projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: clientid handling
[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;
209 dp->dl_stateid.si_boot = boot_time;
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 }
646}
647
1da177e4
LT		 648static inline void
649free_client(struct nfs4_client *clp)
650{
1b1a9b31 651 shutdown_callback_client(clp);
38eb76a5
AA		 652 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
653 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 654 if (clp->cl_cred.cr_group_info)
655 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 656 kfree(clp->cl_principal);
1da177e4
LT		 657 kfree(clp->cl_name.data);
658 kfree(clp);
659}
660
661void
662put_nfs4_client(struct nfs4_client *clp)
663{
664 if (atomic_dec_and_test(&clp->cl_count))
665 free_client(clp);
666}
667
668static void
669expire_client(struct nfs4_client *clp)
670{
671 struct nfs4_stateowner *sop;
672 struct nfs4_delegation *dp;
1da177e4
LT		 673 struct list_head reaplist;
674
675 dprintk("NFSD: expire_client cl_count %d\n",
676 atomic_read(&clp->cl_count));
677
1da177e4
LT		 678 INIT_LIST_HEAD(&reaplist);
679 spin_lock(&recall_lock);
ea1da636
N		 680 while (!list_empty(&clp->cl_delegations)) {
681 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 682 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
683 dp->dl_flock);
ea1da636 684 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 685 list_move(&dp->dl_recall_lru, &reaplist);
686 }
687 spin_unlock(&recall_lock);
688 while (!list_empty(&reaplist)) {
689 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
690 list_del_init(&dp->dl_recall_lru);
691 unhash_delegation(dp);
692 }
693 list_del(&clp->cl_idhash);
694 list_del(&clp->cl_strhash);
695 list_del(&clp->cl_lru);
ea1da636
N		 696 while (!list_empty(&clp->cl_openowners)) {
697 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 698 release_openowner(sop);
1da177e4 699 }
c4bf7868
ME		 700 while (!list_empty(&clp->cl_sessions)) {
701 struct nfsd4_session *ses;
702 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
703 se_perclnt);
704 release_session(ses);
705 }
1da177e4
LT		 706 put_nfs4_client(clp);
707}
708
35bba9a3
BF		 709static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
710{
1da177e4
LT		 711 struct nfs4_client *clp;
712
35bba9a3
BF		 713 clp = alloc_client(name);
714 if (clp == NULL)
715 return NULL;
a55370a3 716 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT		 717 atomic_set(&clp->cl_count, 1);
718 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT		 719 INIT_LIST_HEAD(&clp->cl_idhash);
720 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 721 INIT_LIST_HEAD(&clp->cl_openowners);
722 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 723 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 724 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 725 return clp;
726}
727
35bba9a3
BF		 728static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
729{
730 memcpy(target->cl_verifier.data, source->data,
731 sizeof(target->cl_verifier.data));
1da177e4
LT		 732}
733
35bba9a3
BF		 734static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
735{
1da177e4
LT		 736 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
737 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
738}
739
35bba9a3
BF		 740static void copy_cred(struct svc_cred *target, struct svc_cred *source)
741{
1da177e4
LT		 742 target->cr_uid = source->cr_uid;
743 target->cr_gid = source->cr_gid;
744 target->cr_group_info = source->cr_group_info;
745 get_group_info(target->cr_group_info);
746}
747
35bba9a3 748static int same_name(const char *n1, const char *n2)
599e0a22 749{
a55370a3 750 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 751}
752
753static int
599e0a22
BF		 754same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
755{
756 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 757}
758
759static int
599e0a22
BF		 760same_clid(clientid_t *cl1, clientid_t *cl2)
761{
762 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 763}
764
765/* XXX what about NGROUP */
766static int
599e0a22
BF		 767same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
768{
769 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 770}
771
5ec7b46c
BF		 772static void gen_clid(struct nfs4_client *clp)
773{
774 static u32 current_clientid = 1;
775
1da177e4
LT		 776 clp->cl_clientid.cl_boot = boot_time;
777 clp->cl_clientid.cl_id = current_clientid++;
778}
779
deda2faa
BF		 780static void gen_confirm(struct nfs4_client *clp)
781{
782 static u32 i;
783 u32 *p;
1da177e4 784
1da177e4 785 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 786 *p++ = get_seconds();
787 *p++ = i++;
1da177e4
LT		 788}
789
35bba9a3
BF		 790static int check_name(struct xdr_netobj name)
791{
1da177e4
LT		 792 if (name.len == 0)
793 return 0;
794 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 795 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 796 return 0;
797 }
798 return 1;
799}
800
fd39ca9a 801static void
1da177e4
LT		 802add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
803{
804 unsigned int idhashval;
805
806 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
807 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
808 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
809 list_add_tail(&clp->cl_lru, &client_lru);
810 clp->cl_time = get_seconds();
811}
812
fd39ca9a 813static void
1da177e4
LT		 814move_to_confirmed(struct nfs4_client *clp)
815{
816 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
817 unsigned int strhashval;
818
819 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
820 list_del_init(&clp->cl_strhash);
f116629d 821 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 822 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 823 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
824 renew_client(clp);
825}
826
827static struct nfs4_client *
828find_confirmed_client(clientid_t *clid)
829{
830 struct nfs4_client *clp;
831 unsigned int idhashval = clientid_hashval(clid->cl_id);
832
833 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 834 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 835 return clp;
836 }
837 return NULL;
838}
839
840static struct nfs4_client *
841find_unconfirmed_client(clientid_t *clid)
842{
843 struct nfs4_client *clp;
844 unsigned int idhashval = clientid_hashval(clid->cl_id);
845
846 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 847 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 848 return clp;
849 }
850 return NULL;
851}
852
a1bcecd2
AA		 853/*
854 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
855 * parameter. Matching is based on the fact the at least one of the
856 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
857 *
858 * FIXME: we need to unify the clientid namespaces for nfsv4.x
859 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
860 * and SET_CLIENTID{,_CONFIRM}
861 */
862static inline int
863match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
864{
865 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
866 return use_exchange_id == has_exchange_flags;
867}
868
28ce6054 869static struct nfs4_client *
a1bcecd2
AA		 870find_confirmed_client_by_str(const char *dname, unsigned int hashval,
871 bool use_exchange_id)
28ce6054
N		 872{
873 struct nfs4_client *clp;
874
875 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 876 if (same_name(clp->cl_recdir, dname) &&
877 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 878 return clp;
879 }
880 return NULL;
881}
882
883static struct nfs4_client *
a1bcecd2
AA		 884find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
885 bool use_exchange_id)
28ce6054
N		 886{
887 struct nfs4_client *clp;
888
889 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 890 if (same_name(clp->cl_recdir, dname) &&
891 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 892 return clp;
893 }
894 return NULL;
895}
896
1da177e4
LT		 897/* a helper function for parse_callback */
898static int
899parse_octet(unsigned int *lenp, char **addrp)
900{
901 unsigned int len = *lenp;
902 char *p = *addrp;
903 int n = -1;
904 char c;
905
906 for (;;) {
907 if (!len)
908 break;
909 len--;
910 c = *p++;
911 if (c == '.')
912 break;
913 if ((c < '0') || (c > '9')) {
914 n = -1;
915 break;
916 }
917 if (n < 0)
918 n = 0;
919 n = (n * 10) + (c - '0');
920 if (n > 255) {
921 n = -1;
922 break;
923 }
924 }
925 *lenp = len;
926 *addrp = p;
927 return n;
928}
929
930/* parse and set the setclientid ipv4 callback address */
fd39ca9a 931static int
1da177e4
LT		 932parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
933{
934 int temp = 0;
935 u32 cbaddr = 0;
936 u16 cbport = 0;
937 u32 addrlen = addr_len;
938 char *addr = addr_val;
939 int i, shift;
940
941 /* ipaddress */
942 shift = 24;
943 for(i = 4; i > 0 ; i--) {
944 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
945 return 0;
946 }
947 cbaddr |= (temp << shift);
948 if (shift > 0)
949 shift -= 8;
950 }
951 *cbaddrp = cbaddr;
952
953 /* port */
954 shift = 8;
955 for(i = 2; i > 0 ; i--) {
956 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
957 return 0;
958 }
959 cbport |= (temp << shift);
960 if (shift > 0)
961 shift -= 8;
962 }
963 *cbportp = cbport;
964 return 1;
965}
966
fd39ca9a 967static void
1da177e4
LT		 968gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
969{
970 struct nfs4_callback *cb = &clp->cl_callback;
971
972 /* Currently, we only support tcp for the callback channel */
973 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
974 goto out_err;
975
976 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
977 &cb->cb_addr, &cb->cb_port)))
978 goto out_err;
979 cb->cb_prog = se->se_callback_prog;
980 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 981 return;
982out_err:
849823c5 983 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 984 "will not receive delegations\n",
985 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
986
1da177e4
LT		 987 return;
988}
989
074fe897
AA		 990void
991nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
992{
993 struct nfsd4_compoundres *resp = rqstp->rq_resp;
994
995 resp->cstate.statp = statp;
996}
997
998/*
999 * Dereference the result pages.
1000 */
1001static void
1002nfsd4_release_respages(struct page **respages, short resused)
1003{
1004 int i;
1005
1006 dprintk("--> %s\n", __func__);
1007 for (i = 0; i < resused; i++) {
1008 if (!respages[i])
1009 continue;
1010 put_page(respages[i]);
1011 respages[i] = NULL;
1012 }
1013}
1014
1015static void
1016nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1017{
1018 int i;
1019
1020 for (i = 0; i < count; i++) {
1021 topages[i] = frompages[i];
1022 if (!topages[i])
1023 continue;
1024 get_page(topages[i]);
1025 }
1026}
1027
1028/*
1029 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1030 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1031 * length of the XDR response is less than se_fmaxresp_cached
1032 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1033 * of the reply (e.g. readdir).
1034 *
1035 * Store the base and length of the rq_req.head[0] page
1036 * of the NFSv4.1 data, just past the rpc header.
1037 */
1038void
1039nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1040{
1041 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1042 struct svc_rqst *rqstp = resp->rqstp;
1043 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1044 struct nfsd4_op *op = &args->ops[resp->opcnt];
1045 struct kvec *resv = &rqstp->rq_res.head[0];
1046
1047 dprintk("--> %s entry %p\n", __func__, entry);
1048
1049 /* Don't cache a failed OP_SEQUENCE. */
1050 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1051 return;
bf864a31 1052
074fe897 1053 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1054 entry->ce_opcnt = resp->opcnt;
1055 entry->ce_status = resp->cstate.status;
1056
1057 /*
1058 * Don't need a page to cache just the sequence operation - the slot
1059 * does this for us!
1060 */
1061
1062 if (nfsd4_not_cached(resp)) {
1063 entry->ce_resused = 0;
1064 entry->ce_rpchdrlen = 0;
1065 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1066 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1067 return;
1068 }
074fe897
AA		 1069 entry->ce_resused = rqstp->rq_resused;
1070 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1071 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1072 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1073 entry->ce_resused);
074fe897
AA		 1074 entry->ce_datav.iov_base = resp->cstate.statp;
1075 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1076 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1077 /* Current request rpc header length*/
1078 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1079 (char *)page_address(rqstp->rq_respages[0]);
1080}
1081
1082/*
1083 * We keep the rpc header, but take the nfs reply from the replycache.
1084 */
1085static int
1086nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1087 struct nfsd4_cache_entry *entry)
1088{
1089 struct svc_rqst *rqstp = resp->rqstp;
1090 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1091 int len;
1092
1093 /* Current request rpc header length*/
1094 len = (char *)resp->cstate.statp -
1095 (char *)page_address(rqstp->rq_respages[0]);
1096 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1097 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1098 entry->ce_datav.iov_len);
1099 return 0;
1100 }
1101 /* copy the cached reply nfsd data past the current rpc header */
1102 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1103 entry->ce_datav.iov_len);
1104 resv->iov_len = len + entry->ce_datav.iov_len;
1105 return 1;
1106}
1107
1108/*
1109 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1110 * cached page. Replace any futher replay pages from the cache.
1111 */
1112__be32
bf864a31
AA		 1113nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1114 struct nfsd4_sequence *seq)
074fe897
AA		 1115{
1116 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1117 __be32 status;
1118
1119 dprintk("--> %s entry %p\n", __func__, entry);
1120
bf864a31
AA		 1121 /*
1122 * If this is just the sequence operation, we did not keep
1123 * a page in the cache entry because we can just use the
1124 * slot info stored in struct nfsd4_sequence that was checked
1125 * against the slot in nfsd4_sequence().
1126 *
1127 * This occurs when seq->cachethis is FALSE, or when the client
1128 * session inactivity timer fires and a solo sequence operation
1129 * is sent (lease renewal).
1130 */
1131 if (seq && nfsd4_not_cached(resp)) {
1132 seq->maxslots = resp->cstate.session->se_fnumslots;
1133 return nfs_ok;
1134 }
074fe897
AA		 1135
1136 if (!nfsd41_copy_replay_data(resp, entry)) {
1137 /*
1138 * Not enough room to use the replay rpc header, send the
1139 * cached header. Release all the allocated result pages.
1140 */
1141 svc_free_res_pages(resp->rqstp);
1142 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1143 entry->ce_resused);
1144 } else {
1145 /* Release all but the first allocated result page */
1146
1147 resp->rqstp->rq_resused--;
1148 svc_free_res_pages(resp->rqstp);
1149
1150 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1151 &entry->ce_respages[1],
1152 entry->ce_resused - 1);
1153 }
1154
1155 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1156 resp->opcnt = entry->ce_opcnt;
1157 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1158 status = entry->ce_status;
1159
1160 return status;
1161}
1162
0733d213
AA		 1163/*
1164 * Set the exchange_id flags returned by the server.
1165 */
1166static void
1167nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1168{
1169 /* pNFS is not supported */
1170 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1171
1172 /* Referrals are supported, Migration is not. */
1173 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1174
1175 /* set the wire flags to return to client. */
1176 clid->flags = new->cl_exchange_flags;
1177}
1178
069b6ad4
AA		 1179__be32
1180nfsd4_exchange_id(struct svc_rqst *rqstp,
1181 struct nfsd4_compound_state *cstate,
1182 struct nfsd4_exchange_id *exid)
1183{
0733d213
AA		 1184 struct nfs4_client *unconf, *conf, *new;
1185 int status;
1186 unsigned int strhashval;
1187 char dname[HEXDIR_LEN];
1188 nfs4_verifier verf = exid->verifier;
1189 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1190
1191 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1192 " ip_addr=%u flags %x, spa_how %d\n",
1193 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1194 ip_addr, exid->flags, exid->spa_how);
1195
1196 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1197 return nfserr_inval;
1198
1199 /* Currently only support SP4_NONE */
1200 switch (exid->spa_how) {
1201 case SP4_NONE:
1202 break;
1203 case SP4_SSV:
1204 return nfserr_encr_alg_unsupp;
1205 default:
1206 BUG(); /* checked by xdr code */
1207 case SP4_MACH_CRED:
1208 return nfserr_serverfault; /* no excuse :-/ */
1209 }
1210
1211 status = nfs4_make_rec_clidname(dname, &exid->clname);
1212
1213 if (status)
1214 goto error;
1215
1216 strhashval = clientstr_hashval(dname);
1217
1218 nfs4_lock_state();
1219 status = nfs_ok;
1220
a1bcecd2 1221 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1222 if (conf) {
1223 if (!same_verf(&verf, &conf->cl_verifier)) {
1224 /* 18.35.4 case 8 */
1225 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1226 status = nfserr_not_same;
1227 goto out;
1228 }
1229 /* Client reboot: destroy old state */
1230 expire_client(conf);
1231 goto out_new;
1232 }
1233 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1234 /* 18.35.4 case 9 */
1235 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1236 status = nfserr_perm;
1237 goto out;
1238 }
1239 expire_client(conf);
1240 goto out_new;
1241 }
1242 if (ip_addr != conf->cl_addr &&
1243 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1244 /* Client collision. 18.35.4 case 3 */
1245 status = nfserr_clid_inuse;
1246 goto out;
1247 }
1248 /*
1249 * Set bit when the owner id and verifier map to an already
1250 * confirmed client id (18.35.3).
1251 */
1252 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1253
1254 /*
1255 * Falling into 18.35.4 case 2, possible router replay.
1256 * Leave confirmed record intact and return same result.
1257 */
1258 copy_verf(conf, &verf);
1259 new = conf;
1260 goto out_copy;
1261 } else {
1262 /* 18.35.4 case 7 */
1263 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1264 status = nfserr_noent;
1265 goto out;
1266 }
1267 }
1268
a1bcecd2 1269 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1270 if (unconf) {
1271 /*
1272 * Possible retry or client restart. Per 18.35.4 case 4,
1273 * a new unconfirmed record should be generated regardless
1274 * of whether any properties have changed.
1275 */
1276 expire_client(unconf);
1277 }
1278
1279out_new:
1280 /* Normal case */
1281 new = create_client(exid->clname, dname);
1282 if (new == NULL) {
1283 status = nfserr_resource;
1284 goto out;
1285 }
1286
1287 copy_verf(new, &verf);
1288 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1289 new->cl_addr = ip_addr;
1290 gen_clid(new);
1291 gen_confirm(new);
1292 add_to_unconfirmed(new, strhashval);
1293out_copy:
1294 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1295 exid->clientid.cl_id = new->cl_clientid.cl_id;
1296
38eb76a5
AA		 1297 new->cl_slot.sl_seqid = 0;
1298 exid->seqid = 1;
0733d213
AA		 1299 nfsd4_set_ex_flags(new, exid);
1300
1301 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1302 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1303 status = nfs_ok;
1304
1305out:
1306 nfs4_unlock_state();
1307error:
1308 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1309 return status;
069b6ad4
AA		 1310}
1311
b85d4c01
BH		 1312static int
1313check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1314{
1315 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1316 slot->sl_seqid);
1317
1318 /* The slot is in use, and no response has been sent. */
1319 if (slot->sl_inuse) {
1320 if (seqid == slot->sl_seqid)
1321 return nfserr_jukebox;
1322 else
1323 return nfserr_seq_misordered;
1324 }
1325 /* Normal */
1326 if (likely(seqid == slot->sl_seqid + 1))
1327 return nfs_ok;
1328 /* Replay */
1329 if (seqid == slot->sl_seqid)
1330 return nfserr_replay_cache;
1331 /* Wraparound */
1332 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1333 return nfs_ok;
1334 /* Misordered replay or misordered new request */
1335 return nfserr_seq_misordered;
1336}
1337
069b6ad4
AA		 1338__be32
1339nfsd4_create_session(struct svc_rqst *rqstp,
1340 struct nfsd4_compound_state *cstate,
1341 struct nfsd4_create_session *cr_ses)
1342{
ec6b5d7b 1343 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1344 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1345 struct nfs4_client *conf, *unconf;
38eb76a5 1346 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1347 int status = 0;
1348
1349 nfs4_lock_state();
1350 unconf = find_unconfirmed_client(&cr_ses->clientid);
1351 conf = find_confirmed_client(&cr_ses->clientid);
1352
1353 if (conf) {
38eb76a5
AA		 1354 slot = &conf->cl_slot;
1355 status = check_slot_seqid(cr_ses->seqid, slot);
1356 if (status == nfserr_replay_cache) {
ec6b5d7b 1357 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1358 slot->sl_seqid);
1359 cstate->slot = slot;
1360 cstate->status = status;
1361 /* Return the cached reply status */
bf864a31 1362 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1363 goto out;
1364 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1365 status = nfserr_seq_misordered;
1366 dprintk("Sequence misordered!\n");
1367 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1368 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1369 goto out;
1370 }
38eb76a5 1371 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1372 } else if (unconf) {
1373 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1374 (ip_addr != unconf->cl_addr)) {
1375 status = nfserr_clid_inuse;
1376 goto out;
1377 }
1378
38eb76a5
AA		 1379 slot = &unconf->cl_slot;
1380 status = check_slot_seqid(cr_ses->seqid, slot);
1381 if (status) {
1382 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1383 status = nfserr_seq_misordered;
1384 goto out;
1385 }
1386
38eb76a5 1387 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1388 move_to_confirmed(unconf);
1389
1390 /*
1391 * We do not support RDMA or persistent sessions
1392 */
1393 cr_ses->flags &= ~SESSION4_PERSIST;
1394 cr_ses->flags &= ~SESSION4_RDMA;
1395
1396 conf = unconf;
1397 } else {
1398 status = nfserr_stale_clientid;
1399 goto out;
1400 }
1401
1402 status = alloc_init_session(rqstp, conf, cr_ses);
1403 if (status)
1404 goto out;
1405
ec6b5d7b
AA		 1406 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1407 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1408 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1409
38eb76a5
AA		 1410 slot->sl_inuse = true;
1411 cstate->slot = slot;
bf864a31
AA		 1412 /* Ensure a page is used for the cache */
1413 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1414out:
1415 nfs4_unlock_state();
1416 dprintk("%s returns %d\n", __func__, ntohl(status));
1417 return status;
069b6ad4
AA		 1418}
1419
1420__be32
1421nfsd4_destroy_session(struct svc_rqst *r,
1422 struct nfsd4_compound_state *cstate,
1423 struct nfsd4_destroy_session *sessionid)
1424{
e10e0cfc
BH		 1425 struct nfsd4_session *ses;
1426 u32 status = nfserr_badsession;
1427
1428 /* Notes:
1429 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1430 * - Should we return nfserr_back_chan_busy if waiting for
1431 * callbacks on to-be-destroyed session?
1432 * - Do we need to clear any callback info from previous session?
1433 */
1434
1435 dump_sessionid(__func__, &sessionid->sessionid);
1436 spin_lock(&sessionid_lock);
1437 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1438 if (!ses) {
1439 spin_unlock(&sessionid_lock);
1440 goto out;
1441 }
1442
1443 unhash_session(ses);
1444 spin_unlock(&sessionid_lock);
1445
1446 /* wait for callbacks */
1447 shutdown_callback_client(ses->se_client);
1448 nfsd4_put_session(ses);
1449 status = nfs_ok;
1450out:
1451 dprintk("%s returns %d\n", __func__, ntohl(status));
1452 return status;
069b6ad4
AA		 1453}
1454
1455__be32
b85d4c01 1456nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1457 struct nfsd4_compound_state *cstate,
1458 struct nfsd4_sequence *seq)
1459{
f9bb94c4 1460 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1461 struct nfsd4_session *session;
1462 struct nfsd4_slot *slot;
1463 int status;
1464
f9bb94c4
AA		 1465 if (resp->opcnt != 1)
1466 return nfserr_sequence_pos;
1467
b85d4c01
BH		 1468 spin_lock(&sessionid_lock);
1469 status = nfserr_badsession;
1470 session = find_in_sessionid_hashtbl(&seq->sessionid);
1471 if (!session)
1472 goto out;
1473
1474 status = nfserr_badslot;
1475 if (seq->slotid >= session->se_fnumslots)
1476 goto out;
1477
1478 slot = &session->se_slots[seq->slotid];
1479 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1480
1481 status = check_slot_seqid(seq->seqid, slot);
1482 if (status == nfserr_replay_cache) {
1483 cstate->slot = slot;
1484 cstate->session = session;
da3846a2 1485 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1486 * for nfsd4_svc_encode_compoundres processing */
1487 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1488 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1489 goto replay_cache;
1490 }
1491 if (status)
1492 goto out;
1493
1494 /* Success! bump slot seqid */
1495 slot->sl_inuse = true;
1496 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1497 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1498 /* Always set the cache entry cachethis for solo sequence */
1499 if (nfsd4_is_solo_sequence(resp))
1500 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1501
1502 cstate->slot = slot;
1503 cstate->session = session;
1504
1505replay_cache:
1506 /* Renew the clientid on success and on replay.
1507 * Hold a session reference until done processing the compound:
1508 * nfsd4_put_session called only if the cstate slot is set.
1509 */
1510 renew_client(session->se_client);
1511 nfsd4_get_session(session);
1512out:
1513 spin_unlock(&sessionid_lock);
1514 dprintk("%s: return %d\n", __func__, ntohl(status));
1515 return status;
069b6ad4
AA		 1516}
1517
b37ad28b 1518__be32
b591480b
BF		 1519nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1520 struct nfsd4_setclientid *setclid)
1da177e4 1521{
27459f09 1522 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1523 struct xdr_netobj clname = {
1524 .len = setclid->se_namelen,
1525 .data = setclid->se_name,
1526 };
1527 nfs4_verifier clverifier = setclid->se_verf;
1528 unsigned int strhashval;
28ce6054 1529 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1530 __be32 status;
68e76ad0 1531 char *princ;
a55370a3 1532 char dname[HEXDIR_LEN];
1da177e4 1533
1da177e4 1534 if (!check_name(clname))
73aea4ec 1535 return nfserr_inval;
1da177e4 1536
a55370a3
N		 1537 status = nfs4_make_rec_clidname(dname, &clname);
1538 if (status)
73aea4ec 1539 return status;
a55370a3 1540
1da177e4
LT		 1541 /*
1542 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1543 * We get here on a DRC miss.
1544 */
1545
a55370a3 1546 strhashval = clientstr_hashval(dname);
1da177e4 1547
1da177e4 1548 nfs4_lock_state();
a1bcecd2 1549 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1550 if (conf) {
a186e767 1551 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1552 status = nfserr_clid_inuse;
026722c2
BF		 1553 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1554 dprintk("NFSD: setclientid: string in use by client"
1555 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1556 goto out;
1557 }
1da177e4 1558 }
a186e767
BF		 1559 /*
1560 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1561 * has a description of SETCLIENTID request processing consisting
1562 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1563 */
a1bcecd2 1564 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1565 status = nfserr_resource;
1566 if (!conf) {
a186e767
BF		 1567 /*
1568 * RFC 3530 14.2.33 CASE 4:
1569 * placed first, because it is the normal case
1da177e4
LT		 1570 */
1571 if (unconf)
1572 expire_client(unconf);
a55370a3
N		 1573 new = create_client(clname, dname);
1574 if (new == NULL)
1da177e4 1575 goto out;
1da177e4 1576 gen_clid(new);
599e0a22 1577 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1578 /*
a186e767
BF		 1579 * RFC 3530 14.2.33 CASE 1:
1580 * probable callback update
1da177e4 1581 */
31f4a6c1
N		 1582 if (unconf) {
1583 /* Note this is removing unconfirmed {*x***},
1584 * which is stronger than RFC recommended {vxc**}.
1585 * This has the advantage that there is at most
1586 * one {*x***} in either list at any time.
1587 */
1588 expire_client(unconf);
1da177e4 1589 }
a55370a3
N		 1590 new = create_client(clname, dname);
1591 if (new == NULL)
1da177e4 1592 goto out;
1da177e4 1593 copy_clid(new, conf);
1da177e4
LT		 1594 } else if (!unconf) {
1595 /*
a186e767
BF		 1596 * RFC 3530 14.2.33 CASE 2:
1597 * probable client reboot; state will be removed if
1598 * confirmed.
1da177e4 1599 */
a55370a3
N		 1600 new = create_client(clname, dname);
1601 if (new == NULL)
1da177e4 1602 goto out;
1da177e4 1603 gen_clid(new);
49ba8781 1604 } else {
a186e767
BF		 1605 /*
1606 * RFC 3530 14.2.33 CASE 3:
1607 * probable client reboot; state will be removed if
1608 * confirmed.
1da177e4
LT		 1609 */
1610 expire_client(unconf);
a55370a3
N		 1611 new = create_client(clname, dname);
1612 if (new == NULL)
1da177e4 1613 goto out;
1da177e4 1614 gen_clid(new);
1da177e4 1615 }
c175b83c
BF		 1616 copy_verf(new, &clverifier);
1617 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1618 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1619 princ = svc_gss_principal(rqstp);
1620 if (princ) {
1621 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1622 if (new->cl_principal == NULL) {
1623 free_client(new);
1624 goto out;
1625 }
1626 }
c175b83c
BF		 1627 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1628 gen_confirm(new);
1629 gen_callback(new, setclid);
1630 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1631 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1632 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1633 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1634 status = nfs_ok;
1635out:
1636 nfs4_unlock_state();
1637 return status;
1638}
1639
1640
1641/*
a186e767
BF		 1642 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1643 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1644 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1645 */
b37ad28b 1646__be32
b591480b
BF		 1647nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1648 struct nfsd4_compound_state *cstate,
1649 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1650{
27459f09 1651 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1652 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1653 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1654 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1655 __be32 status;
1da177e4
LT		 1656
1657 if (STALE_CLIENTID(clid))
1658 return nfserr_stale_clientid;
1659 /*
1660 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1661 * We get here on a DRC miss.
1662 */
1663
1664 nfs4_lock_state();
21ab45a4
N		 1665
1666 conf = find_confirmed_client(clid);
1667 unconf = find_unconfirmed_client(clid);
1668
1669 status = nfserr_clid_inuse;
27459f09 1670 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1671 goto out;
27459f09 1672 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1673 goto out;
1674
a186e767
BF		 1675 /*
1676 * section 14.2.34 of RFC 3530 has a description of
1677 * SETCLIENTID_CONFIRM request processing consisting
1678 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1679 */
366e0c1d 1680 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1681 /*
1682 * RFC 3530 14.2.34 CASE 1:
1683 * callback update
1684 */
599e0a22 1685 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1686 status = nfserr_clid_inuse;
1687 else {
1a69c179
N		 1688 /* XXX: We just turn off callbacks until we can handle
1689 * change request correctly. */
cb36d634 1690 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 1691 gen_confirm(conf);
46309029 1692 nfsd4_remove_clid_dir(unconf);
1a69c179 1693 expire_client(unconf);
1da177e4 1694 status = nfs_ok;
1a69c179 1695
1da177e4 1696 }
f3aba4e5 1697 } else if (conf && !unconf) {
a186e767
BF		 1698 /*
1699 * RFC 3530 14.2.34 CASE 2:
1700 * probable retransmitted request; play it safe and
1701 * do nothing.
7c79f737 1702 */
599e0a22 1703 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1704 status = nfserr_clid_inuse;
21ab45a4 1705 else
1da177e4 1706 status = nfs_ok;
7c79f737 1707 } else if (!conf && unconf
599e0a22 1708 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1709 /*
1710 * RFC 3530 14.2.34 CASE 3:
1711 * Normal case; new or rebooted client:
7c79f737 1712 */
599e0a22 1713 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1714 status = nfserr_clid_inuse;
1715 } else {
1a69c179
N		 1716 unsigned int hash =
1717 clientstr_hashval(unconf->cl_recdir);
1718 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1719 hash, false);
1a69c179 1720 if (conf) {
c7b9a459 1721 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1722 expire_client(conf);
1723 }
1da177e4 1724 move_to_confirmed(unconf);
21ab45a4 1725 conf = unconf;
46f8a64b 1726 nfsd4_probe_callback(conf);
1a69c179 1727 status = nfs_ok;
1da177e4 1728 }
599e0a22
BF		 1729 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1730 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1731 &confirm)))) {
a186e767
BF		 1732 /*
1733 * RFC 3530 14.2.34 CASE 4:
1734 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1735 */
1da177e4 1736 status = nfserr_stale_clientid;
7c79f737 1737 } else {
08e8987c
N		 1738 /* check that we have hit one of the cases...*/
1739 status = nfserr_clid_inuse;
1740 }
1da177e4 1741out:
1da177e4
LT		 1742 nfs4_unlock_state();
1743 return status;
1744}
1745
1da177e4
LT		 1746/* OPEN Share state helper functions */
1747static inline struct nfs4_file *
1748alloc_init_file(struct inode *ino)
1749{
1750 struct nfs4_file *fp;
1751 unsigned int hashval = file_hashval(ino);
1752
e60d4398
N		 1753 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1754 if (fp) {
8b671b80 1755 atomic_set(&fp->fi_ref, 1);
1da177e4 1756 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1757 INIT_LIST_HEAD(&fp->fi_stateids);
1758 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1759 spin_lock(&recall_lock);
1da177e4 1760 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1761 spin_unlock(&recall_lock);
1da177e4
LT		 1762 fp->fi_inode = igrab(ino);
1763 fp->fi_id = current_fileid++;
47f9940c 1764 fp->fi_had_conflict = false;
1da177e4
LT		 1765 return fp;
1766 }
1767 return NULL;
1768}
1769
e60d4398 1770static void
e18b890b 1771nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1772{
e60d4398
N		 1773 if (*slab == NULL)
1774 return;
1a1d92c1 1775 kmem_cache_destroy(*slab);
e60d4398 1776 *slab = NULL;
1da177e4
LT		 1777}
1778
e8ff2a84 1779void
1da177e4
LT		 1780nfsd4_free_slabs(void)
1781{
e60d4398
N		 1782 nfsd4_free_slab(&stateowner_slab);
1783 nfsd4_free_slab(&file_slab);
5ac049ac 1784 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1785 nfsd4_free_slab(&deleg_slab);
e60d4398 1786}
1da177e4 1787
e60d4398
N		 1788static int
1789nfsd4_init_slabs(void)
1790{
1791 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1792 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1793 if (stateowner_slab == NULL)
1794 goto out_nomem;
1795 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1796 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1797 if (file_slab == NULL)
1798 goto out_nomem;
5ac049ac 1799 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1800 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1801 if (stateid_slab == NULL)
1802 goto out_nomem;
5b2d21c1 1803 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1804 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1805 if (deleg_slab == NULL)
1806 goto out_nomem;
e60d4398
N		 1807 return 0;
1808out_nomem:
1809 nfsd4_free_slabs();
1810 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1811 return -ENOMEM;
1da177e4
LT		 1812}
1813
1814void
1815nfs4_free_stateowner(struct kref *kref)
1816{
1817 struct nfs4_stateowner *sop =
1818 container_of(kref, struct nfs4_stateowner, so_ref);
1819 kfree(sop->so_owner.data);
1820 kmem_cache_free(stateowner_slab, sop);
1821}
1822
1823static inline struct nfs4_stateowner *
1824alloc_stateowner(struct xdr_netobj *owner)
1825{
1826 struct nfs4_stateowner *sop;
1827
1828 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1829 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1830 memcpy(sop->so_owner.data, owner->data, owner->len);
1831 sop->so_owner.len = owner->len;
1832 kref_init(&sop->so_ref);
1833 return sop;
1834 }
1835 kmem_cache_free(stateowner_slab, sop);
1836 }
1837 return NULL;
1838}
1839
1840static struct nfs4_stateowner *
1841alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1842 struct nfs4_stateowner *sop;
1843 struct nfs4_replay *rp;
1844 unsigned int idhashval;
1845
1846 if (!(sop = alloc_stateowner(&open->op_owner)))
1847 return NULL;
1848 idhashval = ownerid_hashval(current_ownerid);
1849 INIT_LIST_HEAD(&sop->so_idhash);
1850 INIT_LIST_HEAD(&sop->so_strhash);
1851 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1852 INIT_LIST_HEAD(&sop->so_stateids);
1853 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1854 INIT_LIST_HEAD(&sop->so_close_lru);
1855 sop->so_time = 0;
1856 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1857 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1858 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1859 sop->so_is_open_owner = 1;
1860 sop->so_id = current_ownerid++;
1861 sop->so_client = clp;
1862 sop->so_seqid = open->op_seqid;
1863 sop->so_confirmed = 0;
1864 rp = &sop->so_replay;
de1ae286 1865 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1866 rp->rp_buflen = 0;
1867 rp->rp_buf = rp->rp_ibuf;
1868 return sop;
1869}
1870
1da177e4
LT		 1871static inline void
1872init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1873 struct nfs4_stateowner *sop = open->op_stateowner;
1874 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1875
1876 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1877 INIT_LIST_HEAD(&stp->st_perstateowner);
1878 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1879 INIT_LIST_HEAD(&stp->st_perfile);
1880 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1881 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1882 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1883 stp->st_stateowner = sop;
13cd2184 1884 get_nfs4_file(fp);
1da177e4
LT		 1885 stp->st_file = fp;
1886 stp->st_stateid.si_boot = boot_time;
1887 stp->st_stateid.si_stateownerid = sop->so_id;
1888 stp->st_stateid.si_fileid = fp->fi_id;
1889 stp->st_stateid.si_generation = 0;
1890 stp->st_access_bmap = 0;
1891 stp->st_deny_bmap = 0;
1892 __set_bit(open->op_share_access, &stp->st_access_bmap);
1893 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1894 stp->st_openstp = NULL;
1da177e4
LT		 1895}
1896
fd39ca9a 1897static void
1da177e4
LT		 1898move_to_close_lru(struct nfs4_stateowner *sop)
1899{
1900 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1901
358dd55a 1902 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1903 sop->so_time = get_seconds();
1904}
1905
1da177e4 1906static int
599e0a22
BF		 1907same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1908 clientid_t *clid)
1909{
1910 return (sop->so_owner.len == owner->len) &&
1911 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1912 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1913}
1914
1915static struct nfs4_stateowner *
1916find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1917{
1918 struct nfs4_stateowner *so = NULL;
1919
1920 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1921 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1922 return so;
1923 }
1924 return NULL;
1925}
1926
1927/* search file_hashtbl[] for file */
1928static struct nfs4_file *
1929find_file(struct inode *ino)
1930{
1931 unsigned int hashval = file_hashval(ino);
1932 struct nfs4_file *fp;
1933
8b671b80 1934 spin_lock(&recall_lock);
1da177e4 1935 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1936 if (fp->fi_inode == ino) {
1937 get_nfs4_file(fp);
8b671b80 1938 spin_unlock(&recall_lock);
1da177e4 1939 return fp;
13cd2184 1940 }
1da177e4 1941 }
8b671b80 1942 spin_unlock(&recall_lock);
1da177e4
LT		 1943 return NULL;
1944}
1945
8838dc43 1946static inline int access_valid(u32 x)
ba5a6a19 1947{
8838dc43
BF		 1948 if (x < NFS4_SHARE_ACCESS_READ)
1949 return 0;
1950 if (x > NFS4_SHARE_ACCESS_BOTH)
1951 return 0;
1952 return 1;
ba5a6a19
BF		 1953}
1954
8838dc43 1955static inline int deny_valid(u32 x)
ba5a6a19 1956{
8838dc43
BF		 1957 /* Note: unlike access bits, deny bits may be zero. */
1958 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1959}
1da177e4 1960
4f83aa30
BF		 1961/*
1962 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1963 * st_{access,deny}_bmap field of the stateid, in order to track not
1964 * only what share bits are currently in force, but also what
1965 * combinations of share bits previous opens have used. This allows us
1966 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1967 * return an error if the client attempt to downgrade to a combination
1968 * of share bits not explicable by closing some of its previous opens.
1969 *
1970 * XXX: This enforcement is actually incomplete, since we don't keep
1971 * track of access/deny bit combinations; so, e.g., we allow:
1972 *
1973 * OPEN allow read, deny write
1974 * OPEN allow both, deny none
1975 * DOWNGRADE allow read, deny none
1976 *
1977 * which we should reject.
1978 */
fd39ca9a 1979static void
1da177e4
LT		 1980set_access(unsigned int *access, unsigned long bmap) {
1981 int i;
1982
1983 *access = 0;
1984 for (i = 1; i < 4; i++) {
1985 if (test_bit(i, &bmap))
1986 *access |= i;
1987 }
1988}
1989
fd39ca9a 1990static void
1da177e4
LT		 1991set_deny(unsigned int *deny, unsigned long bmap) {
1992 int i;
1993
1994 *deny = 0;
1995 for (i = 0; i < 4; i++) {
1996 if (test_bit(i, &bmap))
1997 *deny |= i ;
1998 }
1999}
2000
2001static int
2002test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2003 unsigned int access, deny;
2004
2005 set_access(&access, stp->st_access_bmap);
2006 set_deny(&deny, stp->st_deny_bmap);
2007 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2008 return 0;
2009 return 1;
2010}
2011
2012/*
2013 * Called to check deny when READ with all zero stateid or
2014 * WRITE with all zero or all one stateid
2015 */
b37ad28b 2016static __be32
1da177e4
LT		 2017nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2018{
2019 struct inode *ino = current_fh->fh_dentry->d_inode;
2020 struct nfs4_file *fp;
2021 struct nfs4_stateid *stp;
b37ad28b 2022 __be32 ret;
1da177e4
LT		 2023
2024 dprintk("NFSD: nfs4_share_conflict\n");
2025
2026 fp = find_file(ino);
13cd2184
N		 2027 if (!fp)
2028 return nfs_ok;
b700949b 2029 ret = nfserr_locked;
1da177e4 2030 /* Search for conflicting share reservations */
13cd2184
N		 2031 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2032 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2033 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2034 goto out;
1da177e4 2035 }
13cd2184
N		 2036 ret = nfs_ok;
2037out:
2038 put_nfs4_file(fp);
2039 return ret;
1da177e4
LT		 2040}
2041
2042static inline void
2043nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2044{
2045 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2046 drop_file_write_access(filp);
1da177e4
LT		 2047 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2048 }
2049}
2050
2051/*
2052 * Recall a delegation
2053 */
2054static int
2055do_recall(void *__dp)
2056{
2057 struct nfs4_delegation *dp = __dp;
2058
47f9940c 2059 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2060 nfsd4_cb_recall(dp);
2061 return 0;
2062}
2063
2064/*
2065 * Spawn a thread to perform a recall on the delegation represented
2066 * by the lease (file_lock)
2067 *
2068 * Called from break_lease() with lock_kernel() held.
2069 * Note: we assume break_lease will only call this *once* for any given
2070 * lease.
2071 */
2072static
2073void nfsd_break_deleg_cb(struct file_lock *fl)
2074{
2075 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2076 struct task_struct *t;
2077
2078 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2079 if (!dp)
2080 return;
2081
2082 /* We're assuming the state code never drops its reference
2083 * without first removing the lease. Since we're in this lease
2084 * callback (and since the lease code is serialized by the kernel
2085 * lock) we know the server hasn't removed the lease yet, we know
2086 * it's safe to take a reference: */
2087 atomic_inc(&dp->dl_count);
cfdcad4d 2088 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2089
2090 spin_lock(&recall_lock);
2091 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2092 spin_unlock(&recall_lock);
2093
2094 /* only place dl_time is set. protected by lock_kernel*/
2095 dp->dl_time = get_seconds();
2096
0272e1fd
BF		 2097 /*
2098 * We don't want the locks code to timeout the lease for us;
2099 * we'll remove it ourself if the delegation isn't returned
2100 * in time.
2101 */
2102 fl->fl_break_time = 0;
1da177e4
LT		 2103
2104 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2105 if (IS_ERR(t)) {
2106 struct nfs4_client *clp = dp->dl_client;
2107
2108 printk(KERN_INFO "NFSD: Callback thread failed for "
2109 "for client (clientid %08x/%08x)\n",
2110 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2111 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2112 nfs4_put_delegation(dp);
2113 }
2114}
2115
2116/*
2117 * The file_lock is being reapd.
2118 *
2119 * Called by locks_free_lock() with lock_kernel() held.
2120 */
2121static
2122void nfsd_release_deleg_cb(struct file_lock *fl)
2123{
2124 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2125
2126 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2127
2128 if (!(fl->fl_flags & FL_LEASE) || !dp)
2129 return;
2130 dp->dl_flock = NULL;
2131}
2132
2133/*
2134 * Set the delegation file_lock back pointer.
2135 *
a9933cea 2136 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2137 */
2138static
2139void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2140{
2141 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2142
2143 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2144 if (!dp)
2145 return;
2146 dp->dl_flock = new;
2147}
2148
2149/*
a9933cea 2150 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2151 */
2152static
2153int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2154{
2155 struct nfs4_delegation *onlistd =
2156 (struct nfs4_delegation *)onlist->fl_owner;
2157 struct nfs4_delegation *tryd =
2158 (struct nfs4_delegation *)try->fl_owner;
2159
2160 if (onlist->fl_lmops != try->fl_lmops)
2161 return 0;
2162
2163 return onlistd->dl_client == tryd->dl_client;
2164}
2165
2166
2167static
2168int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2169{
2170 if (arg & F_UNLCK)
2171 return lease_modify(onlist, arg);
2172 else
2173 return -EAGAIN;
2174}
2175
fd39ca9a 2176static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2177 .fl_break = nfsd_break_deleg_cb,
2178 .fl_release_private = nfsd_release_deleg_cb,
2179 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2180 .fl_mylease = nfsd_same_client_deleg_cb,
2181 .fl_change = nfsd_change_deleg_cb,
2182};
2183
2184
b37ad28b 2185__be32
6668958f
AA		 2186nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2187 struct nfsd4_open *open)
1da177e4 2188{
1da177e4
LT		 2189 clientid_t *clientid = &open->op_clientid;
2190 struct nfs4_client *clp = NULL;
2191 unsigned int strhashval;
2192 struct nfs4_stateowner *sop = NULL;
2193
1da177e4 2194 if (!check_name(open->op_owner))
0f442aa2 2195 return nfserr_inval;
1da177e4
LT		 2196
2197 if (STALE_CLIENTID(&open->op_clientid))
2198 return nfserr_stale_clientid;
2199
2200 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2201 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2202 open->op_stateowner = sop;
2203 if (!sop) {
2204 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2205 clp = find_confirmed_client(clientid);
2206 if (clp == NULL)
0f442aa2
BF		 2207 return nfserr_expired;
2208 goto renew;
1da177e4 2209 }
6668958f
AA		 2210 /* When sessions are used, skip open sequenceid processing */
2211 if (nfsd4_has_session(cstate))
2212 goto renew;
0f442aa2
BF		 2213 if (!sop->so_confirmed) {
2214 /* Replace unconfirmed owners without checking for replay. */
2215 clp = sop->so_client;
f044ff83 2216 release_openowner(sop);
0f442aa2
BF		 2217 open->op_stateowner = NULL;
2218 goto renew;
2219 }
2220 if (open->op_seqid == sop->so_seqid - 1) {
2221 if (sop->so_replay.rp_buflen)
a90b061c 2222 return nfserr_replay_me;
0f442aa2
BF		 2223 /* The original OPEN failed so spectacularly
2224 * that we don't even have replay data saved!
2225 * Therefore, we have no choice but to continue
2226 * processing this OPEN; presumably, we'll
2227 * fail again for the same reason.
2228 */
2229 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2230 goto renew;
2231 }
2232 if (open->op_seqid != sop->so_seqid)
2233 return nfserr_bad_seqid;
1da177e4 2234renew:
0f442aa2
BF		 2235 if (open->op_stateowner == NULL) {
2236 sop = alloc_init_open_stateowner(strhashval, clp, open);
2237 if (sop == NULL)
2238 return nfserr_resource;
2239 open->op_stateowner = sop;
2240 }
2241 list_del_init(&sop->so_close_lru);
1da177e4 2242 renew_client(sop->so_client);
0f442aa2 2243 return nfs_ok;
1da177e4
LT		 2244}
2245
b37ad28b 2246static inline __be32
4a6e43e6
N		 2247nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2248{
2249 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2250 return nfserr_openmode;
2251 else
2252 return nfs_ok;
2253}
2254
52f4fb43
N		 2255static struct nfs4_delegation *
2256find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2257{
2258 struct nfs4_delegation *dp;
2259
ea1da636 2260 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2261 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2262 return dp;
2263 }
2264 return NULL;
2265}
2266
b37ad28b 2267static __be32
567d9829
N		 2268nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2269 struct nfs4_delegation **dp)
2270{
2271 int flags;
b37ad28b 2272 __be32 status = nfserr_bad_stateid;
567d9829
N		 2273
2274 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2275 if (*dp == NULL)
c44c5eeb 2276 goto out;
567d9829
N		 2277 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2278 RD_STATE : WR_STATE;
2279 status = nfs4_check_delegmode(*dp, flags);
2280 if (status)
2281 *dp = NULL;
c44c5eeb
N		 2282out:
2283 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2284 return nfs_ok;
2285 if (status)
2286 return status;
2287 open->op_stateowner->so_confirmed = 1;
2288 return nfs_ok;
567d9829
N		 2289}
2290
b37ad28b 2291static __be32
1da177e4
LT		 2292nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2293{
2294 struct nfs4_stateid *local;
b37ad28b 2295 __be32 status = nfserr_share_denied;
1da177e4
LT		 2296 struct nfs4_stateowner *sop = open->op_stateowner;
2297
8beefa24 2298 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2299 /* ignore lock owners */
2300 if (local->st_stateowner->so_is_open_owner == 0)
2301 continue;
2302 /* remember if we have seen this open owner */
2303 if (local->st_stateowner == sop)
2304 *stpp = local;
2305 /* check for conflicting share reservations */
2306 if (!test_share(local, open))
2307 goto out;
2308 }
2309 status = 0;
2310out:
2311 return status;
2312}
2313
5ac049ac
N		 2314static inline struct nfs4_stateid *
2315nfs4_alloc_stateid(void)
2316{
2317 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2318}
2319
b37ad28b 2320static __be32
1da177e4 2321nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2322 struct nfs4_delegation *dp,
1da177e4
LT		 2323 struct svc_fh *cur_fh, int flags)
2324{
2325 struct nfs4_stateid *stp;
1da177e4 2326
5ac049ac 2327 stp = nfs4_alloc_stateid();
1da177e4
LT		 2328 if (stp == NULL)
2329 return nfserr_resource;
2330
567d9829
N		 2331 if (dp) {
2332 get_file(dp->dl_vfs_file);
2333 stp->st_vfs_file = dp->dl_vfs_file;
2334 } else {
b37ad28b 2335 __be32 status;
567d9829
N		 2336 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2337 &stp->st_vfs_file);
2338 if (status) {
2339 if (status == nfserr_dropit)
2340 status = nfserr_jukebox;
5ac049ac 2341 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2342 return status;
2343 }
1da177e4 2344 }
1da177e4
LT		 2345 *stpp = stp;
2346 return 0;
2347}
2348
b37ad28b 2349static inline __be32
1da177e4
LT		 2350nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2351 struct nfsd4_open *open)
2352{
2353 struct iattr iattr = {
2354 .ia_valid = ATTR_SIZE,
2355 .ia_size = 0,
2356 };
2357 if (!open->op_truncate)
2358 return 0;
2359 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2360 return nfserr_inval;
1da177e4
LT		 2361 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2362}
2363
b37ad28b 2364static __be32
1da177e4
LT		 2365nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2366{
2367 struct file *filp = stp->st_vfs_file;
7eaa36e2 2368 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2369 unsigned int share_access, new_writer;
b37ad28b 2370 __be32 status;
1da177e4
LT		 2371
2372 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2373 new_writer = (~share_access) & open->op_share_access
2374 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2375
6c26d08f 2376 if (new_writer) {
b37ad28b
AV		 2377 int err = get_write_access(inode);
2378 if (err)
2379 return nfserrno(err);
e518f056
BH		 2380 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2381 if (err)
2382 return nfserrno(err);
2383 file_take_write(filp);
6c26d08f 2384 }
1da177e4
LT		 2385 status = nfsd4_truncate(rqstp, cur_fh, open);
2386 if (status) {
6c26d08f
BF		 2387 if (new_writer)
2388 put_write_access(inode);
1da177e4
LT		 2389 return status;
2390 }
2391 /* remember the open */
6c26d08f 2392 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2393 __set_bit(open->op_share_access, &stp->st_access_bmap);
2394 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2395
2396 return nfs_ok;
2397}
2398
2399
1da177e4 2400static void
37515177 2401nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2402{
37515177
N		 2403 open->op_stateowner->so_confirmed = 1;
2404 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2405}
2406
2407/*
2408 * Attempt to hand out a delegation.
2409 */
2410static void
2411nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2412{
2413 struct nfs4_delegation *dp;
2414 struct nfs4_stateowner *sop = stp->st_stateowner;
2415 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2416 struct file_lock fl, *flp = &fl;
2417 int status, flag = 0;
2418
2419 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2420 open->op_recall = 0;
2421 switch (open->op_claim_type) {
2422 case NFS4_OPEN_CLAIM_PREVIOUS:
2423 if (!atomic_read(&cb->cb_set))
2424 open->op_recall = 1;
2425 flag = open->op_delegate_type;
2426 if (flag == NFS4_OPEN_DELEGATE_NONE)
2427 goto out;
2428 break;
2429 case NFS4_OPEN_CLAIM_NULL:
2430 /* Let's not give out any delegations till everyone's
2431 * had the chance to reclaim theirs.... */
af558e33 2432 if (locks_in_grace())
7b190fec
N		 2433 goto out;
2434 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2435 goto out;
2436 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2437 flag = NFS4_OPEN_DELEGATE_WRITE;
2438 else
2439 flag = NFS4_OPEN_DELEGATE_READ;
2440 break;
2441 default:
2442 goto out;
2443 }
1da177e4
LT		 2444
2445 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2446 if (dp == NULL) {
2447 flag = NFS4_OPEN_DELEGATE_NONE;
2448 goto out;
2449 }
2450 locks_init_lock(&fl);
2451 fl.fl_lmops = &nfsd_lease_mng_ops;
2452 fl.fl_flags = FL_LEASE;
9167f501 2453 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2454 fl.fl_end = OFFSET_MAX;
2455 fl.fl_owner = (fl_owner_t)dp;
2456 fl.fl_file = stp->st_vfs_file;
2457 fl.fl_pid = current->tgid;
2458
a9933cea 2459 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2460 * the lock_manager callbacks fl_mylease and fl_change are used
2461 */
9167f501 2462 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2463 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2464 unhash_delegation(dp);
1da177e4
LT		 2465 flag = NFS4_OPEN_DELEGATE_NONE;
2466 goto out;
2467 }
2468
2469 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2470
2471 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2472 dp->dl_stateid.si_boot,
2473 dp->dl_stateid.si_stateownerid,
2474 dp->dl_stateid.si_fileid,
2475 dp->dl_stateid.si_generation);
2476out:
7b190fec
N		 2477 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2478 && flag == NFS4_OPEN_DELEGATE_NONE
2479 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2480 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2481 open->op_delegate_type = flag;
2482}
2483
2484/*
2485 * called with nfs4_lock_state() held.
2486 */
b37ad28b 2487__be32
1da177e4
LT		 2488nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2489{
6668958f 2490 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2491 struct nfs4_file *fp = NULL;
2492 struct inode *ino = current_fh->fh_dentry->d_inode;
2493 struct nfs4_stateid *stp = NULL;
567d9829 2494 struct nfs4_delegation *dp = NULL;
b37ad28b 2495 __be32 status;
1da177e4
LT		 2496
2497 status = nfserr_inval;
ba5a6a19
BF		 2498 if (!access_valid(open->op_share_access)
2499 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2500 goto out;
2501 /*
2502 * Lookup file; if found, lookup stateid and check open request,
2503 * and check for delegations in the process of being recalled.
2504 * If not found, create the nfs4_file struct
2505 */
2506 fp = find_file(ino);
2507 if (fp) {
2508 if ((status = nfs4_check_open(fp, open, &stp)))
2509 goto out;
c44c5eeb
N		 2510 status = nfs4_check_deleg(fp, open, &dp);
2511 if (status)
2512 goto out;
1da177e4 2513 } else {
c44c5eeb
N		 2514 status = nfserr_bad_stateid;
2515 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2516 goto out;
1da177e4
LT		 2517 status = nfserr_resource;
2518 fp = alloc_init_file(ino);
2519 if (fp == NULL)
2520 goto out;
2521 }
2522
2523 /*
2524 * OPEN the file, or upgrade an existing OPEN.
2525 * If truncate fails, the OPEN fails.
2526 */
2527 if (stp) {
2528 /* Stateid was found, this is an OPEN upgrade */
2529 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2530 if (status)
2531 goto out;
444c2c07 2532 update_stateid(&stp->st_stateid);
1da177e4
LT		 2533 } else {
2534 /* Stateid was not found, this is a new OPEN */
2535 int flags = 0;
9ecb6a08 2536 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2537 flags |= NFSD_MAY_READ;
1da177e4 2538 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2539 flags |= NFSD_MAY_WRITE;
567d9829
N		 2540 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2541 if (status)
1da177e4
LT		 2542 goto out;
2543 init_stateid(stp, fp, open);
2544 status = nfsd4_truncate(rqstp, current_fh, open);
2545 if (status) {
2283963f 2546 release_open_stateid(stp);
1da177e4
LT		 2547 goto out;
2548 }
6668958f
AA		 2549 if (nfsd4_has_session(&resp->cstate))
2550 update_stateid(&stp->st_stateid);
1da177e4
LT		 2551 }
2552 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2553
6668958f
AA		 2554 if (nfsd4_has_session(&resp->cstate))
2555 open->op_stateowner->so_confirmed = 1;
2556
1da177e4
LT		 2557 /*
2558 * Attempt to hand out a delegation. No error return, because the
2559 * OPEN succeeds even if we fail.
2560 */
2561 nfs4_open_delegation(current_fh, open, stp);
2562
2563 status = nfs_ok;
2564
2565 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2566 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2567 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2568out:
13cd2184
N		 2569 if (fp)
2570 put_nfs4_file(fp);
37515177
N		 2571 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2572 nfs4_set_claim_prev(open);
1da177e4
LT		 2573 /*
2574 * To finish the open response, we just need to set the rflags.
2575 */
2576 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2577 if (!open->op_stateowner->so_confirmed &&
2578 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2579 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2580
2581 return status;
2582}
2583
b37ad28b 2584__be32
b591480b
BF		 2585nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2586 clientid_t *clid)
1da177e4
LT		 2587{
2588 struct nfs4_client *clp;
b37ad28b 2589 __be32 status;
1da177e4
LT		 2590
2591 nfs4_lock_state();
2592 dprintk("process_renew(%08x/%08x): starting\n",
2593 clid->cl_boot, clid->cl_id);
2594 status = nfserr_stale_clientid;
2595 if (STALE_CLIENTID(clid))
2596 goto out;
2597 clp = find_confirmed_client(clid);
2598 status = nfserr_expired;
2599 if (clp == NULL) {
2600 /* We assume the client took too long to RENEW. */
2601 dprintk("nfsd4_renew: clientid not found!\n");
2602 goto out;
2603 }
2604 renew_client(clp);
2605 status = nfserr_cb_path_down;
ea1da636 2606 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2607 && !atomic_read(&clp->cl_callback.cb_set))
2608 goto out;
2609 status = nfs_ok;
2610out:
2611 nfs4_unlock_state();
2612 return status;
2613}
2614
af558e33
BF		 2615struct lock_manager nfsd4_manager = {
2616};
2617
a76b4319 2618static void
af558e33 2619nfsd4_end_grace(void)
a76b4319
N		 2620{
2621 dprintk("NFSD: end of grace period\n");
c7b9a459 2622 nfsd4_recdir_purge_old();
af558e33 2623 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2624}
2625
fd39ca9a 2626static time_t
1da177e4
LT		 2627nfs4_laundromat(void)
2628{
2629 struct nfs4_client *clp;
2630 struct nfs4_stateowner *sop;
2631 struct nfs4_delegation *dp;
2632 struct list_head *pos, *next, reaplist;
2633 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2634 time_t t, clientid_val = NFSD_LEASE_TIME;
2635 time_t u, test_val = NFSD_LEASE_TIME;
2636
2637 nfs4_lock_state();
2638
2639 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2640 if (locks_in_grace())
2641 nfsd4_end_grace();
1da177e4
LT		 2642 list_for_each_safe(pos, next, &client_lru) {
2643 clp = list_entry(pos, struct nfs4_client, cl_lru);
2644 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2645 t = clp->cl_time - cutoff;
2646 if (clientid_val > t)
2647 clientid_val = t;
2648 break;
2649 }
2650 dprintk("NFSD: purging unused client (clientid %08x)\n",
2651 clp->cl_clientid.cl_id);
c7b9a459 2652 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2653 expire_client(clp);
2654 }
2655 INIT_LIST_HEAD(&reaplist);
2656 spin_lock(&recall_lock);
2657 list_for_each_safe(pos, next, &del_recall_lru) {
2658 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2659 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2660 u = dp->dl_time - cutoff;
2661 if (test_val > u)
2662 test_val = u;
2663 break;
2664 }
2665 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2666 dp, dp->dl_flock);
2667 list_move(&dp->dl_recall_lru, &reaplist);
2668 }
2669 spin_unlock(&recall_lock);
2670 list_for_each_safe(pos, next, &reaplist) {
2671 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2672 list_del_init(&dp->dl_recall_lru);
2673 unhash_delegation(dp);
2674 }
2675 test_val = NFSD_LEASE_TIME;
2676 list_for_each_safe(pos, next, &close_lru) {
2677 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2678 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2679 u = sop->so_time - cutoff;
2680 if (test_val > u)
2681 test_val = u;
2682 break;
2683 }
2684 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2685 sop->so_id);
f044ff83 2686 release_openowner(sop);
1da177e4
LT		 2687 }
2688 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2689 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2690 nfs4_unlock_state();
2691 return clientid_val;
2692}
2693
a254b246
HH		 2694static struct workqueue_struct *laundry_wq;
2695static void laundromat_main(struct work_struct *);
2696static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2697
2698static void
c4028958 2699laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2700{
2701 time_t t;
2702
2703 t = nfs4_laundromat();
2704 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2705 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2706}
2707
fd39ca9a 2708static struct nfs4_stateowner *
f8816512
N		 2709search_close_lru(u32 st_id, int flags)
2710{
1da177e4
LT		 2711 struct nfs4_stateowner *local = NULL;
2712
1da177e4
LT		 2713 if (flags & CLOSE_STATE) {
2714 list_for_each_entry(local, &close_lru, so_close_lru) {
2715 if (local->so_id == st_id)
2716 return local;
2717 }
2718 }
2719 return NULL;
2720}
2721
2722static inline int
2723nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2724{
7eaa36e2 2725 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2726}
2727
2728static int
2729STALE_STATEID(stateid_t *stateid)
2730{
2731 if (stateid->si_boot == boot_time)
2732 return 0;
849823c5 2733 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT		 2734 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2735 stateid->si_generation);
2736 return 1;
2737}
2738
2739static inline int
2740access_permit_read(unsigned long access_bmap)
2741{
2742 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2743 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2744 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2745}
2746
2747static inline int
2748access_permit_write(unsigned long access_bmap)
2749{
2750 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2751 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2752}
2753
2754static
b37ad28b 2755__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2756{
b37ad28b 2757 __be32 status = nfserr_openmode;
1da177e4
LT		 2758
2759 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2760 goto out;
2761 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2762 goto out;
2763 status = nfs_ok;
2764out:
2765 return status;
2766}
2767
b37ad28b 2768static inline __be32
1da177e4
LT		 2769check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2770{
203a8c8e 2771 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2772 return nfs_ok;
af558e33 2773 else if (locks_in_grace()) {
1da177e4
LT		 2774 /* Answer in remaining cases depends on existance of
2775 * conflicting state; so we must wait out the grace period. */
2776 return nfserr_grace;
2777 } else if (flags & WR_STATE)
2778 return nfs4_share_conflict(current_fh,
2779 NFS4_SHARE_DENY_WRITE);
2780 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2781 return nfs4_share_conflict(current_fh,
2782 NFS4_SHARE_DENY_READ);
2783}
2784
2785/*
2786 * Allow READ/WRITE during grace period on recovered state only for files
2787 * that are not able to provide mandatory locking.
2788 */
2789static inline int
18f82731 2790grace_disallows_io(struct inode *inode)
1da177e4 2791{
203a8c8e 2792 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2793}
2794
6668958f 2795static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2796{
6668958f
AA		 2797 /*
2798 * When sessions are used the stateid generation number is ignored
2799 * when it is zero.
2800 */
2801 if ((flags & HAS_SESSION) && in->si_generation == 0)
2802 goto out;
2803
0836f587
BF		 2804 /* If the client sends us a stateid from the future, it's buggy: */
2805 if (in->si_generation > ref->si_generation)
2806 return nfserr_bad_stateid;
2807 /*
2808 * The following, however, can happen. For example, if the
2809 * client sends an open and some IO at the same time, the open
2810 * may bump si_generation while the IO is still in flight.
2811 * Thanks to hard links and renames, the client never knows what
2812 * file an open will affect. So it could avoid that situation
2813 * only by serializing all opens and IO from the same open
2814 * owner. To recover from the old_stateid error, the client
2815 * will just have to retry the IO:
2816 */
2817 if (in->si_generation < ref->si_generation)
2818 return nfserr_old_stateid;
6668958f 2819out:
0836f587
BF		 2820 return nfs_ok;
2821}
2822
3e633079
BF		 2823static int is_delegation_stateid(stateid_t *stateid)
2824{
2825 return stateid->si_fileid == 0;
2826}
2827
1da177e4
LT		 2828/*
2829* Checks for stateid operations
2830*/
b37ad28b 2831__be32
dd453dfd
BH		 2832nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2833 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2834{
2835 struct nfs4_stateid *stp = NULL;
2836 struct nfs4_delegation *dp = NULL;
dd453dfd 2837 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2838 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2839 __be32 status;
1da177e4 2840
1da177e4
LT		 2841 if (filpp)
2842 *filpp = NULL;
2843
18f82731 2844 if (grace_disallows_io(ino))
1da177e4
LT		 2845 return nfserr_grace;
2846
6668958f
AA		 2847 if (nfsd4_has_session(cstate))
2848 flags |= HAS_SESSION;
2849
1da177e4
LT		 2850 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2851 return check_special_stateids(current_fh, stateid, flags);
2852
1da177e4
LT		 2853 status = nfserr_stale_stateid;
2854 if (STALE_STATEID(stateid))
2855 goto out;
2856
1da177e4 2857 status = nfserr_bad_stateid;
3e633079 2858 if (is_delegation_stateid(stateid)) {
a4455be0 2859 dp = find_delegation_stateid(ino, stateid);
819a8f53 2860 if (!dp)
1da177e4 2861 goto out;
6668958f
AA		 2862 status = check_stateid_generation(stateid, &dp->dl_stateid,
2863 flags);
0c2a498f
BF		 2864 if (status)
2865 goto out;
dc9bf700
BF		 2866 status = nfs4_check_delegmode(dp, flags);
2867 if (status)
2868 goto out;
2869 renew_client(dp->dl_client);
dc9bf700
BF		 2870 if (filpp)
2871 *filpp = dp->dl_vfs_file;
1da177e4 2872 } else { /* open or lock stateid */
a4455be0 2873 stp = find_stateid(stateid, flags);
819a8f53 2874 if (!stp)
1da177e4 2875 goto out;
6150ef0d 2876 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2877 goto out;
2878 if (!stp->st_stateowner->so_confirmed)
2879 goto out;
6668958f
AA		 2880 status = check_stateid_generation(stateid, &stp->st_stateid,
2881 flags);
0c2a498f
BF		 2882 if (status)
2883 goto out;
a4455be0
BF		 2884 status = nfs4_check_openmode(stp, flags);
2885 if (status)
1da177e4
LT		 2886 goto out;
2887 renew_client(stp->st_stateowner->so_client);
2888 if (filpp)
2889 *filpp = stp->st_vfs_file;
1da177e4
LT		 2890 }
2891 status = nfs_ok;
2892out:
2893 return status;
2894}
2895
4c4cd222
N		 2896static inline int
2897setlkflg (int type)
2898{
2899 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2900 RD_STATE : WR_STATE;
2901}
1da177e4
LT		 2902
2903/*
2904 * Checks for sequence id mutating operations.
2905 */
b37ad28b 2906static __be32
dd453dfd
BH		 2907nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2908 stateid_t *stateid, int flags,
2909 struct nfs4_stateowner **sopp,
2910 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2911{
1da177e4
LT		 2912 struct nfs4_stateid *stp;
2913 struct nfs4_stateowner *sop;
dd453dfd 2914 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2915 __be32 status;
1da177e4
LT		 2916
2917 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2918 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2919 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2920 stateid->si_generation);
3a4f98bb 2921
1da177e4
LT		 2922 *stpp = NULL;
2923 *sopp = NULL;
2924
1da177e4 2925 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2926 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2927 return nfserr_bad_stateid;
1da177e4
LT		 2928 }
2929
1da177e4 2930 if (STALE_STATEID(stateid))
3a4f98bb 2931 return nfserr_stale_stateid;
6668958f
AA		 2932
2933 if (nfsd4_has_session(cstate))
2934 flags |= HAS_SESSION;
2935
1da177e4
LT		 2936 /*
2937 * We return BAD_STATEID if filehandle doesn't match stateid,
2938 * the confirmed flag is incorrecly set, or the generation
2939 * number is incorrect.
1da177e4 2940 */
f8816512
N		 2941 stp = find_stateid(stateid, flags);
2942 if (stp == NULL) {
2943 /*
2944 * Also, we should make sure this isn't just the result of
2945 * a replayed close:
2946 */
2947 sop = search_close_lru(stateid->si_stateownerid, flags);
2948 if (sop == NULL)
2949 return nfserr_bad_stateid;
2950 *sopp = sop;
2951 goto check_replay;
2952 }
1da177e4 2953
39325bd0
BF		 2954 *stpp = stp;
2955 *sopp = sop = stp->st_stateowner;
2956
4c4cd222 2957 if (lock) {
4c4cd222 2958 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2959 struct nfs4_client *clp = sop->so_client;
4c4cd222 2960 int lkflg = 0;
b37ad28b 2961 __be32 status;
4c4cd222
N		 2962
2963 lkflg = setlkflg(lock->lk_type);
2964
2965 if (lock->lk_is_new) {
599e0a22
BF		 2966 if (!sop->so_is_open_owner)
2967 return nfserr_bad_stateid;
60adfc50
AA		 2968 if (!(flags & HAS_SESSION) &&
2969 !same_clid(&clp->cl_clientid, lockclid))
2970 return nfserr_bad_stateid;
599e0a22
BF		 2971 /* stp is the open stateid */
2972 status = nfs4_check_openmode(stp, lkflg);
2973 if (status)
2974 return status;
2975 } else {
2976 /* stp is the lock stateid */
2977 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2978 if (status)
2979 return status;
4c4cd222 2980 }
1da177e4
LT		 2981 }
2982
f3362737 2983 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2984 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2985 return nfserr_bad_stateid;
1da177e4
LT		 2986 }
2987
1da177e4
LT		 2988 /*
2989 * We now validate the seqid and stateid generation numbers.
2990 * For the moment, we ignore the possibility of
2991 * generation number wraparound.
2992 */
6668958f 2993 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2994 goto check_replay;
2995
3a4f98bb 2996 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2997 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2998 " unconfirmed stateowner!\n");
2999 return nfserr_bad_stateid;
1da177e4 3000 }
3a4f98bb 3001 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3002 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3003 " confirmed yet!\n");
3004 return nfserr_bad_stateid;
1da177e4 3005 }
6668958f 3006 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3007 if (status)
3008 return status;
52fd004e 3009 renew_client(sop->so_client);
3a4f98bb 3010 return nfs_ok;
1da177e4 3011
1da177e4 3012check_replay:
bd9aac52 3013 if (seqid == sop->so_seqid - 1) {
849823c5 3014 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3015 /* indicate replay to calling function */
a90b061c 3016 return nfserr_replay_me;
1da177e4 3017 }
2fdada03 3018 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3019 sop->so_seqid, seqid);
3020 *sopp = NULL;
3021 return nfserr_bad_seqid;
1da177e4
LT		 3022}
3023
b37ad28b 3024__be32
ca364317 3025nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3026 struct nfsd4_open_confirm *oc)
1da177e4 3027{
b37ad28b 3028 __be32 status;
1da177e4
LT		 3029 struct nfs4_stateowner *sop;
3030 struct nfs4_stateid *stp;
3031
3032 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3033 (int)cstate->current_fh.fh_dentry->d_name.len,
3034 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3035
ca364317 3036 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3037 if (status)
3038 return status;
1da177e4
LT		 3039
3040 nfs4_lock_state();
3041
dd453dfd 3042 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3043 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3044 CONFIRM | OPEN_STATE,
1da177e4
LT		 3045 &oc->oc_stateowner, &stp, NULL)))
3046 goto out;
3047
3048 sop = oc->oc_stateowner;
3049 sop->so_confirmed = 1;
3050 update_stateid(&stp->st_stateid);
3051 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3052 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3053 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3054 stp->st_stateid.si_boot,
3055 stp->st_stateid.si_stateownerid,
3056 stp->st_stateid.si_fileid,
3057 stp->st_stateid.si_generation);
c7b9a459
N		 3058
3059 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3060out:
f2327d9a 3061 if (oc->oc_stateowner) {
1da177e4 3062 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3063 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3064 }
1da177e4
LT		 3065 nfs4_unlock_state();
3066 return status;
3067}
3068
3069
3070/*
3071 * unset all bits in union bitmap (bmap) that
3072 * do not exist in share (from successful OPEN_DOWNGRADE)
3073 */
3074static void
3075reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3076{
3077 int i;
3078 for (i = 1; i < 4; i++) {
3079 if ((i & access) != i)
3080 __clear_bit(i, bmap);
3081 }
3082}
3083
3084static void
3085reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3086{
3087 int i;
3088 for (i = 0; i < 4; i++) {
3089 if ((i & deny) != i)
3090 __clear_bit(i, bmap);
3091 }
3092}
3093
b37ad28b 3094__be32
ca364317
BF		 3095nfsd4_open_downgrade(struct svc_rqst *rqstp,
3096 struct nfsd4_compound_state *cstate,
a4f1706a 3097 struct nfsd4_open_downgrade *od)
1da177e4 3098{
b37ad28b 3099 __be32 status;
1da177e4
LT		 3100 struct nfs4_stateid *stp;
3101 unsigned int share_access;
3102
3103 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3104 (int)cstate->current_fh.fh_dentry->d_name.len,
3105 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3106
ba5a6a19
BF		 3107 if (!access_valid(od->od_share_access)
3108 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3109 return nfserr_inval;
3110
3111 nfs4_lock_state();
dd453dfd 3112 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3113 od->od_seqid,
1da177e4 3114 &od->od_stateid,
f3362737 3115 OPEN_STATE,
1da177e4
LT		 3116 &od->od_stateowner, &stp, NULL)))
3117 goto out;
3118
3119 status = nfserr_inval;
3120 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3121 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3122 stp->st_access_bmap, od->od_share_access);
3123 goto out;
3124 }
3125 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3126 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3127 stp->st_deny_bmap, od->od_share_deny);
3128 goto out;
3129 }
3130 set_access(&share_access, stp->st_access_bmap);
3131 nfs4_file_downgrade(stp->st_vfs_file,
3132 share_access & ~od->od_share_access);
3133
3134 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3135 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3136
3137 update_stateid(&stp->st_stateid);
3138 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3139 status = nfs_ok;
3140out:
f2327d9a 3141 if (od->od_stateowner) {
1da177e4 3142 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3143 cstate->replay_owner = od->od_stateowner;
f2327d9a 3144 }
1da177e4
LT		 3145 nfs4_unlock_state();
3146 return status;
3147}
3148
3149/*
3150 * nfs4_unlock_state() called after encode
3151 */
b37ad28b 3152__be32
ca364317 3153nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3154 struct nfsd4_close *close)
1da177e4 3155{
b37ad28b 3156 __be32 status;
1da177e4
LT		 3157 struct nfs4_stateid *stp;
3158
3159 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3160 (int)cstate->current_fh.fh_dentry->d_name.len,
3161 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3162
3163 nfs4_lock_state();
3164 /* check close_lru for replay */
dd453dfd 3165 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3166 close->cl_seqid,
1da177e4 3167 &close->cl_stateid,
f3362737 3168 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3169 &close->cl_stateowner, &stp, NULL)))
3170 goto out;
1da177e4
LT		 3171 status = nfs_ok;
3172 update_stateid(&stp->st_stateid);
3173 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3174
04ef5954 3175 /* release_stateid() calls nfsd_close() if needed */
2283963f 3176 release_open_stateid(stp);
04ef5954
BF		 3177
3178 /* place unused nfs4_stateowners on so_close_lru list to be
3179 * released by the laundromat service after the lease period
3180 * to enable us to handle CLOSE replay
3181 */
3182 if (list_empty(&close->cl_stateowner->so_stateids))
3183 move_to_close_lru(close->cl_stateowner);
1da177e4 3184out:
f2327d9a 3185 if (close->cl_stateowner) {
1da177e4 3186 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3187 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3188 }
1da177e4
LT		 3189 nfs4_unlock_state();
3190 return status;
3191}
3192
b37ad28b 3193__be32
ca364317
BF		 3194nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3195 struct nfsd4_delegreturn *dr)
1da177e4 3196{
203a8c8e
BF		 3197 struct nfs4_delegation *dp;
3198 stateid_t *stateid = &dr->dr_stateid;
3199 struct inode *inode;
b37ad28b 3200 __be32 status;
6668958f 3201 int flags = 0;
1da177e4 3202
ca364317 3203 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3204 return status;
3205 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3206
6668958f
AA		 3207 if (nfsd4_has_session(cstate))
3208 flags |= HAS_SESSION;
1da177e4 3209 nfs4_lock_state();
203a8c8e
BF		 3210 status = nfserr_bad_stateid;
3211 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3212 goto out;
3213 status = nfserr_stale_stateid;
3214 if (STALE_STATEID(stateid))
3215 goto out;
7e0f7cf5 3216 status = nfserr_bad_stateid;
203a8c8e
BF		 3217 if (!is_delegation_stateid(stateid))
3218 goto out;
203a8c8e
BF		 3219 dp = find_delegation_stateid(inode, stateid);
3220 if (!dp)
3221 goto out;
6668958f 3222 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3223 if (status)
3224 goto out;
3225 renew_client(dp->dl_client);
3226
3227 unhash_delegation(dp);
1da177e4 3228out:
203a8c8e
BF		 3229 nfs4_unlock_state();
3230
1da177e4
LT		 3231 return status;
3232}
3233
3234
3235/*
3236 * Lock owner state (byte-range locks)
3237 */
3238#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3239#define LOCK_HASH_BITS 8
3240#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3241#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3242
87df4de8
BH		 3243static inline u64
3244end_offset(u64 start, u64 len)
3245{
3246 u64 end;
3247
3248 end = start + len;
3249 return end >= start ? end: NFS4_MAX_UINT64;
3250}
3251
3252/* last octet in a range */
3253static inline u64
3254last_byte_offset(u64 start, u64 len)
3255{
3256 u64 end;
3257
3258 BUG_ON(!len);
3259 end = start + len;
3260 return end > start ? end - 1: NFS4_MAX_UINT64;
3261}
3262
1da177e4
LT		 3263#define lockownerid_hashval(id) \
3264 ((id) & LOCK_HASH_MASK)
3265
3266static inline unsigned int
3267lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3268 struct xdr_netobj *ownername)
3269{
3270 return (file_hashval(inode) + cl_id
3271 + opaque_hashval(ownername->data, ownername->len))
3272 & LOCK_HASH_MASK;
3273}
3274
3275static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3276static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3277static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3278
fd39ca9a 3279static struct nfs4_stateid *
1da177e4
LT		 3280find_stateid(stateid_t *stid, int flags)
3281{
9346eff0 3282 struct nfs4_stateid *local;
1da177e4
LT		 3283 u32 st_id = stid->si_stateownerid;
3284 u32 f_id = stid->si_fileid;
3285 unsigned int hashval;
3286
3287 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3288 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3289 hashval = stateid_hashval(st_id, f_id);
3290 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3291 if ((local->st_stateid.si_stateownerid == st_id) &&
3292 (local->st_stateid.si_fileid == f_id))
3293 return local;
3294 }
3295 }
9346eff0
KK		 3296
3297 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3298 hashval = stateid_hashval(st_id, f_id);
3299 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3300 if ((local->st_stateid.si_stateownerid == st_id) &&
3301 (local->st_stateid.si_fileid == f_id))
3302 return local;
3303 }
849823c5 3304 }
1da177e4
LT		 3305 return NULL;
3306}
3307
3308static struct nfs4_delegation *
3309find_delegation_stateid(struct inode *ino, stateid_t *stid)
3310{
13cd2184
N		 3311 struct nfs4_file *fp;
3312 struct nfs4_delegation *dl;
1da177e4
LT		 3313
3314 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3315 stid->si_boot, stid->si_stateownerid,
3316 stid->si_fileid, stid->si_generation);
3317
1da177e4 3318 fp = find_file(ino);
13cd2184
N		 3319 if (!fp)
3320 return NULL;
3321 dl = find_delegation_file(fp, stid);
3322 put_nfs4_file(fp);
3323 return dl;
1da177e4
LT		 3324}
3325
3326/*
3327 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3328 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3329 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3330 * locking, this prevents us from being completely protocol-compliant. The
3331 * real solution to this problem is to start using unsigned file offsets in
3332 * the VFS, but this is a very deep change!
3333 */
3334static inline void
3335nfs4_transform_lock_offset(struct file_lock *lock)
3336{
3337 if (lock->fl_start < 0)
3338 lock->fl_start = OFFSET_MAX;
3339 if (lock->fl_end < 0)
3340 lock->fl_end = OFFSET_MAX;
3341}
3342
d5b9026a
N		 3343/* Hack!: For now, we're defining this just so we can use a pointer to it
3344 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3345static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3346};
1da177e4
LT		 3347
3348static inline void
3349nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3350{
d5b9026a
N		 3351 struct nfs4_stateowner *sop;
3352 unsigned int hval;
1da177e4 3353
d5b9026a
N		 3354 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3355 sop = (struct nfs4_stateowner *) fl->fl_owner;
3356 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3357 kref_get(&sop->so_ref);
3358 deny->ld_sop = sop;
3359 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3360 } else {
3361 deny->ld_sop = NULL;
3362 deny->ld_clientid.cl_boot = 0;
3363 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3364 }
3365 deny->ld_start = fl->fl_start;
87df4de8
BH		 3366 deny->ld_length = NFS4_MAX_UINT64;
3367 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3368 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3369 deny->ld_type = NFS4_READ_LT;
3370 if (fl->fl_type != F_RDLCK)
3371 deny->ld_type = NFS4_WRITE_LT;
3372}
3373
1da177e4
LT		 3374static struct nfs4_stateowner *
3375find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3376 struct xdr_netobj *owner)
3377{
3378 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3379 struct nfs4_stateowner *op;
3380
3381 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3382 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3383 return op;
3384 }
3385 return NULL;
3386}
3387
3388/*
3389 * Alloc a lock owner structure.
3390 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3391 * occured.
3392 *
3393 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3394 */
3395
3396static struct nfs4_stateowner *
3397alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3398 struct nfs4_stateowner *sop;
3399 struct nfs4_replay *rp;
3400 unsigned int idhashval;
3401
3402 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3403 return NULL;
3404 idhashval = lockownerid_hashval(current_ownerid);
3405 INIT_LIST_HEAD(&sop->so_idhash);
3406 INIT_LIST_HEAD(&sop->so_strhash);
3407 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3408 INIT_LIST_HEAD(&sop->so_stateids);
3409 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3410 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3411 sop->so_time = 0;
3412 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3413 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3414 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3415 sop->so_is_open_owner = 0;
3416 sop->so_id = current_ownerid++;
3417 sop->so_client = clp;
b59e3c0e
NB		 3418 /* It is the openowner seqid that will be incremented in encode in the
3419 * case of new lockowners; so increment the lock seqid manually: */
3420 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3421 sop->so_confirmed = 1;
3422 rp = &sop->so_replay;
de1ae286 3423 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3424 rp->rp_buflen = 0;
3425 rp->rp_buf = rp->rp_ibuf;
3426 return sop;
3427}
3428
fd39ca9a 3429static struct nfs4_stateid *
1da177e4
LT		 3430alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3431{
3432 struct nfs4_stateid *stp;
3433 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3434
5ac049ac
N		 3435 stp = nfs4_alloc_stateid();
3436 if (stp == NULL)
1da177e4
LT		 3437 goto out;
3438 INIT_LIST_HEAD(&stp->st_hash);
3439 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3440 INIT_LIST_HEAD(&stp->st_perstateowner);
3441 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3442 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3443 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3444 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3445 stp->st_stateowner = sop;
13cd2184 3446 get_nfs4_file(fp);
1da177e4
LT		 3447 stp->st_file = fp;
3448 stp->st_stateid.si_boot = boot_time;
3449 stp->st_stateid.si_stateownerid = sop->so_id;
3450 stp->st_stateid.si_fileid = fp->fi_id;
3451 stp->st_stateid.si_generation = 0;
3452 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3453 stp->st_access_bmap = open_stp->st_access_bmap;
3454 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3455 stp->st_openstp = open_stp;
1da177e4
LT		 3456
3457out:
3458 return stp;
3459}
3460
fd39ca9a 3461static int
1da177e4
LT		 3462check_lock_length(u64 offset, u64 length)
3463{
87df4de8 3464 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3465 LOFF_OVERFLOW(offset, length)));
3466}
3467
3468/*
3469 * LOCK operation
3470 */
b37ad28b 3471__be32
ca364317 3472nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3473 struct nfsd4_lock *lock)
1da177e4 3474{
3e9e3dbe 3475 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3476 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3477 struct nfs4_stateid *lock_stp;
3478 struct file *filp;
3479 struct file_lock file_lock;
8dc7c311 3480 struct file_lock conflock;
b37ad28b 3481 __be32 status = 0;
1da177e4 3482 unsigned int strhashval;
150b3934 3483 unsigned int cmd;
b8dd7b9a 3484 int err;
1da177e4
LT		 3485
3486 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3487 (long long) lock->lk_offset,
3488 (long long) lock->lk_length);
3489
1da177e4
LT		 3490 if (check_lock_length(lock->lk_offset, lock->lk_length))
3491 return nfserr_inval;
3492
ca364317 3493 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3494 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3495 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3496 return status;
3497 }
3498
1da177e4
LT		 3499 nfs4_lock_state();
3500
3501 if (lock->lk_is_new) {
893f8770
N		 3502 /*
3503 * Client indicates that this is a new lockowner.
3504 * Use open owner and open stateid to create lock owner and
3505 * lock stateid.
3506 */
1da177e4
LT		 3507 struct nfs4_stateid *open_stp = NULL;
3508 struct nfs4_file *fp;
3509
3510 status = nfserr_stale_clientid;
60adfc50
AA		 3511 if (!nfsd4_has_session(cstate) &&
3512 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3513 goto out;
1da177e4 3514
1da177e4 3515 /* validate and update open stateid and open seqid */
dd453dfd 3516 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3517 lock->lk_new_open_seqid,
3518 &lock->lk_new_open_stateid,
f3362737 3519 OPEN_STATE,
3a65588a 3520 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3521 lock);
37515177 3522 if (status)
1da177e4 3523 goto out;
3a65588a 3524 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3525 /* create lockowner and lock stateid */
3526 fp = open_stp->st_file;
3527 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3528 open_sop->so_client->cl_clientid.cl_id,
3529 &lock->v.new.owner);
3e9e3dbe
N		 3530 /* XXX: Do we need to check for duplicate stateowners on
3531 * the same file, or should they just be allowed (and
3532 * create new stateids)? */
1da177e4 3533 status = nfserr_resource;
b59e3c0e
NB		 3534 lock_sop = alloc_init_lock_stateowner(strhashval,
3535 open_sop->so_client, open_stp, lock);
3536 if (lock_sop == NULL)
1da177e4 3537 goto out;
b59e3c0e 3538 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3539 if (lock_stp == NULL)
1da177e4 3540 goto out;
1da177e4
LT		 3541 } else {
3542 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3543 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3544 lock->lk_old_lock_seqid,
3545 &lock->lk_old_lock_stateid,
f3362737 3546 LOCK_STATE,
3a65588a 3547 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3548 if (status)
3549 goto out;
3a65588a 3550 lock_sop = lock->lk_replay_owner;
1da177e4 3551 }
3a65588a 3552 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3553 filp = lock_stp->st_vfs_file;
3554
0dd395dc 3555 status = nfserr_grace;
af558e33 3556 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3557 goto out;
3558 status = nfserr_no_grace;
af558e33 3559 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3560 goto out;
3561
1da177e4
LT		 3562 locks_init_lock(&file_lock);
3563 switch (lock->lk_type) {
3564 case NFS4_READ_LT:
3565 case NFS4_READW_LT:
3566 file_lock.fl_type = F_RDLCK;
150b3934 3567 cmd = F_SETLK;
1da177e4
LT		 3568 break;
3569 case NFS4_WRITE_LT:
3570 case NFS4_WRITEW_LT:
3571 file_lock.fl_type = F_WRLCK;
150b3934 3572 cmd = F_SETLK;
1da177e4
LT		 3573 break;
3574 default:
3575 status = nfserr_inval;
3576 goto out;
3577 }
b59e3c0e 3578 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3579 file_lock.fl_pid = current->tgid;
3580 file_lock.fl_file = filp;
3581 file_lock.fl_flags = FL_POSIX;
d5b9026a 3582 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3583
3584 file_lock.fl_start = lock->lk_offset;
87df4de8 3585 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3586 nfs4_transform_lock_offset(&file_lock);
3587
3588 /*
3589 * Try to lock the file in the VFS.
3590 * Note: locks.c uses the BKL to protect the inode's lock list.
3591 */
3592
fd85b817 3593 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3594 switch (-err) {
1da177e4
LT		 3595 case 0: /* success! */
3596 update_stateid(&lock_stp->st_stateid);
3597 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3598 sizeof(stateid_t));
b8dd7b9a 3599 status = 0;
eb76b3fd
AA		 3600 break;
3601 case (EAGAIN): /* conflock holds conflicting lock */
3602 status = nfserr_denied;
3603 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3604 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3605 break;
1da177e4
LT		 3606 case (EDEADLK):
3607 status = nfserr_deadlock;
eb76b3fd 3608 break;
1da177e4 3609 default:
fd85b817 3610 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3611 status = nfserr_resource;
3612 break;
1da177e4 3613 }
1da177e4 3614out:
8a280510 3615 if (status && lock->lk_is_new && lock_sop)
f044ff83 3616 release_lockowner(lock_sop);
3a65588a
BF		 3617 if (lock->lk_replay_owner) {
3618 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3619 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3620 }
1da177e4
LT		 3621 nfs4_unlock_state();
3622 return status;
3623}
3624
55ef1274
BF		 3625/*
3626 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3627 * so we do a temporary open here just to get an open file to pass to
3628 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3629 * inode operation.)
3630 */
3631static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3632{
3633 struct file *file;
3634 int err;
3635
3636 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3637 if (err)
3638 return err;
3639 err = vfs_test_lock(file, lock);
3640 nfsd_close(file);
3641 return err;
3642}
3643
1da177e4
LT		 3644/*
3645 * LOCKT operation
3646 */
b37ad28b 3647__be32
ca364317
BF		 3648nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3649 struct nfsd4_lockt *lockt)
1da177e4
LT		 3650{
3651 struct inode *inode;
1da177e4 3652 struct file_lock file_lock;
fd85b817 3653 int error;
b37ad28b 3654 __be32 status;
1da177e4 3655
af558e33 3656 if (locks_in_grace())
1da177e4
LT		 3657 return nfserr_grace;
3658
3659 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3660 return nfserr_inval;
3661
3662 lockt->lt_stateowner = NULL;
3663 nfs4_lock_state();
3664
3665 status = nfserr_stale_clientid;
60adfc50 3666 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3667 goto out;
1da177e4 3668
ca364317 3669 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3670 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3671 if (status == nfserr_symlink)
3672 status = nfserr_inval;
3673 goto out;
3674 }
3675
ca364317 3676 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3677 locks_init_lock(&file_lock);
3678 switch (lockt->lt_type) {
3679 case NFS4_READ_LT:
3680 case NFS4_READW_LT:
3681 file_lock.fl_type = F_RDLCK;
3682 break;
3683 case NFS4_WRITE_LT:
3684 case NFS4_WRITEW_LT:
3685 file_lock.fl_type = F_WRLCK;
3686 break;
3687 default:
2fdada03 3688 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3689 status = nfserr_inval;
3690 goto out;
3691 }
3692
3693 lockt->lt_stateowner = find_lockstateowner_str(inode,
3694 &lockt->lt_clientid, &lockt->lt_owner);
3695 if (lockt->lt_stateowner)
3696 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3697 file_lock.fl_pid = current->tgid;
3698 file_lock.fl_flags = FL_POSIX;
3699
3700 file_lock.fl_start = lockt->lt_offset;
87df4de8 3701 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3702
3703 nfs4_transform_lock_offset(&file_lock);
3704
1da177e4 3705 status = nfs_ok;
55ef1274 3706 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3707 if (error) {
3708 status = nfserrno(error);
3709 goto out;
3710 }
9d6a8c5c 3711 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3712 status = nfserr_denied;
9d6a8c5c 3713 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3714 }
3715out:
3716 nfs4_unlock_state();
3717 return status;
3718}
3719
b37ad28b 3720__be32
ca364317 3721nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3722 struct nfsd4_locku *locku)
1da177e4
LT		 3723{
3724 struct nfs4_stateid *stp;
3725 struct file *filp = NULL;
3726 struct file_lock file_lock;
b37ad28b 3727 __be32 status;
b8dd7b9a 3728 int err;
1da177e4
LT		 3729
3730 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3731 (long long) locku->lu_offset,
3732 (long long) locku->lu_length);
3733
3734 if (check_lock_length(locku->lu_offset, locku->lu_length))
3735 return nfserr_inval;
3736
3737 nfs4_lock_state();
3738
dd453dfd 3739 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3740 locku->lu_seqid,
3741 &locku->lu_stateid,
f3362737 3742 LOCK_STATE,
1da177e4
LT		 3743 &locku->lu_stateowner, &stp, NULL)))
3744 goto out;
3745
3746 filp = stp->st_vfs_file;
3747 BUG_ON(!filp);
3748 locks_init_lock(&file_lock);
3749 file_lock.fl_type = F_UNLCK;
3750 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3751 file_lock.fl_pid = current->tgid;
3752 file_lock.fl_file = filp;
3753 file_lock.fl_flags = FL_POSIX;
d5b9026a 3754 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3755 file_lock.fl_start = locku->lu_offset;
3756
87df4de8 3757 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3758 nfs4_transform_lock_offset(&file_lock);
3759
3760 /*
3761 * Try to unlock the file in the VFS.
3762 */
fd85b817 3763 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3764 if (err) {
fd85b817 3765 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3766 goto out_nfserr;
3767 }
3768 /*
3769 * OK, unlock succeeded; the only thing left to do is update the stateid.
3770 */
3771 update_stateid(&stp->st_stateid);
3772 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3773
3774out:
f2327d9a 3775 if (locku->lu_stateowner) {
1da177e4 3776 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3777 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3778 }
1da177e4
LT		 3779 nfs4_unlock_state();
3780 return status;
3781
3782out_nfserr:
b8dd7b9a 3783 status = nfserrno(err);
1da177e4
LT		 3784 goto out;
3785}
3786
3787/*
3788 * returns
3789 * 1: locks held by lockowner
3790 * 0: no locks held by lockowner
3791 */
3792static int
3793check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3794{
3795 struct file_lock **flpp;
7eaa36e2 3796 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3797 int status = 0;
3798
3799 lock_kernel();
3800 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3801 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3802 status = 1;
3803 goto out;
796dadfd 3804 }
1da177e4
LT		 3805 }
3806out:
3807 unlock_kernel();
3808 return status;
3809}
3810
b37ad28b 3811__be32
b591480b
BF		 3812nfsd4_release_lockowner(struct svc_rqst *rqstp,
3813 struct nfsd4_compound_state *cstate,
3814 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3815{
3816 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3817 struct nfs4_stateowner *sop;
3818 struct nfs4_stateid *stp;
1da177e4 3819 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3820 struct list_head matches;
3821 int i;
b37ad28b 3822 __be32 status;
1da177e4
LT		 3823
3824 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3825 clid->cl_boot, clid->cl_id);
3826
3827 /* XXX check for lease expiration */
3828
3829 status = nfserr_stale_clientid;
849823c5 3830 if (STALE_CLIENTID(clid))
1da177e4 3831 return status;
1da177e4
LT		 3832
3833 nfs4_lock_state();
3834
3e9e3dbe
N		 3835 status = nfserr_locks_held;
3836 /* XXX: we're doing a linear search through all the lockowners.
3837 * Yipes! For now we'll just hope clients aren't really using
3838 * release_lockowner much, but eventually we have to fix these
3839 * data structures. */
3840 INIT_LIST_HEAD(&matches);
3841 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3842 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3843 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3844 continue;
3845 list_for_each_entry(stp, &sop->so_stateids,
3846 st_perstateowner) {
3847 if (check_for_locks(stp->st_vfs_file, sop))
3848 goto out;
3849 /* Note: so_perclient unused for lockowners,
3850 * so it's OK to fool with here. */
3851 list_add(&sop->so_perclient, &matches);
3852 }
1da177e4 3853 }
3e9e3dbe
N		 3854 }
3855 /* Clients probably won't expect us to return with some (but not all)
3856 * of the lockowner state released; so don't release any until all
3857 * have been checked. */
3858 status = nfs_ok;
0fa822e4
N		 3859 while (!list_empty(&matches)) {
3860 sop = list_entry(matches.next, struct nfs4_stateowner,
3861 so_perclient);
3862 /* unhash_stateowner deletes so_perclient only
3863 * for openowners. */
3864 list_del(&sop->so_perclient);
f044ff83 3865 release_lockowner(sop);
1da177e4
LT		 3866 }
3867out:
3868 nfs4_unlock_state();
3869 return status;
3870}
3871
3872static inline struct nfs4_client_reclaim *
a55370a3 3873alloc_reclaim(void)
1da177e4 3874{
a55370a3 3875 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3876}
3877
c7b9a459 3878int
a1bcecd2 3879nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3880{
3881 unsigned int strhashval = clientstr_hashval(name);
3882 struct nfs4_client *clp;
3883
a1bcecd2 3884 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3885 return clp ? 1 : 0;
3886}
3887
1da177e4
LT		 3888/*
3889 * failure => all reset bets are off, nfserr_no_grace...
3890 */
190e4fbf
N		 3891int
3892nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3893{
3894 unsigned int strhashval;
3895 struct nfs4_client_reclaim *crp = NULL;
3896
a55370a3
N		 3897 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3898 crp = alloc_reclaim();
1da177e4
LT		 3899 if (!crp)
3900 return 0;
a55370a3 3901 strhashval = clientstr_hashval(name);
1da177e4
LT		 3902 INIT_LIST_HEAD(&crp->cr_strhash);
3903 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3904 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3905 reclaim_str_hashtbl_size++;
3906 return 1;
3907}
3908
3909static void
3910nfs4_release_reclaim(void)
3911{
3912 struct nfs4_client_reclaim *crp = NULL;
3913 int i;
3914
1da177e4
LT		 3915 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3916 while (!list_empty(&reclaim_str_hashtbl[i])) {
3917 crp = list_entry(reclaim_str_hashtbl[i].next,
3918 struct nfs4_client_reclaim, cr_strhash);
3919 list_del(&crp->cr_strhash);
1da177e4
LT		 3920 kfree(crp);
3921 reclaim_str_hashtbl_size--;
3922 }
3923 }
3924 BUG_ON(reclaim_str_hashtbl_size);
3925}
3926
3927/*
3928 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3929static struct nfs4_client_reclaim *
1da177e4
LT		 3930nfs4_find_reclaim_client(clientid_t *clid)
3931{
3932 unsigned int strhashval;
3933 struct nfs4_client *clp;
3934 struct nfs4_client_reclaim *crp = NULL;
3935
3936
3937 /* find clientid in conf_id_hashtbl */
3938 clp = find_confirmed_client(clid);
3939 if (clp == NULL)
3940 return NULL;
3941
a55370a3
N		 3942 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3943 clp->cl_name.len, clp->cl_name.data,
3944 clp->cl_recdir);
1da177e4
LT		 3945
3946 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3947 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3948 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3949 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3950 return crp;
3951 }
3952 }
3953 return NULL;
3954}
3955
3956/*
3957* Called from OPEN. Look for clientid in reclaim list.
3958*/
b37ad28b 3959__be32
1da177e4
LT		 3960nfs4_check_open_reclaim(clientid_t *clid)
3961{
dfc83565 3962 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3963}
3964
ac4d8ff2 3965/* initialization to perform at module load time: */
1da177e4 3966
e8ff2a84 3967int
ac4d8ff2 3968nfs4_state_init(void)
1da177e4 3969{
e8ff2a84 3970 int i, status;
1da177e4 3971
e8ff2a84
BF		 3972 status = nfsd4_init_slabs();
3973 if (status)
3974 return status;
1da177e4
LT		 3975 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3976 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3977 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3978 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3979 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3980 }
5282fd72
ME		 3981 for (i = 0; i < SESSION_HASH_SIZE; i++)
3982 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3983 for (i = 0; i < FILE_HASH_SIZE; i++) {
3984 INIT_LIST_HEAD(&file_hashtbl[i]);
3985 }
3986 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3987 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3988 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3989 }
3990 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3991 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3992 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3993 }
3994 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3995 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3996 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3997 }
1da177e4 3998 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3999 INIT_LIST_HEAD(&close_lru);
4000 INIT_LIST_HEAD(&client_lru);
4001 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 4002 for (i = 0; i < CLIENT_HASH_SIZE; i++)
4003 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4004 reclaim_str_hashtbl_size = 0;
e8ff2a84 4005 return 0;
ac4d8ff2
N		 4006}
4007
190e4fbf
N		 4008static void
4009nfsd4_load_reboot_recovery_data(void)
4010{
4011 int status;
4012
0964a3d3
N		 4013 nfs4_lock_state();
4014 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4015 status = nfsd4_recdir_load();
0964a3d3 4016 nfs4_unlock_state();
190e4fbf
N		 4017 if (status)
4018 printk("NFSD: Failure reading reboot recovery data\n");
4019}
4020
9a8db97e
ME		 4021unsigned long
4022get_nfs4_grace_period(void)
4023{
4024 return max(user_lease_time, lease_time) * HZ;
4025}
4026
c2f1a551
MS		 4027/*
4028 * Since the lifetime of a delegation isn't limited to that of an open, a
4029 * client may quite reasonably hang on to a delegation as long as it has
4030 * the inode cached. This becomes an obvious problem the first time a
4031 * client's inode cache approaches the size of the server's total memory.
4032 *
4033 * For now we avoid this problem by imposing a hard limit on the number
4034 * of delegations, which varies according to the server's memory size.
4035 */
4036static void
4037set_max_delegations(void)
4038{
4039 /*
4040 * Allow at most 4 delegations per megabyte of RAM. Quick
4041 * estimates suggest that in the worst case (where every delegation
4042 * is for a different inode), a delegation could take about 1.5K,
4043 * giving a worst case usage of about 6% of memory.
4044 */
4045 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4046}
4047
ac4d8ff2
N		 4048/* initialization to perform when the nfsd service is started: */
4049
4050static void
4051__nfs4_state_start(void)
4052{
9a8db97e 4053 unsigned long grace_time;
ac4d8ff2 4054
1da177e4 4055 boot_time = get_seconds();
af558e33 4056 grace_time = get_nfs4_grace_period();
d99a05ad 4057 lease_time = user_lease_time;
af558e33 4058 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4059 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4060 grace_time/HZ);
58da282b 4061 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4062 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4063 set_max_delegations();
1da177e4
LT		 4064}
4065
e8ff2a84 4066void
76a3550e 4067nfs4_state_start(void)
1da177e4 4068{
1da177e4 4069 if (nfs4_init)
e8ff2a84 4070 return;
190e4fbf 4071 nfsd4_load_reboot_recovery_data();
76a3550e 4072 __nfs4_state_start();
1da177e4 4073 nfs4_init = 1;
e8ff2a84 4074 return;
1da177e4
LT		 4075}
4076
1da177e4
LT		 4077time_t
4078nfs4_lease_time(void)
4079{
4080 return lease_time;
4081}
4082
4083static void
4084__nfs4_state_shutdown(void)
4085{
4086 int i;
4087 struct nfs4_client *clp = NULL;
4088 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4089 struct list_head *pos, *next, reaplist;
4090
1da177e4
LT		 4091 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4092 while (!list_empty(&conf_id_hashtbl[i])) {
4093 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4094 expire_client(clp);
4095 }
4096 while (!list_empty(&unconf_str_hashtbl[i])) {
4097 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4098 expire_client(clp);
4099 }
4100 }
4101 INIT_LIST_HEAD(&reaplist);
4102 spin_lock(&recall_lock);
4103 list_for_each_safe(pos, next, &del_recall_lru) {
4104 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4105 list_move(&dp->dl_recall_lru, &reaplist);
4106 }
4107 spin_unlock(&recall_lock);
4108 list_for_each_safe(pos, next, &reaplist) {
4109 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4110 list_del_init(&dp->dl_recall_lru);
4111 unhash_delegation(dp);
4112 }
4113
190e4fbf 4114 nfsd4_shutdown_recdir();
1da177e4 4115 nfs4_init = 0;
1da177e4
LT		 4116}
4117
4118void
4119nfs4_state_shutdown(void)
4120{
5e8d5c29
N		 4121 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4122 destroy_workqueue(laundry_wq);
2c5e7615 4123 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4124 nfs4_lock_state();
4125 nfs4_release_reclaim();
4126 __nfs4_state_shutdown();
1da177e4
LT		 4127 nfs4_unlock_state();
4128}
4129
3dd98a3b
JL		 4130/*
4131 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4132 * accessed when nfsd is starting.
4133 */
0964a3d3
N		 4134static void
4135nfs4_set_recdir(char *recdir)
4136{
0964a3d3 4137 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4138}
4139
4140/*
4141 * Change the NFSv4 recovery directory to recdir.
4142 */
4143int
4144nfs4_reset_recoverydir(char *recdir)
4145{
4146 int status;
a63bb996 4147 struct path path;
0964a3d3 4148
a63bb996 4149 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4150 if (status)
4151 return status;
4152 status = -ENOTDIR;
a63bb996 4153 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4154 nfs4_set_recdir(recdir);
4155 status = 0;
4156 }
a63bb996 4157 path_put(&path);
0964a3d3
N		 4158 return status;
4159}
4160
3dd98a3b
JL		 4161char *
4162nfs4_recoverydir(void)
4163{
4164 return user_recovery_dirname;
4165}
4166
1da177e4
LT		 4167/*
4168 * Called when leasetime is changed.
4169 *
d99a05ad
N		 4170 * The only way the protocol gives us to handle on-the-fly lease changes is to
4171 * simulate a reboot. Instead of doing that, we just wait till the next time
4172 * we start to register any changes in lease time. If the administrator
4173 * really wants to change the lease time *now*, they can go ahead and bring
4174 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4175 *
4176 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4177 * when nfsd is starting
1da177e4
LT		 4178 */
4179void
4180nfs4_reset_lease(time_t leasetime)
4181{
d99a05ad 4182 user_lease_time = leasetime;
1da177e4 4183}
Linux 6.x block layer and io_uring tree(s)