projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd: make nfs4_client->cl_addr a struct sockaddr_storage
[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
1da177e4
LT		 900/* a helper function for parse_callback */
901static int
902parse_octet(unsigned int *lenp, char **addrp)
903{
904 unsigned int len = *lenp;
905 char *p = *addrp;
906 int n = -1;
907 char c;
908
909 for (;;) {
910 if (!len)
911 break;
912 len--;
913 c = *p++;
914 if (c == '.')
915 break;
916 if ((c < '0') || (c > '9')) {
917 n = -1;
918 break;
919 }
920 if (n < 0)
921 n = 0;
922 n = (n * 10) + (c - '0');
923 if (n > 255) {
924 n = -1;
925 break;
926 }
927 }
928 *lenp = len;
929 *addrp = p;
930 return n;
931}
932
933/* parse and set the setclientid ipv4 callback address */
fd39ca9a 934static int
1da177e4
LT		 935parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
936{
937 int temp = 0;
938 u32 cbaddr = 0;
939 u16 cbport = 0;
940 u32 addrlen = addr_len;
941 char *addr = addr_val;
942 int i, shift;
943
944 /* ipaddress */
945 shift = 24;
946 for(i = 4; i > 0 ; i--) {
947 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
948 return 0;
949 }
950 cbaddr |= (temp << shift);
951 if (shift > 0)
952 shift -= 8;
953 }
954 *cbaddrp = cbaddr;
955
956 /* port */
957 shift = 8;
958 for(i = 2; i > 0 ; i--) {
959 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
960 return 0;
961 }
962 cbport |= (temp << shift);
963 if (shift > 0)
964 shift -= 8;
965 }
966 *cbportp = cbport;
967 return 1;
968}
969
fd39ca9a 970static void
1da177e4
LT		 971gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
972{
c237dc03 973 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT		 974
975 /* Currently, we only support tcp for the callback channel */
976 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
977 goto out_err;
978
979 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
980 &cb->cb_addr, &cb->cb_port)))
981 goto out_err;
ab52ae6d 982 cb->cb_minorversion = 0;
1da177e4
LT		 983 cb->cb_prog = se->se_callback_prog;
984 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 985 return;
986out_err:
849823c5 987 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 988 "will not receive delegations\n",
989 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
990
1da177e4
LT		 991 return;
992}
993
074fe897
AA		 994void
995nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
996{
997 struct nfsd4_compoundres *resp = rqstp->rq_resp;
998
999 resp->cstate.statp = statp;
1000}
1001
1002/*
1003 * Dereference the result pages.
1004 */
1005static void
1006nfsd4_release_respages(struct page **respages, short resused)
1007{
1008 int i;
1009
1010 dprintk("--> %s\n", __func__);
1011 for (i = 0; i < resused; i++) {
1012 if (!respages[i])
1013 continue;
1014 put_page(respages[i]);
1015 respages[i] = NULL;
1016 }
1017}
1018
1019static void
1020nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1021{
1022 int i;
1023
1024 for (i = 0; i < count; i++) {
1025 topages[i] = frompages[i];
1026 if (!topages[i])
1027 continue;
1028 get_page(topages[i]);
1029 }
1030}
1031
1032/*
1033 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1034 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1035 * length of the XDR response is less than se_fmaxresp_cached
1036 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1037 * of the reply (e.g. readdir).
1038 *
1039 * Store the base and length of the rq_req.head[0] page
1040 * of the NFSv4.1 data, just past the rpc header.
1041 */
1042void
1043nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1044{
1045 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1046 struct svc_rqst *rqstp = resp->rqstp;
1047 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1048 struct nfsd4_op *op = &args->ops[resp->opcnt];
1049 struct kvec *resv = &rqstp->rq_res.head[0];
1050
1051 dprintk("--> %s entry %p\n", __func__, entry);
1052
1053 /* Don't cache a failed OP_SEQUENCE. */
1054 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1055 return;
bf864a31 1056
074fe897 1057 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1058 entry->ce_opcnt = resp->opcnt;
1059 entry->ce_status = resp->cstate.status;
1060
1061 /*
1062 * Don't need a page to cache just the sequence operation - the slot
1063 * does this for us!
1064 */
1065
1066 if (nfsd4_not_cached(resp)) {
1067 entry->ce_resused = 0;
1068 entry->ce_rpchdrlen = 0;
1069 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1070 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1071 return;
1072 }
074fe897
AA		 1073 entry->ce_resused = rqstp->rq_resused;
1074 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1075 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1076 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1077 entry->ce_resused);
074fe897
AA		 1078 entry->ce_datav.iov_base = resp->cstate.statp;
1079 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1080 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1081 /* Current request rpc header length*/
1082 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1083 (char *)page_address(rqstp->rq_respages[0]);
1084}
1085
1086/*
1087 * We keep the rpc header, but take the nfs reply from the replycache.
1088 */
1089static int
1090nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1091 struct nfsd4_cache_entry *entry)
1092{
1093 struct svc_rqst *rqstp = resp->rqstp;
1094 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1095 int len;
1096
1097 /* Current request rpc header length*/
1098 len = (char *)resp->cstate.statp -
1099 (char *)page_address(rqstp->rq_respages[0]);
1100 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1101 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1102 entry->ce_datav.iov_len);
1103 return 0;
1104 }
1105 /* copy the cached reply nfsd data past the current rpc header */
1106 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1107 entry->ce_datav.iov_len);
1108 resv->iov_len = len + entry->ce_datav.iov_len;
1109 return 1;
1110}
1111
abfabf8c
AA		 1112/*
1113 * Encode the replay sequence operation from the slot values.
1114 * If cachethis is FALSE encode the uncached rep error on the next
1115 * operation which sets resp->p and increments resp->opcnt for
1116 * nfs4svc_encode_compoundres.
1117 *
1118 */
1119static __be32
1120nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1121 struct nfsd4_compoundres *resp)
1122{
1123 struct nfsd4_op *op;
1124 struct nfsd4_slot *slot = resp->cstate.slot;
1125
1126 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1127 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
1128
1129 /* Encode the replayed sequence operation */
1130 op = &args->ops[resp->opcnt - 1];
1131 nfsd4_encode_operation(resp, op);
1132
1133 /* Return nfserr_retry_uncached_rep in next operation. */
1134 if (args->opcnt > 1 && slot->sl_cache_entry.ce_cachethis == 0) {
1135 op = &args->ops[resp->opcnt++];
1136 op->status = nfserr_retry_uncached_rep;
1137 nfsd4_encode_operation(resp, op);
1138 }
1139 return op->status;
1140}
1141
074fe897
AA		 1142/*
1143 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1144 * cached page. Replace any futher replay pages from the cache.
1145 */
1146__be32
bf864a31
AA		 1147nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1148 struct nfsd4_sequence *seq)
074fe897
AA		 1149{
1150 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1151 __be32 status;
1152
1153 dprintk("--> %s entry %p\n", __func__, entry);
1154
bf864a31
AA		 1155 /*
1156 * If this is just the sequence operation, we did not keep
1157 * a page in the cache entry because we can just use the
1158 * slot info stored in struct nfsd4_sequence that was checked
1159 * against the slot in nfsd4_sequence().
1160 *
1161 * This occurs when seq->cachethis is FALSE, or when the client
1162 * session inactivity timer fires and a solo sequence operation
1163 * is sent (lease renewal).
1164 */
abfabf8c
AA		 1165 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
1166
1167 /* Either returns 0 or nfserr_retry_uncached */
1168 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1169 if (status == nfserr_retry_uncached_rep)
1170 return status;
074fe897
AA		 1171
1172 if (!nfsd41_copy_replay_data(resp, entry)) {
1173 /*
1174 * Not enough room to use the replay rpc header, send the
1175 * cached header. Release all the allocated result pages.
1176 */
1177 svc_free_res_pages(resp->rqstp);
1178 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1179 entry->ce_resused);
1180 } else {
1181 /* Release all but the first allocated result page */
1182
1183 resp->rqstp->rq_resused--;
1184 svc_free_res_pages(resp->rqstp);
1185
1186 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1187 &entry->ce_respages[1],
1188 entry->ce_resused - 1);
1189 }
1190
1191 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1192 resp->opcnt = entry->ce_opcnt;
1193 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1194 status = entry->ce_status;
1195
1196 return status;
1197}
1198
0733d213
AA		 1199/*
1200 * Set the exchange_id flags returned by the server.
1201 */
1202static void
1203nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1204{
1205 /* pNFS is not supported */
1206 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1207
1208 /* Referrals are supported, Migration is not. */
1209 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1210
1211 /* set the wire flags to return to client. */
1212 clid->flags = new->cl_exchange_flags;
1213}
1214
069b6ad4
AA		 1215__be32
1216nfsd4_exchange_id(struct svc_rqst *rqstp,
1217 struct nfsd4_compound_state *cstate,
1218 struct nfsd4_exchange_id *exid)
1219{
0733d213
AA		 1220 struct nfs4_client *unconf, *conf, *new;
1221 int status;
1222 unsigned int strhashval;
1223 char dname[HEXDIR_LEN];
363168b4 1224 char addr_str[INET6_ADDRSTRLEN];
0733d213 1225 nfs4_verifier verf = exid->verifier;
363168b4 1226 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1227
363168b4 1228 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1229 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1230 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1231 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1232 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1233
1234 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1235 return nfserr_inval;
1236
1237 /* Currently only support SP4_NONE */
1238 switch (exid->spa_how) {
1239 case SP4_NONE:
1240 break;
1241 case SP4_SSV:
1242 return nfserr_encr_alg_unsupp;
1243 default:
1244 BUG(); /* checked by xdr code */
1245 case SP4_MACH_CRED:
1246 return nfserr_serverfault; /* no excuse :-/ */
1247 }
1248
1249 status = nfs4_make_rec_clidname(dname, &exid->clname);
1250
1251 if (status)
1252 goto error;
1253
1254 strhashval = clientstr_hashval(dname);
1255
1256 nfs4_lock_state();
1257 status = nfs_ok;
1258
a1bcecd2 1259 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1260 if (conf) {
1261 if (!same_verf(&verf, &conf->cl_verifier)) {
1262 /* 18.35.4 case 8 */
1263 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1264 status = nfserr_not_same;
1265 goto out;
1266 }
1267 /* Client reboot: destroy old state */
1268 expire_client(conf);
1269 goto out_new;
1270 }
1271 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1272 /* 18.35.4 case 9 */
1273 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1274 status = nfserr_perm;
1275 goto out;
1276 }
1277 expire_client(conf);
1278 goto out_new;
1279 }
0733d213
AA		 1280 /*
1281 * Set bit when the owner id and verifier map to an already
1282 * confirmed client id (18.35.3).
1283 */
1284 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1285
1286 /*
1287 * Falling into 18.35.4 case 2, possible router replay.
1288 * Leave confirmed record intact and return same result.
1289 */
1290 copy_verf(conf, &verf);
1291 new = conf;
1292 goto out_copy;
6ddbbbfe
MS		 1293 }
1294
1295 /* 18.35.4 case 7 */
1296 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1297 status = nfserr_noent;
1298 goto out;
0733d213
AA		 1299 }
1300
a1bcecd2 1301 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1302 if (unconf) {
1303 /*
1304 * Possible retry or client restart. Per 18.35.4 case 4,
1305 * a new unconfirmed record should be generated regardless
1306 * of whether any properties have changed.
1307 */
1308 expire_client(unconf);
1309 }
1310
1311out_new:
1312 /* Normal case */
1313 new = create_client(exid->clname, dname);
1314 if (new == NULL) {
1315 status = nfserr_resource;
1316 goto out;
1317 }
1318
1319 copy_verf(new, &verf);
1320 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1321 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1322 gen_clid(new);
1323 gen_confirm(new);
1324 add_to_unconfirmed(new, strhashval);
1325out_copy:
1326 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1327 exid->clientid.cl_id = new->cl_clientid.cl_id;
1328
38eb76a5 1329 exid->seqid = 1;
0733d213
AA		 1330 nfsd4_set_ex_flags(new, exid);
1331
1332 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1333 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1334 status = nfs_ok;
1335
1336out:
1337 nfs4_unlock_state();
1338error:
1339 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1340 return status;
069b6ad4
AA		 1341}
1342
b85d4c01 1343static int
88e588d5 1344check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1345{
88e588d5
AA		 1346 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1347 slot_seqid);
b85d4c01
BH		 1348
1349 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1350 if (slot_inuse) {
1351 if (seqid == slot_seqid)
b85d4c01
BH		 1352 return nfserr_jukebox;
1353 else
1354 return nfserr_seq_misordered;
1355 }
1356 /* Normal */
88e588d5 1357 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1358 return nfs_ok;
1359 /* Replay */
88e588d5 1360 if (seqid == slot_seqid)
b85d4c01
BH		 1361 return nfserr_replay_cache;
1362 /* Wraparound */
88e588d5 1363 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1364 return nfs_ok;
1365 /* Misordered replay or misordered new request */
1366 return nfserr_seq_misordered;
1367}
1368
49557cc7
AA		 1369/*
1370 * Cache the create session result into the create session single DRC
1371 * slot cache by saving the xdr structure. sl_seqid has been set.
1372 * Do this for solo or embedded create session operations.
1373 */
1374static void
1375nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1376 struct nfsd4_clid_slot *slot, int nfserr)
1377{
1378 slot->sl_status = nfserr;
1379 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1380}
1381
1382static __be32
1383nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1384 struct nfsd4_clid_slot *slot)
1385{
1386 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1387 return slot->sl_status;
1388}
1389
069b6ad4
AA		 1390__be32
1391nfsd4_create_session(struct svc_rqst *rqstp,
1392 struct nfsd4_compound_state *cstate,
1393 struct nfsd4_create_session *cr_ses)
1394{
363168b4 1395 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1396 struct nfs4_client *conf, *unconf;
49557cc7 1397 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1398 int status = 0;
1399
1400 nfs4_lock_state();
1401 unconf = find_unconfirmed_client(&cr_ses->clientid);
1402 conf = find_confirmed_client(&cr_ses->clientid);
1403
1404 if (conf) {
49557cc7
AA		 1405 cs_slot = &conf->cl_cs_slot;
1406 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1407 if (status == nfserr_replay_cache) {
ec6b5d7b 1408 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1409 cs_slot->sl_seqid);
38eb76a5 1410 /* Return the cached reply status */
49557cc7 1411 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1412 goto out;
49557cc7 1413 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1414 status = nfserr_seq_misordered;
1415 dprintk("Sequence misordered!\n");
1416 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1417 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1418 goto out;
1419 }
49557cc7 1420 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1421 } else if (unconf) {
1422 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1423 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1424 status = nfserr_clid_inuse;
1425 goto out;
1426 }
1427
49557cc7
AA		 1428 cs_slot = &unconf->cl_cs_slot;
1429 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1430 if (status) {
1431 /* an unconfirmed replay returns misordered */
ec6b5d7b 1432 status = nfserr_seq_misordered;
49557cc7 1433 goto out_cache;
ec6b5d7b
AA		 1434 }
1435
49557cc7 1436 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1437 move_to_confirmed(unconf);
1438
1439 /*
1440 * We do not support RDMA or persistent sessions
1441 */
1442 cr_ses->flags &= ~SESSION4_PERSIST;
1443 cr_ses->flags &= ~SESSION4_RDMA;
1444
1445 conf = unconf;
1446 } else {
1447 status = nfserr_stale_clientid;
1448 goto out;
1449 }
1450
1451 status = alloc_init_session(rqstp, conf, cr_ses);
1452 if (status)
1453 goto out;
1454
ec6b5d7b
AA		 1455 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1456 NFS4_MAX_SESSIONID_LEN);
49557cc7 1457 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1458
49557cc7
AA		 1459out_cache:
1460 /* cache solo and embedded create sessions under the state lock */
1461 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1462out:
1463 nfs4_unlock_state();
1464 dprintk("%s returns %d\n", __func__, ntohl(status));
1465 return status;
069b6ad4
AA		 1466}
1467
1468__be32
1469nfsd4_destroy_session(struct svc_rqst *r,
1470 struct nfsd4_compound_state *cstate,
1471 struct nfsd4_destroy_session *sessionid)
1472{
e10e0cfc
BH		 1473 struct nfsd4_session *ses;
1474 u32 status = nfserr_badsession;
1475
1476 /* Notes:
1477 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1478 * - Should we return nfserr_back_chan_busy if waiting for
1479 * callbacks on to-be-destroyed session?
1480 * - Do we need to clear any callback info from previous session?
1481 */
1482
1483 dump_sessionid(__func__, &sessionid->sessionid);
1484 spin_lock(&sessionid_lock);
1485 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1486 if (!ses) {
1487 spin_unlock(&sessionid_lock);
1488 goto out;
1489 }
1490
1491 unhash_session(ses);
1492 spin_unlock(&sessionid_lock);
1493
1494 /* wait for callbacks */
1495 shutdown_callback_client(ses->se_client);
1496 nfsd4_put_session(ses);
1497 status = nfs_ok;
1498out:
1499 dprintk("%s returns %d\n", __func__, ntohl(status));
1500 return status;
069b6ad4
AA		 1501}
1502
1503__be32
b85d4c01 1504nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1505 struct nfsd4_compound_state *cstate,
1506 struct nfsd4_sequence *seq)
1507{
f9bb94c4 1508 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1509 struct nfsd4_session *session;
1510 struct nfsd4_slot *slot;
1511 int status;
1512
f9bb94c4
AA		 1513 if (resp->opcnt != 1)
1514 return nfserr_sequence_pos;
1515
b85d4c01
BH		 1516 spin_lock(&sessionid_lock);
1517 status = nfserr_badsession;
1518 session = find_in_sessionid_hashtbl(&seq->sessionid);
1519 if (!session)
1520 goto out;
1521
1522 status = nfserr_badslot;
6c18ba9f 1523 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1524 goto out;
1525
1526 slot = &session->se_slots[seq->slotid];
1527 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1528
88e588d5 1529 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1530 if (status == nfserr_replay_cache) {
1531 cstate->slot = slot;
1532 cstate->session = session;
da3846a2 1533 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1534 * for nfsd4_svc_encode_compoundres processing */
1535 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1536 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1537 goto replay_cache;
1538 }
1539 if (status)
1540 goto out;
1541
1542 /* Success! bump slot seqid */
1543 slot->sl_inuse = true;
1544 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1545 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1546 /* Always set the cache entry cachethis for solo sequence */
1547 if (nfsd4_is_solo_sequence(resp))
1548 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1549
1550 cstate->slot = slot;
1551 cstate->session = session;
1552
1553replay_cache:
1554 /* Renew the clientid on success and on replay.
1555 * Hold a session reference until done processing the compound:
1556 * nfsd4_put_session called only if the cstate slot is set.
1557 */
1558 renew_client(session->se_client);
1559 nfsd4_get_session(session);
1560out:
1561 spin_unlock(&sessionid_lock);
1562 dprintk("%s: return %d\n", __func__, ntohl(status));
1563 return status;
069b6ad4
AA		 1564}
1565
b37ad28b 1566__be32
b591480b
BF		 1567nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1568 struct nfsd4_setclientid *setclid)
1da177e4 1569{
363168b4 1570 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1571 struct xdr_netobj clname = {
1572 .len = setclid->se_namelen,
1573 .data = setclid->se_name,
1574 };
1575 nfs4_verifier clverifier = setclid->se_verf;
1576 unsigned int strhashval;
28ce6054 1577 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1578 __be32 status;
68e76ad0 1579 char *princ;
a55370a3 1580 char dname[HEXDIR_LEN];
1da177e4 1581
1da177e4 1582 if (!check_name(clname))
73aea4ec 1583 return nfserr_inval;
1da177e4 1584
a55370a3
N		 1585 status = nfs4_make_rec_clidname(dname, &clname);
1586 if (status)
73aea4ec 1587 return status;
a55370a3 1588
1da177e4
LT		 1589 /*
1590 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1591 * We get here on a DRC miss.
1592 */
1593
a55370a3 1594 strhashval = clientstr_hashval(dname);
1da177e4 1595
1da177e4 1596 nfs4_lock_state();
a1bcecd2 1597 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1598 if (conf) {
a186e767 1599 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1600 status = nfserr_clid_inuse;
026722c2 1601 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1602 char addr_str[INET6_ADDRSTRLEN];
1603 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1604 sizeof(addr_str));
1605 dprintk("NFSD: setclientid: string in use by client "
1606 "at %s\n", addr_str);
1da177e4
LT		 1607 goto out;
1608 }
1da177e4 1609 }
a186e767
BF		 1610 /*
1611 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1612 * has a description of SETCLIENTID request processing consisting
1613 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1614 */
a1bcecd2 1615 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1616 status = nfserr_resource;
1617 if (!conf) {
a186e767
BF		 1618 /*
1619 * RFC 3530 14.2.33 CASE 4:
1620 * placed first, because it is the normal case
1da177e4
LT		 1621 */
1622 if (unconf)
1623 expire_client(unconf);
a55370a3
N		 1624 new = create_client(clname, dname);
1625 if (new == NULL)
1da177e4 1626 goto out;
1da177e4 1627 gen_clid(new);
599e0a22 1628 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1629 /*
a186e767
BF		 1630 * RFC 3530 14.2.33 CASE 1:
1631 * probable callback update
1da177e4 1632 */
31f4a6c1
N		 1633 if (unconf) {
1634 /* Note this is removing unconfirmed {*x***},
1635 * which is stronger than RFC recommended {vxc**}.
1636 * This has the advantage that there is at most
1637 * one {*x***} in either list at any time.
1638 */
1639 expire_client(unconf);
1da177e4 1640 }
a55370a3
N		 1641 new = create_client(clname, dname);
1642 if (new == NULL)
1da177e4 1643 goto out;
1da177e4 1644 copy_clid(new, conf);
1da177e4
LT		 1645 } else if (!unconf) {
1646 /*
a186e767
BF		 1647 * RFC 3530 14.2.33 CASE 2:
1648 * probable client reboot; state will be removed if
1649 * confirmed.
1da177e4 1650 */
a55370a3
N		 1651 new = create_client(clname, dname);
1652 if (new == NULL)
1da177e4 1653 goto out;
1da177e4 1654 gen_clid(new);
49ba8781 1655 } else {
a186e767
BF		 1656 /*
1657 * RFC 3530 14.2.33 CASE 3:
1658 * probable client reboot; state will be removed if
1659 * confirmed.
1da177e4
LT		 1660 */
1661 expire_client(unconf);
a55370a3
N		 1662 new = create_client(clname, dname);
1663 if (new == NULL)
1da177e4 1664 goto out;
1da177e4 1665 gen_clid(new);
1da177e4 1666 }
c175b83c 1667 copy_verf(new, &clverifier);
363168b4 1668 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1669 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1670 princ = svc_gss_principal(rqstp);
1671 if (princ) {
1672 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1673 if (new->cl_principal == NULL) {
1674 free_client(new);
1675 goto out;
1676 }
1677 }
c175b83c
BF		 1678 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1679 gen_confirm(new);
1680 gen_callback(new, setclid);
1681 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1682 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1683 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1684 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1685 status = nfs_ok;
1686out:
1687 nfs4_unlock_state();
1688 return status;
1689}
1690
1691
1692/*
a186e767
BF		 1693 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1694 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1695 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1696 */
b37ad28b 1697__be32
b591480b
BF		 1698nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1699 struct nfsd4_compound_state *cstate,
1700 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1701{
363168b4 1702 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1703 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1704 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1705 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1706 __be32 status;
1da177e4
LT		 1707
1708 if (STALE_CLIENTID(clid))
1709 return nfserr_stale_clientid;
1710 /*
1711 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1712 * We get here on a DRC miss.
1713 */
1714
1715 nfs4_lock_state();
21ab45a4
N		 1716
1717 conf = find_confirmed_client(clid);
1718 unconf = find_unconfirmed_client(clid);
1719
1720 status = nfserr_clid_inuse;
363168b4 1721 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1722 goto out;
363168b4 1723 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1724 goto out;
1725
a186e767
BF		 1726 /*
1727 * section 14.2.34 of RFC 3530 has a description of
1728 * SETCLIENTID_CONFIRM request processing consisting
1729 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1730 */
366e0c1d 1731 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1732 /*
1733 * RFC 3530 14.2.34 CASE 1:
1734 * callback update
1735 */
599e0a22 1736 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1737 status = nfserr_clid_inuse;
1738 else {
1a69c179
N		 1739 /* XXX: We just turn off callbacks until we can handle
1740 * change request correctly. */
c237dc03 1741 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1742 expire_client(unconf);
1da177e4 1743 status = nfs_ok;
1a69c179 1744
1da177e4 1745 }
f3aba4e5 1746 } else if (conf && !unconf) {
a186e767
BF		 1747 /*
1748 * RFC 3530 14.2.34 CASE 2:
1749 * probable retransmitted request; play it safe and
1750 * do nothing.
7c79f737 1751 */
599e0a22 1752 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1753 status = nfserr_clid_inuse;
21ab45a4 1754 else
1da177e4 1755 status = nfs_ok;
7c79f737 1756 } else if (!conf && unconf
599e0a22 1757 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1758 /*
1759 * RFC 3530 14.2.34 CASE 3:
1760 * Normal case; new or rebooted client:
7c79f737 1761 */
599e0a22 1762 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1763 status = nfserr_clid_inuse;
1764 } else {
1a69c179
N		 1765 unsigned int hash =
1766 clientstr_hashval(unconf->cl_recdir);
1767 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1768 hash, false);
1a69c179 1769 if (conf) {
c7b9a459 1770 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1771 expire_client(conf);
1772 }
1da177e4 1773 move_to_confirmed(unconf);
21ab45a4 1774 conf = unconf;
46f8a64b 1775 nfsd4_probe_callback(conf);
1a69c179 1776 status = nfs_ok;
1da177e4 1777 }
599e0a22
BF		 1778 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1779 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1780 &confirm)))) {
a186e767
BF		 1781 /*
1782 * RFC 3530 14.2.34 CASE 4:
1783 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1784 */
1da177e4 1785 status = nfserr_stale_clientid;
7c79f737 1786 } else {
08e8987c
N		 1787 /* check that we have hit one of the cases...*/
1788 status = nfserr_clid_inuse;
1789 }
1da177e4 1790out:
1da177e4
LT		 1791 nfs4_unlock_state();
1792 return status;
1793}
1794
1da177e4
LT		 1795/* OPEN Share state helper functions */
1796static inline struct nfs4_file *
1797alloc_init_file(struct inode *ino)
1798{
1799 struct nfs4_file *fp;
1800 unsigned int hashval = file_hashval(ino);
1801
e60d4398
N		 1802 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1803 if (fp) {
8b671b80 1804 atomic_set(&fp->fi_ref, 1);
1da177e4 1805 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1806 INIT_LIST_HEAD(&fp->fi_stateids);
1807 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1808 spin_lock(&recall_lock);
1da177e4 1809 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1810 spin_unlock(&recall_lock);
1da177e4
LT		 1811 fp->fi_inode = igrab(ino);
1812 fp->fi_id = current_fileid++;
47f9940c 1813 fp->fi_had_conflict = false;
1da177e4
LT		 1814 return fp;
1815 }
1816 return NULL;
1817}
1818
e60d4398 1819static void
e18b890b 1820nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1821{
e60d4398
N		 1822 if (*slab == NULL)
1823 return;
1a1d92c1 1824 kmem_cache_destroy(*slab);
e60d4398 1825 *slab = NULL;
1da177e4
LT		 1826}
1827
e8ff2a84 1828void
1da177e4
LT		 1829nfsd4_free_slabs(void)
1830{
e60d4398
N		 1831 nfsd4_free_slab(&stateowner_slab);
1832 nfsd4_free_slab(&file_slab);
5ac049ac 1833 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1834 nfsd4_free_slab(&deleg_slab);
e60d4398 1835}
1da177e4 1836
e60d4398
N		 1837static int
1838nfsd4_init_slabs(void)
1839{
1840 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1841 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1842 if (stateowner_slab == NULL)
1843 goto out_nomem;
1844 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1845 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1846 if (file_slab == NULL)
1847 goto out_nomem;
5ac049ac 1848 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1849 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1850 if (stateid_slab == NULL)
1851 goto out_nomem;
5b2d21c1 1852 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1853 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1854 if (deleg_slab == NULL)
1855 goto out_nomem;
e60d4398
N		 1856 return 0;
1857out_nomem:
1858 nfsd4_free_slabs();
1859 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1860 return -ENOMEM;
1da177e4
LT		 1861}
1862
1863void
1864nfs4_free_stateowner(struct kref *kref)
1865{
1866 struct nfs4_stateowner *sop =
1867 container_of(kref, struct nfs4_stateowner, so_ref);
1868 kfree(sop->so_owner.data);
1869 kmem_cache_free(stateowner_slab, sop);
1870}
1871
1872static inline struct nfs4_stateowner *
1873alloc_stateowner(struct xdr_netobj *owner)
1874{
1875 struct nfs4_stateowner *sop;
1876
1877 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1878 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1879 memcpy(sop->so_owner.data, owner->data, owner->len);
1880 sop->so_owner.len = owner->len;
1881 kref_init(&sop->so_ref);
1882 return sop;
1883 }
1884 kmem_cache_free(stateowner_slab, sop);
1885 }
1886 return NULL;
1887}
1888
1889static struct nfs4_stateowner *
1890alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1891 struct nfs4_stateowner *sop;
1892 struct nfs4_replay *rp;
1893 unsigned int idhashval;
1894
1895 if (!(sop = alloc_stateowner(&open->op_owner)))
1896 return NULL;
1897 idhashval = ownerid_hashval(current_ownerid);
1898 INIT_LIST_HEAD(&sop->so_idhash);
1899 INIT_LIST_HEAD(&sop->so_strhash);
1900 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1901 INIT_LIST_HEAD(&sop->so_stateids);
1902 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1903 INIT_LIST_HEAD(&sop->so_close_lru);
1904 sop->so_time = 0;
1905 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1906 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1907 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1908 sop->so_is_open_owner = 1;
1909 sop->so_id = current_ownerid++;
1910 sop->so_client = clp;
1911 sop->so_seqid = open->op_seqid;
1912 sop->so_confirmed = 0;
1913 rp = &sop->so_replay;
de1ae286 1914 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1915 rp->rp_buflen = 0;
1916 rp->rp_buf = rp->rp_ibuf;
1917 return sop;
1918}
1919
1da177e4
LT		 1920static inline void
1921init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1922 struct nfs4_stateowner *sop = open->op_stateowner;
1923 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1924
1925 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1926 INIT_LIST_HEAD(&stp->st_perstateowner);
1927 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1928 INIT_LIST_HEAD(&stp->st_perfile);
1929 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1930 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1931 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1932 stp->st_stateowner = sop;
13cd2184 1933 get_nfs4_file(fp);
1da177e4 1934 stp->st_file = fp;
78155ed7 1935 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1936 stp->st_stateid.si_stateownerid = sop->so_id;
1937 stp->st_stateid.si_fileid = fp->fi_id;
1938 stp->st_stateid.si_generation = 0;
1939 stp->st_access_bmap = 0;
1940 stp->st_deny_bmap = 0;
84459a11
AA		 1941 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1942 &stp->st_access_bmap);
1da177e4 1943 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1944 stp->st_openstp = NULL;
1da177e4
LT		 1945}
1946
fd39ca9a 1947static void
1da177e4
LT		 1948move_to_close_lru(struct nfs4_stateowner *sop)
1949{
1950 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1951
358dd55a 1952 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1953 sop->so_time = get_seconds();
1954}
1955
1da177e4 1956static int
599e0a22
BF		 1957same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1958 clientid_t *clid)
1959{
1960 return (sop->so_owner.len == owner->len) &&
1961 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1962 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1963}
1964
1965static struct nfs4_stateowner *
1966find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1967{
1968 struct nfs4_stateowner *so = NULL;
1969
1970 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1971 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1972 return so;
1973 }
1974 return NULL;
1975}
1976
1977/* search file_hashtbl[] for file */
1978static struct nfs4_file *
1979find_file(struct inode *ino)
1980{
1981 unsigned int hashval = file_hashval(ino);
1982 struct nfs4_file *fp;
1983
8b671b80 1984 spin_lock(&recall_lock);
1da177e4 1985 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1986 if (fp->fi_inode == ino) {
1987 get_nfs4_file(fp);
8b671b80 1988 spin_unlock(&recall_lock);
1da177e4 1989 return fp;
13cd2184 1990 }
1da177e4 1991 }
8b671b80 1992 spin_unlock(&recall_lock);
1da177e4
LT		 1993 return NULL;
1994}
1995
d87a8ade 1996static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1997{
d87a8ade 1998 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1999 return 0;
d87a8ade
AA		 2000 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
2001 return 0;
2002 x &= ~NFS4_SHARE_ACCESS_MASK;
2003 if (minorversion && x) {
2004 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
2005 return 0;
2006 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
2007 return 0;
2008 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
2009 }
2010 if (x)
8838dc43
BF		 2011 return 0;
2012 return 1;
ba5a6a19
BF		 2013}
2014
8838dc43 2015static inline int deny_valid(u32 x)
ba5a6a19 2016{
8838dc43
BF		 2017 /* Note: unlike access bits, deny bits may be zero. */
2018 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 2019}
1da177e4 2020
4f83aa30
BF		 2021/*
2022 * We store the NONE, READ, WRITE, and BOTH bits separately in the
2023 * st_{access,deny}_bmap field of the stateid, in order to track not
2024 * only what share bits are currently in force, but also what
2025 * combinations of share bits previous opens have used. This allows us
2026 * to enforce the recommendation of rfc 3530 14.2.19 that the server
2027 * return an error if the client attempt to downgrade to a combination
2028 * of share bits not explicable by closing some of its previous opens.
2029 *
2030 * XXX: This enforcement is actually incomplete, since we don't keep
2031 * track of access/deny bit combinations; so, e.g., we allow:
2032 *
2033 * OPEN allow read, deny write
2034 * OPEN allow both, deny none
2035 * DOWNGRADE allow read, deny none
2036 *
2037 * which we should reject.
2038 */
fd39ca9a 2039static void
1da177e4
LT		 2040set_access(unsigned int *access, unsigned long bmap) {
2041 int i;
2042
2043 *access = 0;
2044 for (i = 1; i < 4; i++) {
2045 if (test_bit(i, &bmap))
2046 *access |= i;
2047 }
2048}
2049
fd39ca9a 2050static void
1da177e4
LT		 2051set_deny(unsigned int *deny, unsigned long bmap) {
2052 int i;
2053
2054 *deny = 0;
2055 for (i = 0; i < 4; i++) {
2056 if (test_bit(i, &bmap))
2057 *deny |= i ;
2058 }
2059}
2060
2061static int
2062test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2063 unsigned int access, deny;
2064
2065 set_access(&access, stp->st_access_bmap);
2066 set_deny(&deny, stp->st_deny_bmap);
2067 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2068 return 0;
2069 return 1;
2070}
2071
2072/*
2073 * Called to check deny when READ with all zero stateid or
2074 * WRITE with all zero or all one stateid
2075 */
b37ad28b 2076static __be32
1da177e4
LT		 2077nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2078{
2079 struct inode *ino = current_fh->fh_dentry->d_inode;
2080 struct nfs4_file *fp;
2081 struct nfs4_stateid *stp;
b37ad28b 2082 __be32 ret;
1da177e4
LT		 2083
2084 dprintk("NFSD: nfs4_share_conflict\n");
2085
2086 fp = find_file(ino);
13cd2184
N		 2087 if (!fp)
2088 return nfs_ok;
b700949b 2089 ret = nfserr_locked;
1da177e4 2090 /* Search for conflicting share reservations */
13cd2184
N		 2091 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2092 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2093 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2094 goto out;
1da177e4 2095 }
13cd2184
N		 2096 ret = nfs_ok;
2097out:
2098 put_nfs4_file(fp);
2099 return ret;
1da177e4
LT		 2100}
2101
2102static inline void
2103nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2104{
2105 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2106 drop_file_write_access(filp);
1da177e4
LT		 2107 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2108 }
2109}
2110
1da177e4
LT		 2111/*
2112 * Spawn a thread to perform a recall on the delegation represented
2113 * by the lease (file_lock)
2114 *
2115 * Called from break_lease() with lock_kernel() held.
2116 * Note: we assume break_lease will only call this *once* for any given
2117 * lease.
2118 */
2119static
2120void nfsd_break_deleg_cb(struct file_lock *fl)
2121{
63e4863f 2122 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2123
2124 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2125 if (!dp)
2126 return;
2127
2128 /* We're assuming the state code never drops its reference
2129 * without first removing the lease. Since we're in this lease
2130 * callback (and since the lease code is serialized by the kernel
2131 * lock) we know the server hasn't removed the lease yet, we know
2132 * it's safe to take a reference: */
2133 atomic_inc(&dp->dl_count);
cfdcad4d 2134 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2135
2136 spin_lock(&recall_lock);
2137 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2138 spin_unlock(&recall_lock);
2139
2140 /* only place dl_time is set. protected by lock_kernel*/
2141 dp->dl_time = get_seconds();
2142
0272e1fd
BF		 2143 /*
2144 * We don't want the locks code to timeout the lease for us;
2145 * we'll remove it ourself if the delegation isn't returned
2146 * in time.
2147 */
2148 fl->fl_break_time = 0;
1da177e4 2149
63e4863f
BF		 2150 dp->dl_file->fi_had_conflict = true;
2151 nfsd4_cb_recall(dp);
1da177e4
LT		 2152}
2153
2154/*
2155 * The file_lock is being reapd.
2156 *
2157 * Called by locks_free_lock() with lock_kernel() held.
2158 */
2159static
2160void nfsd_release_deleg_cb(struct file_lock *fl)
2161{
2162 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2163
2164 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2165
2166 if (!(fl->fl_flags & FL_LEASE) || !dp)
2167 return;
2168 dp->dl_flock = NULL;
2169}
2170
2171/*
2172 * Set the delegation file_lock back pointer.
2173 *
a9933cea 2174 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2175 */
2176static
2177void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2178{
2179 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2180
2181 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2182 if (!dp)
2183 return;
2184 dp->dl_flock = new;
2185}
2186
2187/*
a9933cea 2188 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2189 */
2190static
2191int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2192{
2193 struct nfs4_delegation *onlistd =
2194 (struct nfs4_delegation *)onlist->fl_owner;
2195 struct nfs4_delegation *tryd =
2196 (struct nfs4_delegation *)try->fl_owner;
2197
2198 if (onlist->fl_lmops != try->fl_lmops)
2199 return 0;
2200
2201 return onlistd->dl_client == tryd->dl_client;
2202}
2203
2204
2205static
2206int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2207{
2208 if (arg & F_UNLCK)
2209 return lease_modify(onlist, arg);
2210 else
2211 return -EAGAIN;
2212}
2213
fd39ca9a 2214static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2215 .fl_break = nfsd_break_deleg_cb,
2216 .fl_release_private = nfsd_release_deleg_cb,
2217 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2218 .fl_mylease = nfsd_same_client_deleg_cb,
2219 .fl_change = nfsd_change_deleg_cb,
2220};
2221
2222
b37ad28b 2223__be32
6668958f
AA		 2224nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2225 struct nfsd4_open *open)
1da177e4 2226{
1da177e4
LT		 2227 clientid_t *clientid = &open->op_clientid;
2228 struct nfs4_client *clp = NULL;
2229 unsigned int strhashval;
2230 struct nfs4_stateowner *sop = NULL;
2231
1da177e4 2232 if (!check_name(open->op_owner))
0f442aa2 2233 return nfserr_inval;
1da177e4
LT		 2234
2235 if (STALE_CLIENTID(&open->op_clientid))
2236 return nfserr_stale_clientid;
2237
2238 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2239 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2240 open->op_stateowner = sop;
2241 if (!sop) {
2242 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2243 clp = find_confirmed_client(clientid);
2244 if (clp == NULL)
0f442aa2
BF		 2245 return nfserr_expired;
2246 goto renew;
1da177e4 2247 }
6668958f
AA		 2248 /* When sessions are used, skip open sequenceid processing */
2249 if (nfsd4_has_session(cstate))
2250 goto renew;
0f442aa2
BF		 2251 if (!sop->so_confirmed) {
2252 /* Replace unconfirmed owners without checking for replay. */
2253 clp = sop->so_client;
f044ff83 2254 release_openowner(sop);
0f442aa2
BF		 2255 open->op_stateowner = NULL;
2256 goto renew;
2257 }
2258 if (open->op_seqid == sop->so_seqid - 1) {
2259 if (sop->so_replay.rp_buflen)
a90b061c 2260 return nfserr_replay_me;
0f442aa2
BF		 2261 /* The original OPEN failed so spectacularly
2262 * that we don't even have replay data saved!
2263 * Therefore, we have no choice but to continue
2264 * processing this OPEN; presumably, we'll
2265 * fail again for the same reason.
2266 */
2267 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2268 goto renew;
2269 }
2270 if (open->op_seqid != sop->so_seqid)
2271 return nfserr_bad_seqid;
1da177e4 2272renew:
0f442aa2
BF		 2273 if (open->op_stateowner == NULL) {
2274 sop = alloc_init_open_stateowner(strhashval, clp, open);
2275 if (sop == NULL)
2276 return nfserr_resource;
2277 open->op_stateowner = sop;
2278 }
2279 list_del_init(&sop->so_close_lru);
1da177e4 2280 renew_client(sop->so_client);
0f442aa2 2281 return nfs_ok;
1da177e4
LT		 2282}
2283
b37ad28b 2284static inline __be32
4a6e43e6
N		 2285nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2286{
2287 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2288 return nfserr_openmode;
2289 else
2290 return nfs_ok;
2291}
2292
52f4fb43
N		 2293static struct nfs4_delegation *
2294find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2295{
2296 struct nfs4_delegation *dp;
2297
ea1da636 2298 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2299 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2300 return dp;
2301 }
2302 return NULL;
2303}
2304
b37ad28b 2305static __be32
567d9829
N		 2306nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2307 struct nfs4_delegation **dp)
2308{
2309 int flags;
b37ad28b 2310 __be32 status = nfserr_bad_stateid;
567d9829
N		 2311
2312 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2313 if (*dp == NULL)
c44c5eeb 2314 goto out;
567d9829
N		 2315 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2316 RD_STATE : WR_STATE;
2317 status = nfs4_check_delegmode(*dp, flags);
2318 if (status)
2319 *dp = NULL;
c44c5eeb
N		 2320out:
2321 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2322 return nfs_ok;
2323 if (status)
2324 return status;
2325 open->op_stateowner->so_confirmed = 1;
2326 return nfs_ok;
567d9829
N		 2327}
2328
b37ad28b 2329static __be32
1da177e4
LT		 2330nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2331{
2332 struct nfs4_stateid *local;
b37ad28b 2333 __be32 status = nfserr_share_denied;
1da177e4
LT		 2334 struct nfs4_stateowner *sop = open->op_stateowner;
2335
8beefa24 2336 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2337 /* ignore lock owners */
2338 if (local->st_stateowner->so_is_open_owner == 0)
2339 continue;
2340 /* remember if we have seen this open owner */
2341 if (local->st_stateowner == sop)
2342 *stpp = local;
2343 /* check for conflicting share reservations */
2344 if (!test_share(local, open))
2345 goto out;
2346 }
2347 status = 0;
2348out:
2349 return status;
2350}
2351
5ac049ac
N		 2352static inline struct nfs4_stateid *
2353nfs4_alloc_stateid(void)
2354{
2355 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2356}
2357
b37ad28b 2358static __be32
1da177e4 2359nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2360 struct nfs4_delegation *dp,
1da177e4
LT		 2361 struct svc_fh *cur_fh, int flags)
2362{
2363 struct nfs4_stateid *stp;
1da177e4 2364
5ac049ac 2365 stp = nfs4_alloc_stateid();
1da177e4
LT		 2366 if (stp == NULL)
2367 return nfserr_resource;
2368
567d9829
N		 2369 if (dp) {
2370 get_file(dp->dl_vfs_file);
2371 stp->st_vfs_file = dp->dl_vfs_file;
2372 } else {
b37ad28b 2373 __be32 status;
567d9829
N		 2374 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2375 &stp->st_vfs_file);
2376 if (status) {
2377 if (status == nfserr_dropit)
2378 status = nfserr_jukebox;
5ac049ac 2379 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2380 return status;
2381 }
1da177e4 2382 }
1da177e4
LT		 2383 *stpp = stp;
2384 return 0;
2385}
2386
b37ad28b 2387static inline __be32
1da177e4
LT		 2388nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2389 struct nfsd4_open *open)
2390{
2391 struct iattr iattr = {
2392 .ia_valid = ATTR_SIZE,
2393 .ia_size = 0,
2394 };
2395 if (!open->op_truncate)
2396 return 0;
2397 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2398 return nfserr_inval;
1da177e4
LT		 2399 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2400}
2401
b37ad28b 2402static __be32
1da177e4
LT		 2403nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2404{
2405 struct file *filp = stp->st_vfs_file;
7eaa36e2 2406 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2407 unsigned int share_access, new_writer;
b37ad28b 2408 __be32 status;
1da177e4
LT		 2409
2410 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2411 new_writer = (~share_access) & open->op_share_access
2412 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2413
6c26d08f 2414 if (new_writer) {
b37ad28b
AV		 2415 int err = get_write_access(inode);
2416 if (err)
2417 return nfserrno(err);
e518f056
BH		 2418 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2419 if (err)
2420 return nfserrno(err);
2421 file_take_write(filp);
6c26d08f 2422 }
1da177e4
LT		 2423 status = nfsd4_truncate(rqstp, cur_fh, open);
2424 if (status) {
6c26d08f
BF		 2425 if (new_writer)
2426 put_write_access(inode);
1da177e4
LT		 2427 return status;
2428 }
2429 /* remember the open */
6c26d08f 2430 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2431 __set_bit(open->op_share_access, &stp->st_access_bmap);
2432 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2433
2434 return nfs_ok;
2435}
2436
2437
1da177e4 2438static void
37515177 2439nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2440{
37515177
N		 2441 open->op_stateowner->so_confirmed = 1;
2442 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2443}
2444
2445/*
2446 * Attempt to hand out a delegation.
2447 */
2448static void
2449nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2450{
2451 struct nfs4_delegation *dp;
2452 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2453 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2454 struct file_lock fl, *flp = &fl;
2455 int status, flag = 0;
2456
2457 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2458 open->op_recall = 0;
2459 switch (open->op_claim_type) {
2460 case NFS4_OPEN_CLAIM_PREVIOUS:
2461 if (!atomic_read(&cb->cb_set))
2462 open->op_recall = 1;
2463 flag = open->op_delegate_type;
2464 if (flag == NFS4_OPEN_DELEGATE_NONE)
2465 goto out;
2466 break;
2467 case NFS4_OPEN_CLAIM_NULL:
2468 /* Let's not give out any delegations till everyone's
2469 * had the chance to reclaim theirs.... */
af558e33 2470 if (locks_in_grace())
7b190fec
N		 2471 goto out;
2472 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2473 goto out;
2474 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2475 flag = NFS4_OPEN_DELEGATE_WRITE;
2476 else
2477 flag = NFS4_OPEN_DELEGATE_READ;
2478 break;
2479 default:
2480 goto out;
2481 }
1da177e4
LT		 2482
2483 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2484 if (dp == NULL) {
2485 flag = NFS4_OPEN_DELEGATE_NONE;
2486 goto out;
2487 }
2488 locks_init_lock(&fl);
2489 fl.fl_lmops = &nfsd_lease_mng_ops;
2490 fl.fl_flags = FL_LEASE;
9167f501 2491 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2492 fl.fl_end = OFFSET_MAX;
2493 fl.fl_owner = (fl_owner_t)dp;
2494 fl.fl_file = stp->st_vfs_file;
2495 fl.fl_pid = current->tgid;
2496
a9933cea 2497 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2498 * the lock_manager callbacks fl_mylease and fl_change are used
2499 */
9167f501 2500 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2501 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2502 unhash_delegation(dp);
1da177e4
LT		 2503 flag = NFS4_OPEN_DELEGATE_NONE;
2504 goto out;
2505 }
2506
2507 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2508
2509 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2510 dp->dl_stateid.si_boot,
2511 dp->dl_stateid.si_stateownerid,
2512 dp->dl_stateid.si_fileid,
2513 dp->dl_stateid.si_generation);
2514out:
7b190fec
N		 2515 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2516 && flag == NFS4_OPEN_DELEGATE_NONE
2517 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2518 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2519 open->op_delegate_type = flag;
2520}
2521
2522/*
2523 * called with nfs4_lock_state() held.
2524 */
b37ad28b 2525__be32
1da177e4
LT		 2526nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2527{
6668958f 2528 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2529 struct nfs4_file *fp = NULL;
2530 struct inode *ino = current_fh->fh_dentry->d_inode;
2531 struct nfs4_stateid *stp = NULL;
567d9829 2532 struct nfs4_delegation *dp = NULL;
b37ad28b 2533 __be32 status;
1da177e4
LT		 2534
2535 status = nfserr_inval;
d87a8ade 2536 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2537 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2538 goto out;
2539 /*
2540 * Lookup file; if found, lookup stateid and check open request,
2541 * and check for delegations in the process of being recalled.
2542 * If not found, create the nfs4_file struct
2543 */
2544 fp = find_file(ino);
2545 if (fp) {
2546 if ((status = nfs4_check_open(fp, open, &stp)))
2547 goto out;
c44c5eeb
N		 2548 status = nfs4_check_deleg(fp, open, &dp);
2549 if (status)
2550 goto out;
1da177e4 2551 } else {
c44c5eeb
N		 2552 status = nfserr_bad_stateid;
2553 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2554 goto out;
1da177e4
LT		 2555 status = nfserr_resource;
2556 fp = alloc_init_file(ino);
2557 if (fp == NULL)
2558 goto out;
2559 }
2560
2561 /*
2562 * OPEN the file, or upgrade an existing OPEN.
2563 * If truncate fails, the OPEN fails.
2564 */
2565 if (stp) {
2566 /* Stateid was found, this is an OPEN upgrade */
2567 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2568 if (status)
2569 goto out;
444c2c07 2570 update_stateid(&stp->st_stateid);
1da177e4
LT		 2571 } else {
2572 /* Stateid was not found, this is a new OPEN */
2573 int flags = 0;
9ecb6a08 2574 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2575 flags |= NFSD_MAY_READ;
1da177e4 2576 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2577 flags |= NFSD_MAY_WRITE;
567d9829
N		 2578 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2579 if (status)
1da177e4
LT		 2580 goto out;
2581 init_stateid(stp, fp, open);
2582 status = nfsd4_truncate(rqstp, current_fh, open);
2583 if (status) {
2283963f 2584 release_open_stateid(stp);
1da177e4
LT		 2585 goto out;
2586 }
6668958f
AA		 2587 if (nfsd4_has_session(&resp->cstate))
2588 update_stateid(&stp->st_stateid);
1da177e4
LT		 2589 }
2590 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2591
6668958f
AA		 2592 if (nfsd4_has_session(&resp->cstate))
2593 open->op_stateowner->so_confirmed = 1;
2594
1da177e4
LT		 2595 /*
2596 * Attempt to hand out a delegation. No error return, because the
2597 * OPEN succeeds even if we fail.
2598 */
2599 nfs4_open_delegation(current_fh, open, stp);
2600
2601 status = nfs_ok;
2602
2603 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2604 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2605 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2606out:
13cd2184
N		 2607 if (fp)
2608 put_nfs4_file(fp);
37515177
N		 2609 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2610 nfs4_set_claim_prev(open);
1da177e4
LT		 2611 /*
2612 * To finish the open response, we just need to set the rflags.
2613 */
2614 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2615 if (!open->op_stateowner->so_confirmed &&
2616 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2617 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2618
2619 return status;
2620}
2621
b37ad28b 2622__be32
b591480b
BF		 2623nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2624 clientid_t *clid)
1da177e4
LT		 2625{
2626 struct nfs4_client *clp;
b37ad28b 2627 __be32 status;
1da177e4
LT		 2628
2629 nfs4_lock_state();
2630 dprintk("process_renew(%08x/%08x): starting\n",
2631 clid->cl_boot, clid->cl_id);
2632 status = nfserr_stale_clientid;
2633 if (STALE_CLIENTID(clid))
2634 goto out;
2635 clp = find_confirmed_client(clid);
2636 status = nfserr_expired;
2637 if (clp == NULL) {
2638 /* We assume the client took too long to RENEW. */
2639 dprintk("nfsd4_renew: clientid not found!\n");
2640 goto out;
2641 }
2642 renew_client(clp);
2643 status = nfserr_cb_path_down;
ea1da636 2644 if (!list_empty(&clp->cl_delegations)
c237dc03 2645 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2646 goto out;
2647 status = nfs_ok;
2648out:
2649 nfs4_unlock_state();
2650 return status;
2651}
2652
af558e33
BF		 2653struct lock_manager nfsd4_manager = {
2654};
2655
a76b4319 2656static void
af558e33 2657nfsd4_end_grace(void)
a76b4319
N		 2658{
2659 dprintk("NFSD: end of grace period\n");
c7b9a459 2660 nfsd4_recdir_purge_old();
af558e33 2661 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2662}
2663
fd39ca9a 2664static time_t
1da177e4
LT		 2665nfs4_laundromat(void)
2666{
2667 struct nfs4_client *clp;
2668 struct nfs4_stateowner *sop;
2669 struct nfs4_delegation *dp;
2670 struct list_head *pos, *next, reaplist;
2671 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2672 time_t t, clientid_val = NFSD_LEASE_TIME;
2673 time_t u, test_val = NFSD_LEASE_TIME;
2674
2675 nfs4_lock_state();
2676
2677 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2678 if (locks_in_grace())
2679 nfsd4_end_grace();
1da177e4
LT		 2680 list_for_each_safe(pos, next, &client_lru) {
2681 clp = list_entry(pos, struct nfs4_client, cl_lru);
2682 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2683 t = clp->cl_time - cutoff;
2684 if (clientid_val > t)
2685 clientid_val = t;
2686 break;
2687 }
2688 dprintk("NFSD: purging unused client (clientid %08x)\n",
2689 clp->cl_clientid.cl_id);
c7b9a459 2690 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2691 expire_client(clp);
2692 }
2693 INIT_LIST_HEAD(&reaplist);
2694 spin_lock(&recall_lock);
2695 list_for_each_safe(pos, next, &del_recall_lru) {
2696 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2697 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2698 u = dp->dl_time - cutoff;
2699 if (test_val > u)
2700 test_val = u;
2701 break;
2702 }
2703 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2704 dp, dp->dl_flock);
2705 list_move(&dp->dl_recall_lru, &reaplist);
2706 }
2707 spin_unlock(&recall_lock);
2708 list_for_each_safe(pos, next, &reaplist) {
2709 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2710 list_del_init(&dp->dl_recall_lru);
2711 unhash_delegation(dp);
2712 }
2713 test_val = NFSD_LEASE_TIME;
2714 list_for_each_safe(pos, next, &close_lru) {
2715 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2716 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2717 u = sop->so_time - cutoff;
2718 if (test_val > u)
2719 test_val = u;
2720 break;
2721 }
2722 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2723 sop->so_id);
f044ff83 2724 release_openowner(sop);
1da177e4
LT		 2725 }
2726 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2727 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2728 nfs4_unlock_state();
2729 return clientid_val;
2730}
2731
a254b246
HH		 2732static struct workqueue_struct *laundry_wq;
2733static void laundromat_main(struct work_struct *);
2734static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2735
2736static void
c4028958 2737laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2738{
2739 time_t t;
2740
2741 t = nfs4_laundromat();
2742 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2743 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2744}
2745
fd39ca9a 2746static struct nfs4_stateowner *
f8816512
N		 2747search_close_lru(u32 st_id, int flags)
2748{
1da177e4
LT		 2749 struct nfs4_stateowner *local = NULL;
2750
1da177e4
LT		 2751 if (flags & CLOSE_STATE) {
2752 list_for_each_entry(local, &close_lru, so_close_lru) {
2753 if (local->so_id == st_id)
2754 return local;
2755 }
2756 }
2757 return NULL;
2758}
2759
2760static inline int
2761nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2762{
7eaa36e2 2763 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2764}
2765
2766static int
2767STALE_STATEID(stateid_t *stateid)
2768{
78155ed7
BN		 2769 if (time_after((unsigned long)boot_time,
2770 (unsigned long)stateid->si_boot)) {
2771 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2772 stateid->si_boot, stateid->si_stateownerid,
2773 stateid->si_fileid, stateid->si_generation);
2774 return 1;
2775 }
2776 return 0;
2777}
2778
2779static int
2780EXPIRED_STATEID(stateid_t *stateid)
2781{
2782 if (time_before((unsigned long)boot_time,
2783 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2784 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2785 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2786 stateid->si_boot, stateid->si_stateownerid,
2787 stateid->si_fileid, stateid->si_generation);
2788 return 1;
2789 }
2790 return 0;
2791}
2792
2793static __be32
2794stateid_error_map(stateid_t *stateid)
2795{
2796 if (STALE_STATEID(stateid))
2797 return nfserr_stale_stateid;
2798 if (EXPIRED_STATEID(stateid))
2799 return nfserr_expired;
2800
2801 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2802 stateid->si_boot, stateid->si_stateownerid,
2803 stateid->si_fileid, stateid->si_generation);
2804 return nfserr_bad_stateid;
1da177e4
LT		 2805}
2806
2807static inline int
2808access_permit_read(unsigned long access_bmap)
2809{
2810 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2811 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2812 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2813}
2814
2815static inline int
2816access_permit_write(unsigned long access_bmap)
2817{
2818 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2819 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2820}
2821
2822static
b37ad28b 2823__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2824{
b37ad28b 2825 __be32 status = nfserr_openmode;
1da177e4
LT		 2826
2827 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2828 goto out;
2829 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2830 goto out;
2831 status = nfs_ok;
2832out:
2833 return status;
2834}
2835
b37ad28b 2836static inline __be32
1da177e4
LT		 2837check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2838{
203a8c8e 2839 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2840 return nfs_ok;
af558e33 2841 else if (locks_in_grace()) {
1da177e4
LT		 2842 /* Answer in remaining cases depends on existance of
2843 * conflicting state; so we must wait out the grace period. */
2844 return nfserr_grace;
2845 } else if (flags & WR_STATE)
2846 return nfs4_share_conflict(current_fh,
2847 NFS4_SHARE_DENY_WRITE);
2848 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2849 return nfs4_share_conflict(current_fh,
2850 NFS4_SHARE_DENY_READ);
2851}
2852
2853/*
2854 * Allow READ/WRITE during grace period on recovered state only for files
2855 * that are not able to provide mandatory locking.
2856 */
2857static inline int
18f82731 2858grace_disallows_io(struct inode *inode)
1da177e4 2859{
203a8c8e 2860 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2861}
2862
6668958f 2863static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2864{
6668958f
AA		 2865 /*
2866 * When sessions are used the stateid generation number is ignored
2867 * when it is zero.
2868 */
2869 if ((flags & HAS_SESSION) && in->si_generation == 0)
2870 goto out;
2871
0836f587
BF		 2872 /* If the client sends us a stateid from the future, it's buggy: */
2873 if (in->si_generation > ref->si_generation)
2874 return nfserr_bad_stateid;
2875 /*
2876 * The following, however, can happen. For example, if the
2877 * client sends an open and some IO at the same time, the open
2878 * may bump si_generation while the IO is still in flight.
2879 * Thanks to hard links and renames, the client never knows what
2880 * file an open will affect. So it could avoid that situation
2881 * only by serializing all opens and IO from the same open
2882 * owner. To recover from the old_stateid error, the client
2883 * will just have to retry the IO:
2884 */
2885 if (in->si_generation < ref->si_generation)
2886 return nfserr_old_stateid;
6668958f 2887out:
0836f587
BF		 2888 return nfs_ok;
2889}
2890
3e633079
BF		 2891static int is_delegation_stateid(stateid_t *stateid)
2892{
2893 return stateid->si_fileid == 0;
2894}
2895
1da177e4
LT		 2896/*
2897* Checks for stateid operations
2898*/
b37ad28b 2899__be32
dd453dfd
BH		 2900nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2901 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2902{
2903 struct nfs4_stateid *stp = NULL;
2904 struct nfs4_delegation *dp = NULL;
dd453dfd 2905 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2906 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2907 __be32 status;
1da177e4 2908
1da177e4
LT		 2909 if (filpp)
2910 *filpp = NULL;
2911
18f82731 2912 if (grace_disallows_io(ino))
1da177e4
LT		 2913 return nfserr_grace;
2914
6668958f
AA		 2915 if (nfsd4_has_session(cstate))
2916 flags |= HAS_SESSION;
2917
1da177e4
LT		 2918 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2919 return check_special_stateids(current_fh, stateid, flags);
2920
1da177e4
LT		 2921 status = nfserr_stale_stateid;
2922 if (STALE_STATEID(stateid))
2923 goto out;
2924
1da177e4 2925 status = nfserr_bad_stateid;
3e633079 2926 if (is_delegation_stateid(stateid)) {
a4455be0 2927 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2928 if (!dp) {
2929 status = stateid_error_map(stateid);
1da177e4 2930 goto out;
78155ed7 2931 }
6668958f
AA		 2932 status = check_stateid_generation(stateid, &dp->dl_stateid,
2933 flags);
0c2a498f
BF		 2934 if (status)
2935 goto out;
dc9bf700
BF		 2936 status = nfs4_check_delegmode(dp, flags);
2937 if (status)
2938 goto out;
2939 renew_client(dp->dl_client);
dc9bf700
BF		 2940 if (filpp)
2941 *filpp = dp->dl_vfs_file;
1da177e4 2942 } else { /* open or lock stateid */
a4455be0 2943 stp = find_stateid(stateid, flags);
78155ed7
BN		 2944 if (!stp) {
2945 status = stateid_error_map(stateid);
1da177e4 2946 goto out;
78155ed7 2947 }
6150ef0d 2948 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2949 goto out;
2950 if (!stp->st_stateowner->so_confirmed)
2951 goto out;
6668958f
AA		 2952 status = check_stateid_generation(stateid, &stp->st_stateid,
2953 flags);
0c2a498f
BF		 2954 if (status)
2955 goto out;
a4455be0
BF		 2956 status = nfs4_check_openmode(stp, flags);
2957 if (status)
1da177e4
LT		 2958 goto out;
2959 renew_client(stp->st_stateowner->so_client);
2960 if (filpp)
2961 *filpp = stp->st_vfs_file;
1da177e4
LT		 2962 }
2963 status = nfs_ok;
2964out:
2965 return status;
2966}
2967
4c4cd222
N		 2968static inline int
2969setlkflg (int type)
2970{
2971 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2972 RD_STATE : WR_STATE;
2973}
1da177e4
LT		 2974
2975/*
2976 * Checks for sequence id mutating operations.
2977 */
b37ad28b 2978static __be32
dd453dfd
BH		 2979nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2980 stateid_t *stateid, int flags,
2981 struct nfs4_stateowner **sopp,
2982 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2983{
1da177e4
LT		 2984 struct nfs4_stateid *stp;
2985 struct nfs4_stateowner *sop;
dd453dfd 2986 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2987 __be32 status;
1da177e4
LT		 2988
2989 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2990 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2991 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2992 stateid->si_generation);
3a4f98bb 2993
1da177e4
LT		 2994 *stpp = NULL;
2995 *sopp = NULL;
2996
1da177e4 2997 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2998 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2999 return nfserr_bad_stateid;
1da177e4
LT		 3000 }
3001
1da177e4 3002 if (STALE_STATEID(stateid))
3a4f98bb 3003 return nfserr_stale_stateid;
6668958f
AA		 3004
3005 if (nfsd4_has_session(cstate))
3006 flags |= HAS_SESSION;
3007
1da177e4
LT		 3008 /*
3009 * We return BAD_STATEID if filehandle doesn't match stateid,
3010 * the confirmed flag is incorrecly set, or the generation
3011 * number is incorrect.
1da177e4 3012 */
f8816512
N		 3013 stp = find_stateid(stateid, flags);
3014 if (stp == NULL) {
3015 /*
3016 * Also, we should make sure this isn't just the result of
3017 * a replayed close:
3018 */
3019 sop = search_close_lru(stateid->si_stateownerid, flags);
3020 if (sop == NULL)
78155ed7 3021 return stateid_error_map(stateid);
f8816512
N		 3022 *sopp = sop;
3023 goto check_replay;
3024 }
1da177e4 3025
39325bd0
BF		 3026 *stpp = stp;
3027 *sopp = sop = stp->st_stateowner;
3028
4c4cd222 3029 if (lock) {
4c4cd222 3030 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 3031 struct nfs4_client *clp = sop->so_client;
4c4cd222 3032 int lkflg = 0;
b37ad28b 3033 __be32 status;
4c4cd222
N		 3034
3035 lkflg = setlkflg(lock->lk_type);
3036
3037 if (lock->lk_is_new) {
599e0a22
BF		 3038 if (!sop->so_is_open_owner)
3039 return nfserr_bad_stateid;
60adfc50
AA		 3040 if (!(flags & HAS_SESSION) &&
3041 !same_clid(&clp->cl_clientid, lockclid))
3042 return nfserr_bad_stateid;
599e0a22
BF		 3043 /* stp is the open stateid */
3044 status = nfs4_check_openmode(stp, lkflg);
3045 if (status)
3046 return status;
3047 } else {
3048 /* stp is the lock stateid */
3049 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3050 if (status)
3051 return status;
4c4cd222 3052 }
1da177e4
LT		 3053 }
3054
f3362737 3055 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3056 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3057 return nfserr_bad_stateid;
1da177e4
LT		 3058 }
3059
1da177e4
LT		 3060 /*
3061 * We now validate the seqid and stateid generation numbers.
3062 * For the moment, we ignore the possibility of
3063 * generation number wraparound.
3064 */
6668958f 3065 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3066 goto check_replay;
3067
3a4f98bb 3068 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3069 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3070 " unconfirmed stateowner!\n");
3071 return nfserr_bad_stateid;
1da177e4 3072 }
3a4f98bb 3073 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3074 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3075 " confirmed yet!\n");
3076 return nfserr_bad_stateid;
1da177e4 3077 }
6668958f 3078 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3079 if (status)
3080 return status;
52fd004e 3081 renew_client(sop->so_client);
3a4f98bb 3082 return nfs_ok;
1da177e4 3083
1da177e4 3084check_replay:
bd9aac52 3085 if (seqid == sop->so_seqid - 1) {
849823c5 3086 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3087 /* indicate replay to calling function */
a90b061c 3088 return nfserr_replay_me;
1da177e4 3089 }
2fdada03 3090 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3091 sop->so_seqid, seqid);
3092 *sopp = NULL;
3093 return nfserr_bad_seqid;
1da177e4
LT		 3094}
3095
b37ad28b 3096__be32
ca364317 3097nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3098 struct nfsd4_open_confirm *oc)
1da177e4 3099{
b37ad28b 3100 __be32 status;
1da177e4
LT		 3101 struct nfs4_stateowner *sop;
3102 struct nfs4_stateid *stp;
3103
3104 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3105 (int)cstate->current_fh.fh_dentry->d_name.len,
3106 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3107
ca364317 3108 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3109 if (status)
3110 return status;
1da177e4
LT		 3111
3112 nfs4_lock_state();
3113
dd453dfd 3114 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3115 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3116 CONFIRM | OPEN_STATE,
1da177e4
LT		 3117 &oc->oc_stateowner, &stp, NULL)))
3118 goto out;
3119
3120 sop = oc->oc_stateowner;
3121 sop->so_confirmed = 1;
3122 update_stateid(&stp->st_stateid);
3123 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3124 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3125 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3126 stp->st_stateid.si_boot,
3127 stp->st_stateid.si_stateownerid,
3128 stp->st_stateid.si_fileid,
3129 stp->st_stateid.si_generation);
c7b9a459
N		 3130
3131 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3132out:
f2327d9a 3133 if (oc->oc_stateowner) {
1da177e4 3134 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3135 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3136 }
1da177e4
LT		 3137 nfs4_unlock_state();
3138 return status;
3139}
3140
3141
3142/*
3143 * unset all bits in union bitmap (bmap) that
3144 * do not exist in share (from successful OPEN_DOWNGRADE)
3145 */
3146static void
3147reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3148{
3149 int i;
3150 for (i = 1; i < 4; i++) {
3151 if ((i & access) != i)
3152 __clear_bit(i, bmap);
3153 }
3154}
3155
3156static void
3157reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3158{
3159 int i;
3160 for (i = 0; i < 4; i++) {
3161 if ((i & deny) != i)
3162 __clear_bit(i, bmap);
3163 }
3164}
3165
b37ad28b 3166__be32
ca364317
BF		 3167nfsd4_open_downgrade(struct svc_rqst *rqstp,
3168 struct nfsd4_compound_state *cstate,
a4f1706a 3169 struct nfsd4_open_downgrade *od)
1da177e4 3170{
b37ad28b 3171 __be32 status;
1da177e4
LT		 3172 struct nfs4_stateid *stp;
3173 unsigned int share_access;
3174
3175 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3176 (int)cstate->current_fh.fh_dentry->d_name.len,
3177 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3178
d87a8ade 3179 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3180 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3181 return nfserr_inval;
3182
3183 nfs4_lock_state();
dd453dfd 3184 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3185 od->od_seqid,
1da177e4 3186 &od->od_stateid,
f3362737 3187 OPEN_STATE,
1da177e4
LT		 3188 &od->od_stateowner, &stp, NULL)))
3189 goto out;
3190
3191 status = nfserr_inval;
3192 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3193 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3194 stp->st_access_bmap, od->od_share_access);
3195 goto out;
3196 }
3197 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3198 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3199 stp->st_deny_bmap, od->od_share_deny);
3200 goto out;
3201 }
3202 set_access(&share_access, stp->st_access_bmap);
3203 nfs4_file_downgrade(stp->st_vfs_file,
3204 share_access & ~od->od_share_access);
3205
3206 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3207 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3208
3209 update_stateid(&stp->st_stateid);
3210 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3211 status = nfs_ok;
3212out:
f2327d9a 3213 if (od->od_stateowner) {
1da177e4 3214 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3215 cstate->replay_owner = od->od_stateowner;
f2327d9a 3216 }
1da177e4
LT		 3217 nfs4_unlock_state();
3218 return status;
3219}
3220
3221/*
3222 * nfs4_unlock_state() called after encode
3223 */
b37ad28b 3224__be32
ca364317 3225nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3226 struct nfsd4_close *close)
1da177e4 3227{
b37ad28b 3228 __be32 status;
1da177e4
LT		 3229 struct nfs4_stateid *stp;
3230
3231 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3232 (int)cstate->current_fh.fh_dentry->d_name.len,
3233 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3234
3235 nfs4_lock_state();
3236 /* check close_lru for replay */
dd453dfd 3237 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3238 close->cl_seqid,
1da177e4 3239 &close->cl_stateid,
f3362737 3240 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3241 &close->cl_stateowner, &stp, NULL)))
3242 goto out;
1da177e4
LT		 3243 status = nfs_ok;
3244 update_stateid(&stp->st_stateid);
3245 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3246
04ef5954 3247 /* release_stateid() calls nfsd_close() if needed */
2283963f 3248 release_open_stateid(stp);
04ef5954
BF		 3249
3250 /* place unused nfs4_stateowners on so_close_lru list to be
3251 * released by the laundromat service after the lease period
3252 * to enable us to handle CLOSE replay
3253 */
3254 if (list_empty(&close->cl_stateowner->so_stateids))
3255 move_to_close_lru(close->cl_stateowner);
1da177e4 3256out:
f2327d9a 3257 if (close->cl_stateowner) {
1da177e4 3258 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3259 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3260 }
1da177e4
LT		 3261 nfs4_unlock_state();
3262 return status;
3263}
3264
b37ad28b 3265__be32
ca364317
BF		 3266nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3267 struct nfsd4_delegreturn *dr)
1da177e4 3268{
203a8c8e
BF		 3269 struct nfs4_delegation *dp;
3270 stateid_t *stateid = &dr->dr_stateid;
3271 struct inode *inode;
b37ad28b 3272 __be32 status;
6668958f 3273 int flags = 0;
1da177e4 3274
ca364317 3275 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3276 return status;
3277 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3278
6668958f
AA		 3279 if (nfsd4_has_session(cstate))
3280 flags |= HAS_SESSION;
1da177e4 3281 nfs4_lock_state();
203a8c8e
BF		 3282 status = nfserr_bad_stateid;
3283 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3284 goto out;
3285 status = nfserr_stale_stateid;
3286 if (STALE_STATEID(stateid))
3287 goto out;
7e0f7cf5 3288 status = nfserr_bad_stateid;
203a8c8e
BF		 3289 if (!is_delegation_stateid(stateid))
3290 goto out;
203a8c8e 3291 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3292 if (!dp) {
3293 status = stateid_error_map(stateid);
203a8c8e 3294 goto out;
78155ed7 3295 }
6668958f 3296 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3297 if (status)
3298 goto out;
3299 renew_client(dp->dl_client);
3300
3301 unhash_delegation(dp);
1da177e4 3302out:
203a8c8e
BF		 3303 nfs4_unlock_state();
3304
1da177e4
LT		 3305 return status;
3306}
3307
3308
3309/*
3310 * Lock owner state (byte-range locks)
3311 */
3312#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3313#define LOCK_HASH_BITS 8
3314#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3315#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3316
87df4de8
BH		 3317static inline u64
3318end_offset(u64 start, u64 len)
3319{
3320 u64 end;
3321
3322 end = start + len;
3323 return end >= start ? end: NFS4_MAX_UINT64;
3324}
3325
3326/* last octet in a range */
3327static inline u64
3328last_byte_offset(u64 start, u64 len)
3329{
3330 u64 end;
3331
3332 BUG_ON(!len);
3333 end = start + len;
3334 return end > start ? end - 1: NFS4_MAX_UINT64;
3335}
3336
1da177e4
LT		 3337#define lockownerid_hashval(id) \
3338 ((id) & LOCK_HASH_MASK)
3339
3340static inline unsigned int
3341lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3342 struct xdr_netobj *ownername)
3343{
3344 return (file_hashval(inode) + cl_id
3345 + opaque_hashval(ownername->data, ownername->len))
3346 & LOCK_HASH_MASK;
3347}
3348
3349static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3350static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3351static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3352
fd39ca9a 3353static struct nfs4_stateid *
1da177e4
LT		 3354find_stateid(stateid_t *stid, int flags)
3355{
9346eff0 3356 struct nfs4_stateid *local;
1da177e4
LT		 3357 u32 st_id = stid->si_stateownerid;
3358 u32 f_id = stid->si_fileid;
3359 unsigned int hashval;
3360
3361 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3362 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3363 hashval = stateid_hashval(st_id, f_id);
3364 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3365 if ((local->st_stateid.si_stateownerid == st_id) &&
3366 (local->st_stateid.si_fileid == f_id))
3367 return local;
3368 }
3369 }
9346eff0
KK		 3370
3371 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3372 hashval = stateid_hashval(st_id, f_id);
3373 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3374 if ((local->st_stateid.si_stateownerid == st_id) &&
3375 (local->st_stateid.si_fileid == f_id))
3376 return local;
3377 }
849823c5 3378 }
1da177e4
LT		 3379 return NULL;
3380}
3381
3382static struct nfs4_delegation *
3383find_delegation_stateid(struct inode *ino, stateid_t *stid)
3384{
13cd2184
N		 3385 struct nfs4_file *fp;
3386 struct nfs4_delegation *dl;
1da177e4
LT		 3387
3388 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3389 stid->si_boot, stid->si_stateownerid,
3390 stid->si_fileid, stid->si_generation);
3391
1da177e4 3392 fp = find_file(ino);
13cd2184
N		 3393 if (!fp)
3394 return NULL;
3395 dl = find_delegation_file(fp, stid);
3396 put_nfs4_file(fp);
3397 return dl;
1da177e4
LT		 3398}
3399
3400/*
3401 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3402 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3403 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3404 * locking, this prevents us from being completely protocol-compliant. The
3405 * real solution to this problem is to start using unsigned file offsets in
3406 * the VFS, but this is a very deep change!
3407 */
3408static inline void
3409nfs4_transform_lock_offset(struct file_lock *lock)
3410{
3411 if (lock->fl_start < 0)
3412 lock->fl_start = OFFSET_MAX;
3413 if (lock->fl_end < 0)
3414 lock->fl_end = OFFSET_MAX;
3415}
3416
d5b9026a
N		 3417/* Hack!: For now, we're defining this just so we can use a pointer to it
3418 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3419static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3420};
1da177e4
LT		 3421
3422static inline void
3423nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3424{
d5b9026a
N		 3425 struct nfs4_stateowner *sop;
3426 unsigned int hval;
1da177e4 3427
d5b9026a
N		 3428 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3429 sop = (struct nfs4_stateowner *) fl->fl_owner;
3430 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3431 kref_get(&sop->so_ref);
3432 deny->ld_sop = sop;
3433 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3434 } else {
3435 deny->ld_sop = NULL;
3436 deny->ld_clientid.cl_boot = 0;
3437 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3438 }
3439 deny->ld_start = fl->fl_start;
87df4de8
BH		 3440 deny->ld_length = NFS4_MAX_UINT64;
3441 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3442 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3443 deny->ld_type = NFS4_READ_LT;
3444 if (fl->fl_type != F_RDLCK)
3445 deny->ld_type = NFS4_WRITE_LT;
3446}
3447
1da177e4
LT		 3448static struct nfs4_stateowner *
3449find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3450 struct xdr_netobj *owner)
3451{
3452 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3453 struct nfs4_stateowner *op;
3454
3455 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3456 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3457 return op;
3458 }
3459 return NULL;
3460}
3461
3462/*
3463 * Alloc a lock owner structure.
3464 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3465 * occured.
3466 *
3467 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3468 */
3469
3470static struct nfs4_stateowner *
3471alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3472 struct nfs4_stateowner *sop;
3473 struct nfs4_replay *rp;
3474 unsigned int idhashval;
3475
3476 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3477 return NULL;
3478 idhashval = lockownerid_hashval(current_ownerid);
3479 INIT_LIST_HEAD(&sop->so_idhash);
3480 INIT_LIST_HEAD(&sop->so_strhash);
3481 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3482 INIT_LIST_HEAD(&sop->so_stateids);
3483 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3484 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3485 sop->so_time = 0;
3486 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3487 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3488 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3489 sop->so_is_open_owner = 0;
3490 sop->so_id = current_ownerid++;
3491 sop->so_client = clp;
b59e3c0e
NB		 3492 /* It is the openowner seqid that will be incremented in encode in the
3493 * case of new lockowners; so increment the lock seqid manually: */
3494 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3495 sop->so_confirmed = 1;
3496 rp = &sop->so_replay;
de1ae286 3497 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3498 rp->rp_buflen = 0;
3499 rp->rp_buf = rp->rp_ibuf;
3500 return sop;
3501}
3502
fd39ca9a 3503static struct nfs4_stateid *
1da177e4
LT		 3504alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3505{
3506 struct nfs4_stateid *stp;
3507 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3508
5ac049ac
N		 3509 stp = nfs4_alloc_stateid();
3510 if (stp == NULL)
1da177e4
LT		 3511 goto out;
3512 INIT_LIST_HEAD(&stp->st_hash);
3513 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3514 INIT_LIST_HEAD(&stp->st_perstateowner);
3515 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3516 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3517 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3518 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3519 stp->st_stateowner = sop;
13cd2184 3520 get_nfs4_file(fp);
1da177e4 3521 stp->st_file = fp;
78155ed7 3522 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3523 stp->st_stateid.si_stateownerid = sop->so_id;
3524 stp->st_stateid.si_fileid = fp->fi_id;
3525 stp->st_stateid.si_generation = 0;
3526 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3527 stp->st_access_bmap = open_stp->st_access_bmap;
3528 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3529 stp->st_openstp = open_stp;
1da177e4
LT		 3530
3531out:
3532 return stp;
3533}
3534
fd39ca9a 3535static int
1da177e4
LT		 3536check_lock_length(u64 offset, u64 length)
3537{
87df4de8 3538 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3539 LOFF_OVERFLOW(offset, length)));
3540}
3541
3542/*
3543 * LOCK operation
3544 */
b37ad28b 3545__be32
ca364317 3546nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3547 struct nfsd4_lock *lock)
1da177e4 3548{
3e9e3dbe 3549 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3550 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3551 struct nfs4_stateid *lock_stp;
3552 struct file *filp;
3553 struct file_lock file_lock;
8dc7c311 3554 struct file_lock conflock;
b37ad28b 3555 __be32 status = 0;
1da177e4 3556 unsigned int strhashval;
150b3934 3557 unsigned int cmd;
b8dd7b9a 3558 int err;
1da177e4
LT		 3559
3560 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3561 (long long) lock->lk_offset,
3562 (long long) lock->lk_length);
3563
1da177e4
LT		 3564 if (check_lock_length(lock->lk_offset, lock->lk_length))
3565 return nfserr_inval;
3566
ca364317 3567 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3568 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3569 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3570 return status;
3571 }
3572
1da177e4
LT		 3573 nfs4_lock_state();
3574
3575 if (lock->lk_is_new) {
893f8770
N		 3576 /*
3577 * Client indicates that this is a new lockowner.
3578 * Use open owner and open stateid to create lock owner and
3579 * lock stateid.
3580 */
1da177e4
LT		 3581 struct nfs4_stateid *open_stp = NULL;
3582 struct nfs4_file *fp;
3583
3584 status = nfserr_stale_clientid;
60adfc50
AA		 3585 if (!nfsd4_has_session(cstate) &&
3586 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3587 goto out;
1da177e4 3588
1da177e4 3589 /* validate and update open stateid and open seqid */
dd453dfd 3590 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3591 lock->lk_new_open_seqid,
3592 &lock->lk_new_open_stateid,
f3362737 3593 OPEN_STATE,
3a65588a 3594 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3595 lock);
37515177 3596 if (status)
1da177e4 3597 goto out;
3a65588a 3598 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3599 /* create lockowner and lock stateid */
3600 fp = open_stp->st_file;
3601 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3602 open_sop->so_client->cl_clientid.cl_id,
3603 &lock->v.new.owner);
3e9e3dbe
N		 3604 /* XXX: Do we need to check for duplicate stateowners on
3605 * the same file, or should they just be allowed (and
3606 * create new stateids)? */
1da177e4 3607 status = nfserr_resource;
b59e3c0e
NB		 3608 lock_sop = alloc_init_lock_stateowner(strhashval,
3609 open_sop->so_client, open_stp, lock);
3610 if (lock_sop == NULL)
1da177e4 3611 goto out;
b59e3c0e 3612 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3613 if (lock_stp == NULL)
1da177e4 3614 goto out;
1da177e4
LT		 3615 } else {
3616 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3617 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3618 lock->lk_old_lock_seqid,
3619 &lock->lk_old_lock_stateid,
f3362737 3620 LOCK_STATE,
3a65588a 3621 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3622 if (status)
3623 goto out;
3a65588a 3624 lock_sop = lock->lk_replay_owner;
1da177e4 3625 }
3a65588a 3626 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3627 filp = lock_stp->st_vfs_file;
3628
0dd395dc 3629 status = nfserr_grace;
af558e33 3630 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3631 goto out;
3632 status = nfserr_no_grace;
af558e33 3633 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3634 goto out;
3635
1da177e4
LT		 3636 locks_init_lock(&file_lock);
3637 switch (lock->lk_type) {
3638 case NFS4_READ_LT:
3639 case NFS4_READW_LT:
3640 file_lock.fl_type = F_RDLCK;
150b3934 3641 cmd = F_SETLK;
1da177e4
LT		 3642 break;
3643 case NFS4_WRITE_LT:
3644 case NFS4_WRITEW_LT:
3645 file_lock.fl_type = F_WRLCK;
150b3934 3646 cmd = F_SETLK;
1da177e4
LT		 3647 break;
3648 default:
3649 status = nfserr_inval;
3650 goto out;
3651 }
b59e3c0e 3652 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3653 file_lock.fl_pid = current->tgid;
3654 file_lock.fl_file = filp;
3655 file_lock.fl_flags = FL_POSIX;
d5b9026a 3656 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3657
3658 file_lock.fl_start = lock->lk_offset;
87df4de8 3659 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3660 nfs4_transform_lock_offset(&file_lock);
3661
3662 /*
3663 * Try to lock the file in the VFS.
3664 * Note: locks.c uses the BKL to protect the inode's lock list.
3665 */
3666
fd85b817 3667 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3668 switch (-err) {
1da177e4
LT		 3669 case 0: /* success! */
3670 update_stateid(&lock_stp->st_stateid);
3671 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3672 sizeof(stateid_t));
b8dd7b9a 3673 status = 0;
eb76b3fd
AA		 3674 break;
3675 case (EAGAIN): /* conflock holds conflicting lock */
3676 status = nfserr_denied;
3677 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3678 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3679 break;
1da177e4
LT		 3680 case (EDEADLK):
3681 status = nfserr_deadlock;
eb76b3fd 3682 break;
1da177e4 3683 default:
fd85b817 3684 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3685 status = nfserr_resource;
3686 break;
1da177e4 3687 }
1da177e4 3688out:
8a280510 3689 if (status && lock->lk_is_new && lock_sop)
f044ff83 3690 release_lockowner(lock_sop);
3a65588a
BF		 3691 if (lock->lk_replay_owner) {
3692 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3693 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3694 }
1da177e4
LT		 3695 nfs4_unlock_state();
3696 return status;
3697}
3698
55ef1274
BF		 3699/*
3700 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3701 * so we do a temporary open here just to get an open file to pass to
3702 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3703 * inode operation.)
3704 */
3705static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3706{
3707 struct file *file;
3708 int err;
3709
3710 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3711 if (err)
3712 return err;
3713 err = vfs_test_lock(file, lock);
3714 nfsd_close(file);
3715 return err;
3716}
3717
1da177e4
LT		 3718/*
3719 * LOCKT operation
3720 */
b37ad28b 3721__be32
ca364317
BF		 3722nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3723 struct nfsd4_lockt *lockt)
1da177e4
LT		 3724{
3725 struct inode *inode;
1da177e4 3726 struct file_lock file_lock;
fd85b817 3727 int error;
b37ad28b 3728 __be32 status;
1da177e4 3729
af558e33 3730 if (locks_in_grace())
1da177e4
LT		 3731 return nfserr_grace;
3732
3733 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3734 return nfserr_inval;
3735
3736 lockt->lt_stateowner = NULL;
3737 nfs4_lock_state();
3738
3739 status = nfserr_stale_clientid;
60adfc50 3740 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3741 goto out;
1da177e4 3742
ca364317 3743 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3744 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3745 if (status == nfserr_symlink)
3746 status = nfserr_inval;
3747 goto out;
3748 }
3749
ca364317 3750 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3751 locks_init_lock(&file_lock);
3752 switch (lockt->lt_type) {
3753 case NFS4_READ_LT:
3754 case NFS4_READW_LT:
3755 file_lock.fl_type = F_RDLCK;
3756 break;
3757 case NFS4_WRITE_LT:
3758 case NFS4_WRITEW_LT:
3759 file_lock.fl_type = F_WRLCK;
3760 break;
3761 default:
2fdada03 3762 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3763 status = nfserr_inval;
3764 goto out;
3765 }
3766
3767 lockt->lt_stateowner = find_lockstateowner_str(inode,
3768 &lockt->lt_clientid, &lockt->lt_owner);
3769 if (lockt->lt_stateowner)
3770 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3771 file_lock.fl_pid = current->tgid;
3772 file_lock.fl_flags = FL_POSIX;
3773
3774 file_lock.fl_start = lockt->lt_offset;
87df4de8 3775 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3776
3777 nfs4_transform_lock_offset(&file_lock);
3778
1da177e4 3779 status = nfs_ok;
55ef1274 3780 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3781 if (error) {
3782 status = nfserrno(error);
3783 goto out;
3784 }
9d6a8c5c 3785 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3786 status = nfserr_denied;
9d6a8c5c 3787 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3788 }
3789out:
3790 nfs4_unlock_state();
3791 return status;
3792}
3793
b37ad28b 3794__be32
ca364317 3795nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3796 struct nfsd4_locku *locku)
1da177e4
LT		 3797{
3798 struct nfs4_stateid *stp;
3799 struct file *filp = NULL;
3800 struct file_lock file_lock;
b37ad28b 3801 __be32 status;
b8dd7b9a 3802 int err;
1da177e4
LT		 3803
3804 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3805 (long long) locku->lu_offset,
3806 (long long) locku->lu_length);
3807
3808 if (check_lock_length(locku->lu_offset, locku->lu_length))
3809 return nfserr_inval;
3810
3811 nfs4_lock_state();
3812
dd453dfd 3813 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3814 locku->lu_seqid,
3815 &locku->lu_stateid,
f3362737 3816 LOCK_STATE,
1da177e4
LT		 3817 &locku->lu_stateowner, &stp, NULL)))
3818 goto out;
3819
3820 filp = stp->st_vfs_file;
3821 BUG_ON(!filp);
3822 locks_init_lock(&file_lock);
3823 file_lock.fl_type = F_UNLCK;
3824 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3825 file_lock.fl_pid = current->tgid;
3826 file_lock.fl_file = filp;
3827 file_lock.fl_flags = FL_POSIX;
d5b9026a 3828 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3829 file_lock.fl_start = locku->lu_offset;
3830
87df4de8 3831 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3832 nfs4_transform_lock_offset(&file_lock);
3833
3834 /*
3835 * Try to unlock the file in the VFS.
3836 */
fd85b817 3837 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3838 if (err) {
fd85b817 3839 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3840 goto out_nfserr;
3841 }
3842 /*
3843 * OK, unlock succeeded; the only thing left to do is update the stateid.
3844 */
3845 update_stateid(&stp->st_stateid);
3846 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3847
3848out:
f2327d9a 3849 if (locku->lu_stateowner) {
1da177e4 3850 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3851 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3852 }
1da177e4
LT		 3853 nfs4_unlock_state();
3854 return status;
3855
3856out_nfserr:
b8dd7b9a 3857 status = nfserrno(err);
1da177e4
LT		 3858 goto out;
3859}
3860
3861/*
3862 * returns
3863 * 1: locks held by lockowner
3864 * 0: no locks held by lockowner
3865 */
3866static int
3867check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3868{
3869 struct file_lock **flpp;
7eaa36e2 3870 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3871 int status = 0;
3872
3873 lock_kernel();
3874 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3875 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3876 status = 1;
3877 goto out;
796dadfd 3878 }
1da177e4
LT		 3879 }
3880out:
3881 unlock_kernel();
3882 return status;
3883}
3884
b37ad28b 3885__be32
b591480b
BF		 3886nfsd4_release_lockowner(struct svc_rqst *rqstp,
3887 struct nfsd4_compound_state *cstate,
3888 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3889{
3890 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3891 struct nfs4_stateowner *sop;
3892 struct nfs4_stateid *stp;
1da177e4 3893 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3894 struct list_head matches;
3895 int i;
b37ad28b 3896 __be32 status;
1da177e4
LT		 3897
3898 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3899 clid->cl_boot, clid->cl_id);
3900
3901 /* XXX check for lease expiration */
3902
3903 status = nfserr_stale_clientid;
849823c5 3904 if (STALE_CLIENTID(clid))
1da177e4 3905 return status;
1da177e4
LT		 3906
3907 nfs4_lock_state();
3908
3e9e3dbe
N		 3909 status = nfserr_locks_held;
3910 /* XXX: we're doing a linear search through all the lockowners.
3911 * Yipes! For now we'll just hope clients aren't really using
3912 * release_lockowner much, but eventually we have to fix these
3913 * data structures. */
3914 INIT_LIST_HEAD(&matches);
3915 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3916 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3917 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3918 continue;
3919 list_for_each_entry(stp, &sop->so_stateids,
3920 st_perstateowner) {
3921 if (check_for_locks(stp->st_vfs_file, sop))
3922 goto out;
3923 /* Note: so_perclient unused for lockowners,
3924 * so it's OK to fool with here. */
3925 list_add(&sop->so_perclient, &matches);
3926 }
1da177e4 3927 }
3e9e3dbe
N		 3928 }
3929 /* Clients probably won't expect us to return with some (but not all)
3930 * of the lockowner state released; so don't release any until all
3931 * have been checked. */
3932 status = nfs_ok;
0fa822e4
N		 3933 while (!list_empty(&matches)) {
3934 sop = list_entry(matches.next, struct nfs4_stateowner,
3935 so_perclient);
3936 /* unhash_stateowner deletes so_perclient only
3937 * for openowners. */
3938 list_del(&sop->so_perclient);
f044ff83 3939 release_lockowner(sop);
1da177e4
LT		 3940 }
3941out:
3942 nfs4_unlock_state();
3943 return status;
3944}
3945
3946static inline struct nfs4_client_reclaim *
a55370a3 3947alloc_reclaim(void)
1da177e4 3948{
a55370a3 3949 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3950}
3951
c7b9a459 3952int
a1bcecd2 3953nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3954{
3955 unsigned int strhashval = clientstr_hashval(name);
3956 struct nfs4_client *clp;
3957
a1bcecd2 3958 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3959 return clp ? 1 : 0;
3960}
3961
1da177e4
LT		 3962/*
3963 * failure => all reset bets are off, nfserr_no_grace...
3964 */
190e4fbf
N		 3965int
3966nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3967{
3968 unsigned int strhashval;
3969 struct nfs4_client_reclaim *crp = NULL;
3970
a55370a3
N		 3971 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3972 crp = alloc_reclaim();
1da177e4
LT		 3973 if (!crp)
3974 return 0;
a55370a3 3975 strhashval = clientstr_hashval(name);
1da177e4
LT		 3976 INIT_LIST_HEAD(&crp->cr_strhash);
3977 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3978 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3979 reclaim_str_hashtbl_size++;
3980 return 1;
3981}
3982
3983static void
3984nfs4_release_reclaim(void)
3985{
3986 struct nfs4_client_reclaim *crp = NULL;
3987 int i;
3988
1da177e4
LT		 3989 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3990 while (!list_empty(&reclaim_str_hashtbl[i])) {
3991 crp = list_entry(reclaim_str_hashtbl[i].next,
3992 struct nfs4_client_reclaim, cr_strhash);
3993 list_del(&crp->cr_strhash);
1da177e4
LT		 3994 kfree(crp);
3995 reclaim_str_hashtbl_size--;
3996 }
3997 }
3998 BUG_ON(reclaim_str_hashtbl_size);
3999}
4000
4001/*
4002 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 4003static struct nfs4_client_reclaim *
1da177e4
LT		 4004nfs4_find_reclaim_client(clientid_t *clid)
4005{
4006 unsigned int strhashval;
4007 struct nfs4_client *clp;
4008 struct nfs4_client_reclaim *crp = NULL;
4009
4010
4011 /* find clientid in conf_id_hashtbl */
4012 clp = find_confirmed_client(clid);
4013 if (clp == NULL)
4014 return NULL;
4015
a55370a3
N		 4016 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4017 clp->cl_name.len, clp->cl_name.data,
4018 clp->cl_recdir);
1da177e4
LT		 4019
4020 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 4021 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 4022 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 4023 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 4024 return crp;
4025 }
4026 }
4027 return NULL;
4028}
4029
4030/*
4031* Called from OPEN. Look for clientid in reclaim list.
4032*/
b37ad28b 4033__be32
1da177e4
LT		 4034nfs4_check_open_reclaim(clientid_t *clid)
4035{
dfc83565 4036 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 4037}
4038
ac4d8ff2 4039/* initialization to perform at module load time: */
1da177e4 4040
e8ff2a84 4041int
ac4d8ff2 4042nfs4_state_init(void)
1da177e4 4043{
e8ff2a84 4044 int i, status;
1da177e4 4045
e8ff2a84
BF		 4046 status = nfsd4_init_slabs();
4047 if (status)
4048 return status;
1da177e4
LT		 4049 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4050 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4051 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4052 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4053 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4054 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4055 }
5282fd72
ME		 4056 for (i = 0; i < SESSION_HASH_SIZE; i++)
4057 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4058 for (i = 0; i < FILE_HASH_SIZE; i++) {
4059 INIT_LIST_HEAD(&file_hashtbl[i]);
4060 }
4061 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4062 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4063 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4064 }
4065 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4066 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4067 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4068 }
4069 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4070 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4071 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4072 }
1da177e4 4073 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4074 INIT_LIST_HEAD(&close_lru);
4075 INIT_LIST_HEAD(&client_lru);
4076 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4077 reclaim_str_hashtbl_size = 0;
e8ff2a84 4078 return 0;
ac4d8ff2
N		 4079}
4080
190e4fbf
N		 4081static void
4082nfsd4_load_reboot_recovery_data(void)
4083{
4084 int status;
4085
0964a3d3
N		 4086 nfs4_lock_state();
4087 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4088 status = nfsd4_recdir_load();
0964a3d3 4089 nfs4_unlock_state();
190e4fbf
N		 4090 if (status)
4091 printk("NFSD: Failure reading reboot recovery data\n");
4092}
4093
9a8db97e
ME		 4094unsigned long
4095get_nfs4_grace_period(void)
4096{
4097 return max(user_lease_time, lease_time) * HZ;
4098}
4099
c2f1a551
MS		 4100/*
4101 * Since the lifetime of a delegation isn't limited to that of an open, a
4102 * client may quite reasonably hang on to a delegation as long as it has
4103 * the inode cached. This becomes an obvious problem the first time a
4104 * client's inode cache approaches the size of the server's total memory.
4105 *
4106 * For now we avoid this problem by imposing a hard limit on the number
4107 * of delegations, which varies according to the server's memory size.
4108 */
4109static void
4110set_max_delegations(void)
4111{
4112 /*
4113 * Allow at most 4 delegations per megabyte of RAM. Quick
4114 * estimates suggest that in the worst case (where every delegation
4115 * is for a different inode), a delegation could take about 1.5K,
4116 * giving a worst case usage of about 6% of memory.
4117 */
4118 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4119}
4120
ac4d8ff2
N		 4121/* initialization to perform when the nfsd service is started: */
4122
4123static void
4124__nfs4_state_start(void)
4125{
9a8db97e 4126 unsigned long grace_time;
ac4d8ff2 4127
1da177e4 4128 boot_time = get_seconds();
af558e33 4129 grace_time = get_nfs4_grace_period();
d99a05ad 4130 lease_time = user_lease_time;
af558e33 4131 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4132 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4133 grace_time/HZ);
58da282b 4134 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4135 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4136 set_max_delegations();
1da177e4
LT		 4137}
4138
e8ff2a84 4139void
76a3550e 4140nfs4_state_start(void)
1da177e4 4141{
1da177e4 4142 if (nfs4_init)
e8ff2a84 4143 return;
190e4fbf 4144 nfsd4_load_reboot_recovery_data();
76a3550e 4145 __nfs4_state_start();
1da177e4 4146 nfs4_init = 1;
e8ff2a84 4147 return;
1da177e4
LT		 4148}
4149
1da177e4
LT		 4150time_t
4151nfs4_lease_time(void)
4152{
4153 return lease_time;
4154}
4155
4156static void
4157__nfs4_state_shutdown(void)
4158{
4159 int i;
4160 struct nfs4_client *clp = NULL;
4161 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4162 struct list_head *pos, *next, reaplist;
4163
1da177e4
LT		 4164 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4165 while (!list_empty(&conf_id_hashtbl[i])) {
4166 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4167 expire_client(clp);
4168 }
4169 while (!list_empty(&unconf_str_hashtbl[i])) {
4170 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4171 expire_client(clp);
4172 }
4173 }
4174 INIT_LIST_HEAD(&reaplist);
4175 spin_lock(&recall_lock);
4176 list_for_each_safe(pos, next, &del_recall_lru) {
4177 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4178 list_move(&dp->dl_recall_lru, &reaplist);
4179 }
4180 spin_unlock(&recall_lock);
4181 list_for_each_safe(pos, next, &reaplist) {
4182 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4183 list_del_init(&dp->dl_recall_lru);
4184 unhash_delegation(dp);
4185 }
4186
190e4fbf 4187 nfsd4_shutdown_recdir();
1da177e4 4188 nfs4_init = 0;
1da177e4
LT		 4189}
4190
4191void
4192nfs4_state_shutdown(void)
4193{
5e8d5c29
N		 4194 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4195 destroy_workqueue(laundry_wq);
2c5e7615 4196 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4197 nfs4_lock_state();
4198 nfs4_release_reclaim();
4199 __nfs4_state_shutdown();
1da177e4
LT		 4200 nfs4_unlock_state();
4201}
4202
3dd98a3b
JL		 4203/*
4204 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4205 * accessed when nfsd is starting.
4206 */
0964a3d3
N		 4207static void
4208nfs4_set_recdir(char *recdir)
4209{
0964a3d3 4210 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4211}
4212
4213/*
4214 * Change the NFSv4 recovery directory to recdir.
4215 */
4216int
4217nfs4_reset_recoverydir(char *recdir)
4218{
4219 int status;
a63bb996 4220 struct path path;
0964a3d3 4221
a63bb996 4222 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4223 if (status)
4224 return status;
4225 status = -ENOTDIR;
a63bb996 4226 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4227 nfs4_set_recdir(recdir);
4228 status = 0;
4229 }
a63bb996 4230 path_put(&path);
0964a3d3
N		 4231 return status;
4232}
4233
3dd98a3b
JL		 4234char *
4235nfs4_recoverydir(void)
4236{
4237 return user_recovery_dirname;
4238}
4239
1da177e4
LT		 4240/*
4241 * Called when leasetime is changed.
4242 *
d99a05ad
N		 4243 * The only way the protocol gives us to handle on-the-fly lease changes is to
4244 * simulate a reboot. Instead of doing that, we just wait till the next time
4245 * we start to register any changes in lease time. If the administrator
4246 * really wants to change the lease time *now*, they can go ahead and bring
4247 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4248 *
4249 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4250 * when nfsd is starting
1da177e4
LT		 4251 */
4252void
4253nfs4_reset_lease(time_t leasetime)
4254{
d99a05ad 4255 user_lease_time = leasetime;
1da177e4 4256}
Linux 6.x block layer and io_uring tree(s)