projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFSD: Fix a bug in the NFSv4 'supported attrs' mandatory attribute
[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>
363168b4 58#include <linux/sunrpc/clnt.h>
1da177e4
LT		 59
60#define NFSDDBG_FACILITY NFSDDBG_PROC
61
62/* Globals */
63static time_t lease_time = 90; /* default lease time */
d99a05ad 64static time_t user_lease_time = 90;
fd39ca9a 65static time_t boot_time;
1da177e4
LT		 66static u32 current_ownerid = 1;
67static u32 current_fileid = 1;
68static u32 current_delegid = 1;
69static u32 nfs4_init;
fd39ca9a
N		 70static stateid_t zerostateid; /* bits all 0 */
71static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 72static u64 current_sessionid = 1;
fd39ca9a
N		 73
74#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
75#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 76
1da177e4 77/* forward declarations */
fd39ca9a 78static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 79static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 80static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
81static void nfs4_set_recdir(char *recdir);
1da177e4 82
8b671b80
BF		 83/* Locking: */
84
85/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 86static DEFINE_MUTEX(client_mutex);
1da177e4 87
8b671b80
BF		 88/*
89 * Currently used for the del_recall_lru and file hash table. In an
90 * effort to decrease the scope of the client_mutex, this spinlock may
91 * eventually cover more:
92 */
93static DEFINE_SPINLOCK(recall_lock);
94
e18b890b
CL		 95static struct kmem_cache *stateowner_slab = NULL;
96static struct kmem_cache *file_slab = NULL;
97static struct kmem_cache *stateid_slab = NULL;
98static struct kmem_cache *deleg_slab = NULL;
e60d4398 99
1da177e4
LT		 100void
101nfs4_lock_state(void)
102{
353ab6e9 103 mutex_lock(&client_mutex);
1da177e4
LT		 104}
105
106void
107nfs4_unlock_state(void)
108{
353ab6e9 109 mutex_unlock(&client_mutex);
1da177e4
LT		 110}
111
112static inline u32
113opaque_hashval(const void *ptr, int nbytes)
114{
115 unsigned char *cptr = (unsigned char *) ptr;
116
117 u32 x = 0;
118 while (nbytes--) {
119 x *= 37;
120 x += *cptr++;
121 }
122 return x;
123}
124
1da177e4
LT		 125static struct list_head del_recall_lru;
126
13cd2184
N		 127static inline void
128put_nfs4_file(struct nfs4_file *fi)
129{
8b671b80
BF		 130 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
131 list_del(&fi->fi_hash);
132 spin_unlock(&recall_lock);
133 iput(fi->fi_inode);
134 kmem_cache_free(file_slab, fi);
135 }
13cd2184
N		 136}
137
138static inline void
139get_nfs4_file(struct nfs4_file *fi)
140{
8b671b80 141 atomic_inc(&fi->fi_ref);
13cd2184
N		 142}
143
ef0f3390 144static int num_delegations;
c2f1a551 145unsigned int max_delegations;
ef0f3390
N		 146
147/*
148 * Open owner state (share locks)
149 */
150
151/* hash tables for nfs4_stateowner */
152#define OWNER_HASH_BITS 8
153#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
154#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155
156#define ownerid_hashval(id) \
157 ((id) & OWNER_HASH_MASK)
158#define ownerstr_hashval(clientid, ownername) \
159 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160
161static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
162static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
163
164/* hash table for nfs4_file */
165#define FILE_HASH_BITS 8
166#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
167#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
168/* hash table for (open)nfs4_stateid */
169#define STATEID_HASH_BITS 10
170#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
171#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172
173#define file_hashval(x) \
174 hash_ptr(x, FILE_HASH_BITS)
175#define stateid_hashval(owner_id, file_id) \
176 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177
178static struct list_head file_hashtbl[FILE_HASH_SIZE];
179static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
180
1da177e4
LT		 181static struct nfs4_delegation *
182alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
183{
184 struct nfs4_delegation *dp;
185 struct nfs4_file *fp = stp->st_file;
c237dc03 186 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 187
188 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 189 if (fp->fi_had_conflict)
190 return NULL;
c2f1a551 191 if (num_delegations > max_delegations)
ef0f3390 192 return NULL;
5b2d21c1
N		 193 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
194 if (dp == NULL)
1da177e4 195 return dp;
ef0f3390 196 num_delegations++;
ea1da636
N		 197 INIT_LIST_HEAD(&dp->dl_perfile);
198 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 199 INIT_LIST_HEAD(&dp->dl_recall_lru);
200 dp->dl_client = clp;
13cd2184 201 get_nfs4_file(fp);
1da177e4
LT		 202 dp->dl_file = fp;
203 dp->dl_flock = NULL;
204 get_file(stp->st_vfs_file);
205 dp->dl_vfs_file = stp->st_vfs_file;
206 dp->dl_type = type;
b53d40c5 207 dp->dl_ident = cb->cb_ident;
78155ed7 208 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
6c02eaa1 212 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 213 dp->dl_time = 0;
214 atomic_set(&dp->dl_count, 1);
ea1da636
N		 215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 217 return dp;
218}
219
220void
221nfs4_put_delegation(struct nfs4_delegation *dp)
222{
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 225 put_nfs4_file(dp->dl_file);
5b2d21c1 226 kmem_cache_free(deleg_slab, dp);
ef0f3390 227 num_delegations--;
1da177e4
LT		 228 }
229}
230
231/* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
235 */
236static void
237nfs4_close_delegation(struct nfs4_delegation *dp)
238{
239 struct file *filp = dp->dl_vfs_file;
240
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
c907132d 245 if (dp->dl_flock)
a9933cea 246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 247 nfsd_close(filp);
1da177e4
LT		 248}
249
250/* Called under the state lock. */
251static void
252unhash_delegation(struct nfs4_delegation *dp)
253{
ea1da636
N		 254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
261}
262
263/*
264 * SETCLIENTID state
265 */
266
267/* Hash tables for nfs4_clientid state */
268#define CLIENT_HASH_BITS 4
269#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271
272#define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 274#define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 276/*
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
279 *
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
282 *
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
284 * setclientid info.
285 *
286 * client_lru holds client queue ordered by nfs4_client.cl_time
287 * for lease renewal.
288 *
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
291 */
292static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293static int reclaim_str_hashtbl_size = 0;
294static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head client_lru;
299static struct list_head close_lru;
300
2283963f
BF		 301static void unhash_generic_stateid(struct nfs4_stateid *stp)
302{
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
306}
307
308static void free_generic_stateid(struct nfs4_stateid *stp)
309{
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
312}
313
314static void release_lock_stateid(struct nfs4_stateid *stp)
315{
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
319}
320
f044ff83
BF		 321static void unhash_lockowner(struct nfs4_stateowner *sop)
322{
323 struct nfs4_stateid *stp;
324
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
332 }
333}
334
335static void release_lockowner(struct nfs4_stateowner *sop)
336{
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
339}
340
f1d110ca
BF		 341static void
342release_stateid_lockowners(struct nfs4_stateid *open_stp)
343{
344 struct nfs4_stateowner *lock_sop;
345
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 351 release_lockowner(lock_sop);
f1d110ca
BF		 352 }
353}
354
2283963f
BF		 355static void release_open_stateid(struct nfs4_stateid *stp)
356{
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
361}
362
f044ff83 363static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 364{
365 struct nfs4_stateid *stp;
366
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
f044ff83
BF		 369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 371 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
f1d110ca
BF		 375 }
376}
377
f044ff83 378static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 379{
f044ff83 380 unhash_openowner(sop);
f1d110ca
BF		 381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
383}
384
5282fd72
ME		 385static DEFINE_SPINLOCK(sessionid_lock);
386#define SESSION_HASH_SIZE 512
387static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
388
389static inline int
390hash_sessionid(struct nfs4_sessionid *sessionid)
391{
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
393
394 return sid->sequence % SESSION_HASH_SIZE;
395}
396
397static inline void
398dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
399{
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
402}
403
ec6b5d7b
AA		 404static void
405gen_sessionid(struct nfsd4_session *ses)
406{
407 struct nfs4_client *clp = ses->se_client;
408 struct nfsd4_sessionid *sid;
409
410 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
411 sid->clientid = clp->cl_clientid;
412 sid->sequence = current_sessionid++;
413 sid->reserved = 0;
414}
415
416/*
417 * Give the client the number of slots it requests bound by
0c193054 418 * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
ec6b5d7b 419 *
0c193054
AA		 420 * If we run out of reserved DRC memory we should (up to a point) re-negotiate
421 * active sessions and reduce their slot usage to make rooom for new
422 * connections. For now we just fail the create session.
ec6b5d7b
AA		 423 */
424static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
425{
0c193054 426 int mem;
ec6b5d7b 427
5d77ddfb
AA		 428 if (fchan->maxreqs < 1)
429 return nfserr_inval;
430 else if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
431 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
432
0c193054 433 mem = fchan->maxreqs * NFSD_SLOT_CACHE_SIZE;
b101ebbc 434
4bd9b0f4 435 spin_lock(&nfsd_drc_lock);
0c193054
AA		 436 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem)
437 mem = ((nfsd_drc_max_mem - nfsd_drc_mem_used) /
438 NFSD_SLOT_CACHE_SIZE) * NFSD_SLOT_CACHE_SIZE;
439 nfsd_drc_mem_used += mem;
4bd9b0f4 440 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 441
0c193054 442 fchan->maxreqs = mem / NFSD_SLOT_CACHE_SIZE;
b101ebbc
AA		 443 if (fchan->maxreqs == 0)
444 return nfserr_resource;
445 return 0;
ec6b5d7b
AA		 446}
447
448/*
449 * fchan holds the client values on input, and the server values on output
450 */
451static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 452 struct nfsd4_channel_attrs *session_fchan,
453 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 454{
455 int status = 0;
456 __u32 maxcount = svc_max_payload(rqstp);
457
458 /* headerpadsz set to zero in encode routine */
459
460 /* Use the client's max request and max response size if possible */
461 if (fchan->maxreq_sz > maxcount)
462 fchan->maxreq_sz = maxcount;
6c18ba9f 463 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 464
465 if (fchan->maxresp_sz > maxcount)
466 fchan->maxresp_sz = maxcount;
6c18ba9f 467 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 468
0c193054 469 session_fchan->maxresp_cached = NFSD_SLOT_CACHE_SIZE;
6c18ba9f 470 fchan->maxresp_cached = session_fchan->maxresp_cached;
ec6b5d7b
AA		 471
472 /* Use the client's maxops if possible */
473 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
474 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 475 session_fchan->maxops = fchan->maxops;
ec6b5d7b 476
ec6b5d7b
AA		 477 /* FIXME: Error means no more DRC pages so the server should
478 * recover pages from existing sessions. For now fail session
479 * creation.
480 */
481 status = set_forechannel_maxreqs(fchan);
482
6c18ba9f 483 session_fchan->maxreqs = fchan->maxreqs;
ec6b5d7b
AA		 484 return status;
485}
486
487static int
488alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
489 struct nfsd4_create_session *cses)
490{
491 struct nfsd4_session *new, tmp;
492 int idx, status = nfserr_resource, slotsize;
493
494 memset(&tmp, 0, sizeof(tmp));
495
496 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 497 tmp.se_bchannel = cses->back_channel;
498 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
499 &cses->fore_channel);
ec6b5d7b
AA		 500 if (status)
501 goto out;
502
503 /* allocate struct nfsd4_session and slot table in one piece */
6c18ba9f 504 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot);
ec6b5d7b
AA		 505 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
506 if (!new)
507 goto out;
508
509 memcpy(new, &tmp, sizeof(*new));
510
511 new->se_client = clp;
512 gen_sessionid(new);
513 idx = hash_sessionid(&new->se_sessionid);
514 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
515 NFS4_MAX_SESSIONID_LEN);
516
517 new->se_flags = cses->flags;
518 kref_init(&new->se_ref);
519 spin_lock(&sessionid_lock);
520 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
521 list_add(&new->se_perclnt, &clp->cl_sessions);
522 spin_unlock(&sessionid_lock);
523
524 status = nfs_ok;
525out:
526 return status;
527}
528
5282fd72
ME		 529/* caller must hold sessionid_lock */
530static struct nfsd4_session *
531find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
532{
533 struct nfsd4_session *elem;
534 int idx;
535
536 dump_sessionid(__func__, sessionid);
537 idx = hash_sessionid(sessionid);
538 dprintk("%s: idx is %d\n", __func__, idx);
539 /* Search in the appropriate list */
540 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
541 dump_sessionid("list traversal", &elem->se_sessionid);
542 if (!memcmp(elem->se_sessionid.data, sessionid->data,
543 NFS4_MAX_SESSIONID_LEN)) {
544 return elem;
545 }
546 }
547
548 dprintk("%s: session not found\n", __func__);
549 return NULL;
550}
551
552/* caller must hold sessionid_lock */
7116ed6b 553static void
5282fd72 554unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 555{
556 list_del(&ses->se_hash);
557 list_del(&ses->se_perclnt);
5282fd72
ME		 558}
559
560static void
561release_session(struct nfsd4_session *ses)
562{
563 spin_lock(&sessionid_lock);
564 unhash_session(ses);
565 spin_unlock(&sessionid_lock);
7116ed6b
AA		 566 nfsd4_put_session(ses);
567}
568
14778a13
AA		 569static void nfsd4_release_respages(struct page **respages, short resused);
570
7116ed6b
AA		 571void
572free_session(struct kref *kref)
573{
574 struct nfsd4_session *ses;
14778a13 575 int i;
7116ed6b
AA		 576
577 ses = container_of(kref, struct nfsd4_session, se_ref);
6c18ba9f 578 for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
14778a13
AA		 579 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
580 nfsd4_release_respages(e->ce_respages, e->ce_resused);
581 }
be98d1bb 582 spin_lock(&nfsd_drc_lock);
0c193054 583 nfsd_drc_mem_used -= ses->se_fchannel.maxreqs * NFSD_SLOT_CACHE_SIZE;
be98d1bb 584 spin_unlock(&nfsd_drc_lock);
7116ed6b
AA		 585 kfree(ses);
586}
587
1da177e4
LT		 588static inline void
589renew_client(struct nfs4_client *clp)
590{
591 /*
592 * Move client to the end to the LRU list.
593 */
594 dprintk("renewing client (clientid %08x/%08x)\n",
595 clp->cl_clientid.cl_boot,
596 clp->cl_clientid.cl_id);
597 list_move_tail(&clp->cl_lru, &client_lru);
598 clp->cl_time = get_seconds();
599}
600
601/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
602static int
603STALE_CLIENTID(clientid_t *clid)
604{
605 if (clid->cl_boot == boot_time)
606 return 0;
60adfc50
AA		 607 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
608 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 609 return 1;
610}
611
612/*
613 * XXX Should we use a slab cache ?
614 * This type of memory management is somewhat inefficient, but we use it
615 * anyway since SETCLIENTID is not a common operation.
616 */
35bba9a3 617static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 618{
619 struct nfs4_client *clp;
620
35bba9a3
BF		 621 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
622 if (clp == NULL)
623 return NULL;
624 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
625 if (clp->cl_name.data == NULL) {
626 kfree(clp);
627 return NULL;
1da177e4 628 }
35bba9a3
BF		 629 memcpy(clp->cl_name.data, name.data, name.len);
630 clp->cl_name.len = name.len;
1da177e4
LT		 631 return clp;
632}
633
1b1a9b31
BF		 634static void
635shutdown_callback_client(struct nfs4_client *clp)
636{
c237dc03 637 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 638
1b1a9b31 639 if (clnt) {
404ec117
BF		 640 /*
641 * Callback threads take a reference on the client, so there
642 * should be no outstanding callbacks at this point.
643 */
c237dc03 644 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 645 rpc_shutdown_client(clnt);
646 }
c237dc03
BF		 647 if (clp->cl_cb_conn.cb_cred) {
648 put_rpccred(clp->cl_cb_conn.cb_cred);
649 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 650 }
1b1a9b31
BF		 651}
652
1da177e4
LT		 653static inline void
654free_client(struct nfs4_client *clp)
655{
1b1a9b31 656 shutdown_callback_client(clp);
1da177e4
LT		 657 if (clp->cl_cred.cr_group_info)
658 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 659 kfree(clp->cl_principal);
1da177e4
LT		 660 kfree(clp->cl_name.data);
661 kfree(clp);
662}
663
664void
665put_nfs4_client(struct nfs4_client *clp)
666{
667 if (atomic_dec_and_test(&clp->cl_count))
668 free_client(clp);
669}
670
671static void
672expire_client(struct nfs4_client *clp)
673{
674 struct nfs4_stateowner *sop;
675 struct nfs4_delegation *dp;
1da177e4
LT		 676 struct list_head reaplist;
677
678 dprintk("NFSD: expire_client cl_count %d\n",
679 atomic_read(&clp->cl_count));
680
1da177e4
LT		 681 INIT_LIST_HEAD(&reaplist);
682 spin_lock(&recall_lock);
ea1da636
N		 683 while (!list_empty(&clp->cl_delegations)) {
684 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 685 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
686 dp->dl_flock);
ea1da636 687 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 688 list_move(&dp->dl_recall_lru, &reaplist);
689 }
690 spin_unlock(&recall_lock);
691 while (!list_empty(&reaplist)) {
692 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
693 list_del_init(&dp->dl_recall_lru);
694 unhash_delegation(dp);
695 }
696 list_del(&clp->cl_idhash);
697 list_del(&clp->cl_strhash);
698 list_del(&clp->cl_lru);
ea1da636
N		 699 while (!list_empty(&clp->cl_openowners)) {
700 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 701 release_openowner(sop);
1da177e4 702 }
c4bf7868
ME		 703 while (!list_empty(&clp->cl_sessions)) {
704 struct nfsd4_session *ses;
705 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
706 se_perclnt);
707 release_session(ses);
708 }
1da177e4
LT		 709 put_nfs4_client(clp);
710}
711
35bba9a3
BF		 712static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
713{
1da177e4
LT		 714 struct nfs4_client *clp;
715
35bba9a3
BF		 716 clp = alloc_client(name);
717 if (clp == NULL)
718 return NULL;
a55370a3 719 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 720 atomic_set(&clp->cl_count, 1);
c237dc03 721 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 722 INIT_LIST_HEAD(&clp->cl_idhash);
723 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 724 INIT_LIST_HEAD(&clp->cl_openowners);
725 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 726 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 727 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 728 return clp;
729}
730
35bba9a3
BF		 731static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
732{
733 memcpy(target->cl_verifier.data, source->data,
734 sizeof(target->cl_verifier.data));
1da177e4
LT		 735}
736
35bba9a3
BF		 737static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
738{
1da177e4
LT		 739 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
740 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
741}
742
35bba9a3
BF		 743static void copy_cred(struct svc_cred *target, struct svc_cred *source)
744{
1da177e4
LT		 745 target->cr_uid = source->cr_uid;
746 target->cr_gid = source->cr_gid;
747 target->cr_group_info = source->cr_group_info;
748 get_group_info(target->cr_group_info);
749}
750
35bba9a3 751static int same_name(const char *n1, const char *n2)
599e0a22 752{
a55370a3 753 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 754}
755
756static int
599e0a22
BF		 757same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
758{
759 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 760}
761
762static int
599e0a22
BF		 763same_clid(clientid_t *cl1, clientid_t *cl2)
764{
765 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 766}
767
768/* XXX what about NGROUP */
769static int
599e0a22
BF		 770same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
771{
772 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 773}
774
5ec7b46c
BF		 775static void gen_clid(struct nfs4_client *clp)
776{
777 static u32 current_clientid = 1;
778
1da177e4
LT		 779 clp->cl_clientid.cl_boot = boot_time;
780 clp->cl_clientid.cl_id = current_clientid++;
781}
782
deda2faa
BF		 783static void gen_confirm(struct nfs4_client *clp)
784{
785 static u32 i;
786 u32 *p;
1da177e4 787
1da177e4 788 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 789 *p++ = get_seconds();
790 *p++ = i++;
1da177e4
LT		 791}
792
35bba9a3
BF		 793static int check_name(struct xdr_netobj name)
794{
1da177e4
LT		 795 if (name.len == 0)
796 return 0;
797 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 798 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 799 return 0;
800 }
801 return 1;
802}
803
fd39ca9a 804static void
1da177e4
LT		 805add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
806{
807 unsigned int idhashval;
808
809 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
810 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
811 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
812 list_add_tail(&clp->cl_lru, &client_lru);
813 clp->cl_time = get_seconds();
814}
815
fd39ca9a 816static void
1da177e4
LT		 817move_to_confirmed(struct nfs4_client *clp)
818{
819 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
820 unsigned int strhashval;
821
822 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
823 list_del_init(&clp->cl_strhash);
f116629d 824 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 825 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 826 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
827 renew_client(clp);
828}
829
830static struct nfs4_client *
831find_confirmed_client(clientid_t *clid)
832{
833 struct nfs4_client *clp;
834 unsigned int idhashval = clientid_hashval(clid->cl_id);
835
836 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 837 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 838 return clp;
839 }
840 return NULL;
841}
842
843static struct nfs4_client *
844find_unconfirmed_client(clientid_t *clid)
845{
846 struct nfs4_client *clp;
847 unsigned int idhashval = clientid_hashval(clid->cl_id);
848
849 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 850 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 851 return clp;
852 }
853 return NULL;
854}
855
a1bcecd2
AA		 856/*
857 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
858 * parameter. Matching is based on the fact the at least one of the
859 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
860 *
861 * FIXME: we need to unify the clientid namespaces for nfsv4.x
862 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
863 * and SET_CLIENTID{,_CONFIRM}
864 */
865static inline int
866match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
867{
868 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
869 return use_exchange_id == has_exchange_flags;
870}
871
28ce6054 872static struct nfs4_client *
a1bcecd2
AA		 873find_confirmed_client_by_str(const char *dname, unsigned int hashval,
874 bool use_exchange_id)
28ce6054
N		 875{
876 struct nfs4_client *clp;
877
878 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 879 if (same_name(clp->cl_recdir, dname) &&
880 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 881 return clp;
882 }
883 return NULL;
884}
885
886static struct nfs4_client *
a1bcecd2
AA		 887find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
888 bool use_exchange_id)
28ce6054
N		 889{
890 struct nfs4_client *clp;
891
892 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 893 if (same_name(clp->cl_recdir, dname) &&
894 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 895 return clp;
896 }
897 return NULL;
898}
899
fd39ca9a 900static void
fbf4665f 901gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 902{
c237dc03 903 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 904 unsigned short expected_family;
905
906 /* Currently, we only support tcp and tcp6 for the callback channel */
907 if (se->se_callback_netid_len == 3 &&
908 !memcmp(se->se_callback_netid_val, "tcp", 3))
909 expected_family = AF_INET;
910 else if (se->se_callback_netid_len == 4 &&
911 !memcmp(se->se_callback_netid_val, "tcp6", 4))
912 expected_family = AF_INET6;
913 else
1da177e4
LT		 914 goto out_err;
915
aa9a4ec7
JL		 916 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
917 se->se_callback_addr_len,
918 (struct sockaddr *) &cb->cb_addr,
919 sizeof(cb->cb_addr));
920
7077ecba 921 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 922 goto out_err;
aa9a4ec7 923
fbf4665f
JL		 924 if (cb->cb_addr.ss_family == AF_INET6)
925 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
926
ab52ae6d 927 cb->cb_minorversion = 0;
1da177e4
LT		 928 cb->cb_prog = se->se_callback_prog;
929 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 930 return;
931out_err:
aa9a4ec7
JL		 932 cb->cb_addr.ss_family = AF_UNSPEC;
933 cb->cb_addrlen = 0;
849823c5 934 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 935 "will not receive delegations\n",
936 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
937
1da177e4
LT		 938 return;
939}
940
074fe897
AA		 941void
942nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
943{
944 struct nfsd4_compoundres *resp = rqstp->rq_resp;
945
946 resp->cstate.statp = statp;
947}
948
949/*
950 * Dereference the result pages.
951 */
952static void
953nfsd4_release_respages(struct page **respages, short resused)
954{
955 int i;
956
957 dprintk("--> %s\n", __func__);
958 for (i = 0; i < resused; i++) {
959 if (!respages[i])
960 continue;
961 put_page(respages[i]);
962 respages[i] = NULL;
963 }
964}
965
966static void
967nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
968{
969 int i;
970
971 for (i = 0; i < count; i++) {
972 topages[i] = frompages[i];
973 if (!topages[i])
974 continue;
975 get_page(topages[i]);
976 }
977}
978
979/*
980 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
981 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
982 * length of the XDR response is less than se_fmaxresp_cached
983 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
984 * of the reply (e.g. readdir).
985 *
986 * Store the base and length of the rq_req.head[0] page
987 * of the NFSv4.1 data, just past the rpc header.
988 */
989void
990nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
991{
992 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
993 struct svc_rqst *rqstp = resp->rqstp;
074fe897
AA		 994 struct kvec *resv = &rqstp->rq_res.head[0];
995
996 dprintk("--> %s entry %p\n", __func__, entry);
997
074fe897 998 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 999 entry->ce_opcnt = resp->opcnt;
1000 entry->ce_status = resp->cstate.status;
1001
1002 /*
1003 * Don't need a page to cache just the sequence operation - the slot
1004 * does this for us!
1005 */
1006
1007 if (nfsd4_not_cached(resp)) {
1008 entry->ce_resused = 0;
1009 entry->ce_rpchdrlen = 0;
1010 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1011 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1012 return;
1013 }
074fe897
AA		 1014 entry->ce_resused = rqstp->rq_resused;
1015 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1016 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1017 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1018 entry->ce_resused);
074fe897
AA		 1019 entry->ce_datav.iov_base = resp->cstate.statp;
1020 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1021 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1022 /* Current request rpc header length*/
1023 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1024 (char *)page_address(rqstp->rq_respages[0]);
1025}
1026
1027/*
1028 * We keep the rpc header, but take the nfs reply from the replycache.
1029 */
1030static int
1031nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1032 struct nfsd4_cache_entry *entry)
1033{
1034 struct svc_rqst *rqstp = resp->rqstp;
1035 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1036 int len;
1037
1038 /* Current request rpc header length*/
1039 len = (char *)resp->cstate.statp -
1040 (char *)page_address(rqstp->rq_respages[0]);
1041 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1042 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1043 entry->ce_datav.iov_len);
1044 return 0;
1045 }
1046 /* copy the cached reply nfsd data past the current rpc header */
1047 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1048 entry->ce_datav.iov_len);
1049 resv->iov_len = len + entry->ce_datav.iov_len;
1050 return 1;
1051}
1052
abfabf8c
AA		 1053/*
1054 * Encode the replay sequence operation from the slot values.
1055 * If cachethis is FALSE encode the uncached rep error on the next
1056 * operation which sets resp->p and increments resp->opcnt for
1057 * nfs4svc_encode_compoundres.
1058 *
1059 */
1060static __be32
1061nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1062 struct nfsd4_compoundres *resp)
1063{
1064 struct nfsd4_op *op;
1065 struct nfsd4_slot *slot = resp->cstate.slot;
1066
1067 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1068 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
1069
1070 /* Encode the replayed sequence operation */
1071 op = &args->ops[resp->opcnt - 1];
1072 nfsd4_encode_operation(resp, op);
1073
1074 /* Return nfserr_retry_uncached_rep in next operation. */
1075 if (args->opcnt > 1 && slot->sl_cache_entry.ce_cachethis == 0) {
1076 op = &args->ops[resp->opcnt++];
1077 op->status = nfserr_retry_uncached_rep;
1078 nfsd4_encode_operation(resp, op);
1079 }
1080 return op->status;
1081}
1082
074fe897
AA		 1083/*
1084 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1085 * cached page. Replace any futher replay pages from the cache.
1086 */
1087__be32
bf864a31
AA		 1088nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1089 struct nfsd4_sequence *seq)
074fe897
AA		 1090{
1091 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1092 __be32 status;
1093
1094 dprintk("--> %s entry %p\n", __func__, entry);
1095
bf864a31
AA		 1096 /*
1097 * If this is just the sequence operation, we did not keep
1098 * a page in the cache entry because we can just use the
1099 * slot info stored in struct nfsd4_sequence that was checked
1100 * against the slot in nfsd4_sequence().
1101 *
1102 * This occurs when seq->cachethis is FALSE, or when the client
1103 * session inactivity timer fires and a solo sequence operation
1104 * is sent (lease renewal).
1105 */
abfabf8c
AA		 1106 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
1107
1108 /* Either returns 0 or nfserr_retry_uncached */
1109 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1110 if (status == nfserr_retry_uncached_rep)
1111 return status;
074fe897
AA		 1112
1113 if (!nfsd41_copy_replay_data(resp, entry)) {
1114 /*
1115 * Not enough room to use the replay rpc header, send the
1116 * cached header. Release all the allocated result pages.
1117 */
1118 svc_free_res_pages(resp->rqstp);
1119 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1120 entry->ce_resused);
1121 } else {
1122 /* Release all but the first allocated result page */
1123
1124 resp->rqstp->rq_resused--;
1125 svc_free_res_pages(resp->rqstp);
1126
1127 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1128 &entry->ce_respages[1],
1129 entry->ce_resused - 1);
1130 }
1131
1132 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1133 resp->opcnt = entry->ce_opcnt;
1134 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1135 status = entry->ce_status;
1136
1137 return status;
1138}
1139
0733d213
AA		 1140/*
1141 * Set the exchange_id flags returned by the server.
1142 */
1143static void
1144nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1145{
1146 /* pNFS is not supported */
1147 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1148
1149 /* Referrals are supported, Migration is not. */
1150 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1151
1152 /* set the wire flags to return to client. */
1153 clid->flags = new->cl_exchange_flags;
1154}
1155
069b6ad4
AA		 1156__be32
1157nfsd4_exchange_id(struct svc_rqst *rqstp,
1158 struct nfsd4_compound_state *cstate,
1159 struct nfsd4_exchange_id *exid)
1160{
0733d213
AA		 1161 struct nfs4_client *unconf, *conf, *new;
1162 int status;
1163 unsigned int strhashval;
1164 char dname[HEXDIR_LEN];
363168b4 1165 char addr_str[INET6_ADDRSTRLEN];
0733d213 1166 nfs4_verifier verf = exid->verifier;
363168b4 1167 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1168
363168b4 1169 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1170 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1171 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1172 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1173 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1174
1175 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1176 return nfserr_inval;
1177
1178 /* Currently only support SP4_NONE */
1179 switch (exid->spa_how) {
1180 case SP4_NONE:
1181 break;
1182 case SP4_SSV:
1183 return nfserr_encr_alg_unsupp;
1184 default:
1185 BUG(); /* checked by xdr code */
1186 case SP4_MACH_CRED:
1187 return nfserr_serverfault; /* no excuse :-/ */
1188 }
1189
1190 status = nfs4_make_rec_clidname(dname, &exid->clname);
1191
1192 if (status)
1193 goto error;
1194
1195 strhashval = clientstr_hashval(dname);
1196
1197 nfs4_lock_state();
1198 status = nfs_ok;
1199
a1bcecd2 1200 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1201 if (conf) {
1202 if (!same_verf(&verf, &conf->cl_verifier)) {
1203 /* 18.35.4 case 8 */
1204 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1205 status = nfserr_not_same;
1206 goto out;
1207 }
1208 /* Client reboot: destroy old state */
1209 expire_client(conf);
1210 goto out_new;
1211 }
1212 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1213 /* 18.35.4 case 9 */
1214 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1215 status = nfserr_perm;
1216 goto out;
1217 }
1218 expire_client(conf);
1219 goto out_new;
1220 }
0733d213
AA		 1221 /*
1222 * Set bit when the owner id and verifier map to an already
1223 * confirmed client id (18.35.3).
1224 */
1225 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1226
1227 /*
1228 * Falling into 18.35.4 case 2, possible router replay.
1229 * Leave confirmed record intact and return same result.
1230 */
1231 copy_verf(conf, &verf);
1232 new = conf;
1233 goto out_copy;
6ddbbbfe
MS		 1234 }
1235
1236 /* 18.35.4 case 7 */
1237 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1238 status = nfserr_noent;
1239 goto out;
0733d213
AA		 1240 }
1241
a1bcecd2 1242 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1243 if (unconf) {
1244 /*
1245 * Possible retry or client restart. Per 18.35.4 case 4,
1246 * a new unconfirmed record should be generated regardless
1247 * of whether any properties have changed.
1248 */
1249 expire_client(unconf);
1250 }
1251
1252out_new:
1253 /* Normal case */
1254 new = create_client(exid->clname, dname);
1255 if (new == NULL) {
1256 status = nfserr_resource;
1257 goto out;
1258 }
1259
1260 copy_verf(new, &verf);
1261 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1262 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1263 gen_clid(new);
1264 gen_confirm(new);
1265 add_to_unconfirmed(new, strhashval);
1266out_copy:
1267 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1268 exid->clientid.cl_id = new->cl_clientid.cl_id;
1269
38eb76a5 1270 exid->seqid = 1;
0733d213
AA		 1271 nfsd4_set_ex_flags(new, exid);
1272
1273 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1274 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1275 status = nfs_ok;
1276
1277out:
1278 nfs4_unlock_state();
1279error:
1280 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1281 return status;
069b6ad4
AA		 1282}
1283
b85d4c01 1284static int
88e588d5 1285check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1286{
88e588d5
AA		 1287 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1288 slot_seqid);
b85d4c01
BH		 1289
1290 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1291 if (slot_inuse) {
1292 if (seqid == slot_seqid)
b85d4c01
BH		 1293 return nfserr_jukebox;
1294 else
1295 return nfserr_seq_misordered;
1296 }
1297 /* Normal */
88e588d5 1298 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1299 return nfs_ok;
1300 /* Replay */
88e588d5 1301 if (seqid == slot_seqid)
b85d4c01
BH		 1302 return nfserr_replay_cache;
1303 /* Wraparound */
88e588d5 1304 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1305 return nfs_ok;
1306 /* Misordered replay or misordered new request */
1307 return nfserr_seq_misordered;
1308}
1309
49557cc7
AA		 1310/*
1311 * Cache the create session result into the create session single DRC
1312 * slot cache by saving the xdr structure. sl_seqid has been set.
1313 * Do this for solo or embedded create session operations.
1314 */
1315static void
1316nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1317 struct nfsd4_clid_slot *slot, int nfserr)
1318{
1319 slot->sl_status = nfserr;
1320 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1321}
1322
1323static __be32
1324nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1325 struct nfsd4_clid_slot *slot)
1326{
1327 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1328 return slot->sl_status;
1329}
1330
069b6ad4
AA		 1331__be32
1332nfsd4_create_session(struct svc_rqst *rqstp,
1333 struct nfsd4_compound_state *cstate,
1334 struct nfsd4_create_session *cr_ses)
1335{
363168b4 1336 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1337 struct nfs4_client *conf, *unconf;
49557cc7 1338 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1339 int status = 0;
1340
1341 nfs4_lock_state();
1342 unconf = find_unconfirmed_client(&cr_ses->clientid);
1343 conf = find_confirmed_client(&cr_ses->clientid);
1344
1345 if (conf) {
49557cc7
AA		 1346 cs_slot = &conf->cl_cs_slot;
1347 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1348 if (status == nfserr_replay_cache) {
ec6b5d7b 1349 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1350 cs_slot->sl_seqid);
38eb76a5 1351 /* Return the cached reply status */
49557cc7 1352 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1353 goto out;
49557cc7 1354 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1355 status = nfserr_seq_misordered;
1356 dprintk("Sequence misordered!\n");
1357 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1358 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1359 goto out;
1360 }
49557cc7 1361 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1362 } else if (unconf) {
1363 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1364 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1365 status = nfserr_clid_inuse;
1366 goto out;
1367 }
1368
49557cc7
AA		 1369 cs_slot = &unconf->cl_cs_slot;
1370 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1371 if (status) {
1372 /* an unconfirmed replay returns misordered */
ec6b5d7b 1373 status = nfserr_seq_misordered;
49557cc7 1374 goto out_cache;
ec6b5d7b
AA		 1375 }
1376
49557cc7 1377 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1378 move_to_confirmed(unconf);
1379
1380 /*
1381 * We do not support RDMA or persistent sessions
1382 */
1383 cr_ses->flags &= ~SESSION4_PERSIST;
1384 cr_ses->flags &= ~SESSION4_RDMA;
1385
1386 conf = unconf;
1387 } else {
1388 status = nfserr_stale_clientid;
1389 goto out;
1390 }
1391
1392 status = alloc_init_session(rqstp, conf, cr_ses);
1393 if (status)
1394 goto out;
1395
ec6b5d7b
AA		 1396 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1397 NFS4_MAX_SESSIONID_LEN);
49557cc7 1398 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1399
49557cc7
AA		 1400out_cache:
1401 /* cache solo and embedded create sessions under the state lock */
1402 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1403out:
1404 nfs4_unlock_state();
1405 dprintk("%s returns %d\n", __func__, ntohl(status));
1406 return status;
069b6ad4
AA		 1407}
1408
1409__be32
1410nfsd4_destroy_session(struct svc_rqst *r,
1411 struct nfsd4_compound_state *cstate,
1412 struct nfsd4_destroy_session *sessionid)
1413{
e10e0cfc
BH		 1414 struct nfsd4_session *ses;
1415 u32 status = nfserr_badsession;
1416
1417 /* Notes:
1418 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1419 * - Should we return nfserr_back_chan_busy if waiting for
1420 * callbacks on to-be-destroyed session?
1421 * - Do we need to clear any callback info from previous session?
1422 */
1423
1424 dump_sessionid(__func__, &sessionid->sessionid);
1425 spin_lock(&sessionid_lock);
1426 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1427 if (!ses) {
1428 spin_unlock(&sessionid_lock);
1429 goto out;
1430 }
1431
1432 unhash_session(ses);
1433 spin_unlock(&sessionid_lock);
1434
1435 /* wait for callbacks */
1436 shutdown_callback_client(ses->se_client);
1437 nfsd4_put_session(ses);
1438 status = nfs_ok;
1439out:
1440 dprintk("%s returns %d\n", __func__, ntohl(status));
1441 return status;
069b6ad4
AA		 1442}
1443
1444__be32
b85d4c01 1445nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1446 struct nfsd4_compound_state *cstate,
1447 struct nfsd4_sequence *seq)
1448{
f9bb94c4 1449 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1450 struct nfsd4_session *session;
1451 struct nfsd4_slot *slot;
1452 int status;
1453
f9bb94c4
AA		 1454 if (resp->opcnt != 1)
1455 return nfserr_sequence_pos;
1456
b85d4c01
BH		 1457 spin_lock(&sessionid_lock);
1458 status = nfserr_badsession;
1459 session = find_in_sessionid_hashtbl(&seq->sessionid);
1460 if (!session)
1461 goto out;
1462
1463 status = nfserr_badslot;
6c18ba9f 1464 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1465 goto out;
1466
1467 slot = &session->se_slots[seq->slotid];
1468 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1469
88e588d5 1470 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1471 if (status == nfserr_replay_cache) {
1472 cstate->slot = slot;
1473 cstate->session = session;
da3846a2 1474 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1475 * for nfsd4_svc_encode_compoundres processing */
1476 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1477 cstate->status = nfserr_replay_cache;
aaf84eb9 1478 goto out;
b85d4c01
BH		 1479 }
1480 if (status)
1481 goto out;
1482
1483 /* Success! bump slot seqid */
1484 slot->sl_inuse = true;
1485 slot->sl_seqid = seq->seqid;
bf864a31 1486 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
b85d4c01
BH		 1487
1488 cstate->slot = slot;
1489 cstate->session = session;
1490
aaf84eb9 1491 /* Hold a session reference until done processing the compound:
b85d4c01
BH		 1492 * nfsd4_put_session called only if the cstate slot is set.
1493 */
b85d4c01
BH		 1494 nfsd4_get_session(session);
1495out:
1496 spin_unlock(&sessionid_lock);
aaf84eb9
BH		 1497 /* Renew the clientid on success and on replay */
1498 if (cstate->session) {
1499 nfs4_lock_state();
1500 renew_client(session->se_client);
1501 nfs4_unlock_state();
1502 }
b85d4c01
BH		 1503 dprintk("%s: return %d\n", __func__, ntohl(status));
1504 return status;
069b6ad4
AA		 1505}
1506
b37ad28b 1507__be32
b591480b
BF		 1508nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1509 struct nfsd4_setclientid *setclid)
1da177e4 1510{
363168b4 1511 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1512 struct xdr_netobj clname = {
1513 .len = setclid->se_namelen,
1514 .data = setclid->se_name,
1515 };
1516 nfs4_verifier clverifier = setclid->se_verf;
1517 unsigned int strhashval;
28ce6054 1518 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1519 __be32 status;
68e76ad0 1520 char *princ;
a55370a3 1521 char dname[HEXDIR_LEN];
1da177e4 1522
1da177e4 1523 if (!check_name(clname))
73aea4ec 1524 return nfserr_inval;
1da177e4 1525
a55370a3
N		 1526 status = nfs4_make_rec_clidname(dname, &clname);
1527 if (status)
73aea4ec 1528 return status;
a55370a3 1529
1da177e4
LT		 1530 /*
1531 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1532 * We get here on a DRC miss.
1533 */
1534
a55370a3 1535 strhashval = clientstr_hashval(dname);
1da177e4 1536
1da177e4 1537 nfs4_lock_state();
a1bcecd2 1538 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1539 if (conf) {
a186e767 1540 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1541 status = nfserr_clid_inuse;
026722c2 1542 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1543 char addr_str[INET6_ADDRSTRLEN];
1544 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1545 sizeof(addr_str));
1546 dprintk("NFSD: setclientid: string in use by client "
1547 "at %s\n", addr_str);
1da177e4
LT		 1548 goto out;
1549 }
1da177e4 1550 }
a186e767
BF		 1551 /*
1552 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1553 * has a description of SETCLIENTID request processing consisting
1554 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1555 */
a1bcecd2 1556 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1557 status = nfserr_resource;
1558 if (!conf) {
a186e767
BF		 1559 /*
1560 * RFC 3530 14.2.33 CASE 4:
1561 * placed first, because it is the normal case
1da177e4
LT		 1562 */
1563 if (unconf)
1564 expire_client(unconf);
a55370a3
N		 1565 new = create_client(clname, dname);
1566 if (new == NULL)
1da177e4 1567 goto out;
1da177e4 1568 gen_clid(new);
599e0a22 1569 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1570 /*
a186e767
BF		 1571 * RFC 3530 14.2.33 CASE 1:
1572 * probable callback update
1da177e4 1573 */
31f4a6c1
N		 1574 if (unconf) {
1575 /* Note this is removing unconfirmed {*x***},
1576 * which is stronger than RFC recommended {vxc**}.
1577 * This has the advantage that there is at most
1578 * one {*x***} in either list at any time.
1579 */
1580 expire_client(unconf);
1da177e4 1581 }
a55370a3
N		 1582 new = create_client(clname, dname);
1583 if (new == NULL)
1da177e4 1584 goto out;
1da177e4 1585 copy_clid(new, conf);
1da177e4
LT		 1586 } else if (!unconf) {
1587 /*
a186e767
BF		 1588 * RFC 3530 14.2.33 CASE 2:
1589 * probable client reboot; state will be removed if
1590 * confirmed.
1da177e4 1591 */
a55370a3
N		 1592 new = create_client(clname, dname);
1593 if (new == NULL)
1da177e4 1594 goto out;
1da177e4 1595 gen_clid(new);
49ba8781 1596 } else {
a186e767
BF		 1597 /*
1598 * RFC 3530 14.2.33 CASE 3:
1599 * probable client reboot; state will be removed if
1600 * confirmed.
1da177e4
LT		 1601 */
1602 expire_client(unconf);
a55370a3
N		 1603 new = create_client(clname, dname);
1604 if (new == NULL)
1da177e4 1605 goto out;
1da177e4 1606 gen_clid(new);
1da177e4 1607 }
c175b83c 1608 copy_verf(new, &clverifier);
363168b4 1609 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1610 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1611 princ = svc_gss_principal(rqstp);
1612 if (princ) {
1613 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1614 if (new->cl_principal == NULL) {
1615 free_client(new);
1616 goto out;
1617 }
1618 }
c175b83c
BF		 1619 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1620 gen_confirm(new);
fbf4665f 1621 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1622 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1623 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1624 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1625 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1626 status = nfs_ok;
1627out:
1628 nfs4_unlock_state();
1629 return status;
1630}
1631
1632
1633/*
a186e767
BF		 1634 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1635 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1636 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1637 */
b37ad28b 1638__be32
b591480b
BF		 1639nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1640 struct nfsd4_compound_state *cstate,
1641 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1642{
363168b4 1643 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1644 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1645 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1646 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1647 __be32 status;
1da177e4
LT		 1648
1649 if (STALE_CLIENTID(clid))
1650 return nfserr_stale_clientid;
1651 /*
1652 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1653 * We get here on a DRC miss.
1654 */
1655
1656 nfs4_lock_state();
21ab45a4
N		 1657
1658 conf = find_confirmed_client(clid);
1659 unconf = find_unconfirmed_client(clid);
1660
1661 status = nfserr_clid_inuse;
363168b4 1662 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1663 goto out;
363168b4 1664 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1665 goto out;
1666
a186e767
BF		 1667 /*
1668 * section 14.2.34 of RFC 3530 has a description of
1669 * SETCLIENTID_CONFIRM request processing consisting
1670 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1671 */
366e0c1d 1672 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1673 /*
1674 * RFC 3530 14.2.34 CASE 1:
1675 * callback update
1676 */
599e0a22 1677 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1678 status = nfserr_clid_inuse;
1679 else {
1a69c179
N		 1680 /* XXX: We just turn off callbacks until we can handle
1681 * change request correctly. */
c237dc03 1682 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1683 expire_client(unconf);
1da177e4 1684 status = nfs_ok;
1a69c179 1685
1da177e4 1686 }
f3aba4e5 1687 } else if (conf && !unconf) {
a186e767
BF		 1688 /*
1689 * RFC 3530 14.2.34 CASE 2:
1690 * probable retransmitted request; play it safe and
1691 * do nothing.
7c79f737 1692 */
599e0a22 1693 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1694 status = nfserr_clid_inuse;
21ab45a4 1695 else
1da177e4 1696 status = nfs_ok;
7c79f737 1697 } else if (!conf && unconf
599e0a22 1698 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1699 /*
1700 * RFC 3530 14.2.34 CASE 3:
1701 * Normal case; new or rebooted client:
7c79f737 1702 */
599e0a22 1703 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1704 status = nfserr_clid_inuse;
1705 } else {
1a69c179
N		 1706 unsigned int hash =
1707 clientstr_hashval(unconf->cl_recdir);
1708 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1709 hash, false);
1a69c179 1710 if (conf) {
c7b9a459 1711 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1712 expire_client(conf);
1713 }
1da177e4 1714 move_to_confirmed(unconf);
21ab45a4 1715 conf = unconf;
46f8a64b 1716 nfsd4_probe_callback(conf);
1a69c179 1717 status = nfs_ok;
1da177e4 1718 }
599e0a22
BF		 1719 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1720 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1721 &confirm)))) {
a186e767
BF		 1722 /*
1723 * RFC 3530 14.2.34 CASE 4:
1724 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1725 */
1da177e4 1726 status = nfserr_stale_clientid;
7c79f737 1727 } else {
08e8987c
N		 1728 /* check that we have hit one of the cases...*/
1729 status = nfserr_clid_inuse;
1730 }
1da177e4 1731out:
1da177e4
LT		 1732 nfs4_unlock_state();
1733 return status;
1734}
1735
1da177e4
LT		 1736/* OPEN Share state helper functions */
1737static inline struct nfs4_file *
1738alloc_init_file(struct inode *ino)
1739{
1740 struct nfs4_file *fp;
1741 unsigned int hashval = file_hashval(ino);
1742
e60d4398
N		 1743 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1744 if (fp) {
8b671b80 1745 atomic_set(&fp->fi_ref, 1);
1da177e4 1746 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1747 INIT_LIST_HEAD(&fp->fi_stateids);
1748 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1749 spin_lock(&recall_lock);
1da177e4 1750 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1751 spin_unlock(&recall_lock);
1da177e4
LT		 1752 fp->fi_inode = igrab(ino);
1753 fp->fi_id = current_fileid++;
47f9940c 1754 fp->fi_had_conflict = false;
1da177e4
LT		 1755 return fp;
1756 }
1757 return NULL;
1758}
1759
e60d4398 1760static void
e18b890b 1761nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1762{
e60d4398
N		 1763 if (*slab == NULL)
1764 return;
1a1d92c1 1765 kmem_cache_destroy(*slab);
e60d4398 1766 *slab = NULL;
1da177e4
LT		 1767}
1768
e8ff2a84 1769void
1da177e4
LT		 1770nfsd4_free_slabs(void)
1771{
e60d4398
N		 1772 nfsd4_free_slab(&stateowner_slab);
1773 nfsd4_free_slab(&file_slab);
5ac049ac 1774 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1775 nfsd4_free_slab(&deleg_slab);
e60d4398 1776}
1da177e4 1777
e60d4398
N		 1778static int
1779nfsd4_init_slabs(void)
1780{
1781 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1782 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1783 if (stateowner_slab == NULL)
1784 goto out_nomem;
1785 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1786 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1787 if (file_slab == NULL)
1788 goto out_nomem;
5ac049ac 1789 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1790 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1791 if (stateid_slab == NULL)
1792 goto out_nomem;
5b2d21c1 1793 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1794 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1795 if (deleg_slab == NULL)
1796 goto out_nomem;
e60d4398
N		 1797 return 0;
1798out_nomem:
1799 nfsd4_free_slabs();
1800 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1801 return -ENOMEM;
1da177e4
LT		 1802}
1803
1804void
1805nfs4_free_stateowner(struct kref *kref)
1806{
1807 struct nfs4_stateowner *sop =
1808 container_of(kref, struct nfs4_stateowner, so_ref);
1809 kfree(sop->so_owner.data);
1810 kmem_cache_free(stateowner_slab, sop);
1811}
1812
1813static inline struct nfs4_stateowner *
1814alloc_stateowner(struct xdr_netobj *owner)
1815{
1816 struct nfs4_stateowner *sop;
1817
1818 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1819 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1820 memcpy(sop->so_owner.data, owner->data, owner->len);
1821 sop->so_owner.len = owner->len;
1822 kref_init(&sop->so_ref);
1823 return sop;
1824 }
1825 kmem_cache_free(stateowner_slab, sop);
1826 }
1827 return NULL;
1828}
1829
1830static struct nfs4_stateowner *
1831alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1832 struct nfs4_stateowner *sop;
1833 struct nfs4_replay *rp;
1834 unsigned int idhashval;
1835
1836 if (!(sop = alloc_stateowner(&open->op_owner)))
1837 return NULL;
1838 idhashval = ownerid_hashval(current_ownerid);
1839 INIT_LIST_HEAD(&sop->so_idhash);
1840 INIT_LIST_HEAD(&sop->so_strhash);
1841 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1842 INIT_LIST_HEAD(&sop->so_stateids);
1843 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1844 INIT_LIST_HEAD(&sop->so_close_lru);
1845 sop->so_time = 0;
1846 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1847 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1848 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1849 sop->so_is_open_owner = 1;
1850 sop->so_id = current_ownerid++;
1851 sop->so_client = clp;
1852 sop->so_seqid = open->op_seqid;
1853 sop->so_confirmed = 0;
1854 rp = &sop->so_replay;
de1ae286 1855 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1856 rp->rp_buflen = 0;
1857 rp->rp_buf = rp->rp_ibuf;
1858 return sop;
1859}
1860
1da177e4
LT		 1861static inline void
1862init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1863 struct nfs4_stateowner *sop = open->op_stateowner;
1864 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1865
1866 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1867 INIT_LIST_HEAD(&stp->st_perstateowner);
1868 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1869 INIT_LIST_HEAD(&stp->st_perfile);
1870 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1871 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1872 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1873 stp->st_stateowner = sop;
13cd2184 1874 get_nfs4_file(fp);
1da177e4 1875 stp->st_file = fp;
78155ed7 1876 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1877 stp->st_stateid.si_stateownerid = sop->so_id;
1878 stp->st_stateid.si_fileid = fp->fi_id;
1879 stp->st_stateid.si_generation = 0;
1880 stp->st_access_bmap = 0;
1881 stp->st_deny_bmap = 0;
84459a11
AA		 1882 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1883 &stp->st_access_bmap);
1da177e4 1884 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1885 stp->st_openstp = NULL;
1da177e4
LT		 1886}
1887
fd39ca9a 1888static void
1da177e4
LT		 1889move_to_close_lru(struct nfs4_stateowner *sop)
1890{
1891 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1892
358dd55a 1893 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1894 sop->so_time = get_seconds();
1895}
1896
1da177e4 1897static int
599e0a22
BF		 1898same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1899 clientid_t *clid)
1900{
1901 return (sop->so_owner.len == owner->len) &&
1902 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1903 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1904}
1905
1906static struct nfs4_stateowner *
1907find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1908{
1909 struct nfs4_stateowner *so = NULL;
1910
1911 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1912 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1913 return so;
1914 }
1915 return NULL;
1916}
1917
1918/* search file_hashtbl[] for file */
1919static struct nfs4_file *
1920find_file(struct inode *ino)
1921{
1922 unsigned int hashval = file_hashval(ino);
1923 struct nfs4_file *fp;
1924
8b671b80 1925 spin_lock(&recall_lock);
1da177e4 1926 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1927 if (fp->fi_inode == ino) {
1928 get_nfs4_file(fp);
8b671b80 1929 spin_unlock(&recall_lock);
1da177e4 1930 return fp;
13cd2184 1931 }
1da177e4 1932 }
8b671b80 1933 spin_unlock(&recall_lock);
1da177e4
LT		 1934 return NULL;
1935}
1936
d87a8ade 1937static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1938{
d87a8ade 1939 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1940 return 0;
d87a8ade
AA		 1941 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1942 return 0;
1943 x &= ~NFS4_SHARE_ACCESS_MASK;
1944 if (minorversion && x) {
1945 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1946 return 0;
1947 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1948 return 0;
1949 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1950 }
1951 if (x)
8838dc43
BF		 1952 return 0;
1953 return 1;
ba5a6a19
BF		 1954}
1955
8838dc43 1956static inline int deny_valid(u32 x)
ba5a6a19 1957{
8838dc43
BF		 1958 /* Note: unlike access bits, deny bits may be zero. */
1959 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1960}
1da177e4 1961
4f83aa30
BF		 1962/*
1963 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1964 * st_{access,deny}_bmap field of the stateid, in order to track not
1965 * only what share bits are currently in force, but also what
1966 * combinations of share bits previous opens have used. This allows us
1967 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1968 * return an error if the client attempt to downgrade to a combination
1969 * of share bits not explicable by closing some of its previous opens.
1970 *
1971 * XXX: This enforcement is actually incomplete, since we don't keep
1972 * track of access/deny bit combinations; so, e.g., we allow:
1973 *
1974 * OPEN allow read, deny write
1975 * OPEN allow both, deny none
1976 * DOWNGRADE allow read, deny none
1977 *
1978 * which we should reject.
1979 */
fd39ca9a 1980static void
1da177e4
LT		 1981set_access(unsigned int *access, unsigned long bmap) {
1982 int i;
1983
1984 *access = 0;
1985 for (i = 1; i < 4; i++) {
1986 if (test_bit(i, &bmap))
1987 *access |= i;
1988 }
1989}
1990
fd39ca9a 1991static void
1da177e4
LT		 1992set_deny(unsigned int *deny, unsigned long bmap) {
1993 int i;
1994
1995 *deny = 0;
1996 for (i = 0; i < 4; i++) {
1997 if (test_bit(i, &bmap))
1998 *deny |= i ;
1999 }
2000}
2001
2002static int
2003test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2004 unsigned int access, deny;
2005
2006 set_access(&access, stp->st_access_bmap);
2007 set_deny(&deny, stp->st_deny_bmap);
2008 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2009 return 0;
2010 return 1;
2011}
2012
2013/*
2014 * Called to check deny when READ with all zero stateid or
2015 * WRITE with all zero or all one stateid
2016 */
b37ad28b 2017static __be32
1da177e4
LT		 2018nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2019{
2020 struct inode *ino = current_fh->fh_dentry->d_inode;
2021 struct nfs4_file *fp;
2022 struct nfs4_stateid *stp;
b37ad28b 2023 __be32 ret;
1da177e4
LT		 2024
2025 dprintk("NFSD: nfs4_share_conflict\n");
2026
2027 fp = find_file(ino);
13cd2184
N		 2028 if (!fp)
2029 return nfs_ok;
b700949b 2030 ret = nfserr_locked;
1da177e4 2031 /* Search for conflicting share reservations */
13cd2184
N		 2032 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2033 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2034 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2035 goto out;
1da177e4 2036 }
13cd2184
N		 2037 ret = nfs_ok;
2038out:
2039 put_nfs4_file(fp);
2040 return ret;
1da177e4
LT		 2041}
2042
2043static inline void
2044nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2045{
2046 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2047 drop_file_write_access(filp);
1da177e4
LT		 2048 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2049 }
2050}
2051
1da177e4
LT		 2052/*
2053 * Spawn a thread to perform a recall on the delegation represented
2054 * by the lease (file_lock)
2055 *
2056 * Called from break_lease() with lock_kernel() held.
2057 * Note: we assume break_lease will only call this *once* for any given
2058 * lease.
2059 */
2060static
2061void nfsd_break_deleg_cb(struct file_lock *fl)
2062{
63e4863f 2063 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2064
2065 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2066 if (!dp)
2067 return;
2068
2069 /* We're assuming the state code never drops its reference
2070 * without first removing the lease. Since we're in this lease
2071 * callback (and since the lease code is serialized by the kernel
2072 * lock) we know the server hasn't removed the lease yet, we know
2073 * it's safe to take a reference: */
2074 atomic_inc(&dp->dl_count);
cfdcad4d 2075 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2076
2077 spin_lock(&recall_lock);
2078 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2079 spin_unlock(&recall_lock);
2080
2081 /* only place dl_time is set. protected by lock_kernel*/
2082 dp->dl_time = get_seconds();
2083
0272e1fd
BF		 2084 /*
2085 * We don't want the locks code to timeout the lease for us;
2086 * we'll remove it ourself if the delegation isn't returned
2087 * in time.
2088 */
2089 fl->fl_break_time = 0;
1da177e4 2090
63e4863f
BF		 2091 dp->dl_file->fi_had_conflict = true;
2092 nfsd4_cb_recall(dp);
1da177e4
LT		 2093}
2094
2095/*
2096 * The file_lock is being reapd.
2097 *
2098 * Called by locks_free_lock() with lock_kernel() held.
2099 */
2100static
2101void nfsd_release_deleg_cb(struct file_lock *fl)
2102{
2103 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2104
2105 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2106
2107 if (!(fl->fl_flags & FL_LEASE) || !dp)
2108 return;
2109 dp->dl_flock = NULL;
2110}
2111
2112/*
2113 * Set the delegation file_lock back pointer.
2114 *
a9933cea 2115 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2116 */
2117static
2118void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2119{
2120 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2121
2122 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2123 if (!dp)
2124 return;
2125 dp->dl_flock = new;
2126}
2127
2128/*
a9933cea 2129 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2130 */
2131static
2132int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2133{
2134 struct nfs4_delegation *onlistd =
2135 (struct nfs4_delegation *)onlist->fl_owner;
2136 struct nfs4_delegation *tryd =
2137 (struct nfs4_delegation *)try->fl_owner;
2138
2139 if (onlist->fl_lmops != try->fl_lmops)
2140 return 0;
2141
2142 return onlistd->dl_client == tryd->dl_client;
2143}
2144
2145
2146static
2147int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2148{
2149 if (arg & F_UNLCK)
2150 return lease_modify(onlist, arg);
2151 else
2152 return -EAGAIN;
2153}
2154
fd39ca9a 2155static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2156 .fl_break = nfsd_break_deleg_cb,
2157 .fl_release_private = nfsd_release_deleg_cb,
2158 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2159 .fl_mylease = nfsd_same_client_deleg_cb,
2160 .fl_change = nfsd_change_deleg_cb,
2161};
2162
2163
b37ad28b 2164__be32
6668958f
AA		 2165nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2166 struct nfsd4_open *open)
1da177e4 2167{
1da177e4
LT		 2168 clientid_t *clientid = &open->op_clientid;
2169 struct nfs4_client *clp = NULL;
2170 unsigned int strhashval;
2171 struct nfs4_stateowner *sop = NULL;
2172
1da177e4 2173 if (!check_name(open->op_owner))
0f442aa2 2174 return nfserr_inval;
1da177e4
LT		 2175
2176 if (STALE_CLIENTID(&open->op_clientid))
2177 return nfserr_stale_clientid;
2178
2179 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2180 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2181 open->op_stateowner = sop;
2182 if (!sop) {
2183 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2184 clp = find_confirmed_client(clientid);
2185 if (clp == NULL)
0f442aa2
BF		 2186 return nfserr_expired;
2187 goto renew;
1da177e4 2188 }
6668958f
AA		 2189 /* When sessions are used, skip open sequenceid processing */
2190 if (nfsd4_has_session(cstate))
2191 goto renew;
0f442aa2
BF		 2192 if (!sop->so_confirmed) {
2193 /* Replace unconfirmed owners without checking for replay. */
2194 clp = sop->so_client;
f044ff83 2195 release_openowner(sop);
0f442aa2
BF		 2196 open->op_stateowner = NULL;
2197 goto renew;
2198 }
2199 if (open->op_seqid == sop->so_seqid - 1) {
2200 if (sop->so_replay.rp_buflen)
a90b061c 2201 return nfserr_replay_me;
0f442aa2
BF		 2202 /* The original OPEN failed so spectacularly
2203 * that we don't even have replay data saved!
2204 * Therefore, we have no choice but to continue
2205 * processing this OPEN; presumably, we'll
2206 * fail again for the same reason.
2207 */
2208 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2209 goto renew;
2210 }
2211 if (open->op_seqid != sop->so_seqid)
2212 return nfserr_bad_seqid;
1da177e4 2213renew:
0f442aa2
BF		 2214 if (open->op_stateowner == NULL) {
2215 sop = alloc_init_open_stateowner(strhashval, clp, open);
2216 if (sop == NULL)
2217 return nfserr_resource;
2218 open->op_stateowner = sop;
2219 }
2220 list_del_init(&sop->so_close_lru);
1da177e4 2221 renew_client(sop->so_client);
0f442aa2 2222 return nfs_ok;
1da177e4
LT		 2223}
2224
b37ad28b 2225static inline __be32
4a6e43e6
N		 2226nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2227{
2228 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2229 return nfserr_openmode;
2230 else
2231 return nfs_ok;
2232}
2233
52f4fb43
N		 2234static struct nfs4_delegation *
2235find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2236{
2237 struct nfs4_delegation *dp;
2238
ea1da636 2239 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2240 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2241 return dp;
2242 }
2243 return NULL;
2244}
2245
b37ad28b 2246static __be32
567d9829
N		 2247nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2248 struct nfs4_delegation **dp)
2249{
2250 int flags;
b37ad28b 2251 __be32 status = nfserr_bad_stateid;
567d9829
N		 2252
2253 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2254 if (*dp == NULL)
c44c5eeb 2255 goto out;
567d9829
N		 2256 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2257 RD_STATE : WR_STATE;
2258 status = nfs4_check_delegmode(*dp, flags);
2259 if (status)
2260 *dp = NULL;
c44c5eeb
N		 2261out:
2262 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2263 return nfs_ok;
2264 if (status)
2265 return status;
2266 open->op_stateowner->so_confirmed = 1;
2267 return nfs_ok;
567d9829
N		 2268}
2269
b37ad28b 2270static __be32
1da177e4
LT		 2271nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2272{
2273 struct nfs4_stateid *local;
b37ad28b 2274 __be32 status = nfserr_share_denied;
1da177e4
LT		 2275 struct nfs4_stateowner *sop = open->op_stateowner;
2276
8beefa24 2277 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2278 /* ignore lock owners */
2279 if (local->st_stateowner->so_is_open_owner == 0)
2280 continue;
2281 /* remember if we have seen this open owner */
2282 if (local->st_stateowner == sop)
2283 *stpp = local;
2284 /* check for conflicting share reservations */
2285 if (!test_share(local, open))
2286 goto out;
2287 }
2288 status = 0;
2289out:
2290 return status;
2291}
2292
5ac049ac
N		 2293static inline struct nfs4_stateid *
2294nfs4_alloc_stateid(void)
2295{
2296 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2297}
2298
b37ad28b 2299static __be32
1da177e4 2300nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2301 struct nfs4_delegation *dp,
1da177e4
LT		 2302 struct svc_fh *cur_fh, int flags)
2303{
2304 struct nfs4_stateid *stp;
1da177e4 2305
5ac049ac 2306 stp = nfs4_alloc_stateid();
1da177e4
LT		 2307 if (stp == NULL)
2308 return nfserr_resource;
2309
567d9829
N		 2310 if (dp) {
2311 get_file(dp->dl_vfs_file);
2312 stp->st_vfs_file = dp->dl_vfs_file;
2313 } else {
b37ad28b 2314 __be32 status;
567d9829
N		 2315 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2316 &stp->st_vfs_file);
2317 if (status) {
2318 if (status == nfserr_dropit)
2319 status = nfserr_jukebox;
5ac049ac 2320 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2321 return status;
2322 }
1da177e4 2323 }
1da177e4
LT		 2324 *stpp = stp;
2325 return 0;
2326}
2327
b37ad28b 2328static inline __be32
1da177e4
LT		 2329nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2330 struct nfsd4_open *open)
2331{
2332 struct iattr iattr = {
2333 .ia_valid = ATTR_SIZE,
2334 .ia_size = 0,
2335 };
2336 if (!open->op_truncate)
2337 return 0;
2338 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2339 return nfserr_inval;
1da177e4
LT		 2340 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2341}
2342
b37ad28b 2343static __be32
1da177e4
LT		 2344nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2345{
2346 struct file *filp = stp->st_vfs_file;
7eaa36e2 2347 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2348 unsigned int share_access, new_writer;
b37ad28b 2349 __be32 status;
1da177e4
LT		 2350
2351 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2352 new_writer = (~share_access) & open->op_share_access
2353 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2354
6c26d08f 2355 if (new_writer) {
b37ad28b
AV		 2356 int err = get_write_access(inode);
2357 if (err)
2358 return nfserrno(err);
e518f056
BH		 2359 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2360 if (err)
2361 return nfserrno(err);
2362 file_take_write(filp);
6c26d08f 2363 }
1da177e4
LT		 2364 status = nfsd4_truncate(rqstp, cur_fh, open);
2365 if (status) {
6c26d08f
BF		 2366 if (new_writer)
2367 put_write_access(inode);
1da177e4
LT		 2368 return status;
2369 }
2370 /* remember the open */
6c26d08f 2371 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2372 __set_bit(open->op_share_access, &stp->st_access_bmap);
2373 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2374
2375 return nfs_ok;
2376}
2377
2378
1da177e4 2379static void
37515177 2380nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2381{
37515177
N		 2382 open->op_stateowner->so_confirmed = 1;
2383 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2384}
2385
2386/*
2387 * Attempt to hand out a delegation.
2388 */
2389static void
2390nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2391{
2392 struct nfs4_delegation *dp;
2393 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2394 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2395 struct file_lock fl, *flp = &fl;
2396 int status, flag = 0;
2397
2398 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2399 open->op_recall = 0;
2400 switch (open->op_claim_type) {
2401 case NFS4_OPEN_CLAIM_PREVIOUS:
2402 if (!atomic_read(&cb->cb_set))
2403 open->op_recall = 1;
2404 flag = open->op_delegate_type;
2405 if (flag == NFS4_OPEN_DELEGATE_NONE)
2406 goto out;
2407 break;
2408 case NFS4_OPEN_CLAIM_NULL:
2409 /* Let's not give out any delegations till everyone's
2410 * had the chance to reclaim theirs.... */
af558e33 2411 if (locks_in_grace())
7b190fec
N		 2412 goto out;
2413 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2414 goto out;
2415 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2416 flag = NFS4_OPEN_DELEGATE_WRITE;
2417 else
2418 flag = NFS4_OPEN_DELEGATE_READ;
2419 break;
2420 default:
2421 goto out;
2422 }
1da177e4
LT		 2423
2424 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2425 if (dp == NULL) {
2426 flag = NFS4_OPEN_DELEGATE_NONE;
2427 goto out;
2428 }
2429 locks_init_lock(&fl);
2430 fl.fl_lmops = &nfsd_lease_mng_ops;
2431 fl.fl_flags = FL_LEASE;
9167f501 2432 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2433 fl.fl_end = OFFSET_MAX;
2434 fl.fl_owner = (fl_owner_t)dp;
2435 fl.fl_file = stp->st_vfs_file;
2436 fl.fl_pid = current->tgid;
2437
a9933cea 2438 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2439 * the lock_manager callbacks fl_mylease and fl_change are used
2440 */
9167f501 2441 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2442 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2443 unhash_delegation(dp);
1da177e4
LT		 2444 flag = NFS4_OPEN_DELEGATE_NONE;
2445 goto out;
2446 }
2447
2448 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2449
2450 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2451 dp->dl_stateid.si_boot,
2452 dp->dl_stateid.si_stateownerid,
2453 dp->dl_stateid.si_fileid,
2454 dp->dl_stateid.si_generation);
2455out:
7b190fec
N		 2456 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2457 && flag == NFS4_OPEN_DELEGATE_NONE
2458 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2459 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2460 open->op_delegate_type = flag;
2461}
2462
2463/*
2464 * called with nfs4_lock_state() held.
2465 */
b37ad28b 2466__be32
1da177e4
LT		 2467nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2468{
6668958f 2469 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2470 struct nfs4_file *fp = NULL;
2471 struct inode *ino = current_fh->fh_dentry->d_inode;
2472 struct nfs4_stateid *stp = NULL;
567d9829 2473 struct nfs4_delegation *dp = NULL;
b37ad28b 2474 __be32 status;
1da177e4
LT		 2475
2476 status = nfserr_inval;
d87a8ade 2477 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2478 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2479 goto out;
2480 /*
2481 * Lookup file; if found, lookup stateid and check open request,
2482 * and check for delegations in the process of being recalled.
2483 * If not found, create the nfs4_file struct
2484 */
2485 fp = find_file(ino);
2486 if (fp) {
2487 if ((status = nfs4_check_open(fp, open, &stp)))
2488 goto out;
c44c5eeb
N		 2489 status = nfs4_check_deleg(fp, open, &dp);
2490 if (status)
2491 goto out;
1da177e4 2492 } else {
c44c5eeb
N		 2493 status = nfserr_bad_stateid;
2494 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2495 goto out;
1da177e4
LT		 2496 status = nfserr_resource;
2497 fp = alloc_init_file(ino);
2498 if (fp == NULL)
2499 goto out;
2500 }
2501
2502 /*
2503 * OPEN the file, or upgrade an existing OPEN.
2504 * If truncate fails, the OPEN fails.
2505 */
2506 if (stp) {
2507 /* Stateid was found, this is an OPEN upgrade */
2508 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2509 if (status)
2510 goto out;
444c2c07 2511 update_stateid(&stp->st_stateid);
1da177e4
LT		 2512 } else {
2513 /* Stateid was not found, this is a new OPEN */
2514 int flags = 0;
9ecb6a08 2515 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2516 flags |= NFSD_MAY_READ;
1da177e4 2517 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2518 flags |= NFSD_MAY_WRITE;
567d9829
N		 2519 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2520 if (status)
1da177e4
LT		 2521 goto out;
2522 init_stateid(stp, fp, open);
2523 status = nfsd4_truncate(rqstp, current_fh, open);
2524 if (status) {
2283963f 2525 release_open_stateid(stp);
1da177e4
LT		 2526 goto out;
2527 }
6668958f
AA		 2528 if (nfsd4_has_session(&resp->cstate))
2529 update_stateid(&stp->st_stateid);
1da177e4
LT		 2530 }
2531 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2532
6668958f
AA		 2533 if (nfsd4_has_session(&resp->cstate))
2534 open->op_stateowner->so_confirmed = 1;
2535
1da177e4
LT		 2536 /*
2537 * Attempt to hand out a delegation. No error return, because the
2538 * OPEN succeeds even if we fail.
2539 */
2540 nfs4_open_delegation(current_fh, open, stp);
2541
2542 status = nfs_ok;
2543
2544 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2545 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2546 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2547out:
13cd2184
N		 2548 if (fp)
2549 put_nfs4_file(fp);
37515177
N		 2550 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2551 nfs4_set_claim_prev(open);
1da177e4
LT		 2552 /*
2553 * To finish the open response, we just need to set the rflags.
2554 */
2555 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2556 if (!open->op_stateowner->so_confirmed &&
2557 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2558 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2559
2560 return status;
2561}
2562
b37ad28b 2563__be32
b591480b
BF		 2564nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2565 clientid_t *clid)
1da177e4
LT		 2566{
2567 struct nfs4_client *clp;
b37ad28b 2568 __be32 status;
1da177e4
LT		 2569
2570 nfs4_lock_state();
2571 dprintk("process_renew(%08x/%08x): starting\n",
2572 clid->cl_boot, clid->cl_id);
2573 status = nfserr_stale_clientid;
2574 if (STALE_CLIENTID(clid))
2575 goto out;
2576 clp = find_confirmed_client(clid);
2577 status = nfserr_expired;
2578 if (clp == NULL) {
2579 /* We assume the client took too long to RENEW. */
2580 dprintk("nfsd4_renew: clientid not found!\n");
2581 goto out;
2582 }
2583 renew_client(clp);
2584 status = nfserr_cb_path_down;
ea1da636 2585 if (!list_empty(&clp->cl_delegations)
c237dc03 2586 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2587 goto out;
2588 status = nfs_ok;
2589out:
2590 nfs4_unlock_state();
2591 return status;
2592}
2593
af558e33
BF		 2594struct lock_manager nfsd4_manager = {
2595};
2596
a76b4319 2597static void
af558e33 2598nfsd4_end_grace(void)
a76b4319
N		 2599{
2600 dprintk("NFSD: end of grace period\n");
c7b9a459 2601 nfsd4_recdir_purge_old();
af558e33 2602 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2603}
2604
fd39ca9a 2605static time_t
1da177e4
LT		 2606nfs4_laundromat(void)
2607{
2608 struct nfs4_client *clp;
2609 struct nfs4_stateowner *sop;
2610 struct nfs4_delegation *dp;
2611 struct list_head *pos, *next, reaplist;
2612 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2613 time_t t, clientid_val = NFSD_LEASE_TIME;
2614 time_t u, test_val = NFSD_LEASE_TIME;
2615
2616 nfs4_lock_state();
2617
2618 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2619 if (locks_in_grace())
2620 nfsd4_end_grace();
1da177e4
LT		 2621 list_for_each_safe(pos, next, &client_lru) {
2622 clp = list_entry(pos, struct nfs4_client, cl_lru);
2623 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2624 t = clp->cl_time - cutoff;
2625 if (clientid_val > t)
2626 clientid_val = t;
2627 break;
2628 }
2629 dprintk("NFSD: purging unused client (clientid %08x)\n",
2630 clp->cl_clientid.cl_id);
c7b9a459 2631 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2632 expire_client(clp);
2633 }
2634 INIT_LIST_HEAD(&reaplist);
2635 spin_lock(&recall_lock);
2636 list_for_each_safe(pos, next, &del_recall_lru) {
2637 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2638 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2639 u = dp->dl_time - cutoff;
2640 if (test_val > u)
2641 test_val = u;
2642 break;
2643 }
2644 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2645 dp, dp->dl_flock);
2646 list_move(&dp->dl_recall_lru, &reaplist);
2647 }
2648 spin_unlock(&recall_lock);
2649 list_for_each_safe(pos, next, &reaplist) {
2650 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2651 list_del_init(&dp->dl_recall_lru);
2652 unhash_delegation(dp);
2653 }
2654 test_val = NFSD_LEASE_TIME;
2655 list_for_each_safe(pos, next, &close_lru) {
2656 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2657 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2658 u = sop->so_time - cutoff;
2659 if (test_val > u)
2660 test_val = u;
2661 break;
2662 }
2663 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2664 sop->so_id);
f044ff83 2665 release_openowner(sop);
1da177e4
LT		 2666 }
2667 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2668 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2669 nfs4_unlock_state();
2670 return clientid_val;
2671}
2672
a254b246
HH		 2673static struct workqueue_struct *laundry_wq;
2674static void laundromat_main(struct work_struct *);
2675static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2676
2677static void
c4028958 2678laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2679{
2680 time_t t;
2681
2682 t = nfs4_laundromat();
2683 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2684 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2685}
2686
fd39ca9a 2687static struct nfs4_stateowner *
f8816512
N		 2688search_close_lru(u32 st_id, int flags)
2689{
1da177e4
LT		 2690 struct nfs4_stateowner *local = NULL;
2691
1da177e4
LT		 2692 if (flags & CLOSE_STATE) {
2693 list_for_each_entry(local, &close_lru, so_close_lru) {
2694 if (local->so_id == st_id)
2695 return local;
2696 }
2697 }
2698 return NULL;
2699}
2700
2701static inline int
2702nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2703{
7eaa36e2 2704 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2705}
2706
2707static int
2708STALE_STATEID(stateid_t *stateid)
2709{
78155ed7
BN		 2710 if (time_after((unsigned long)boot_time,
2711 (unsigned long)stateid->si_boot)) {
2712 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2713 stateid->si_boot, stateid->si_stateownerid,
2714 stateid->si_fileid, stateid->si_generation);
2715 return 1;
2716 }
2717 return 0;
2718}
2719
2720static int
2721EXPIRED_STATEID(stateid_t *stateid)
2722{
2723 if (time_before((unsigned long)boot_time,
2724 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2725 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2726 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2727 stateid->si_boot, stateid->si_stateownerid,
2728 stateid->si_fileid, stateid->si_generation);
2729 return 1;
2730 }
2731 return 0;
2732}
2733
2734static __be32
2735stateid_error_map(stateid_t *stateid)
2736{
2737 if (STALE_STATEID(stateid))
2738 return nfserr_stale_stateid;
2739 if (EXPIRED_STATEID(stateid))
2740 return nfserr_expired;
2741
2742 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2743 stateid->si_boot, stateid->si_stateownerid,
2744 stateid->si_fileid, stateid->si_generation);
2745 return nfserr_bad_stateid;
1da177e4
LT		 2746}
2747
2748static inline int
2749access_permit_read(unsigned long access_bmap)
2750{
2751 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2752 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2753 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2754}
2755
2756static inline int
2757access_permit_write(unsigned long access_bmap)
2758{
2759 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2760 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2761}
2762
2763static
b37ad28b 2764__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2765{
b37ad28b 2766 __be32 status = nfserr_openmode;
1da177e4
LT		 2767
2768 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2769 goto out;
2770 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2771 goto out;
2772 status = nfs_ok;
2773out:
2774 return status;
2775}
2776
b37ad28b 2777static inline __be32
1da177e4
LT		 2778check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2779{
203a8c8e 2780 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2781 return nfs_ok;
af558e33 2782 else if (locks_in_grace()) {
1da177e4
LT		 2783 /* Answer in remaining cases depends on existance of
2784 * conflicting state; so we must wait out the grace period. */
2785 return nfserr_grace;
2786 } else if (flags & WR_STATE)
2787 return nfs4_share_conflict(current_fh,
2788 NFS4_SHARE_DENY_WRITE);
2789 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2790 return nfs4_share_conflict(current_fh,
2791 NFS4_SHARE_DENY_READ);
2792}
2793
2794/*
2795 * Allow READ/WRITE during grace period on recovered state only for files
2796 * that are not able to provide mandatory locking.
2797 */
2798static inline int
18f82731 2799grace_disallows_io(struct inode *inode)
1da177e4 2800{
203a8c8e 2801 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2802}
2803
6668958f 2804static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2805{
6668958f
AA		 2806 /*
2807 * When sessions are used the stateid generation number is ignored
2808 * when it is zero.
2809 */
2810 if ((flags & HAS_SESSION) && in->si_generation == 0)
2811 goto out;
2812
0836f587
BF		 2813 /* If the client sends us a stateid from the future, it's buggy: */
2814 if (in->si_generation > ref->si_generation)
2815 return nfserr_bad_stateid;
2816 /*
2817 * The following, however, can happen. For example, if the
2818 * client sends an open and some IO at the same time, the open
2819 * may bump si_generation while the IO is still in flight.
2820 * Thanks to hard links and renames, the client never knows what
2821 * file an open will affect. So it could avoid that situation
2822 * only by serializing all opens and IO from the same open
2823 * owner. To recover from the old_stateid error, the client
2824 * will just have to retry the IO:
2825 */
2826 if (in->si_generation < ref->si_generation)
2827 return nfserr_old_stateid;
6668958f 2828out:
0836f587
BF		 2829 return nfs_ok;
2830}
2831
3e633079
BF		 2832static int is_delegation_stateid(stateid_t *stateid)
2833{
2834 return stateid->si_fileid == 0;
2835}
2836
1da177e4
LT		 2837/*
2838* Checks for stateid operations
2839*/
b37ad28b 2840__be32
dd453dfd
BH		 2841nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2842 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2843{
2844 struct nfs4_stateid *stp = NULL;
2845 struct nfs4_delegation *dp = NULL;
dd453dfd 2846 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2847 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2848 __be32 status;
1da177e4 2849
1da177e4
LT		 2850 if (filpp)
2851 *filpp = NULL;
2852
18f82731 2853 if (grace_disallows_io(ino))
1da177e4
LT		 2854 return nfserr_grace;
2855
6668958f
AA		 2856 if (nfsd4_has_session(cstate))
2857 flags |= HAS_SESSION;
2858
1da177e4
LT		 2859 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2860 return check_special_stateids(current_fh, stateid, flags);
2861
1da177e4
LT		 2862 status = nfserr_stale_stateid;
2863 if (STALE_STATEID(stateid))
2864 goto out;
2865
1da177e4 2866 status = nfserr_bad_stateid;
3e633079 2867 if (is_delegation_stateid(stateid)) {
a4455be0 2868 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2869 if (!dp) {
2870 status = stateid_error_map(stateid);
1da177e4 2871 goto out;
78155ed7 2872 }
6668958f
AA		 2873 status = check_stateid_generation(stateid, &dp->dl_stateid,
2874 flags);
0c2a498f
BF		 2875 if (status)
2876 goto out;
dc9bf700
BF		 2877 status = nfs4_check_delegmode(dp, flags);
2878 if (status)
2879 goto out;
2880 renew_client(dp->dl_client);
dc9bf700
BF		 2881 if (filpp)
2882 *filpp = dp->dl_vfs_file;
1da177e4 2883 } else { /* open or lock stateid */
a4455be0 2884 stp = find_stateid(stateid, flags);
78155ed7
BN		 2885 if (!stp) {
2886 status = stateid_error_map(stateid);
1da177e4 2887 goto out;
78155ed7 2888 }
6150ef0d 2889 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2890 goto out;
2891 if (!stp->st_stateowner->so_confirmed)
2892 goto out;
6668958f
AA		 2893 status = check_stateid_generation(stateid, &stp->st_stateid,
2894 flags);
0c2a498f
BF		 2895 if (status)
2896 goto out;
a4455be0
BF		 2897 status = nfs4_check_openmode(stp, flags);
2898 if (status)
1da177e4
LT		 2899 goto out;
2900 renew_client(stp->st_stateowner->so_client);
2901 if (filpp)
2902 *filpp = stp->st_vfs_file;
1da177e4
LT		 2903 }
2904 status = nfs_ok;
2905out:
2906 return status;
2907}
2908
4c4cd222
N		 2909static inline int
2910setlkflg (int type)
2911{
2912 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2913 RD_STATE : WR_STATE;
2914}
1da177e4
LT		 2915
2916/*
2917 * Checks for sequence id mutating operations.
2918 */
b37ad28b 2919static __be32
dd453dfd
BH		 2920nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2921 stateid_t *stateid, int flags,
2922 struct nfs4_stateowner **sopp,
2923 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2924{
1da177e4
LT		 2925 struct nfs4_stateid *stp;
2926 struct nfs4_stateowner *sop;
dd453dfd 2927 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2928 __be32 status;
1da177e4
LT		 2929
2930 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2931 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2932 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2933 stateid->si_generation);
3a4f98bb 2934
1da177e4
LT		 2935 *stpp = NULL;
2936 *sopp = NULL;
2937
1da177e4 2938 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2939 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2940 return nfserr_bad_stateid;
1da177e4
LT		 2941 }
2942
1da177e4 2943 if (STALE_STATEID(stateid))
3a4f98bb 2944 return nfserr_stale_stateid;
6668958f
AA		 2945
2946 if (nfsd4_has_session(cstate))
2947 flags |= HAS_SESSION;
2948
1da177e4
LT		 2949 /*
2950 * We return BAD_STATEID if filehandle doesn't match stateid,
2951 * the confirmed flag is incorrecly set, or the generation
2952 * number is incorrect.
1da177e4 2953 */
f8816512
N		 2954 stp = find_stateid(stateid, flags);
2955 if (stp == NULL) {
2956 /*
2957 * Also, we should make sure this isn't just the result of
2958 * a replayed close:
2959 */
2960 sop = search_close_lru(stateid->si_stateownerid, flags);
2961 if (sop == NULL)
78155ed7 2962 return stateid_error_map(stateid);
f8816512
N		 2963 *sopp = sop;
2964 goto check_replay;
2965 }
1da177e4 2966
39325bd0
BF		 2967 *stpp = stp;
2968 *sopp = sop = stp->st_stateowner;
2969
4c4cd222 2970 if (lock) {
4c4cd222 2971 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2972 struct nfs4_client *clp = sop->so_client;
4c4cd222 2973 int lkflg = 0;
b37ad28b 2974 __be32 status;
4c4cd222
N		 2975
2976 lkflg = setlkflg(lock->lk_type);
2977
2978 if (lock->lk_is_new) {
599e0a22
BF		 2979 if (!sop->so_is_open_owner)
2980 return nfserr_bad_stateid;
60adfc50
AA		 2981 if (!(flags & HAS_SESSION) &&
2982 !same_clid(&clp->cl_clientid, lockclid))
2983 return nfserr_bad_stateid;
599e0a22
BF		 2984 /* stp is the open stateid */
2985 status = nfs4_check_openmode(stp, lkflg);
2986 if (status)
2987 return status;
2988 } else {
2989 /* stp is the lock stateid */
2990 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2991 if (status)
2992 return status;
4c4cd222 2993 }
1da177e4
LT		 2994 }
2995
f3362737 2996 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2997 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2998 return nfserr_bad_stateid;
1da177e4
LT		 2999 }
3000
1da177e4
LT		 3001 /*
3002 * We now validate the seqid and stateid generation numbers.
3003 * For the moment, we ignore the possibility of
3004 * generation number wraparound.
3005 */
6668958f 3006 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3007 goto check_replay;
3008
3a4f98bb 3009 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3010 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3011 " unconfirmed stateowner!\n");
3012 return nfserr_bad_stateid;
1da177e4 3013 }
3a4f98bb 3014 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3015 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3016 " confirmed yet!\n");
3017 return nfserr_bad_stateid;
1da177e4 3018 }
6668958f 3019 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3020 if (status)
3021 return status;
52fd004e 3022 renew_client(sop->so_client);
3a4f98bb 3023 return nfs_ok;
1da177e4 3024
1da177e4 3025check_replay:
bd9aac52 3026 if (seqid == sop->so_seqid - 1) {
849823c5 3027 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3028 /* indicate replay to calling function */
a90b061c 3029 return nfserr_replay_me;
1da177e4 3030 }
2fdada03 3031 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3032 sop->so_seqid, seqid);
3033 *sopp = NULL;
3034 return nfserr_bad_seqid;
1da177e4
LT		 3035}
3036
b37ad28b 3037__be32
ca364317 3038nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3039 struct nfsd4_open_confirm *oc)
1da177e4 3040{
b37ad28b 3041 __be32 status;
1da177e4
LT		 3042 struct nfs4_stateowner *sop;
3043 struct nfs4_stateid *stp;
3044
3045 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3046 (int)cstate->current_fh.fh_dentry->d_name.len,
3047 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3048
ca364317 3049 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3050 if (status)
3051 return status;
1da177e4
LT		 3052
3053 nfs4_lock_state();
3054
dd453dfd 3055 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3056 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3057 CONFIRM | OPEN_STATE,
1da177e4
LT		 3058 &oc->oc_stateowner, &stp, NULL)))
3059 goto out;
3060
3061 sop = oc->oc_stateowner;
3062 sop->so_confirmed = 1;
3063 update_stateid(&stp->st_stateid);
3064 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3065 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3066 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3067 stp->st_stateid.si_boot,
3068 stp->st_stateid.si_stateownerid,
3069 stp->st_stateid.si_fileid,
3070 stp->st_stateid.si_generation);
c7b9a459
N		 3071
3072 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3073out:
f2327d9a 3074 if (oc->oc_stateowner) {
1da177e4 3075 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3076 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3077 }
1da177e4
LT		 3078 nfs4_unlock_state();
3079 return status;
3080}
3081
3082
3083/*
3084 * unset all bits in union bitmap (bmap) that
3085 * do not exist in share (from successful OPEN_DOWNGRADE)
3086 */
3087static void
3088reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3089{
3090 int i;
3091 for (i = 1; i < 4; i++) {
3092 if ((i & access) != i)
3093 __clear_bit(i, bmap);
3094 }
3095}
3096
3097static void
3098reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3099{
3100 int i;
3101 for (i = 0; i < 4; i++) {
3102 if ((i & deny) != i)
3103 __clear_bit(i, bmap);
3104 }
3105}
3106
b37ad28b 3107__be32
ca364317
BF		 3108nfsd4_open_downgrade(struct svc_rqst *rqstp,
3109 struct nfsd4_compound_state *cstate,
a4f1706a 3110 struct nfsd4_open_downgrade *od)
1da177e4 3111{
b37ad28b 3112 __be32 status;
1da177e4
LT		 3113 struct nfs4_stateid *stp;
3114 unsigned int share_access;
3115
3116 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3117 (int)cstate->current_fh.fh_dentry->d_name.len,
3118 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3119
d87a8ade 3120 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3121 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3122 return nfserr_inval;
3123
3124 nfs4_lock_state();
dd453dfd 3125 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3126 od->od_seqid,
1da177e4 3127 &od->od_stateid,
f3362737 3128 OPEN_STATE,
1da177e4
LT		 3129 &od->od_stateowner, &stp, NULL)))
3130 goto out;
3131
3132 status = nfserr_inval;
3133 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3134 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3135 stp->st_access_bmap, od->od_share_access);
3136 goto out;
3137 }
3138 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3139 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3140 stp->st_deny_bmap, od->od_share_deny);
3141 goto out;
3142 }
3143 set_access(&share_access, stp->st_access_bmap);
3144 nfs4_file_downgrade(stp->st_vfs_file,
3145 share_access & ~od->od_share_access);
3146
3147 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3148 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3149
3150 update_stateid(&stp->st_stateid);
3151 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3152 status = nfs_ok;
3153out:
f2327d9a 3154 if (od->od_stateowner) {
1da177e4 3155 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3156 cstate->replay_owner = od->od_stateowner;
f2327d9a 3157 }
1da177e4
LT		 3158 nfs4_unlock_state();
3159 return status;
3160}
3161
3162/*
3163 * nfs4_unlock_state() called after encode
3164 */
b37ad28b 3165__be32
ca364317 3166nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3167 struct nfsd4_close *close)
1da177e4 3168{
b37ad28b 3169 __be32 status;
1da177e4
LT		 3170 struct nfs4_stateid *stp;
3171
3172 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3173 (int)cstate->current_fh.fh_dentry->d_name.len,
3174 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3175
3176 nfs4_lock_state();
3177 /* check close_lru for replay */
dd453dfd 3178 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3179 close->cl_seqid,
1da177e4 3180 &close->cl_stateid,
f3362737 3181 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3182 &close->cl_stateowner, &stp, NULL)))
3183 goto out;
1da177e4
LT		 3184 status = nfs_ok;
3185 update_stateid(&stp->st_stateid);
3186 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3187
04ef5954 3188 /* release_stateid() calls nfsd_close() if needed */
2283963f 3189 release_open_stateid(stp);
04ef5954
BF		 3190
3191 /* place unused nfs4_stateowners on so_close_lru list to be
3192 * released by the laundromat service after the lease period
3193 * to enable us to handle CLOSE replay
3194 */
3195 if (list_empty(&close->cl_stateowner->so_stateids))
3196 move_to_close_lru(close->cl_stateowner);
1da177e4 3197out:
f2327d9a 3198 if (close->cl_stateowner) {
1da177e4 3199 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3200 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3201 }
1da177e4
LT		 3202 nfs4_unlock_state();
3203 return status;
3204}
3205
b37ad28b 3206__be32
ca364317
BF		 3207nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3208 struct nfsd4_delegreturn *dr)
1da177e4 3209{
203a8c8e
BF		 3210 struct nfs4_delegation *dp;
3211 stateid_t *stateid = &dr->dr_stateid;
3212 struct inode *inode;
b37ad28b 3213 __be32 status;
6668958f 3214 int flags = 0;
1da177e4 3215
ca364317 3216 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3217 return status;
3218 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3219
6668958f
AA		 3220 if (nfsd4_has_session(cstate))
3221 flags |= HAS_SESSION;
1da177e4 3222 nfs4_lock_state();
203a8c8e
BF		 3223 status = nfserr_bad_stateid;
3224 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3225 goto out;
3226 status = nfserr_stale_stateid;
3227 if (STALE_STATEID(stateid))
3228 goto out;
7e0f7cf5 3229 status = nfserr_bad_stateid;
203a8c8e
BF		 3230 if (!is_delegation_stateid(stateid))
3231 goto out;
203a8c8e 3232 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3233 if (!dp) {
3234 status = stateid_error_map(stateid);
203a8c8e 3235 goto out;
78155ed7 3236 }
6668958f 3237 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3238 if (status)
3239 goto out;
3240 renew_client(dp->dl_client);
3241
3242 unhash_delegation(dp);
1da177e4 3243out:
203a8c8e
BF		 3244 nfs4_unlock_state();
3245
1da177e4
LT		 3246 return status;
3247}
3248
3249
3250/*
3251 * Lock owner state (byte-range locks)
3252 */
3253#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3254#define LOCK_HASH_BITS 8
3255#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3256#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3257
87df4de8
BH		 3258static inline u64
3259end_offset(u64 start, u64 len)
3260{
3261 u64 end;
3262
3263 end = start + len;
3264 return end >= start ? end: NFS4_MAX_UINT64;
3265}
3266
3267/* last octet in a range */
3268static inline u64
3269last_byte_offset(u64 start, u64 len)
3270{
3271 u64 end;
3272
3273 BUG_ON(!len);
3274 end = start + len;
3275 return end > start ? end - 1: NFS4_MAX_UINT64;
3276}
3277
1da177e4
LT		 3278#define lockownerid_hashval(id) \
3279 ((id) & LOCK_HASH_MASK)
3280
3281static inline unsigned int
3282lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3283 struct xdr_netobj *ownername)
3284{
3285 return (file_hashval(inode) + cl_id
3286 + opaque_hashval(ownername->data, ownername->len))
3287 & LOCK_HASH_MASK;
3288}
3289
3290static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3291static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3292static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3293
fd39ca9a 3294static struct nfs4_stateid *
1da177e4
LT		 3295find_stateid(stateid_t *stid, int flags)
3296{
9346eff0 3297 struct nfs4_stateid *local;
1da177e4
LT		 3298 u32 st_id = stid->si_stateownerid;
3299 u32 f_id = stid->si_fileid;
3300 unsigned int hashval;
3301
3302 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3303 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3304 hashval = stateid_hashval(st_id, f_id);
3305 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3306 if ((local->st_stateid.si_stateownerid == st_id) &&
3307 (local->st_stateid.si_fileid == f_id))
3308 return local;
3309 }
3310 }
9346eff0
KK		 3311
3312 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3313 hashval = stateid_hashval(st_id, f_id);
3314 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3315 if ((local->st_stateid.si_stateownerid == st_id) &&
3316 (local->st_stateid.si_fileid == f_id))
3317 return local;
3318 }
849823c5 3319 }
1da177e4
LT		 3320 return NULL;
3321}
3322
3323static struct nfs4_delegation *
3324find_delegation_stateid(struct inode *ino, stateid_t *stid)
3325{
13cd2184
N		 3326 struct nfs4_file *fp;
3327 struct nfs4_delegation *dl;
1da177e4
LT		 3328
3329 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3330 stid->si_boot, stid->si_stateownerid,
3331 stid->si_fileid, stid->si_generation);
3332
1da177e4 3333 fp = find_file(ino);
13cd2184
N		 3334 if (!fp)
3335 return NULL;
3336 dl = find_delegation_file(fp, stid);
3337 put_nfs4_file(fp);
3338 return dl;
1da177e4
LT		 3339}
3340
3341/*
3342 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3343 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3344 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3345 * locking, this prevents us from being completely protocol-compliant. The
3346 * real solution to this problem is to start using unsigned file offsets in
3347 * the VFS, but this is a very deep change!
3348 */
3349static inline void
3350nfs4_transform_lock_offset(struct file_lock *lock)
3351{
3352 if (lock->fl_start < 0)
3353 lock->fl_start = OFFSET_MAX;
3354 if (lock->fl_end < 0)
3355 lock->fl_end = OFFSET_MAX;
3356}
3357
d5b9026a
N		 3358/* Hack!: For now, we're defining this just so we can use a pointer to it
3359 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3360static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3361};
1da177e4
LT		 3362
3363static inline void
3364nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3365{
d5b9026a
N		 3366 struct nfs4_stateowner *sop;
3367 unsigned int hval;
1da177e4 3368
d5b9026a
N		 3369 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3370 sop = (struct nfs4_stateowner *) fl->fl_owner;
3371 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3372 kref_get(&sop->so_ref);
3373 deny->ld_sop = sop;
3374 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3375 } else {
3376 deny->ld_sop = NULL;
3377 deny->ld_clientid.cl_boot = 0;
3378 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3379 }
3380 deny->ld_start = fl->fl_start;
87df4de8
BH		 3381 deny->ld_length = NFS4_MAX_UINT64;
3382 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3383 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3384 deny->ld_type = NFS4_READ_LT;
3385 if (fl->fl_type != F_RDLCK)
3386 deny->ld_type = NFS4_WRITE_LT;
3387}
3388
1da177e4
LT		 3389static struct nfs4_stateowner *
3390find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3391 struct xdr_netobj *owner)
3392{
3393 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3394 struct nfs4_stateowner *op;
3395
3396 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3397 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3398 return op;
3399 }
3400 return NULL;
3401}
3402
3403/*
3404 * Alloc a lock owner structure.
3405 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3406 * occured.
3407 *
3408 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3409 */
3410
3411static struct nfs4_stateowner *
3412alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3413 struct nfs4_stateowner *sop;
3414 struct nfs4_replay *rp;
3415 unsigned int idhashval;
3416
3417 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3418 return NULL;
3419 idhashval = lockownerid_hashval(current_ownerid);
3420 INIT_LIST_HEAD(&sop->so_idhash);
3421 INIT_LIST_HEAD(&sop->so_strhash);
3422 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3423 INIT_LIST_HEAD(&sop->so_stateids);
3424 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3425 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3426 sop->so_time = 0;
3427 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3428 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3429 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3430 sop->so_is_open_owner = 0;
3431 sop->so_id = current_ownerid++;
3432 sop->so_client = clp;
b59e3c0e
NB		 3433 /* It is the openowner seqid that will be incremented in encode in the
3434 * case of new lockowners; so increment the lock seqid manually: */
3435 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3436 sop->so_confirmed = 1;
3437 rp = &sop->so_replay;
de1ae286 3438 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3439 rp->rp_buflen = 0;
3440 rp->rp_buf = rp->rp_ibuf;
3441 return sop;
3442}
3443
fd39ca9a 3444static struct nfs4_stateid *
1da177e4
LT		 3445alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3446{
3447 struct nfs4_stateid *stp;
3448 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3449
5ac049ac
N		 3450 stp = nfs4_alloc_stateid();
3451 if (stp == NULL)
1da177e4
LT		 3452 goto out;
3453 INIT_LIST_HEAD(&stp->st_hash);
3454 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3455 INIT_LIST_HEAD(&stp->st_perstateowner);
3456 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3457 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3458 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3459 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3460 stp->st_stateowner = sop;
13cd2184 3461 get_nfs4_file(fp);
1da177e4 3462 stp->st_file = fp;
78155ed7 3463 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3464 stp->st_stateid.si_stateownerid = sop->so_id;
3465 stp->st_stateid.si_fileid = fp->fi_id;
3466 stp->st_stateid.si_generation = 0;
3467 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3468 stp->st_access_bmap = open_stp->st_access_bmap;
3469 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3470 stp->st_openstp = open_stp;
1da177e4
LT		 3471
3472out:
3473 return stp;
3474}
3475
fd39ca9a 3476static int
1da177e4
LT		 3477check_lock_length(u64 offset, u64 length)
3478{
87df4de8 3479 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3480 LOFF_OVERFLOW(offset, length)));
3481}
3482
3483/*
3484 * LOCK operation
3485 */
b37ad28b 3486__be32
ca364317 3487nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3488 struct nfsd4_lock *lock)
1da177e4 3489{
3e9e3dbe 3490 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3491 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3492 struct nfs4_stateid *lock_stp;
3493 struct file *filp;
3494 struct file_lock file_lock;
8dc7c311 3495 struct file_lock conflock;
b37ad28b 3496 __be32 status = 0;
1da177e4 3497 unsigned int strhashval;
150b3934 3498 unsigned int cmd;
b8dd7b9a 3499 int err;
1da177e4
LT		 3500
3501 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3502 (long long) lock->lk_offset,
3503 (long long) lock->lk_length);
3504
1da177e4
LT		 3505 if (check_lock_length(lock->lk_offset, lock->lk_length))
3506 return nfserr_inval;
3507
ca364317 3508 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3509 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3510 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3511 return status;
3512 }
3513
1da177e4
LT		 3514 nfs4_lock_state();
3515
3516 if (lock->lk_is_new) {
893f8770
N		 3517 /*
3518 * Client indicates that this is a new lockowner.
3519 * Use open owner and open stateid to create lock owner and
3520 * lock stateid.
3521 */
1da177e4
LT		 3522 struct nfs4_stateid *open_stp = NULL;
3523 struct nfs4_file *fp;
3524
3525 status = nfserr_stale_clientid;
60adfc50
AA		 3526 if (!nfsd4_has_session(cstate) &&
3527 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3528 goto out;
1da177e4 3529
1da177e4 3530 /* validate and update open stateid and open seqid */
dd453dfd 3531 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3532 lock->lk_new_open_seqid,
3533 &lock->lk_new_open_stateid,
f3362737 3534 OPEN_STATE,
3a65588a 3535 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3536 lock);
37515177 3537 if (status)
1da177e4 3538 goto out;
3a65588a 3539 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3540 /* create lockowner and lock stateid */
3541 fp = open_stp->st_file;
3542 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3543 open_sop->so_client->cl_clientid.cl_id,
3544 &lock->v.new.owner);
3e9e3dbe
N		 3545 /* XXX: Do we need to check for duplicate stateowners on
3546 * the same file, or should they just be allowed (and
3547 * create new stateids)? */
1da177e4 3548 status = nfserr_resource;
b59e3c0e
NB		 3549 lock_sop = alloc_init_lock_stateowner(strhashval,
3550 open_sop->so_client, open_stp, lock);
3551 if (lock_sop == NULL)
1da177e4 3552 goto out;
b59e3c0e 3553 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3554 if (lock_stp == NULL)
1da177e4 3555 goto out;
1da177e4
LT		 3556 } else {
3557 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3558 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3559 lock->lk_old_lock_seqid,
3560 &lock->lk_old_lock_stateid,
f3362737 3561 LOCK_STATE,
3a65588a 3562 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3563 if (status)
3564 goto out;
3a65588a 3565 lock_sop = lock->lk_replay_owner;
1da177e4 3566 }
3a65588a 3567 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3568 filp = lock_stp->st_vfs_file;
3569
0dd395dc 3570 status = nfserr_grace;
af558e33 3571 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3572 goto out;
3573 status = nfserr_no_grace;
af558e33 3574 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3575 goto out;
3576
1da177e4
LT		 3577 locks_init_lock(&file_lock);
3578 switch (lock->lk_type) {
3579 case NFS4_READ_LT:
3580 case NFS4_READW_LT:
3581 file_lock.fl_type = F_RDLCK;
150b3934 3582 cmd = F_SETLK;
1da177e4
LT		 3583 break;
3584 case NFS4_WRITE_LT:
3585 case NFS4_WRITEW_LT:
3586 file_lock.fl_type = F_WRLCK;
150b3934 3587 cmd = F_SETLK;
1da177e4
LT		 3588 break;
3589 default:
3590 status = nfserr_inval;
3591 goto out;
3592 }
b59e3c0e 3593 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3594 file_lock.fl_pid = current->tgid;
3595 file_lock.fl_file = filp;
3596 file_lock.fl_flags = FL_POSIX;
d5b9026a 3597 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3598
3599 file_lock.fl_start = lock->lk_offset;
87df4de8 3600 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3601 nfs4_transform_lock_offset(&file_lock);
3602
3603 /*
3604 * Try to lock the file in the VFS.
3605 * Note: locks.c uses the BKL to protect the inode's lock list.
3606 */
3607
fd85b817 3608 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3609 switch (-err) {
1da177e4
LT		 3610 case 0: /* success! */
3611 update_stateid(&lock_stp->st_stateid);
3612 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3613 sizeof(stateid_t));
b8dd7b9a 3614 status = 0;
eb76b3fd
AA		 3615 break;
3616 case (EAGAIN): /* conflock holds conflicting lock */
3617 status = nfserr_denied;
3618 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3619 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3620 break;
1da177e4
LT		 3621 case (EDEADLK):
3622 status = nfserr_deadlock;
eb76b3fd 3623 break;
1da177e4 3624 default:
fd85b817 3625 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3626 status = nfserr_resource;
3627 break;
1da177e4 3628 }
1da177e4 3629out:
8a280510 3630 if (status && lock->lk_is_new && lock_sop)
f044ff83 3631 release_lockowner(lock_sop);
3a65588a
BF		 3632 if (lock->lk_replay_owner) {
3633 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3634 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3635 }
1da177e4
LT		 3636 nfs4_unlock_state();
3637 return status;
3638}
3639
55ef1274
BF		 3640/*
3641 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3642 * so we do a temporary open here just to get an open file to pass to
3643 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3644 * inode operation.)
3645 */
3646static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3647{
3648 struct file *file;
3649 int err;
3650
3651 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3652 if (err)
3653 return err;
3654 err = vfs_test_lock(file, lock);
3655 nfsd_close(file);
3656 return err;
3657}
3658
1da177e4
LT		 3659/*
3660 * LOCKT operation
3661 */
b37ad28b 3662__be32
ca364317
BF		 3663nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3664 struct nfsd4_lockt *lockt)
1da177e4
LT		 3665{
3666 struct inode *inode;
1da177e4 3667 struct file_lock file_lock;
fd85b817 3668 int error;
b37ad28b 3669 __be32 status;
1da177e4 3670
af558e33 3671 if (locks_in_grace())
1da177e4
LT		 3672 return nfserr_grace;
3673
3674 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3675 return nfserr_inval;
3676
3677 lockt->lt_stateowner = NULL;
3678 nfs4_lock_state();
3679
3680 status = nfserr_stale_clientid;
60adfc50 3681 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3682 goto out;
1da177e4 3683
ca364317 3684 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3685 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3686 if (status == nfserr_symlink)
3687 status = nfserr_inval;
3688 goto out;
3689 }
3690
ca364317 3691 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3692 locks_init_lock(&file_lock);
3693 switch (lockt->lt_type) {
3694 case NFS4_READ_LT:
3695 case NFS4_READW_LT:
3696 file_lock.fl_type = F_RDLCK;
3697 break;
3698 case NFS4_WRITE_LT:
3699 case NFS4_WRITEW_LT:
3700 file_lock.fl_type = F_WRLCK;
3701 break;
3702 default:
2fdada03 3703 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3704 status = nfserr_inval;
3705 goto out;
3706 }
3707
3708 lockt->lt_stateowner = find_lockstateowner_str(inode,
3709 &lockt->lt_clientid, &lockt->lt_owner);
3710 if (lockt->lt_stateowner)
3711 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3712 file_lock.fl_pid = current->tgid;
3713 file_lock.fl_flags = FL_POSIX;
3714
3715 file_lock.fl_start = lockt->lt_offset;
87df4de8 3716 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3717
3718 nfs4_transform_lock_offset(&file_lock);
3719
1da177e4 3720 status = nfs_ok;
55ef1274 3721 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3722 if (error) {
3723 status = nfserrno(error);
3724 goto out;
3725 }
9d6a8c5c 3726 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3727 status = nfserr_denied;
9d6a8c5c 3728 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3729 }
3730out:
3731 nfs4_unlock_state();
3732 return status;
3733}
3734
b37ad28b 3735__be32
ca364317 3736nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3737 struct nfsd4_locku *locku)
1da177e4
LT		 3738{
3739 struct nfs4_stateid *stp;
3740 struct file *filp = NULL;
3741 struct file_lock file_lock;
b37ad28b 3742 __be32 status;
b8dd7b9a 3743 int err;
1da177e4
LT		 3744
3745 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3746 (long long) locku->lu_offset,
3747 (long long) locku->lu_length);
3748
3749 if (check_lock_length(locku->lu_offset, locku->lu_length))
3750 return nfserr_inval;
3751
3752 nfs4_lock_state();
3753
dd453dfd 3754 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3755 locku->lu_seqid,
3756 &locku->lu_stateid,
f3362737 3757 LOCK_STATE,
1da177e4
LT		 3758 &locku->lu_stateowner, &stp, NULL)))
3759 goto out;
3760
3761 filp = stp->st_vfs_file;
3762 BUG_ON(!filp);
3763 locks_init_lock(&file_lock);
3764 file_lock.fl_type = F_UNLCK;
3765 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3766 file_lock.fl_pid = current->tgid;
3767 file_lock.fl_file = filp;
3768 file_lock.fl_flags = FL_POSIX;
d5b9026a 3769 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3770 file_lock.fl_start = locku->lu_offset;
3771
87df4de8 3772 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3773 nfs4_transform_lock_offset(&file_lock);
3774
3775 /*
3776 * Try to unlock the file in the VFS.
3777 */
fd85b817 3778 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3779 if (err) {
fd85b817 3780 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3781 goto out_nfserr;
3782 }
3783 /*
3784 * OK, unlock succeeded; the only thing left to do is update the stateid.
3785 */
3786 update_stateid(&stp->st_stateid);
3787 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3788
3789out:
f2327d9a 3790 if (locku->lu_stateowner) {
1da177e4 3791 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3792 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3793 }
1da177e4
LT		 3794 nfs4_unlock_state();
3795 return status;
3796
3797out_nfserr:
b8dd7b9a 3798 status = nfserrno(err);
1da177e4
LT		 3799 goto out;
3800}
3801
3802/*
3803 * returns
3804 * 1: locks held by lockowner
3805 * 0: no locks held by lockowner
3806 */
3807static int
3808check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3809{
3810 struct file_lock **flpp;
7eaa36e2 3811 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3812 int status = 0;
3813
3814 lock_kernel();
3815 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3816 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3817 status = 1;
3818 goto out;
796dadfd 3819 }
1da177e4
LT		 3820 }
3821out:
3822 unlock_kernel();
3823 return status;
3824}
3825
b37ad28b 3826__be32
b591480b
BF		 3827nfsd4_release_lockowner(struct svc_rqst *rqstp,
3828 struct nfsd4_compound_state *cstate,
3829 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3830{
3831 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3832 struct nfs4_stateowner *sop;
3833 struct nfs4_stateid *stp;
1da177e4 3834 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3835 struct list_head matches;
3836 int i;
b37ad28b 3837 __be32 status;
1da177e4
LT		 3838
3839 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3840 clid->cl_boot, clid->cl_id);
3841
3842 /* XXX check for lease expiration */
3843
3844 status = nfserr_stale_clientid;
849823c5 3845 if (STALE_CLIENTID(clid))
1da177e4 3846 return status;
1da177e4
LT		 3847
3848 nfs4_lock_state();
3849
3e9e3dbe
N		 3850 status = nfserr_locks_held;
3851 /* XXX: we're doing a linear search through all the lockowners.
3852 * Yipes! For now we'll just hope clients aren't really using
3853 * release_lockowner much, but eventually we have to fix these
3854 * data structures. */
3855 INIT_LIST_HEAD(&matches);
3856 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3857 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3858 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3859 continue;
3860 list_for_each_entry(stp, &sop->so_stateids,
3861 st_perstateowner) {
3862 if (check_for_locks(stp->st_vfs_file, sop))
3863 goto out;
3864 /* Note: so_perclient unused for lockowners,
3865 * so it's OK to fool with here. */
3866 list_add(&sop->so_perclient, &matches);
3867 }
1da177e4 3868 }
3e9e3dbe
N		 3869 }
3870 /* Clients probably won't expect us to return with some (but not all)
3871 * of the lockowner state released; so don't release any until all
3872 * have been checked. */
3873 status = nfs_ok;
0fa822e4
N		 3874 while (!list_empty(&matches)) {
3875 sop = list_entry(matches.next, struct nfs4_stateowner,
3876 so_perclient);
3877 /* unhash_stateowner deletes so_perclient only
3878 * for openowners. */
3879 list_del(&sop->so_perclient);
f044ff83 3880 release_lockowner(sop);
1da177e4
LT		 3881 }
3882out:
3883 nfs4_unlock_state();
3884 return status;
3885}
3886
3887static inline struct nfs4_client_reclaim *
a55370a3 3888alloc_reclaim(void)
1da177e4 3889{
a55370a3 3890 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3891}
3892
c7b9a459 3893int
a1bcecd2 3894nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3895{
3896 unsigned int strhashval = clientstr_hashval(name);
3897 struct nfs4_client *clp;
3898
a1bcecd2 3899 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3900 return clp ? 1 : 0;
3901}
3902
1da177e4
LT		 3903/*
3904 * failure => all reset bets are off, nfserr_no_grace...
3905 */
190e4fbf
N		 3906int
3907nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3908{
3909 unsigned int strhashval;
3910 struct nfs4_client_reclaim *crp = NULL;
3911
a55370a3
N		 3912 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3913 crp = alloc_reclaim();
1da177e4
LT		 3914 if (!crp)
3915 return 0;
a55370a3 3916 strhashval = clientstr_hashval(name);
1da177e4
LT		 3917 INIT_LIST_HEAD(&crp->cr_strhash);
3918 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3919 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3920 reclaim_str_hashtbl_size++;
3921 return 1;
3922}
3923
3924static void
3925nfs4_release_reclaim(void)
3926{
3927 struct nfs4_client_reclaim *crp = NULL;
3928 int i;
3929
1da177e4
LT		 3930 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3931 while (!list_empty(&reclaim_str_hashtbl[i])) {
3932 crp = list_entry(reclaim_str_hashtbl[i].next,
3933 struct nfs4_client_reclaim, cr_strhash);
3934 list_del(&crp->cr_strhash);
1da177e4
LT		 3935 kfree(crp);
3936 reclaim_str_hashtbl_size--;
3937 }
3938 }
3939 BUG_ON(reclaim_str_hashtbl_size);
3940}
3941
3942/*
3943 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3944static struct nfs4_client_reclaim *
1da177e4
LT		 3945nfs4_find_reclaim_client(clientid_t *clid)
3946{
3947 unsigned int strhashval;
3948 struct nfs4_client *clp;
3949 struct nfs4_client_reclaim *crp = NULL;
3950
3951
3952 /* find clientid in conf_id_hashtbl */
3953 clp = find_confirmed_client(clid);
3954 if (clp == NULL)
3955 return NULL;
3956
a55370a3
N		 3957 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3958 clp->cl_name.len, clp->cl_name.data,
3959 clp->cl_recdir);
1da177e4
LT		 3960
3961 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3962 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3963 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3964 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3965 return crp;
3966 }
3967 }
3968 return NULL;
3969}
3970
3971/*
3972* Called from OPEN. Look for clientid in reclaim list.
3973*/
b37ad28b 3974__be32
1da177e4
LT		 3975nfs4_check_open_reclaim(clientid_t *clid)
3976{
dfc83565 3977 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3978}
3979
ac4d8ff2 3980/* initialization to perform at module load time: */
1da177e4 3981
e8ff2a84 3982int
ac4d8ff2 3983nfs4_state_init(void)
1da177e4 3984{
e8ff2a84 3985 int i, status;
1da177e4 3986
e8ff2a84
BF		 3987 status = nfsd4_init_slabs();
3988 if (status)
3989 return status;
1da177e4
LT		 3990 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3991 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3992 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3993 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3994 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3995 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3996 }
5282fd72
ME		 3997 for (i = 0; i < SESSION_HASH_SIZE; i++)
3998 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3999 for (i = 0; i < FILE_HASH_SIZE; i++) {
4000 INIT_LIST_HEAD(&file_hashtbl[i]);
4001 }
4002 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4003 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4004 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4005 }
4006 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4007 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4008 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4009 }
4010 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4011 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4012 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4013 }
1da177e4 4014 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4015 INIT_LIST_HEAD(&close_lru);
4016 INIT_LIST_HEAD(&client_lru);
4017 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4018 reclaim_str_hashtbl_size = 0;
e8ff2a84 4019 return 0;
ac4d8ff2
N		 4020}
4021
190e4fbf
N		 4022static void
4023nfsd4_load_reboot_recovery_data(void)
4024{
4025 int status;
4026
0964a3d3
N		 4027 nfs4_lock_state();
4028 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4029 status = nfsd4_recdir_load();
0964a3d3 4030 nfs4_unlock_state();
190e4fbf
N		 4031 if (status)
4032 printk("NFSD: Failure reading reboot recovery data\n");
4033}
4034
9a8db97e
ME		 4035unsigned long
4036get_nfs4_grace_period(void)
4037{
4038 return max(user_lease_time, lease_time) * HZ;
4039}
4040
c2f1a551
MS		 4041/*
4042 * Since the lifetime of a delegation isn't limited to that of an open, a
4043 * client may quite reasonably hang on to a delegation as long as it has
4044 * the inode cached. This becomes an obvious problem the first time a
4045 * client's inode cache approaches the size of the server's total memory.
4046 *
4047 * For now we avoid this problem by imposing a hard limit on the number
4048 * of delegations, which varies according to the server's memory size.
4049 */
4050static void
4051set_max_delegations(void)
4052{
4053 /*
4054 * Allow at most 4 delegations per megabyte of RAM. Quick
4055 * estimates suggest that in the worst case (where every delegation
4056 * is for a different inode), a delegation could take about 1.5K,
4057 * giving a worst case usage of about 6% of memory.
4058 */
4059 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4060}
4061
ac4d8ff2
N		 4062/* initialization to perform when the nfsd service is started: */
4063
4064static void
4065__nfs4_state_start(void)
4066{
9a8db97e 4067 unsigned long grace_time;
ac4d8ff2 4068
1da177e4 4069 boot_time = get_seconds();
af558e33 4070 grace_time = get_nfs4_grace_period();
d99a05ad 4071 lease_time = user_lease_time;
af558e33 4072 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4073 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4074 grace_time/HZ);
58da282b 4075 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4076 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4077 set_max_delegations();
1da177e4
LT		 4078}
4079
e8ff2a84 4080void
76a3550e 4081nfs4_state_start(void)
1da177e4 4082{
1da177e4 4083 if (nfs4_init)
e8ff2a84 4084 return;
190e4fbf 4085 nfsd4_load_reboot_recovery_data();
76a3550e 4086 __nfs4_state_start();
1da177e4 4087 nfs4_init = 1;
e8ff2a84 4088 return;
1da177e4
LT		 4089}
4090
1da177e4
LT		 4091time_t
4092nfs4_lease_time(void)
4093{
4094 return lease_time;
4095}
4096
4097static void
4098__nfs4_state_shutdown(void)
4099{
4100 int i;
4101 struct nfs4_client *clp = NULL;
4102 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4103 struct list_head *pos, *next, reaplist;
4104
1da177e4
LT		 4105 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4106 while (!list_empty(&conf_id_hashtbl[i])) {
4107 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4108 expire_client(clp);
4109 }
4110 while (!list_empty(&unconf_str_hashtbl[i])) {
4111 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4112 expire_client(clp);
4113 }
4114 }
4115 INIT_LIST_HEAD(&reaplist);
4116 spin_lock(&recall_lock);
4117 list_for_each_safe(pos, next, &del_recall_lru) {
4118 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4119 list_move(&dp->dl_recall_lru, &reaplist);
4120 }
4121 spin_unlock(&recall_lock);
4122 list_for_each_safe(pos, next, &reaplist) {
4123 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4124 list_del_init(&dp->dl_recall_lru);
4125 unhash_delegation(dp);
4126 }
4127
190e4fbf 4128 nfsd4_shutdown_recdir();
1da177e4 4129 nfs4_init = 0;
1da177e4
LT		 4130}
4131
4132void
4133nfs4_state_shutdown(void)
4134{
5e8d5c29
N		 4135 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4136 destroy_workqueue(laundry_wq);
2c5e7615 4137 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4138 nfs4_lock_state();
4139 nfs4_release_reclaim();
4140 __nfs4_state_shutdown();
1da177e4
LT		 4141 nfs4_unlock_state();
4142}
4143
3dd98a3b
JL		 4144/*
4145 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4146 * accessed when nfsd is starting.
4147 */
0964a3d3
N		 4148static void
4149nfs4_set_recdir(char *recdir)
4150{
0964a3d3 4151 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4152}
4153
4154/*
4155 * Change the NFSv4 recovery directory to recdir.
4156 */
4157int
4158nfs4_reset_recoverydir(char *recdir)
4159{
4160 int status;
a63bb996 4161 struct path path;
0964a3d3 4162
a63bb996 4163 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4164 if (status)
4165 return status;
4166 status = -ENOTDIR;
a63bb996 4167 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4168 nfs4_set_recdir(recdir);
4169 status = 0;
4170 }
a63bb996 4171 path_put(&path);
0964a3d3
N		 4172 return status;
4173}
4174
3dd98a3b
JL		 4175char *
4176nfs4_recoverydir(void)
4177{
4178 return user_recovery_dirname;
4179}
4180
1da177e4
LT		 4181/*
4182 * Called when leasetime is changed.
4183 *
d99a05ad
N		 4184 * The only way the protocol gives us to handle on-the-fly lease changes is to
4185 * simulate a reboot. Instead of doing that, we just wait till the next time
4186 * we start to register any changes in lease time. If the administrator
4187 * really wants to change the lease time *now*, they can go ahead and bring
4188 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4189 *
4190 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4191 * when nfsd is starting
1da177e4
LT		 4192 */
4193void
4194nfs4_reset_lease(time_t leasetime)
4195{
d99a05ad 4196 user_lease_time = leasetime;
1da177e4 4197}
Linux 6.x block layer and io_uring tree(s)