
1242 }
1243 expire_client(conf);
1244 goto out_new;
1245 }
1246 if (ip_addr != conf->cl_addr &&
1247 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1248 /* Client collision. 18.35.4 case 3 */
1249 status = nfserr_clid_inuse;
1250 goto out;
1251 }
1252 /*
1253 * Set bit when the owner id and verifier map to an already
1254 * confirmed client id (18.35.3).
1255 */
1256 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1257
1258 /*
1259 * Falling into 18.35.4 case 2, possible router replay.
1260 * Leave confirmed record intact and return same result.
1261 */
1262 copy_verf(conf, &verf);
1263 new = conf;
1264 goto out_copy;
1265 } else {
1266 /* 18.35.4 case 7 */
1267 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1268 status = nfserr_noent;
1269 goto out;
1270 }
1271 }
1272
a1bcecd2 1273 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1274 if (unconf) {
1275 /*
1276 * Possible retry or client restart. Per 18.35.4 case 4,
1277 * a new unconfirmed record should be generated regardless
1278 * of whether any properties have changed.
1279 */
1280 expire_client(unconf);
1281 }
1282
1283out_new:
1284 /* Normal case */
1285 new = create_client(exid->clname, dname);
1286 if (new == NULL) {
1287 status = nfserr_resource;
1288 goto out;
1289 }
1290
1291 copy_verf(new, &verf);
1292 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1293 new->cl_addr = ip_addr;
1294 gen_clid(new);
1295 gen_confirm(new);
1296 add_to_unconfirmed(new, strhashval);
1297out_copy:
1298 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1299 exid->clientid.cl_id = new->cl_clientid.cl_id;
1300
38eb76a5
AA		 1301 new->cl_slot.sl_seqid = 0;
1302 exid->seqid = 1;
0733d213
AA		 1303 nfsd4_set_ex_flags(new, exid);
1304
1305 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1306 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1307 status = nfs_ok;
1308
1309out:
1310 nfs4_unlock_state();
1311error:
1312 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1313 return status;
069b6ad4
AA		 1314}
1315
b85d4c01
BH		 1316static int
1317check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1318{
1319 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1320 slot->sl_seqid);
1321
1322 /* The slot is in use, and no response has been sent. */
1323 if (slot->sl_inuse) {
1324 if (seqid == slot->sl_seqid)
1325 return nfserr_jukebox;
1326 else
1327 return nfserr_seq_misordered;
1328 }
1329 /* Normal */
1330 if (likely(seqid == slot->sl_seqid + 1))
1331 return nfs_ok;
1332 /* Replay */
1333 if (seqid == slot->sl_seqid)
1334 return nfserr_replay_cache;
1335 /* Wraparound */
1336 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1337 return nfs_ok;
1338 /* Misordered replay or misordered new request */
1339 return nfserr_seq_misordered;
1340}
1341
069b6ad4
AA		 1342__be32
1343nfsd4_create_session(struct svc_rqst *rqstp,
1344 struct nfsd4_compound_state *cstate,
1345 struct nfsd4_create_session *cr_ses)
1346{
ec6b5d7b 1347 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1348 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1349 struct nfs4_client *conf, *unconf;
38eb76a5 1350 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1351 int status = 0;
1352
1353 nfs4_lock_state();
1354 unconf = find_unconfirmed_client(&cr_ses->clientid);
1355 conf = find_confirmed_client(&cr_ses->clientid);
1356
1357 if (conf) {
38eb76a5
AA		 1358 slot = &conf->cl_slot;
1359 status = check_slot_seqid(cr_ses->seqid, slot);
1360 if (status == nfserr_replay_cache) {
ec6b5d7b 1361 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1362 slot->sl_seqid);
1363 cstate->slot = slot;
1364 cstate->status = status;
1365 /* Return the cached reply status */
bf864a31 1366 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1367 goto out;
1368 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1369 status = nfserr_seq_misordered;
1370 dprintk("Sequence misordered!\n");
1371 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1372 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1373 goto out;
1374 }
38eb76a5 1375 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1376 } else if (unconf) {
1377 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1378 (ip_addr != unconf->cl_addr)) {
1379 status = nfserr_clid_inuse;
1380 goto out;
1381 }
1382
38eb76a5
AA		 1383 slot = &unconf->cl_slot;
1384 status = check_slot_seqid(cr_ses->seqid, slot);
1385 if (status) {
1386 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1387 status = nfserr_seq_misordered;
1388 goto out;
1389 }
1390
38eb76a5 1391 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1392 move_to_confirmed(unconf);
1393
1394 /*
1395 * We do not support RDMA or persistent sessions
1396 */
1397 cr_ses->flags &= ~SESSION4_PERSIST;
1398 cr_ses->flags &= ~SESSION4_RDMA;
1399
1400 conf = unconf;
1401 } else {
1402 status = nfserr_stale_clientid;
1403 goto out;
1404 }
1405
1406 status = alloc_init_session(rqstp, conf, cr_ses);
1407 if (status)
1408 goto out;
1409
ec6b5d7b
AA		 1410 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1411 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1412 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1413
38eb76a5
AA		 1414 slot->sl_inuse = true;
1415 cstate->slot = slot;
bf864a31
AA		 1416 /* Ensure a page is used for the cache */
1417 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1418out:
1419 nfs4_unlock_state();
1420 dprintk("%s returns %d\n", __func__, ntohl(status));
1421 return status;
069b6ad4
AA		 1422}
1423
1424__be32
1425nfsd4_destroy_session(struct svc_rqst *r,
1426 struct nfsd4_compound_state *cstate,
1427 struct nfsd4_destroy_session *sessionid)
1428{
e10e0cfc
BH		 1429 struct nfsd4_session *ses;
1430 u32 status = nfserr_badsession;
1431
1432 /* Notes:
1433 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1434 * - Should we return nfserr_back_chan_busy if waiting for
1435 * callbacks on to-be-destroyed session?
1436 * - Do we need to clear any callback info from previous session?
1437 */
1438
1439 dump_sessionid(__func__, &sessionid->sessionid);
1440 spin_lock(&sessionid_lock);
1441 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1442 if (!ses) {
1443 spin_unlock(&sessionid_lock);
1444 goto out;
1445 }
1446
1447 unhash_session(ses);
1448 spin_unlock(&sessionid_lock);
1449
1450 /* wait for callbacks */
1451 shutdown_callback_client(ses->se_client);
1452 nfsd4_put_session(ses);
1453 status = nfs_ok;
1454out:
1455 dprintk("%s returns %d\n", __func__, ntohl(status));
1456 return status;
069b6ad4
AA		 1457}
1458
1459__be32
b85d4c01 1460nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1461 struct nfsd4_compound_state *cstate,
1462 struct nfsd4_sequence *seq)
1463{
f9bb94c4 1464 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1465 struct nfsd4_session *session;
1466 struct nfsd4_slot *slot;
1467 int status;
1468
f9bb94c4
AA		 1469 if (resp->opcnt != 1)
1470 return nfserr_sequence_pos;
1471
b85d4c01
BH		 1472 spin_lock(&sessionid_lock);
1473 status = nfserr_badsession;
1474 session = find_in_sessionid_hashtbl(&seq->sessionid);
1475 if (!session)
1476 goto out;
1477
1478 status = nfserr_badslot;
1479 if (seq->slotid >= session->se_fnumslots)
1480 goto out;
1481
1482 slot = &session->se_slots[seq->slotid];
1483 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1484
1485 status = check_slot_seqid(seq->seqid, slot);
1486 if (status == nfserr_replay_cache) {
1487 cstate->slot = slot;
1488 cstate->session = session;
da3846a2 1489 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1490 * for nfsd4_svc_encode_compoundres processing */
1491 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1492 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1493 goto replay_cache;
1494 }
1495 if (status)
1496 goto out;
1497
1498 /* Success! bump slot seqid */
1499 slot->sl_inuse = true;
1500 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1501 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1502 /* Always set the cache entry cachethis for solo sequence */
1503 if (nfsd4_is_solo_sequence(resp))
1504 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1505
1506 cstate->slot = slot;
1507 cstate->session = session;
1508
1509replay_cache:
1510 /* Renew the clientid on success and on replay.
1511 * Hold a session reference until done processing the compound:
1512 * nfsd4_put_session called only if the cstate slot is set.
1513 */
1514 renew_client(session->se_client);
1515 nfsd4_get_session(session);
1516out:
1517 spin_unlock(&sessionid_lock);
1518 dprintk("%s: return %d\n", __func__, ntohl(status));
1519 return status;
069b6ad4
AA		 1520}
1521
b37ad28b 1522__be32
b591480b
BF		 1523nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1524 struct nfsd4_setclientid *setclid)
1da177e4 1525{
27459f09 1526 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1527 struct xdr_netobj clname = {
1528 .len = setclid->se_namelen,
1529 .data = setclid->se_name,
1530 };
1531 nfs4_verifier clverifier = setclid->se_verf;
1532 unsigned int strhashval;
28ce6054 1533 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1534 __be32 status;
68e76ad0 1535 char *princ;
a55370a3 1536 char dname[HEXDIR_LEN];
1da177e4 1537
1da177e4 1538 if (!check_name(clname))
73aea4ec 1539 return nfserr_inval;
1da177e4 1540
a55370a3
N		 1541 status = nfs4_make_rec_clidname(dname, &clname);
1542 if (status)
73aea4ec 1543 return status;
a55370a3 1544
1da177e4
LT		 1545 /*
1546 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1547 * We get here on a DRC miss.
1548 */
1549
a55370a3 1550 strhashval = clientstr_hashval(dname);
1da177e4 1551
1da177e4 1552 nfs4_lock_state();
a1bcecd2 1553 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1554 if (conf) {
a186e767 1555 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1556 status = nfserr_clid_inuse;
026722c2
BF		 1557 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1558 dprintk("NFSD: setclientid: string in use by client"
1559 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1560 goto out;
1561 }
1da177e4 1562 }
a186e767
BF		 1563 /*
1564 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1565 * has a description of SETCLIENTID request processing consisting
1566 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1567 */
a1bcecd2 1568 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1569 status = nfserr_resource;
1570 if (!conf) {
a186e767
BF		 1571 /*
1572 * RFC 3530 14.2.33 CASE 4:
1573 * placed first, because it is the normal case
1da177e4
LT		 1574 */
1575 if (unconf)
1576 expire_client(unconf);
a55370a3
N		 1577 new = create_client(clname, dname);
1578 if (new == NULL)
1da177e4 1579 goto out;
1da177e4 1580 gen_clid(new);
599e0a22 1581 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1582 /*
a186e767
BF		 1583 * RFC 3530 14.2.33 CASE 1:
1584 * probable callback update
1da177e4 1585 */
31f4a6c1
N		 1586 if (unconf) {
1587 /* Note this is removing unconfirmed {*x***},
1588 * which is stronger than RFC recommended {vxc**}.
1589 * This has the advantage that there is at most
1590 * one {*x***} in either list at any time.
1591 */
1592 expire_client(unconf);
1da177e4 1593 }
a55370a3
N		 1594 new = create_client(clname, dname);
1595 if (new == NULL)
1da177e4 1596 goto out;
1da177e4 1597 copy_clid(new, conf);
1da177e4
LT		 1598 } else if (!unconf) {
1599 /*
a186e767
BF		 1600 * RFC 3530 14.2.33 CASE 2:
1601 * probable client reboot; state will be removed if
1602 * confirmed.
1da177e4 1603 */
a55370a3
N		 1604 new = create_client(clname, dname);
1605 if (new == NULL)
1da177e4 1606 goto out;
1da177e4 1607 gen_clid(new);
49ba8781 1608 } else {
a186e767
BF		 1609 /*
1610 * RFC 3530 14.2.33 CASE 3:
1611 * probable client reboot; state will be removed if
1612 * confirmed.
1da177e4
LT		 1613 */
1614 expire_client(unconf);
a55370a3
N		 1615 new = create_client(clname, dname);
1616 if (new == NULL)
1da177e4 1617 goto out;
1da177e4 1618 gen_clid(new);
1da177e4 1619 }
c175b83c
BF		 1620 copy_verf(new, &clverifier);
1621 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1622 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1623 princ = svc_gss_principal(rqstp);
1624 if (princ) {
1625 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1626 if (new->cl_principal == NULL) {
1627 free_client(new);
1628 goto out;
1629 }
1630 }
c175b83c
BF		 1631 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1632 gen_confirm(new);
1633 gen_callback(new, setclid);
1634 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1635 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1636 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1637 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1638 status = nfs_ok;
1639out:
1640 nfs4_unlock_state();
1641 return status;
1642}
1643
1644
1645/*
a186e767
BF		 1646 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1647 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1648 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1649 */
b37ad28b 1650__be32
b591480b
BF		 1651nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1652 struct nfsd4_compound_state *cstate,
1653 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1654{
27459f09 1655 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1656 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1657 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1658 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1659 __be32 status;
1da177e4
LT		 1660
1661 if (STALE_CLIENTID(clid))
1662 return nfserr_stale_clientid;
1663 /*
1664 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1665 * We get here on a DRC miss.
1666 */
1667
1668 nfs4_lock_state();
21ab45a4
N		 1669
1670 conf = find_confirmed_client(clid);
1671 unconf = find_unconfirmed_client(clid);
1672
1673 status = nfserr_clid_inuse;
27459f09 1674 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1675 goto out;
27459f09 1676 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1677 goto out;
1678
a186e767
BF		 1679 /*
1680 * section 14.2.34 of RFC 3530 has a description of
1681 * SETCLIENTID_CONFIRM request processing consisting
1682 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1683 */
366e0c1d 1684 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1685 /*
1686 * RFC 3530 14.2.34 CASE 1:
1687 * callback update
1688 */
599e0a22 1689 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1690 status = nfserr_clid_inuse;
1691 else {
1a69c179
N		 1692 /* XXX: We just turn off callbacks until we can handle
1693 * change request correctly. */
cb36d634 1694 atomic_set(&conf->cl_callback.cb_set, 0);
1a69c179 1695 expire_client(unconf);
1da177e4 1696 status = nfs_ok;
1a69c179 1697
1da177e4 1698 }
f3aba4e5 1699 } else if (conf && !unconf) {
a186e767
BF		 1700 /*
1701 * RFC 3530 14.2.34 CASE 2:
1702 * probable retransmitted request; play it safe and
1703 * do nothing.
7c79f737 1704 */
599e0a22 1705 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1706 status = nfserr_clid_inuse;
21ab45a4 1707 else
1da177e4 1708 status = nfs_ok;
7c79f737 1709 } else if (!conf && unconf
599e0a22 1710 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1711 /*
1712 * RFC 3530 14.2.34 CASE 3:
1713 * Normal case; new or rebooted client:
7c79f737 1714 */
599e0a22 1715 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1716 status = nfserr_clid_inuse;
1717 } else {
1a69c179
N		 1718 unsigned int hash =
1719 clientstr_hashval(unconf->cl_recdir);
1720 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1721 hash, false);
1a69c179 1722 if (conf) {
c7b9a459 1723 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1724 expire_client(conf);
1725 }
1da177e4 1726 move_to_confirmed(unconf);
21ab45a4 1727 conf = unconf;
46f8a64b 1728 nfsd4_probe_callback(conf);
1a69c179 1729 status = nfs_ok;
1da177e4 1730 }
599e0a22
BF		 1731 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1732 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1733 &confirm)))) {
a186e767
BF		 1734 /*
1735 * RFC 3530 14.2.34 CASE 4:
1736 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1737 */
1da177e4 1738 status = nfserr_stale_clientid;
7c79f737 1739 } else {
08e8987c
N		 1740 /* check that we have hit one of the cases...*/
1741 status = nfserr_clid_inuse;
1742 }
1da177e4 1743out:
1da177e4
LT		 1744 nfs4_unlock_state();
1745 return status;
1746}
1747
1da177e4
LT		 1748/* OPEN Share state helper functions */
1749static inline struct nfs4_file *
1750alloc_init_file(struct inode *ino)
1751{
1752 struct nfs4_file *fp;
1753 unsigned int hashval = file_hashval(ino);
1754
e60d4398
N		 1755 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1756 if (fp) {
8b671b80 1757 atomic_set(&fp->fi_ref, 1);
1da177e4 1758 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1759 INIT_LIST_HEAD(&fp->fi_stateids);
1760 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1761 spin_lock(&recall_lock);
1da177e4 1762 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1763 spin_unlock(&recall_lock);
1da177e4
LT		 1764 fp->fi_inode = igrab(ino);
1765 fp->fi_id = current_fileid++;
47f9940c 1766 fp->fi_had_conflict = false;
1da177e4
LT		 1767 return fp;
1768 }
1769 return NULL;
1770}
1771
e60d4398 1772static void
e18b890b 1773nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1774{
e60d4398
N		 1775 if (*slab == NULL)
1776 return;
1a1d92c1 1777 kmem_cache_destroy(*slab);
e60d4398 1778 *slab = NULL;
1da177e4
LT		 1779}
1780
e8ff2a84 1781void
1da177e4
LT		 1782nfsd4_free_slabs(void)
1783{
e60d4398
N		 1784 nfsd4_free_slab(&stateowner_slab);
1785 nfsd4_free_slab(&file_slab);
5ac049ac 1786 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1787 nfsd4_free_slab(&deleg_slab);
e60d4398 1788}
1da177e4 1789
e60d4398
N		 1790static int
1791nfsd4_init_slabs(void)
1792{
1793 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1794 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1795 if (stateowner_slab == NULL)
1796 goto out_nomem;
1797 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1798 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1799 if (file_slab == NULL)
1800 goto out_nomem;
5ac049ac 1801 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1802 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1803 if (stateid_slab == NULL)
1804 goto out_nomem;
5b2d21c1 1805 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1806 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1807 if (deleg_slab == NULL)
1808 goto out_nomem;
e60d4398
N		 1809 return 0;
1810out_nomem:
1811 nfsd4_free_slabs();
1812 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1813 return -ENOMEM;
1da177e4
LT		 1814}
1815
1816void
1817nfs4_free_stateowner(struct kref *kref)
1818{
1819 struct nfs4_stateowner *sop =
1820 container_of(kref, struct nfs4_stateowner, so_ref);
1821 kfree(sop->so_owner.data);
1822 kmem_cache_free(stateowner_slab, sop);
1823}
1824
1825static inline struct nfs4_stateowner *
1826alloc_stateowner(struct xdr_netobj *owner)
1827{
1828 struct nfs4_stateowner *sop;
1829
1830 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1831 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1832 memcpy(sop->so_owner.data, owner->data, owner->len);
1833 sop->so_owner.len = owner->len;
1834 kref_init(&sop->so_ref);
1835 return sop;
1836 }
1837 kmem_cache_free(stateowner_slab, sop);
1838 }
1839 return NULL;
1840}
1841
1842static struct nfs4_stateowner *
1843alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1844 struct nfs4_stateowner *sop;
1845 struct nfs4_replay *rp;
1846 unsigned int idhashval;
1847
1848 if (!(sop = alloc_stateowner(&open->op_owner)))
1849 return NULL;
1850 idhashval = ownerid_hashval(current_ownerid);
1851 INIT_LIST_HEAD(&sop->so_idhash);
1852 INIT_LIST_HEAD(&sop->so_strhash);
1853 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1854 INIT_LIST_HEAD(&sop->so_stateids);
1855 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1856 INIT_LIST_HEAD(&sop->so_close_lru);
1857 sop->so_time = 0;
1858 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1859 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1860 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1861 sop->so_is_open_owner = 1;
1862 sop->so_id = current_ownerid++;
1863 sop->so_client = clp;
1864 sop->so_seqid = open->op_seqid;
1865 sop->so_confirmed = 0;
1866 rp = &sop->so_replay;
de1ae286 1867 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1868 rp->rp_buflen = 0;
1869 rp->rp_buf = rp->rp_ibuf;
1870 return sop;
1871}
1872
1da177e4
LT		 1873static inline void
1874init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1875 struct nfs4_stateowner *sop = open->op_stateowner;
1876 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1877
1878 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1879 INIT_LIST_HEAD(&stp->st_perstateowner);
1880 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1881 INIT_LIST_HEAD(&stp->st_perfile);
1882 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1883 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1884 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1885 stp->st_stateowner = sop;
13cd2184 1886 get_nfs4_file(fp);
1da177e4 1887 stp->st_file = fp;
78155ed7 1888 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1889 stp->st_stateid.si_stateownerid = sop->so_id;
1890 stp->st_stateid.si_fileid = fp->fi_id;
1891 stp->st_stateid.si_generation = 0;
1892 stp->st_access_bmap = 0;
1893 stp->st_deny_bmap = 0;
84459a11
AA		 1894 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1895 &stp->st_access_bmap);
1da177e4 1896 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1897 stp->st_openstp = NULL;
1da177e4
LT		 1898}
1899
fd39ca9a 1900static void
1da177e4
LT		 1901move_to_close_lru(struct nfs4_stateowner *sop)
1902{
1903 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1904
358dd55a 1905 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1906 sop->so_time = get_seconds();
1907}
1908
1da177e4 1909static int
599e0a22
BF		 1910same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1911 clientid_t *clid)
1912{
1913 return (sop->so_owner.len == owner->len) &&
1914 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1915 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1916}
1917
1918static struct nfs4_stateowner *
1919find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1920{
1921 struct nfs4_stateowner *so = NULL;
1922
1923 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1924 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1925 return so;
1926 }
1927 return NULL;
1928}
1929
1930/* search file_hashtbl[] for file */
1931static struct nfs4_file *
1932find_file(struct inode *ino)
1933{
1934 unsigned int hashval = file_hashval(ino);
1935 struct nfs4_file *fp;
1936
8b671b80 1937 spin_lock(&recall_lock);
1da177e4 1938 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1939 if (fp->fi_inode == ino) {
1940 get_nfs4_file(fp);
8b671b80 1941 spin_unlock(&recall_lock);
1da177e4 1942 return fp;
13cd2184 1943 }
1da177e4 1944 }
8b671b80 1945 spin_unlock(&recall_lock);
1da177e4
LT		 1946 return NULL;
1947}
1948
d87a8ade 1949static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1950{
d87a8ade 1951 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1952 return 0;
d87a8ade
AA		 1953 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1954 return 0;
1955 x &= ~NFS4_SHARE_ACCESS_MASK;
1956 if (minorversion && x) {
1957 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1958 return 0;
1959 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1960 return 0;
1961 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1962 }
1963 if (x)
8838dc43
BF		 1964 return 0;
1965 return 1;
ba5a6a19
BF		 1966}
1967
8838dc43 1968static inline int deny_valid(u32 x)
ba5a6a19 1969{
8838dc43
BF		 1970 /* Note: unlike access bits, deny bits may be zero. */
1971 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1972}
1da177e4 1973
4f83aa30
BF		 1974/*
1975 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1976 * st_{access,deny}_bmap field of the stateid, in order to track not
1977 * only what share bits are currently in force, but also what
1978 * combinations of share bits previous opens have used. This allows us
1979 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1980 * return an error if the client attempt to downgrade to a combination
1981 * of share bits not explicable by closing some of its previous opens.
1982 *
1983 * XXX: This enforcement is actually incomplete, since we don't keep
1984 * track of access/deny bit combinations; so, e.g., we allow:
1985 *
1986 * OPEN allow read, deny write
1987 * OPEN allow both, deny none
1988 * DOWNGRADE allow read, deny none
1989 *
1990 * which we should reject.
1991 */
fd39ca9a 1992static void
1da177e4
LT		 1993set_access(unsigned int *access, unsigned long bmap) {
1994 int i;
1995
1996 *access = 0;
1997 for (i = 1; i < 4; i++) {
1998 if (test_bit(i, &bmap))
1999 *access |= i;
2000 }
2001}
2002
fd39ca9a 2003static void
1da177e4
LT		 2004set_deny(unsigned int *deny, unsigned long bmap) {
2005 int i;
2006
2007 *deny = 0;
2008 for (i = 0; i < 4; i++) {
2009 if (test_bit(i, &bmap))
2010 *deny |= i ;
2011 }
2012}
2013
2014static int
2015test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2016 unsigned int access, deny;
2017
2018 set_access(&access, stp->st_access_bmap);
2019 set_deny(&deny, stp->st_deny_bmap);
2020 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2021 return 0;
2022 return 1;
2023}
2024
2025/*
2026 * Called to check deny when READ with all zero stateid or
2027 * WRITE with all zero or all one stateid
2028 */
b37ad28b 2029static __be32
1da177e4
LT		 2030nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2031{
2032 struct inode *ino = current_fh->fh_dentry->d_inode;
2033 struct nfs4_file *fp;
2034 struct nfs4_stateid *stp;
b37ad28b 2035 __be32 ret;
1da177e4
LT		 2036
2037 dprintk("NFSD: nfs4_share_conflict\n");
2038
2039 fp = find_file(ino);
13cd2184
N		 2040 if (!fp)
2041 return nfs_ok;
b700949b 2042 ret = nfserr_locked;
1da177e4 2043 /* Search for conflicting share reservations */
13cd2184
N		 2044 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2045 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2046 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2047 goto out;
1da177e4 2048 }
13cd2184
N		 2049 ret = nfs_ok;
2050out:
2051 put_nfs4_file(fp);
2052 return ret;
1da177e4
LT		 2053}
2054
2055static inline void
2056nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2057{
2058 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2059 drop_file_write_access(filp);
1da177e4
LT		 2060 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2061 }
2062}
2063
2064/*
2065 * Recall a delegation
2066 */
2067static int
2068do_recall(void *__dp)
2069{
2070 struct nfs4_delegation *dp = __dp;
2071
47f9940c 2072 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2073 nfsd4_cb_recall(dp);
2074 return 0;
2075}
2076
2077/*
2078 * Spawn a thread to perform a recall on the delegation represented
2079 * by the lease (file_lock)
2080 *
2081 * Called from break_lease() with lock_kernel() held.
2082 * Note: we assume break_lease will only call this *once* for any given
2083 * lease.
2084 */
2085static
2086void nfsd_break_deleg_cb(struct file_lock *fl)
2087{
2088 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2089 struct task_struct *t;
2090
2091 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2092 if (!dp)
2093 return;
2094
2095 /* We're assuming the state code never drops its reference
2096 * without first removing the lease. Since we're in this lease
2097 * callback (and since the lease code is serialized by the kernel
2098 * lock) we know the server hasn't removed the lease yet, we know
2099 * it's safe to take a reference: */
2100 atomic_inc(&dp->dl_count);
cfdcad4d 2101 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2102
2103 spin_lock(&recall_lock);
2104 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2105 spin_unlock(&recall_lock);
2106
2107 /* only place dl_time is set. protected by lock_kernel*/
2108 dp->dl_time = get_seconds();
2109
0272e1fd
BF		 2110 /*
2111 * We don't want the locks code to timeout the lease for us;
2112 * we'll remove it ourself if the delegation isn't returned
2113 * in time.
2114 */
2115 fl->fl_break_time = 0;
1da177e4
LT		 2116
2117 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2118 if (IS_ERR(t)) {
2119 struct nfs4_client *clp = dp->dl_client;
2120
2121 printk(KERN_INFO "NFSD: Callback thread failed for "
2122 "for client (clientid %08x/%08x)\n",
2123 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2124 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2125 nfs4_put_delegation(dp);
2126 }
2127}
2128
2129/*
2130 * The file_lock is being reapd.
2131 *
2132 * Called by locks_free_lock() with lock_kernel() held.
2133 */
2134static
2135void nfsd_release_deleg_cb(struct file_lock *fl)
2136{
2137 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2138
2139 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2140
2141 if (!(fl->fl_flags & FL_LEASE) || !dp)
2142 return;
2143 dp->dl_flock = NULL;
2144}
2145
2146/*
2147 * Set the delegation file_lock back pointer.
2148 *
a9933cea 2149 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2150 */
2151static
2152void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2153{
2154 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2155
2156 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2157 if (!dp)
2158 return;
2159 dp->dl_flock = new;
2160}
2161
2162/*
a9933cea 2163 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2164 */
2165static
2166int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2167{
2168 struct nfs4_delegation *onlistd =
2169 (struct nfs4_delegation *)onlist->fl_owner;
2170 struct nfs4_delegation *tryd =
2171 (struct nfs4_delegation *)try->fl_owner;
2172
2173 if (onlist->fl_lmops != try->fl_lmops)
2174 return 0;
2175
2176 return onlistd->dl_client == tryd->dl_client;
2177}
2178
2179
2180static
2181int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2182{
2183 if (arg & F_UNLCK)
2184 return lease_modify(onlist, arg);
2185 else
2186 return -EAGAIN;
2187}
2188
fd39ca9a 2189static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2190 .fl_break = nfsd_break_deleg_cb,
2191 .fl_release_private = nfsd_release_deleg_cb,
2192 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2193 .fl_mylease = nfsd_same_client_deleg_cb,
2194 .fl_change = nfsd_change_deleg_cb,
2195};
2196
2197
b37ad28b 2198__be32
6668958f
AA		 2199nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2200 struct nfsd4_open *open)
1da177e4 2201{
1da177e4
LT		 2202 clientid_t *clientid = &open->op_clientid;
2203 struct nfs4_client *clp = NULL;
2204 unsigned int strhashval;
2205 struct nfs4_stateowner *sop = NULL;
2206
1da177e4 2207 if (!check_name(open->op_owner))
0f442aa2 2208 return nfserr_inval;
1da177e4
LT		 2209
2210 if (STALE_CLIENTID(&open->op_clientid))
2211 return nfserr_stale_clientid;
2212
2213 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2214 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2215 open->op_stateowner = sop;
2216 if (!sop) {
2217 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2218 clp = find_confirmed_client(clientid);
2219 if (clp == NULL)
0f442aa2
BF		 2220 return nfserr_expired;
2221 goto renew;
1da177e4 2222 }
6668958f
AA		 2223 /* When sessions are used, skip open sequenceid processing */
2224 if (nfsd4_has_session(cstate))
2225 goto renew;
0f442aa2
BF		 2226 if (!sop->so_confirmed) {
2227 /* Replace unconfirmed owners without checking for replay. */
2228 clp = sop->so_client;
f044ff83 2229 release_openowner(sop);
0f442aa2
BF		 2230 open->op_stateowner = NULL;
2231 goto renew;
2232 }
2233 if (open->op_seqid == sop->so_seqid - 1) {
2234 if (sop->so_replay.rp_buflen)
a90b061c 2235 return nfserr_replay_me;
0f442aa2
BF		 2236 /* The original OPEN failed so spectacularly
2237 * that we don't even have replay data saved!
2238 * Therefore, we have no choice but to continue
2239 * processing this OPEN; presumably, we'll
2240 * fail again for the same reason.
2241 */
2242 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2243 goto renew;
2244 }
2245 if (open->op_seqid != sop->so_seqid)
2246 return nfserr_bad_seqid;
1da177e4 2247renew:
0f442aa2
BF		 2248 if (open->op_stateowner == NULL) {
2249 sop = alloc_init_open_stateowner(strhashval, clp, open);
2250 if (sop == NULL)
2251 return nfserr_resource;
2252 open->op_stateowner = sop;
2253 }
2254 list_del_init(&sop->so_close_lru);
1da177e4 2255 renew_client(sop->so_client);
0f442aa2 2256 return nfs_ok;
1da177e4
LT		 2257}
2258
b37ad28b 2259static inline __be32
4a6e43e6
N		 2260nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2261{
2262 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2263 return nfserr_openmode;
2264 else
2265 return nfs_ok;
2266}
2267
52f4fb43
N		 2268static struct nfs4_delegation *
2269find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2270{
2271 struct nfs4_delegation *dp;
2272
ea1da636 2273 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2274 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2275 return dp;
2276 }
2277 return NULL;
2278}
2279
b37ad28b 2280static __be32
567d9829
N		 2281nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2282 struct nfs4_delegation **dp)
2283{
2284 int flags;
b37ad28b 2285 __be32 status = nfserr_bad_stateid;
567d9829
N		 2286
2287 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2288 if (*dp == NULL)
c44c5eeb 2289 goto out;
567d9829
N		 2290 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2291 RD_STATE : WR_STATE;
2292 status = nfs4_check_delegmode(*dp, flags);
2293 if (status)
2294 *dp = NULL;
c44c5eeb
N		 2295out:
2296 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2297 return nfs_ok;
2298 if (status)
2299 return status;
2300 open->op_stateowner->so_confirmed = 1;
2301 return nfs_ok;
567d9829
N		 2302}
2303
b37ad28b 2304static __be32
1da177e4
LT		 2305nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2306{
2307 struct nfs4_stateid *local;
b37ad28b 2308 __be32 status = nfserr_share_denied;
1da177e4
LT		 2309 struct nfs4_stateowner *sop = open->op_stateowner;
2310
8beefa24 2311 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2312 /* ignore lock owners */
2313 if (local->st_stateowner->so_is_open_owner == 0)
2314 continue;
2315 /* remember if we have seen this open owner */
2316 if (local->st_stateowner == sop)
2317 *stpp = local;
2318 /* check for conflicting share reservations */
2319 if (!test_share(local, open))
2320 goto out;
2321 }
2322 status = 0;
2323out:
2324 return status;
2325}
2326
5ac049ac
N		 2327static inline struct nfs4_stateid *
2328nfs4_alloc_stateid(void)
2329{
2330 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2331}
2332
b37ad28b 2333static __be32
1da177e4 2334nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2335 struct nfs4_delegation *dp,
1da177e4
LT		 2336 struct svc_fh *cur_fh, int flags)
2337{
2338 struct nfs4_stateid *stp;
1da177e4 2339
5ac049ac 2340 stp = nfs4_alloc_stateid();
1da177e4
LT		 2341 if (stp == NULL)
2342 return nfserr_resource;
2343
567d9829
N		 2344 if (dp) {
2345 get_file(dp->dl_vfs_file);
2346 stp->st_vfs_file = dp->dl_vfs_file;
2347 } else {
b37ad28b 2348 __be32 status;
567d9829
N		 2349 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2350 &stp->st_vfs_file);
2351 if (status) {
2352 if (status == nfserr_dropit)
2353 status = nfserr_jukebox;
5ac049ac 2354 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2355 return status;
2356 }
1da177e4 2357 }
1da177e4
LT		 2358 *stpp = stp;
2359 return 0;
2360}
2361
b37ad28b 2362static inline __be32
1da177e4
LT		 2363nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2364 struct nfsd4_open *open)
2365{
2366 struct iattr iattr = {
2367 .ia_valid = ATTR_SIZE,
2368 .ia_size = 0,
2369 };
2370 if (!open->op_truncate)
2371 return 0;
2372 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2373 return nfserr_inval;
1da177e4
LT		 2374 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2375}
2376
b37ad28b 2377static __be32
1da177e4
LT		 2378nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2379{
2380 struct file *filp = stp->st_vfs_file;
7eaa36e2 2381 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2382 unsigned int share_access, new_writer;
b37ad28b 2383 __be32 status;
1da177e4
LT		 2384
2385 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2386 new_writer = (~share_access) & open->op_share_access
2387 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2388
6c26d08f 2389 if (new_writer) {
b37ad28b
AV		 2390 int err = get_write_access(inode);
2391 if (err)
2392 return nfserrno(err);
e518f056
BH		 2393 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2394 if (err)
2395 return nfserrno(err);
2396 file_take_write(filp);
6c26d08f 2397 }
1da177e4
LT		 2398 status = nfsd4_truncate(rqstp, cur_fh, open);
2399 if (status) {
6c26d08f
BF		 2400 if (new_writer)
2401 put_write_access(inode);
1da177e4
LT		 2402 return status;
2403 }
2404 /* remember the open */
6c26d08f 2405 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2406 __set_bit(open->op_share_access, &stp->st_access_bmap);
2407 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2408
2409 return nfs_ok;
2410}
2411
2412
1da177e4 2413static void
37515177 2414nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2415{
37515177
N		 2416 open->op_stateowner->so_confirmed = 1;
2417 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2418}
2419
2420/*
2421 * Attempt to hand out a delegation.
2422 */
2423static void
2424nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2425{
2426 struct nfs4_delegation *dp;
2427 struct nfs4_stateowner *sop = stp->st_stateowner;
2428 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2429 struct file_lock fl, *flp = &fl;
2430 int status, flag = 0;
2431
2432 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2433 open->op_recall = 0;
2434 switch (open->op_claim_type) {
2435 case NFS4_OPEN_CLAIM_PREVIOUS:
2436 if (!atomic_read(&cb->cb_set))
2437 open->op_recall = 1;
2438 flag = open->op_delegate_type;
2439 if (flag == NFS4_OPEN_DELEGATE_NONE)
2440 goto out;
2441 break;
2442 case NFS4_OPEN_CLAIM_NULL:
2443 /* Let's not give out any delegations till everyone's
2444 * had the chance to reclaim theirs.... */
af558e33 2445 if (locks_in_grace())
7b190fec
N		 2446 goto out;
2447 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2448 goto out;
2449 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2450 flag = NFS4_OPEN_DELEGATE_WRITE;
2451 else
2452 flag = NFS4_OPEN_DELEGATE_READ;
2453 break;
2454 default:
2455 goto out;
2456 }
1da177e4
LT		 2457
2458 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2459 if (dp == NULL) {
2460 flag = NFS4_OPEN_DELEGATE_NONE;
2461 goto out;
2462 }
2463 locks_init_lock(&fl);
2464 fl.fl_lmops = &nfsd_lease_mng_ops;
2465 fl.fl_flags = FL_LEASE;
9167f501 2466 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2467 fl.fl_end = OFFSET_MAX;
2468 fl.fl_owner = (fl_owner_t)dp;
2469 fl.fl_file = stp->st_vfs_file;
2470 fl.fl_pid = current->tgid;
2471
a9933cea 2472 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2473 * the lock_manager callbacks fl_mylease and fl_change are used
2474 */
9167f501 2475 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2476 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2477 unhash_delegation(dp);
1da177e4
LT		 2478 flag = NFS4_OPEN_DELEGATE_NONE;
2479 goto out;
2480 }
2481
2482 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2483
2484 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2485 dp->dl_stateid.si_boot,
2486 dp->dl_stateid.si_stateownerid,
2487 dp->dl_stateid.si_fileid,
2488 dp->dl_stateid.si_generation);
2489out:
7b190fec
N		 2490 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2491 && flag == NFS4_OPEN_DELEGATE_NONE
2492 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2493 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2494 open->op_delegate_type = flag;
2495}
2496
2497/*
2498 * called with nfs4_lock_state() held.
2499 */
b37ad28b 2500__be32
1da177e4
LT		 2501nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2502{
6668958f 2503 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2504 struct nfs4_file *fp = NULL;
2505 struct inode *ino = current_fh->fh_dentry->d_inode;
2506 struct nfs4_stateid *stp = NULL;
567d9829 2507 struct nfs4_delegation *dp = NULL;
b37ad28b 2508 __be32 status;
1da177e4
LT		 2509
2510 status = nfserr_inval;
d87a8ade 2511 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2512 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2513 goto out;
2514 /*
2515 * Lookup file; if found, lookup stateid and check open request,
2516 * and check for delegations in the process of being recalled.
2517 * If not found, create the nfs4_file struct
2518 */
2519 fp = find_file(ino);
2520 if (fp) {
2521 if ((status = nfs4_check_open(fp, open, &stp)))
2522 goto out;
c44c5eeb
N		 2523 status = nfs4_check_deleg(fp, open, &dp);
2524 if (status)
2525 goto out;
1da177e4 2526 } else {
c44c5eeb
N		 2527 status = nfserr_bad_stateid;
2528 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2529 goto out;
1da177e4
LT		 2530 status = nfserr_resource;
2531 fp = alloc_init_file(ino);
2532 if (fp == NULL)
2533 goto out;
2534 }
2535
2536 /*
2537 * OPEN the file, or upgrade an existing OPEN.
2538 * If truncate fails, the OPEN fails.
2539 */
2540 if (stp) {
2541 /* Stateid was found, this is an OPEN upgrade */
2542 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2543 if (status)
2544 goto out;
444c2c07 2545 update_stateid(&stp->st_stateid);
1da177e4
LT		 2546 } else {
2547 /* Stateid was not found, this is a new OPEN */
2548 int flags = 0;
9ecb6a08 2549 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2550 flags |= NFSD_MAY_READ;
1da177e4 2551 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2552 flags |= NFSD_MAY_WRITE;
567d9829
N		 2553 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2554 if (status)
1da177e4
LT		 2555 goto out;
2556 init_stateid(stp, fp, open);
2557 status = nfsd4_truncate(rqstp, current_fh, open);
2558 if (status) {
2283963f 2559 release_open_stateid(stp);
1da177e4
LT		 2560 goto out;
2561 }
6668958f
AA		 2562 if (nfsd4_has_session(&resp->cstate))
2563 update_stateid(&stp->st_stateid);
1da177e4
LT		 2564 }
2565 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2566
6668958f
AA		 2567 if (nfsd4_has_session(&resp->cstate))
2568 open->op_stateowner->so_confirmed = 1;
2569
1da177e4
LT		 2570 /*
2571 * Attempt to hand out a delegation. No error return, because the
2572 * OPEN succeeds even if we fail.
2573 */
2574 nfs4_open_delegation(current_fh, open, stp);
2575
2576 status = nfs_ok;
2577
2578 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2579 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2580 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2581out:
13cd2184
N		 2582 if (fp)
2583 put_nfs4_file(fp);
37515177
N		 2584 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2585 nfs4_set_claim_prev(open);
1da177e4
LT		 2586 /*
2587 * To finish the open response, we just need to set the rflags.
2588 */
2589 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2590 if (!open->op_stateowner->so_confirmed &&
2591 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2592 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2593
2594 return status;
2595}
2596
b37ad28b 2597__be32
b591480b
BF		 2598nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2599 clientid_t *clid)
1da177e4
LT		 2600{
2601 struct nfs4_client *clp;
b37ad28b 2602 __be32 status;
1da177e4
LT		 2603
2604 nfs4_lock_state();
2605 dprintk("process_renew(%08x/%08x): starting\n",
2606 clid->cl_boot, clid->cl_id);
2607 status = nfserr_stale_clientid;
2608 if (STALE_CLIENTID(clid))
2609 goto out;
2610 clp = find_confirmed_client(clid);
2611 status = nfserr_expired;
2612 if (clp == NULL) {
2613 /* We assume the client took too long to RENEW. */
2614 dprintk("nfsd4_renew: clientid not found!\n");
2615 goto out;
2616 }
2617 renew_client(clp);
2618 status = nfserr_cb_path_down;
ea1da636 2619 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2620 && !atomic_read(&clp->cl_callback.cb_set))
2621 goto out;
2622 status = nfs_ok;
2623out:
2624 nfs4_unlock_state();
2625 return status;
2626}
2627
af558e33
BF		 2628struct lock_manager nfsd4_manager = {
2629};
2630
a76b4319 2631static void
af558e33 2632nfsd4_end_grace(void)
a76b4319
N		 2633{
2634 dprintk("NFSD: end of grace period\n");
c7b9a459 2635 nfsd4_recdir_purge_old();
af558e33 2636 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2637}
2638
fd39ca9a 2639static time_t
1da177e4
LT		 2640nfs4_laundromat(void)
2641{
2642 struct nfs4_client *clp;
2643 struct nfs4_stateowner *sop;
2644 struct nfs4_delegation *dp;
2645 struct list_head *pos, *next, reaplist;
2646 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2647 time_t t, clientid_val = NFSD_LEASE_TIME;
2648 time_t u, test_val = NFSD_LEASE_TIME;
2649
2650 nfs4_lock_state();
2651
2652 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2653 if (locks_in_grace())
2654 nfsd4_end_grace();
1da177e4
LT		 2655 list_for_each_safe(pos, next, &client_lru) {
2656 clp = list_entry(pos, struct nfs4_client, cl_lru);
2657 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2658 t = clp->cl_time - cutoff;
2659 if (clientid_val > t)
2660 clientid_val = t;
2661 break;
2662 }
2663 dprintk("NFSD: purging unused client (clientid %08x)\n",
2664 clp->cl_clientid.cl_id);
c7b9a459 2665 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2666 expire_client(clp);
2667 }
2668 INIT_LIST_HEAD(&reaplist);
2669 spin_lock(&recall_lock);
2670 list_for_each_safe(pos, next, &del_recall_lru) {
2671 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2672 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2673 u = dp->dl_time - cutoff;
2674 if (test_val > u)
2675 test_val = u;
2676 break;
2677 }
2678 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2679 dp, dp->dl_flock);
2680 list_move(&dp->dl_recall_lru, &reaplist);
2681 }
2682 spin_unlock(&recall_lock);
2683 list_for_each_safe(pos, next, &reaplist) {
2684 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2685 list_del_init(&dp->dl_recall_lru);
2686 unhash_delegation(dp);
2687 }
2688 test_val = NFSD_LEASE_TIME;
2689 list_for_each_safe(pos, next, &close_lru) {
2690 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2691 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2692 u = sop->so_time - cutoff;
2693 if (test_val > u)
2694 test_val = u;
2695 break;
2696 }
2697 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2698 sop->so_id);
f044ff83 2699 release_openowner(sop);
1da177e4
LT		 2700 }
2701 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2702 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2703 nfs4_unlock_state();
2704 return clientid_val;
2705}
2706
a254b246
HH		 2707static struct workqueue_struct *laundry_wq;
2708static void laundromat_main(struct work_struct *);
2709static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2710
2711static void
c4028958 2712laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2713{
2714 time_t t;
2715
2716 t = nfs4_laundromat();
2717 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2718 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2719}
2720
fd39ca9a 2721static struct nfs4_stateowner *
f8816512
N		 2722search_close_lru(u32 st_id, int flags)
2723{
1da177e4
LT		 2724 struct nfs4_stateowner *local = NULL;
2725
1da177e4
LT		 2726 if (flags & CLOSE_STATE) {
2727 list_for_each_entry(local, &close_lru, so_close_lru) {
2728 if (local->so_id == st_id)
2729 return local;
2730 }
2731 }
2732 return NULL;
2733}
2734
2735static inline int
2736nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2737{
7eaa36e2 2738 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2739}
2740
2741static int
2742STALE_STATEID(stateid_t *stateid)
2743{
78155ed7
BN		 2744 if (time_after((unsigned long)boot_time,
2745 (unsigned long)stateid->si_boot)) {
2746 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2747 stateid->si_boot, stateid->si_stateownerid,
2748 stateid->si_fileid, stateid->si_generation);
2749 return 1;
2750 }
2751 return 0;
2752}
2753
2754static int
2755EXPIRED_STATEID(stateid_t *stateid)
2756{
2757 if (time_before((unsigned long)boot_time,
2758 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2759 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2760 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2761 stateid->si_boot, stateid->si_stateownerid,
2762 stateid->si_fileid, stateid->si_generation);
2763 return 1;
2764 }
2765 return 0;
2766}
2767
2768static __be32
2769stateid_error_map(stateid_t *stateid)
2770{
2771 if (STALE_STATEID(stateid))
2772 return nfserr_stale_stateid;
2773 if (EXPIRED_STATEID(stateid))
2774 return nfserr_expired;
2775
2776 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2777 stateid->si_boot, stateid->si_stateownerid,
2778 stateid->si_fileid, stateid->si_generation);
2779 return nfserr_bad_stateid;
1da177e4
LT		 2780}
2781
2782static inline int
2783access_permit_read(unsigned long access_bmap)
2784{
2785 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2786 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2787 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2788}
2789
2790static inline int
2791access_permit_write(unsigned long access_bmap)
2792{
2793 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2794 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2795}
2796
2797static
b37ad28b 2798__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2799{
b37ad28b 2800 __be32 status = nfserr_openmode;
1da177e4
LT		 2801
2802 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2803 goto out;
2804 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2805 goto out;
2806 status = nfs_ok;
2807out:
2808 return status;
2809}
2810
b37ad28b 2811static inline __be32
1da177e4
LT		 2812check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2813{
203a8c8e 2814 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2815 return nfs_ok;
af558e33 2816 else if (locks_in_grace()) {
1da177e4
LT		 2817 /* Answer in remaining cases depends on existance of
2818 * conflicting state; so we must wait out the grace period. */
2819 return nfserr_grace;
2820 } else if (flags & WR_STATE)
2821 return nfs4_share_conflict(current_fh,
2822 NFS4_SHARE_DENY_WRITE);
2823 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2824 return nfs4_share_conflict(current_fh,
2825 NFS4_SHARE_DENY_READ);
2826}
2827
2828/*
2829 * Allow READ/WRITE during grace period on recovered state only for files
2830 * that are not able to provide mandatory locking.
2831 */
2832static inline int
18f82731 2833grace_disallows_io(struct inode *inode)
1da177e4 2834{
203a8c8e 2835 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2836}
2837
6668958f 2838static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2839{
6668958f
AA		 2840 /*
2841 * When sessions are used the stateid generation number is ignored
2842 * when it is zero.
2843 */
2844 if ((flags & HAS_SESSION) && in->si_generation == 0)
2845 goto out;
2846
0836f587
BF		 2847 /* If the client sends us a stateid from the future, it's buggy: */
2848 if (in->si_generation > ref->si_generation)
2849 return nfserr_bad_stateid;
2850 /*
2851 * The following, however, can happen. For example, if the
2852 * client sends an open and some IO at the same time, the open
2853 * may bump si_generation while the IO is still in flight.
2854 * Thanks to hard links and renames, the client never knows what
2855 * file an open will affect. So it could avoid that situation
2856 * only by serializing all opens and IO from the same open
2857 * owner. To recover from the old_stateid error, the client
2858 * will just have to retry the IO:
2859 */
2860 if (in->si_generation < ref->si_generation)
2861 return nfserr_old_stateid;
6668958f 2862out:
0836f587
BF		 2863 return nfs_ok;
2864}
2865
3e633079
BF		 2866static int is_delegation_stateid(stateid_t *stateid)
2867{
2868 return stateid->si_fileid == 0;
2869}
2870
1da177e4
LT		 2871/*
2872* Checks for stateid operations
2873*/
b37ad28b 2874__be32
dd453dfd
BH		 2875nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2876 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2877{
2878 struct nfs4_stateid *stp = NULL;
2879 struct nfs4_delegation *dp = NULL;
dd453dfd 2880 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2881 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2882 __be32 status;
1da177e4 2883
1da177e4
LT		 2884 if (filpp)
2885 *filpp = NULL;
2886
18f82731 2887 if (grace_disallows_io(ino))
1da177e4
LT		 2888 return nfserr_grace;
2889
6668958f
AA		 2890 if (nfsd4_has_session(cstate))
2891 flags |= HAS_SESSION;
2892
1da177e4
LT		 2893 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2894 return check_special_stateids(current_fh, stateid, flags);
2895
1da177e4
LT		 2896 status = nfserr_stale_stateid;
2897 if (STALE_STATEID(stateid))
2898 goto out;
2899
1da177e4 2900 status = nfserr_bad_stateid;
3e633079 2901 if (is_delegation_stateid(stateid)) {
a4455be0 2902 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2903 if (!dp) {
2904 status = stateid_error_map(stateid);
1da177e4 2905 goto out;
78155ed7 2906 }
6668958f
AA		 2907 status = check_stateid_generation(stateid, &dp->dl_stateid,
2908 flags);
0c2a498f
BF		 2909 if (status)
2910 goto out;
dc9bf700
BF		 2911 status = nfs4_check_delegmode(dp, flags);
2912 if (status)
2913 goto out;
2914 renew_client(dp->dl_client);
dc9bf700
BF		 2915 if (filpp)
2916 *filpp = dp->dl_vfs_file;
1da177e4 2917 } else { /* open or lock stateid */
a4455be0 2918 stp = find_stateid(stateid, flags);
78155ed7
BN		 2919 if (!stp) {
2920 status = stateid_error_map(stateid);
1da177e4 2921 goto out;
78155ed7 2922 }
6150ef0d 2923 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2924 goto out;
2925 if (!stp->st_stateowner->so_confirmed)
2926 goto out;
6668958f
AA		 2927 status = check_stateid_generation(stateid, &stp->st_stateid,
2928 flags);
0c2a498f
BF		 2929 if (status)
2930 goto out;
a4455be0
BF		 2931 status = nfs4_check_openmode(stp, flags);
2932 if (status)
1da177e4
LT		 2933 goto out;
2934 renew_client(stp->st_stateowner->so_client);
2935 if (filpp)
2936 *filpp = stp->st_vfs_file;
1da177e4
LT		 2937 }
2938 status = nfs_ok;
2939out:
2940 return status;
2941}
2942
4c4cd222
N		 2943static inline int
2944setlkflg (int type)
2945{
2946 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2947 RD_STATE : WR_STATE;
2948}
1da177e4
LT		 2949
2950/*
2951 * Checks for sequence id mutating operations.
2952 */
b37ad28b 2953static __be32
dd453dfd
BH		 2954nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2955 stateid_t *stateid, int flags,
2956 struct nfs4_stateowner **sopp,
2957 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2958{
1da177e4
LT		 2959 struct nfs4_stateid *stp;
2960 struct nfs4_stateowner *sop;
dd453dfd 2961 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2962 __be32 status;
1da177e4
LT		 2963
2964 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2965 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2966 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2967 stateid->si_generation);
3a4f98bb 2968
1da177e4
LT		 2969 *stpp = NULL;
2970 *sopp = NULL;
2971
1da177e4 2972 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2973 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2974 return nfserr_bad_stateid;
1da177e4
LT		 2975 }
2976
1da177e4 2977 if (STALE_STATEID(stateid))
3a4f98bb 2978 return nfserr_stale_stateid;
6668958f
AA		 2979
2980 if (nfsd4_has_session(cstate))
2981 flags |= HAS_SESSION;
2982
1da177e4
LT		 2983 /*
2984 * We return BAD_STATEID if filehandle doesn't match stateid,
2985 * the confirmed flag is incorrecly set, or the generation
2986 * number is incorrect.
1da177e4 2987 */
f8816512
N		 2988 stp = find_stateid(stateid, flags);
2989 if (stp == NULL) {
2990 /*
2991 * Also, we should make sure this isn't just the result of
2992 * a replayed close:
2993 */
2994 sop = search_close_lru(stateid->si_stateownerid, flags);
2995 if (sop == NULL)
78155ed7 2996 return stateid_error_map(stateid);
f8816512
N		 2997 *sopp = sop;
2998 goto check_replay;
2999 }
1da177e4 3000
39325bd0
BF		 3001 *stpp = stp;
3002 *sopp = sop = stp->st_stateowner;
3003
4c4cd222 3004 if (lock) {
4c4cd222 3005 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 3006 struct nfs4_client *clp = sop->so_client;
4c4cd222 3007 int lkflg = 0;
b37ad28b 3008 __be32 status;
4c4cd222
N		 3009
3010 lkflg = setlkflg(lock->lk_type);
3011
3012 if (lock->lk_is_new) {
599e0a22
BF		 3013 if (!sop->so_is_open_owner)
3014 return nfserr_bad_stateid;
60adfc50
AA		 3015 if (!(flags & HAS_SESSION) &&
3016 !same_clid(&clp->cl_clientid, lockclid))
3017 return nfserr_bad_stateid;
599e0a22
BF		 3018 /* stp is the open stateid */
3019 status = nfs4_check_openmode(stp, lkflg);
3020 if (status)
3021 return status;
3022 } else {
3023 /* stp is the lock stateid */
3024 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3025 if (status)
3026 return status;
4c4cd222 3027 }
1da177e4
LT		 3028 }
3029
f3362737 3030 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3031 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3032 return nfserr_bad_stateid;
1da177e4
LT		 3033 }
3034
1da177e4
LT		 3035 /*
3036 * We now validate the seqid and stateid generation numbers.
3037 * For the moment, we ignore the possibility of
3038 * generation number wraparound.
3039 */
6668958f 3040 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3041 goto check_replay;
3042
3a4f98bb 3043 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3044 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3045 " unconfirmed stateowner!\n");
3046 return nfserr_bad_stateid;
1da177e4 3047 }
3a4f98bb 3048 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3049 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3050 " confirmed yet!\n");
3051 return nfserr_bad_stateid;
1da177e4 3052 }
6668958f 3053 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3054 if (status)
3055 return status;
52fd004e 3056 renew_client(sop->so_client);
3a4f98bb 3057 return nfs_ok;
1da177e4 3058
1da177e4 3059check_replay:
bd9aac52 3060 if (seqid == sop->so_seqid - 1) {
849823c5 3061 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3062 /* indicate replay to calling function */
a90b061c 3063 return nfserr_replay_me;
1da177e4 3064 }
2fdada03 3065 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3066 sop->so_seqid, seqid);
3067 *sopp = NULL;
3068 return nfserr_bad_seqid;
1da177e4
LT		 3069}
3070
b37ad28b 3071__be32
ca364317 3072nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3073 struct nfsd4_open_confirm *oc)
1da177e4 3074{
b37ad28b 3075 __be32 status;
1da177e4
LT		 3076 struct nfs4_stateowner *sop;
3077 struct nfs4_stateid *stp;
3078
3079 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3080 (int)cstate->current_fh.fh_dentry->d_name.len,
3081 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3082
ca364317 3083 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3084 if (status)
3085 return status;
1da177e4
LT		 3086
3087 nfs4_lock_state();
3088
dd453dfd 3089 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3090 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3091 CONFIRM | OPEN_STATE,
1da177e4
LT		 3092 &oc->oc_stateowner, &stp, NULL)))
3093 goto out;
3094
3095 sop = oc->oc_stateowner;
3096 sop->so_confirmed = 1;
3097 update_stateid(&stp->st_stateid);
3098 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3099 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3100 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3101 stp->st_stateid.si_boot,
3102 stp->st_stateid.si_stateownerid,
3103 stp->st_stateid.si_fileid,
3104 stp->st_stateid.si_generation);
c7b9a459
N		 3105
3106 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3107out:
f2327d9a 3108 if (oc->oc_stateowner) {
1da177e4 3109 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3110 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3111 }
1da177e4
LT		 3112 nfs4_unlock_state();
3113 return status;
3114}
3115
3116
3117/*
3118 * unset all bits in union bitmap (bmap) that
3119 * do not exist in share (from successful OPEN_DOWNGRADE)
3120 */
3121static void
3122reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3123{
3124 int i;
3125 for (i = 1; i < 4; i++) {
3126 if ((i & access) != i)
3127 __clear_bit(i, bmap);
3128 }
3129}
3130
3131static void
3132reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3133{
3134 int i;
3135 for (i = 0; i < 4; i++) {
3136 if ((i & deny) != i)
3137 __clear_bit(i, bmap);
3138 }
3139}
3140
b37ad28b 3141__be32
ca364317
BF		 3142nfsd4_open_downgrade(struct svc_rqst *rqstp,
3143 struct nfsd4_compound_state *cstate,
a4f1706a 3144 struct nfsd4_open_downgrade *od)
1da177e4 3145{
b37ad28b 3146 __be32 status;
1da177e4
LT		 3147 struct nfs4_stateid *stp;
3148 unsigned int share_access;
3149
3150 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3151 (int)cstate->current_fh.fh_dentry->d_name.len,
3152 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3153
d87a8ade 3154 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3155 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3156 return nfserr_inval;
3157
3158 nfs4_lock_state();
dd453dfd 3159 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3160 od->od_seqid,
1da177e4 3161 &od->od_stateid,
f3362737 3162 OPEN_STATE,
1da177e4
LT		 3163 &od->od_stateowner, &stp, NULL)))
3164 goto out;
3165
3166 status = nfserr_inval;
3167 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3168 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3169 stp->st_access_bmap, od->od_share_access);
3170 goto out;
3171 }
3172 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3173 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3174 stp->st_deny_bmap, od->od_share_deny);
3175 goto out;
3176 }
3177 set_access(&share_access, stp->st_access_bmap);
3178 nfs4_file_downgrade(stp->st_vfs_file,
3179 share_access & ~od->od_share_access);
3180
3181 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3182 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3183
3184 update_stateid(&stp->st_stateid);
3185 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3186 status = nfs_ok;
3187out:
f2327d9a 3188 if (od->od_stateowner) {
1da177e4 3189 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3190 cstate->replay_owner = od->od_stateowner;
f2327d9a 3191 }
1da177e4
LT		 3192 nfs4_unlock_state();
3193 return status;
3194}
3195
3196/*
3197 * nfs4_unlock_state() called after encode
3198 */
b37ad28b 3199__be32
ca364317 3200nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3201 struct nfsd4_close *close)
1da177e4 3202{
b37ad28b 3203 __be32 status;
1da177e4
LT		 3204 struct nfs4_stateid *stp;
3205
3206 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3207 (int)cstate->current_fh.fh_dentry->d_name.len,
3208 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3209
3210 nfs4_lock_state();
3211 /* check close_lru for replay */
dd453dfd 3212 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3213 close->cl_seqid,
1da177e4 3214 &close->cl_stateid,
f3362737 3215 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3216 &close->cl_stateowner, &stp, NULL)))
3217 goto out;
1da177e4
LT		 3218 status = nfs_ok;
3219 update_stateid(&stp->st_stateid);
3220 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3221
04ef5954 3222 /* release_stateid() calls nfsd_close() if needed */
2283963f 3223 release_open_stateid(stp);
04ef5954
BF		 3224
3225 /* place unused nfs4_stateowners on so_close_lru list to be
3226 * released by the laundromat service after the lease period
3227 * to enable us to handle CLOSE replay
3228 */
3229 if (list_empty(&close->cl_stateowner->so_stateids))
3230 move_to_close_lru(close->cl_stateowner);
1da177e4 3231out:
f2327d9a 3232 if (close->cl_stateowner) {
1da177e4 3233 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3234 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3235 }
1da177e4
LT		 3236 nfs4_unlock_state();
3237 return status;
3238}
3239
b37ad28b 3240__be32
ca364317
BF		 3241nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3242 struct nfsd4_delegreturn *dr)
1da177e4 3243{
203a8c8e
BF		 3244 struct nfs4_delegation *dp;
3245 stateid_t *stateid = &dr->dr_stateid;
3246 struct inode *inode;
b37ad28b 3247 __be32 status;
6668958f 3248 int flags = 0;
1da177e4 3249
ca364317 3250 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3251 return status;
3252 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3253
6668958f
AA		 3254 if (nfsd4_has_session(cstate))
3255 flags |= HAS_SESSION;
1da177e4 3256 nfs4_lock_state();
203a8c8e
BF		 3257 status = nfserr_bad_stateid;
3258 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3259 goto out;
3260 status = nfserr_stale_stateid;
3261 if (STALE_STATEID(stateid))
3262 goto out;
7e0f7cf5 3263 status = nfserr_bad_stateid;
203a8c8e
BF		 3264 if (!is_delegation_stateid(stateid))
3265 goto out;
203a8c8e 3266 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3267 if (!dp) {
3268 status = stateid_error_map(stateid);
203a8c8e 3269 goto out;
78155ed7 3270 }
6668958f 3271 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3272 if (status)
3273 goto out;
3274 renew_client(dp->dl_client);
3275
3276 unhash_delegation(dp);
1da177e4 3277out:
203a8c8e
BF		 3278 nfs4_unlock_state();
3279
1da177e4
LT		 3280 return status;
3281}
3282
3283
3284/*
3285 * Lock owner state (byte-range locks)
3286 */
3287#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3288#define LOCK_HASH_BITS 8
3289#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3290#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3291
87df4de8
BH		 3292static inline u64
3293end_offset(u64 start, u64 len)
3294{
3295 u64 end;
3296
3297 end = start + len;
3298 return end >= start ? end: NFS4_MAX_UINT64;
3299}
3300
3301/* last octet in a range */
3302static inline u64
3303last_byte_offset(u64 start, u64 len)
3304{
3305 u64 end;
3306
3307 BUG_ON(!len);
3308 end = start + len;
3309 return end > start ? end - 1: NFS4_MAX_UINT64;
3310}
3311
1da177e4
LT		 3312#define lockownerid_hashval(id) \
3313 ((id) & LOCK_HASH_MASK)
3314
3315static inline unsigned int
3316lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3317 struct xdr_netobj *ownername)
3318{
3319 return (file_hashval(inode) + cl_id
3320 + opaque_hashval(ownername->data, ownername->len))
3321 & LOCK_HASH_MASK;
3322}
3323
3324static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3325static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3326static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3327
fd39ca9a 3328static struct nfs4_stateid *
1da177e4
LT		 3329find_stateid(stateid_t *stid, int flags)
3330{
9346eff0 3331 struct nfs4_stateid *local;
1da177e4
LT		 3332 u32 st_id = stid->si_stateownerid;
3333 u32 f_id = stid->si_fileid;
3334 unsigned int hashval;
3335
3336 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3337 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3338 hashval = stateid_hashval(st_id, f_id);
3339 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3340 if ((local->st_stateid.si_stateownerid == st_id) &&
3341 (local->st_stateid.si_fileid == f_id))
3342 return local;
3343 }
3344 }
9346eff0
KK		 3345
3346 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3347 hashval = stateid_hashval(st_id, f_id);
3348 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3349 if ((local->st_stateid.si_stateownerid == st_id) &&
3350 (local->st_stateid.si_fileid == f_id))
3351 return local;
3352 }
849823c5 3353 }
1da177e4
LT		 3354 return NULL;
3355}
3356
3357static struct nfs4_delegation *
3358find_delegation_stateid(struct inode *ino, stateid_t *stid)
3359{
13cd2184
N		 3360 struct nfs4_file *fp;
3361 struct nfs4_delegation *dl;
1da177e4
LT		 3362
3363 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3364 stid->si_boot, stid->si_stateownerid,
3365 stid->si_fileid, stid->si_generation);
3366
1da177e4 3367 fp = find_file(ino);
13cd2184
N		 3368 if (!fp)
3369 return NULL;
3370 dl = find_delegation_file(fp, stid);
3371 put_nfs4_file(fp);
3372 return dl;
1da177e4
LT		 3373}
3374
3375/*
3376 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3377 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3378 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3379 * locking, this prevents us from being completely protocol-compliant. The
3380 * real solution to this problem is to start using unsigned file offsets in
3381 * the VFS, but this is a very deep change!
3382 */
3383static inline void
3384nfs4_transform_lock_offset(struct file_lock *lock)
3385{
3386 if (lock->fl_start < 0)
3387 lock->fl_start = OFFSET_MAX;
3388 if (lock->fl_end < 0)
3389 lock->fl_end = OFFSET_MAX;
3390}
3391
d5b9026a
N		 3392/* Hack!: For now, we're defining this just so we can use a pointer to it
3393 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3394static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3395};
1da177e4
LT		 3396
3397static inline void
3398nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3399{
d5b9026a
N		 3400 struct nfs4_stateowner *sop;
3401 unsigned int hval;
1da177e4 3402
d5b9026a
N		 3403 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3404 sop = (struct nfs4_stateowner *) fl->fl_owner;
3405 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3406 kref_get(&sop->so_ref);
3407 deny->ld_sop = sop;
3408 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3409 } else {
3410 deny->ld_sop = NULL;
3411 deny->ld_clientid.cl_boot = 0;
3412 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3413 }
3414 deny->ld_start = fl->fl_start;
87df4de8
BH		 3415 deny->ld_length = NFS4_MAX_UINT64;
3416 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3417 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3418 deny->ld_type = NFS4_READ_LT;
3419 if (fl->fl_type != F_RDLCK)
3420 deny->ld_type = NFS4_WRITE_LT;
3421}
3422
1da177e4
LT		 3423static struct nfs4_stateowner *
3424find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3425 struct xdr_netobj *owner)
3426{
3427 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3428 struct nfs4_stateowner *op;
3429
3430 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3431 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3432 return op;
3433 }
3434 return NULL;
3435}
3436
3437/*
3438 * Alloc a lock owner structure.
3439 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3440 * occured.
3441 *
3442 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3443 */
3444
3445static struct nfs4_stateowner *
3446alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3447 struct nfs4_stateowner *sop;
3448 struct nfs4_replay *rp;
3449 unsigned int idhashval;
3450
3451 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3452 return NULL;
3453 idhashval = lockownerid_hashval(current_ownerid);
3454 INIT_LIST_HEAD(&sop->so_idhash);
3455 INIT_LIST_HEAD(&sop->so_strhash);
3456 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3457 INIT_LIST_HEAD(&sop->so_stateids);
3458 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3459 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3460 sop->so_time = 0;
3461 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3462 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3463 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3464 sop->so_is_open_owner = 0;
3465 sop->so_id = current_ownerid++;
3466 sop->so_client = clp;
b59e3c0e
NB		 3467 /* It is the openowner seqid that will be incremented in encode in the
3468 * case of new lockowners; so increment the lock seqid manually: */
3469 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3470 sop->so_confirmed = 1;
3471 rp = &sop->so_replay;
de1ae286 3472 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3473 rp->rp_buflen = 0;
3474 rp->rp_buf = rp->rp_ibuf;
3475 return sop;
3476}
3477
fd39ca9a 3478static struct nfs4_stateid *
1da177e4
LT		 3479alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3480{
3481 struct nfs4_stateid *stp;
3482 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3483
5ac049ac
N		 3484 stp = nfs4_alloc_stateid();
3485 if (stp == NULL)
1da177e4
LT		 3486 goto out;
3487 INIT_LIST_HEAD(&stp->st_hash);
3488 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3489 INIT_LIST_HEAD(&stp->st_perstateowner);
3490 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3491 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3492 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3493 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3494 stp->st_stateowner = sop;
13cd2184 3495 get_nfs4_file(fp);
1da177e4 3496 stp->st_file = fp;
78155ed7 3497 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3498 stp->st_stateid.si_stateownerid = sop->so_id;
3499 stp->st_stateid.si_fileid = fp->fi_id;
3500 stp->st_stateid.si_generation = 0;
3501 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3502 stp->st_access_bmap = open_stp->st_access_bmap;
3503 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3504 stp->st_openstp = open_stp;
1da177e4
LT		 3505
3506out:
3507 return stp;
3508}
3509
fd39ca9a 3510static int
1da177e4
LT		 3511check_lock_length(u64 offset, u64 length)
3512{
87df4de8 3513 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3514 LOFF_OVERFLOW(offset, length)));
3515}
3516
3517/*
3518 * LOCK operation
3519 */
b37ad28b 3520__be32
ca364317 3521nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3522 struct nfsd4_lock *lock)
1da177e4 3523{
3e9e3dbe 3524 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3525 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3526 struct nfs4_stateid *lock_stp;
3527 struct file *filp;
3528 struct file_lock file_lock;
8dc7c311 3529 struct file_lock conflock;
b37ad28b 3530 __be32 status = 0;
1da177e4 3531 unsigned int strhashval;
150b3934 3532 unsigned int cmd;
b8dd7b9a 3533 int err;
1da177e4
LT		 3534
3535 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3536 (long long) lock->lk_offset,
3537 (long long) lock->lk_length);
3538
1da177e4
LT		 3539 if (check_lock_length(lock->lk_offset, lock->lk_length))
3540 return nfserr_inval;
3541
ca364317 3542 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3543 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3544 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3545 return status;
3546 }
3547
1da177e4
LT		 3548 nfs4_lock_state();
3549
3550 if (lock->lk_is_new) {
893f8770
N		 3551 /*
3552 * Client indicates that this is a new lockowner.
3553 * Use open owner and open stateid to create lock owner and
3554 * lock stateid.
3555 */
1da177e4
LT		 3556 struct nfs4_stateid *open_stp = NULL;
3557 struct nfs4_file *fp;
3558
3559 status = nfserr_stale_clientid;
60adfc50
AA		 3560 if (!nfsd4_has_session(cstate) &&
3561 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3562 goto out;
1da177e4 3563
1da177e4 3564 /* validate and update open stateid and open seqid */
dd453dfd 3565 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3566 lock->lk_new_open_seqid,
3567 &lock->lk_new_open_stateid,
f3362737 3568 OPEN_STATE,
3a65588a 3569 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3570 lock);
37515177 3571 if (status)
1da177e4 3572 goto out;
3a65588a 3573 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3574 /* create lockowner and lock stateid */
3575 fp = open_stp->st_file;
3576 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3577 open_sop->so_client->cl_clientid.cl_id,
3578 &lock->v.new.owner);
3e9e3dbe
N		 3579 /* XXX: Do we need to check for duplicate stateowners on
3580 * the same file, or should they just be allowed (and
3581 * create new stateids)? */
1da177e4 3582 status = nfserr_resource;
b59e3c0e
NB		 3583 lock_sop = alloc_init_lock_stateowner(strhashval,
3584 open_sop->so_client, open_stp, lock);
3585 if (lock_sop == NULL)
1da177e4 3586 goto out;
b59e3c0e 3587 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3588 if (lock_stp == NULL)
1da177e4 3589 goto out;
1da177e4
LT		 3590 } else {
3591 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3592 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3593 lock->lk_old_lock_seqid,
3594 &lock->lk_old_lock_stateid,
f3362737 3595 LOCK_STATE,
3a65588a 3596 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3597 if (status)
3598 goto out;
3a65588a 3599 lock_sop = lock->lk_replay_owner;
1da177e4 3600 }
3a65588a 3601 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3602 filp = lock_stp->st_vfs_file;
3603
0dd395dc 3604 status = nfserr_grace;
af558e33 3605 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3606 goto out;
3607 status = nfserr_no_grace;
af558e33 3608 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3609 goto out;
3610
1da177e4
LT		 3611 locks_init_lock(&file_lock);
3612 switch (lock->lk_type) {
3613 case NFS4_READ_LT:
3614 case NFS4_READW_LT:
3615 file_lock.fl_type = F_RDLCK;
150b3934 3616 cmd = F_SETLK;
1da177e4
LT		 3617 break;
3618 case NFS4_WRITE_LT:
3619 case NFS4_WRITEW_LT:
3620 file_lock.fl_type = F_WRLCK;
150b3934 3621 cmd = F_SETLK;
1da177e4
LT		 3622 break;
3623 default:
3624 status = nfserr_inval;
3625 goto out;
3626 }
b59e3c0e 3627 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3628 file_lock.fl_pid = current->tgid;
3629 file_lock.fl_file = filp;
3630 file_lock.fl_flags = FL_POSIX;
d5b9026a 3631 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3632
3633 file_lock.fl_start = lock->lk_offset;
87df4de8 3634 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3635 nfs4_transform_lock_offset(&file_lock);
3636
3637 /*
3638 * Try to lock the file in the VFS.
3639 * Note: locks.c uses the BKL to protect the inode's lock list.
3640 */
3641
fd85b817 3642 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3643 switch (-err) {
1da177e4
LT		 3644 case 0: /* success! */
3645 update_stateid(&lock_stp->st_stateid);
3646 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3647 sizeof(stateid_t));
b8dd7b9a 3648 status = 0;
eb76b3fd
AA		 3649 break;
3650 case (EAGAIN): /* conflock holds conflicting lock */
3651 status = nfserr_denied;
3652 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3653 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3654 break;
1da177e4
LT		 3655 case (EDEADLK):
3656 status = nfserr_deadlock;
eb76b3fd 3657 break;
1da177e4 3658 default:
fd85b817 3659 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3660 status = nfserr_resource;
3661 break;
1da177e4 3662 }
1da177e4 3663out:
8a280510 3664 if (status && lock->lk_is_new && lock_sop)
f044ff83 3665 release_lockowner(lock_sop);
3a65588a
BF		 3666 if (lock->lk_replay_owner) {
3667 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3668 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3669 }
1da177e4
LT		 3670 nfs4_unlock_state();
3671 return status;
3672}
3673
55ef1274
BF		 3674/*
3675 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3676 * so we do a temporary open here just to get an open file to pass to
3677 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3678 * inode operation.)
3679 */
3680static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3681{
3682 struct file *file;
3683 int err;
3684
3685 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3686 if (err)
3687 return err;
3688 err = vfs_test_lock(file, lock);
3689 nfsd_close(file);
3690 return err;
3691}
3692
1da177e4
LT		 3693/*
3694 * LOCKT operation
3695 */
b37ad28b 3696__be32
ca364317
BF		 3697nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3698 struct nfsd4_lockt *lockt)
1da177e4
LT		 3699{
3700 struct inode *inode;
1da177e4 3701 struct file_lock file_lock;
fd85b817 3702 int error;
b37ad28b 3703 __be32 status;
1da177e4 3704
af558e33 3705 if (locks_in_grace())
1da177e4
LT		 3706 return nfserr_grace;
3707
3708 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3709 return nfserr_inval;
3710
3711 lockt->lt_stateowner = NULL;
3712 nfs4_lock_state();
3713
3714 status = nfserr_stale_clientid;
60adfc50 3715 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3716 goto out;
1da177e4 3717
ca364317 3718 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3719 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3720 if (status == nfserr_symlink)
3721 status = nfserr_inval;
3722 goto out;
3723 }
3724
ca364317 3725 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3726 locks_init_lock(&file_lock);
3727 switch (lockt->lt_type) {
3728 case NFS4_READ_LT:
3729 case NFS4_READW_LT:
3730 file_lock.fl_type = F_RDLCK;
3731 break;
3732 case NFS4_WRITE_LT:
3733 case NFS4_WRITEW_LT:
3734 file_lock.fl_type = F_WRLCK;
3735 break;
3736 default:
2fdada03 3737 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3738 status = nfserr_inval;
3739 goto out;
3740 }
3741
3742 lockt->lt_stateowner = find_lockstateowner_str(inode,
3743 &lockt->lt_clientid, &lockt->lt_owner);
3744 if (lockt->lt_stateowner)
3745 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3746 file_lock.fl_pid = current->tgid;
3747 file_lock.fl_flags = FL_POSIX;
3748
3749 file_lock.fl_start = lockt->lt_offset;
87df4de8 3750 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3751
3752 nfs4_transform_lock_offset(&file_lock);
3753
1da177e4 3754 status = nfs_ok;
55ef1274 3755 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3756 if (error) {
3757 status = nfserrno(error);
3758 goto out;
3759 }
9d6a8c5c 3760 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3761 status = nfserr_denied;
9d6a8c5c 3762 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3763 }
3764out:
3765 nfs4_unlock_state();
3766 return status;
3767}
3768
b37ad28b 3769__be32
ca364317 3770nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3771 struct nfsd4_locku *locku)
1da177e4
LT		 3772{
3773 struct nfs4_stateid *stp;
3774 struct file *filp = NULL;
3775 struct file_lock file_lock;
b37ad28b 3776 __be32 status;
b8dd7b9a 3777 int err;
1da177e4
LT		 3778
3779 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3780 (long long) locku->lu_offset,
3781 (long long) locku->lu_length);
3782
3783 if (check_lock_length(locku->lu_offset, locku->lu_length))
3784 return nfserr_inval;
3785
3786 nfs4_lock_state();
3787
dd453dfd 3788 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3789 locku->lu_seqid,
3790 &locku->lu_stateid,
f3362737 3791 LOCK_STATE,
1da177e4
LT		 3792 &locku->lu_stateowner, &stp, NULL)))
3793 goto out;
3794
3795 filp = stp->st_vfs_file;
3796 BUG_ON(!filp);
3797 locks_init_lock(&file_lock);
3798 file_lock.fl_type = F_UNLCK;
3799 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3800 file_lock.fl_pid = current->tgid;
3801 file_lock.fl_file = filp;
3802 file_lock.fl_flags = FL_POSIX;
d5b9026a 3803 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3804 file_lock.fl_start = locku->lu_offset;
3805
87df4de8 3806 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3807 nfs4_transform_lock_offset(&file_lock);
3808
3809 /*
3810 * Try to unlock the file in the VFS.
3811 */
fd85b817 3812 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3813 if (err) {
fd85b817 3814 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3815 goto out_nfserr;
3816 }
3817 /*
3818 * OK, unlock succeeded; the only thing left to do is update the stateid.
3819 */
3820 update_stateid(&stp->st_stateid);
3821 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3822
3823out:
f2327d9a 3824 if (locku->lu_stateowner) {
1da177e4 3825 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3826 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3827 }
1da177e4
LT		 3828 nfs4_unlock_state();
3829 return status;
3830
3831out_nfserr:
b8dd7b9a 3832 status = nfserrno(err);
1da177e4
LT		 3833 goto out;
3834}
3835
3836/*
3837 * returns
3838 * 1: locks held by lockowner
3839 * 0: no locks held by lockowner
3840 */
3841static int
3842check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3843{
3844 struct file_lock **flpp;
7eaa36e2 3845 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3846 int status = 0;
3847
3848 lock_kernel();
3849 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3850 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3851 status = 1;
3852 goto out;
796dadfd 3853 }
1da177e4
LT		 3854 }
3855out:
3856 unlock_kernel();
3857 return status;
3858}
3859
b37ad28b 3860__be32
b591480b
BF		 3861nfsd4_release_lockowner(struct svc_rqst *rqstp,
3862 struct nfsd4_compound_state *cstate,
3863 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3864{
3865 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3866 struct nfs4_stateowner *sop;
3867 struct nfs4_stateid *stp;
1da177e4 3868 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3869 struct list_head matches;
3870 int i;
b37ad28b 3871 __be32 status;
1da177e4
LT		 3872
3873 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3874 clid->cl_boot, clid->cl_id);
3875
3876 /* XXX check for lease expiration */
3877
3878 status = nfserr_stale_clientid;
849823c5 3879 if (STALE_CLIENTID(clid))
1da177e4 3880 return status;
1da177e4
LT		 3881
3882 nfs4_lock_state();
3883
3e9e3dbe
N		 3884 status = nfserr_locks_held;
3885 /* XXX: we're doing a linear search through all the lockowners.
3886 * Yipes! For now we'll just hope clients aren't really using
3887 * release_lockowner much, but eventually we have to fix these
3888 * data structures. */
3889 INIT_LIST_HEAD(&matches);
3890 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3891 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3892 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3893 continue;
3894 list_for_each_entry(stp, &sop->so_stateids,
3895 st_perstateowner) {
3896 if (check_for_locks(stp->st_vfs_file, sop))
3897 goto out;
3898 /* Note: so_perclient unused for lockowners,
3899 * so it's OK to fool with here. */
3900 list_add(&sop->so_perclient, &matches);
3901 }
1da177e4 3902 }
3e9e3dbe
N		 3903 }
3904 /* Clients probably won't expect us to return with some (but not all)
3905 * of the lockowner state released; so don't release any until all
3906 * have been checked. */
3907 status = nfs_ok;
0fa822e4
N		 3908 while (!list_empty(&matches)) {
3909 sop = list_entry(matches.next, struct nfs4_stateowner,
3910 so_perclient);
3911 /* unhash_stateowner deletes so_perclient only
3912 * for openowners. */
3913 list_del(&sop->so_perclient);
f044ff83 3914 release_lockowner(sop);
1da177e4
LT		 3915 }
3916out:
3917 nfs4_unlock_state();
3918 return status;
3919}
3920
3921static inline struct nfs4_client_reclaim *
a55370a3 3922alloc_reclaim(void)
1da177e4 3923{
a55370a3 3924 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3925}
3926
c7b9a459 3927int
a1bcecd2 3928nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3929{
3930 unsigned int strhashval = clientstr_hashval(name);
3931 struct nfs4_client *clp;
3932
a1bcecd2 3933 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3934 return clp ? 1 : 0;
3935}
3936
1da177e4
LT		 3937/*
3938 * failure => all reset bets are off, nfserr_no_grace...
3939 */
190e4fbf
N		 3940int
3941nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3942{
3943 unsigned int strhashval;
3944 struct nfs4_client_reclaim *crp = NULL;
3945
a55370a3
N		 3946 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3947 crp = alloc_reclaim();
1da177e4
LT		 3948 if (!crp)
3949 return 0;
a55370a3 3950 strhashval = clientstr_hashval(name);
1da177e4
LT		 3951 INIT_LIST_HEAD(&crp->cr_strhash);
3952 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3953 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3954 reclaim_str_hashtbl_size++;
3955 return 1;
3956}
3957
3958static void
3959nfs4_release_reclaim(void)
3960{
3961 struct nfs4_client_reclaim *crp = NULL;
3962 int i;
3963
1da177e4
LT		 3964 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3965 while (!list_empty(&reclaim_str_hashtbl[i])) {
3966 crp = list_entry(reclaim_str_hashtbl[i].next,
3967 struct nfs4_client_reclaim, cr_strhash);
3968 list_del(&crp->cr_strhash);
1da177e4
LT		 3969 kfree(crp);
3970 reclaim_str_hashtbl_size--;
3971 }
3972 }
3973 BUG_ON(reclaim_str_hashtbl_size);
3974}
3975
3976/*
3977 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3978static struct nfs4_client_reclaim *
1da177e4
LT		 3979nfs4_find_reclaim_client(clientid_t *clid)
3980{
3981 unsigned int strhashval;
3982 struct nfs4_client *clp;
3983 struct nfs4_client_reclaim *crp = NULL;
3984
3985
3986 /* find clientid in conf_id_hashtbl */
3987 clp = find_confirmed_client(clid);
3988 if (clp == NULL)
3989 return NULL;
3990
a55370a3
N		 3991 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3992 clp->cl_name.len, clp->cl_name.data,
3993 clp->cl_recdir);
1da177e4
LT		 3994
3995 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3996 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3997 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3998 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3999 return crp;
4000 }
4001 }
4002 return NULL;
4003}
4004
4005/*
4006* Called from OPEN. Look for clientid in reclaim list.
4007*/
b37ad28b 4008__be32
1da177e4
LT		 4009nfs4_check_open_reclaim(clientid_t *clid)
4010{
dfc83565 4011 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 4012}
4013
ac4d8ff2 4014/* initialization to perform at module load time: */
1da177e4 4015
e8ff2a84 4016int
ac4d8ff2 4017nfs4_state_init(void)
1da177e4 4018{
e8ff2a84 4019 int i, status;
1da177e4 4020
e8ff2a84
BF		 4021 status = nfsd4_init_slabs();
4022 if (status)
4023 return status;
1da177e4
LT		 4024 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4025 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4026 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4027 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4028 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4029 }
5282fd72
ME		 4030 for (i = 0; i < SESSION_HASH_SIZE; i++)
4031 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4032 for (i = 0; i < FILE_HASH_SIZE; i++) {
4033 INIT_LIST_HEAD(&file_hashtbl[i]);
4034 }
4035 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4036 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4037 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4038 }
4039 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4040 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4041 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4042 }
4043 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4044 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4045 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4046 }
1da177e4 4047 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4048 INIT_LIST_HEAD(&close_lru);
4049 INIT_LIST_HEAD(&client_lru);
4050 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 4051 for (i = 0; i < CLIENT_HASH_SIZE; i++)
4052 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4053 reclaim_str_hashtbl_size = 0;
e8ff2a84 4054 return 0;
ac4d8ff2
N		 4055}
4056
190e4fbf
N		 4057static void
4058nfsd4_load_reboot_recovery_data(void)
4059{
4060 int status;
4061
0964a3d3
N		 4062 nfs4_lock_state();
4063 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4064 status = nfsd4_recdir_load();
0964a3d3 4065 nfs4_unlock_state();
190e4fbf
N		 4066 if (status)
4067 printk("NFSD: Failure reading reboot recovery data\n");
4068}
4069
9a8db97e
ME		 4070unsigned long
4071get_nfs4_grace_period(void)
4072{
4073 return max(user_lease_time, lease_time) * HZ;
4074}
4075
c2f1a551
MS		 4076/*
4077 * Since the lifetime of a delegation isn't limited to that of an open, a
4078 * client may quite reasonably hang on to a delegation as long as it has
4079 * the inode cached. This becomes an obvious problem the first time a
4080 * client's inode cache approaches the size of the server's total memory.
4081 *
4082 * For now we avoid this problem by imposing a hard limit on the number
4083 * of delegations, which varies according to the server's memory size.
4084 */
4085static void
4086set_max_delegations(void)
4087{
4088 /*
4089 * Allow at most 4 delegations per megabyte of RAM. Quick
4090 * estimates suggest that in the worst case (where every delegation
4091 * is for a different inode), a delegation could take about 1.5K,
4092 * giving a worst case usage of about 6% of memory.
4093 */
4094 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4095}
4096
ac4d8ff2
N		 4097/* initialization to perform when the nfsd service is started: */
4098
4099static void
4100__nfs4_state_start(void)
4101{
9a8db97e 4102 unsigned long grace_time;
ac4d8ff2 4103
1da177e4 4104 boot_time = get_seconds();
af558e33 4105 grace_time = get_nfs4_grace_period();
d99a05ad 4106 lease_time = user_lease_time;
af558e33 4107 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4108 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4109 grace_time/HZ);
58da282b 4110 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4111 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4112 set_max_delegations();
1da177e4
LT		 4113}
4114
e8ff2a84 4115void
76a3550e 4116nfs4_state_start(void)
1da177e4 4117{
1da177e4 4118 if (nfs4_init)
e8ff2a84 4119 return;
190e4fbf 4120 nfsd4_load_reboot_recovery_data();
76a3550e 4121 __nfs4_state_start();
1da177e4 4122 nfs4_init = 1;
e8ff2a84 4123 return;
1da177e4
LT		 4124}
4125
1da177e4
LT		 4126time_t
4127nfs4_lease_time(void)
4128{
4129 return lease_time;
4130}
4131
4132static void
4133__nfs4_state_shutdown(void)
4134{
4135 int i;
4136 struct nfs4_client *clp = NULL;
4137 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4138 struct list_head *pos, *next, reaplist;
4139
1da177e4
LT		 4140 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4141 while (!list_empty(&conf_id_hashtbl[i])) {
4142 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4143 expire_client(clp);
4144 }
4145 while (!list_empty(&unconf_str_hashtbl[i])) {
4146 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4147 expire_client(clp);
4148 }
4149 }
4150 INIT_LIST_HEAD(&reaplist);
4151 spin_lock(&recall_lock);
4152 list_for_each_safe(pos, next, &del_recall_lru) {
4153 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4154 list_move(&dp->dl_recall_lru, &reaplist);
4155 }
4156 spin_unlock(&recall_lock);
4157 list_for_each_safe(pos, next, &reaplist) {
4158 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4159 list_del_init(&dp->dl_recall_lru);
4160 unhash_delegation(dp);
4161 }
4162
190e4fbf 4163 nfsd4_shutdown_recdir();
1da177e4 4164 nfs4_init = 0;
1da177e4
LT		 4165}
4166
4167void
4168nfs4_state_shutdown(void)
4169{
5e8d5c29
N		 4170 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4171 destroy_workqueue(laundry_wq);
2c5e7615 4172 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4173 nfs4_lock_state();
4174 nfs4_release_reclaim();
4175 __nfs4_state_shutdown();
1da177e4
LT		 4176 nfs4_unlock_state();
4177}
4178
3dd98a3b
JL		 4179/*
4180 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4181 * accessed when nfsd is starting.
4182 */
0964a3d3
N		 4183static void
4184nfs4_set_recdir(char *recdir)
4185{
0964a3d3 4186 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4187}
4188
4189/*
4190 * Change the NFSv4 recovery directory to recdir.
4191 */
4192int
4193nfs4_reset_recoverydir(char *recdir)
4194{
4195 int status;
a63bb996 4196 struct path path;
0964a3d3 4197
a63bb996 4198 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4199 if (status)
4200 return status;
4201 status = -ENOTDIR;
a63bb996 4202 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4203 nfs4_set_recdir(recdir);
4204 status = 0;
4205 }
a63bb996 4206 path_put(&path);
0964a3d3
N		 4207 return status;
4208}
4209
3dd98a3b
JL		 4210char *
4211nfs4_recoverydir(void)
4212{
4213 return user_recovery_dirname;
4214}
4215
1da177e4
LT		 4216/*
4217 * Called when leasetime is changed.
4218 *
d99a05ad
N		 4219 * The only way the protocol gives us to handle on-the-fly lease changes is to
4220 * simulate a reboot. Instead of doing that, we just wait till the next time
4221 * we start to register any changes in lease time. If the administrator
4222 * really wants to change the lease time *now*, they can go ahead and bring
4223 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4224 *
4225 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4226 * when nfsd is starting
1da177e4
LT		 4227 */
4228void
4229nfs4_reset_lease(time_t leasetime)
4230{
d99a05ad 4231 user_lease_time = leasetime;
1da177e4 4232}
Linux 6.x block layer and io_uring tree(s)