nfs: move more to Documentation/filesystems/nfs
[deliverable/linux.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT
1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT
45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
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/*
a649637c
AA
417 * The protocol defines ca_maxresponssize_cached to include the size of
418 * the rpc header, but all we need to cache is the data starting after
419 * the end of the initial SEQUENCE operation--the rest we regenerate
420 * each time. Therefore we can advertise a ca_maxresponssize_cached
421 * value that is the number of bytes in our cache plus a few additional
422 * bytes. In order to stay on the safe side, and not promise more than
423 * we can cache, those additional bytes must be the minimum possible: 24
424 * bytes of rpc header (xid through accept state, with AUTH_NULL
425 * verifier), 12 for the compound header (with zero-length tag), and 44
426 * for the SEQUENCE op response:
427 */
428#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
429
430/*
431 * Give the client the number of ca_maxresponsesize_cached slots it
432 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
433 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
434 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 435 *
a649637c
AA
436 * If we run out of reserved DRC memory we should (up to a point)
437 * re-negotiate active sessions and reduce their slot usage to make
438 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 439 */
a649637c 440static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 441{
a649637c 442 int mem, size = fchan->maxresp_cached;
ec6b5d7b 443
5d77ddfb
AA
444 if (fchan->maxreqs < 1)
445 return nfserr_inval;
5d77ddfb 446
a649637c
AA
447 if (size < NFSD_MIN_HDR_SEQ_SZ)
448 size = NFSD_MIN_HDR_SEQ_SZ;
449 size -= NFSD_MIN_HDR_SEQ_SZ;
450 if (size > NFSD_SLOT_CACHE_SIZE)
451 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 452
a649637c
AA
453 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
454 mem = fchan->maxreqs * size;
455 if (mem > NFSD_MAX_MEM_PER_SESSION) {
456 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
457 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
458 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
459 mem = fchan->maxreqs * size;
460 }
ec6b5d7b 461
4bd9b0f4 462 spin_lock(&nfsd_drc_lock);
a649637c
AA
463 /* bound the total session drc memory ussage */
464 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
465 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
466 mem = fchan->maxreqs * size;
467 }
0c193054 468 nfsd_drc_mem_used += mem;
4bd9b0f4 469 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 470
b101ebbc 471 if (fchan->maxreqs == 0)
bdac86e2 472 return nfserr_serverfault;
a649637c
AA
473
474 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 475 return 0;
ec6b5d7b
AA
476}
477
478/*
479 * fchan holds the client values on input, and the server values on output
ddc04fd4 480 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA
481 */
482static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB
483 struct nfsd4_channel_attrs *session_fchan,
484 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA
485{
486 int status = 0;
ddc04fd4 487 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA
488
489 /* headerpadsz set to zero in encode routine */
490
491 /* Use the client's max request and max response size if possible */
492 if (fchan->maxreq_sz > maxcount)
493 fchan->maxreq_sz = maxcount;
6c18ba9f 494 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA
495
496 if (fchan->maxresp_sz > maxcount)
497 fchan->maxresp_sz = maxcount;
6c18ba9f 498 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 499
ec6b5d7b
AA
500 /* Use the client's maxops if possible */
501 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
502 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 503 session_fchan->maxops = fchan->maxops;
ec6b5d7b 504
ec6b5d7b
AA
505 /* FIXME: Error means no more DRC pages so the server should
506 * recover pages from existing sessions. For now fail session
507 * creation.
508 */
a649637c 509 status = set_forechannel_drc_size(fchan);
ec6b5d7b 510
a649637c 511 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 512 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA
513
514 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA
515 return status;
516}
517
557ce264
AA
518static void
519free_session_slots(struct nfsd4_session *ses)
520{
521 int i;
522
523 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
524 kfree(ses->se_slots[i]);
525}
526
efe0cb6d
BF
527/*
528 * We don't actually need to cache the rpc and session headers, so we
529 * can allocate a little less for each slot:
530 */
531static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
532{
533 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
534}
535
ec6b5d7b
AA
536static int
537alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
538 struct nfsd4_create_session *cses)
539{
540 struct nfsd4_session *new, tmp;
557ce264
AA
541 struct nfsd4_slot *sp;
542 int idx, slotsize, cachesize, i;
543 int status;
ec6b5d7b
AA
544
545 memset(&tmp, 0, sizeof(tmp));
546
547 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB
548 tmp.se_bchannel = cses->back_channel;
549 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
550 &cses->fore_channel);
ec6b5d7b
AA
551 if (status)
552 goto out;
553
557ce264
AA
554 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
555 + sizeof(struct nfsd4_session) > PAGE_SIZE);
556
557 status = nfserr_serverfault;
558 /* allocate struct nfsd4_session and slot table pointers in one piece */
559 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA
560 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
561 if (!new)
562 goto out;
563
564 memcpy(new, &tmp, sizeof(*new));
565
557ce264 566 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 567 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA
568 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
569 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
570 if (!sp)
571 goto out_free;
572 new->se_slots[i] = sp;
573 }
574
ec6b5d7b
AA
575 new->se_client = clp;
576 gen_sessionid(new);
577 idx = hash_sessionid(&new->se_sessionid);
578 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
579 NFS4_MAX_SESSIONID_LEN);
580
581 new->se_flags = cses->flags;
582 kref_init(&new->se_ref);
583 spin_lock(&sessionid_lock);
584 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
585 list_add(&new->se_perclnt, &clp->cl_sessions);
586 spin_unlock(&sessionid_lock);
587
588 status = nfs_ok;
589out:
590 return status;
557ce264
AA
591out_free:
592 free_session_slots(new);
593 kfree(new);
594 goto out;
ec6b5d7b
AA
595}
596
5282fd72
ME
597/* caller must hold sessionid_lock */
598static struct nfsd4_session *
599find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
600{
601 struct nfsd4_session *elem;
602 int idx;
603
604 dump_sessionid(__func__, sessionid);
605 idx = hash_sessionid(sessionid);
606 dprintk("%s: idx is %d\n", __func__, idx);
607 /* Search in the appropriate list */
608 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
609 dump_sessionid("list traversal", &elem->se_sessionid);
610 if (!memcmp(elem->se_sessionid.data, sessionid->data,
611 NFS4_MAX_SESSIONID_LEN)) {
612 return elem;
613 }
614 }
615
616 dprintk("%s: session not found\n", __func__);
617 return NULL;
618}
619
620/* caller must hold sessionid_lock */
7116ed6b 621static void
5282fd72 622unhash_session(struct nfsd4_session *ses)
7116ed6b
AA
623{
624 list_del(&ses->se_hash);
625 list_del(&ses->se_perclnt);
5282fd72
ME
626}
627
628static void
629release_session(struct nfsd4_session *ses)
630{
631 spin_lock(&sessionid_lock);
632 unhash_session(ses);
633 spin_unlock(&sessionid_lock);
7116ed6b
AA
634 nfsd4_put_session(ses);
635}
636
637void
638free_session(struct kref *kref)
639{
640 struct nfsd4_session *ses;
dd829c45 641 int mem;
7116ed6b
AA
642
643 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 644 spin_lock(&nfsd_drc_lock);
efe0cb6d 645 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 646 nfsd_drc_mem_used -= mem;
be98d1bb 647 spin_unlock(&nfsd_drc_lock);
557ce264 648 free_session_slots(ses);
7116ed6b
AA
649 kfree(ses);
650}
651
1da177e4
LT
652static inline void
653renew_client(struct nfs4_client *clp)
654{
655 /*
656 * Move client to the end to the LRU list.
657 */
658 dprintk("renewing client (clientid %08x/%08x)\n",
659 clp->cl_clientid.cl_boot,
660 clp->cl_clientid.cl_id);
661 list_move_tail(&clp->cl_lru, &client_lru);
662 clp->cl_time = get_seconds();
663}
664
665/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
666static int
667STALE_CLIENTID(clientid_t *clid)
668{
669 if (clid->cl_boot == boot_time)
670 return 0;
60adfc50
AA
671 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
672 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT
673 return 1;
674}
675
676/*
677 * XXX Should we use a slab cache ?
678 * This type of memory management is somewhat inefficient, but we use it
679 * anyway since SETCLIENTID is not a common operation.
680 */
35bba9a3 681static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT
682{
683 struct nfs4_client *clp;
684
35bba9a3
BF
685 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
686 if (clp == NULL)
687 return NULL;
688 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
689 if (clp->cl_name.data == NULL) {
690 kfree(clp);
691 return NULL;
1da177e4 692 }
35bba9a3
BF
693 memcpy(clp->cl_name.data, name.data, name.len);
694 clp->cl_name.len = name.len;
1da177e4
LT
695 return clp;
696}
697
1b1a9b31
BF
698static void
699shutdown_callback_client(struct nfs4_client *clp)
700{
c237dc03 701 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 702
1b1a9b31 703 if (clnt) {
404ec117
BF
704 /*
705 * Callback threads take a reference on the client, so there
706 * should be no outstanding callbacks at this point.
707 */
c237dc03 708 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF
709 rpc_shutdown_client(clnt);
710 }
711}
712
1da177e4
LT
713static inline void
714free_client(struct nfs4_client *clp)
715{
1b1a9b31 716 shutdown_callback_client(clp);
38524ab3
AA
717 if (clp->cl_cb_xprt)
718 svc_xprt_put(clp->cl_cb_xprt);
1da177e4
LT
719 if (clp->cl_cred.cr_group_info)
720 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 721 kfree(clp->cl_principal);
1da177e4
LT
722 kfree(clp->cl_name.data);
723 kfree(clp);
724}
725
726void
727put_nfs4_client(struct nfs4_client *clp)
728{
729 if (atomic_dec_and_test(&clp->cl_count))
730 free_client(clp);
731}
732
733static void
734expire_client(struct nfs4_client *clp)
735{
736 struct nfs4_stateowner *sop;
737 struct nfs4_delegation *dp;
1da177e4
LT
738 struct list_head reaplist;
739
740 dprintk("NFSD: expire_client cl_count %d\n",
741 atomic_read(&clp->cl_count));
742
1da177e4
LT
743 INIT_LIST_HEAD(&reaplist);
744 spin_lock(&recall_lock);
ea1da636
N
745 while (!list_empty(&clp->cl_delegations)) {
746 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT
747 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
748 dp->dl_flock);
ea1da636 749 list_del_init(&dp->dl_perclnt);
1da177e4
LT
750 list_move(&dp->dl_recall_lru, &reaplist);
751 }
752 spin_unlock(&recall_lock);
753 while (!list_empty(&reaplist)) {
754 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
755 list_del_init(&dp->dl_recall_lru);
756 unhash_delegation(dp);
757 }
758 list_del(&clp->cl_idhash);
759 list_del(&clp->cl_strhash);
760 list_del(&clp->cl_lru);
ea1da636
N
761 while (!list_empty(&clp->cl_openowners)) {
762 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 763 release_openowner(sop);
1da177e4 764 }
c4bf7868
ME
765 while (!list_empty(&clp->cl_sessions)) {
766 struct nfsd4_session *ses;
767 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
768 se_perclnt);
769 release_session(ses);
770 }
1da177e4
LT
771 put_nfs4_client(clp);
772}
773
35bba9a3
BF
774static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
775{
776 memcpy(target->cl_verifier.data, source->data,
777 sizeof(target->cl_verifier.data));
1da177e4
LT
778}
779
35bba9a3
BF
780static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
781{
1da177e4
LT
782 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
783 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
784}
785
35bba9a3
BF
786static void copy_cred(struct svc_cred *target, struct svc_cred *source)
787{
1da177e4
LT
788 target->cr_uid = source->cr_uid;
789 target->cr_gid = source->cr_gid;
790 target->cr_group_info = source->cr_group_info;
791 get_group_info(target->cr_group_info);
792}
793
35bba9a3 794static int same_name(const char *n1, const char *n2)
599e0a22 795{
a55370a3 796 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT
797}
798
799static int
599e0a22
BF
800same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
801{
802 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT
803}
804
805static int
599e0a22
BF
806same_clid(clientid_t *cl1, clientid_t *cl2)
807{
808 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT
809}
810
811/* XXX what about NGROUP */
812static int
599e0a22
BF
813same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
814{
815 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT
816}
817
5ec7b46c
BF
818static void gen_clid(struct nfs4_client *clp)
819{
820 static u32 current_clientid = 1;
821
1da177e4
LT
822 clp->cl_clientid.cl_boot = boot_time;
823 clp->cl_clientid.cl_id = current_clientid++;
824}
825
deda2faa
BF
826static void gen_confirm(struct nfs4_client *clp)
827{
828 static u32 i;
829 u32 *p;
1da177e4 830
1da177e4 831 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF
832 *p++ = get_seconds();
833 *p++ = i++;
1da177e4
LT
834}
835
b09333c4
RL
836static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
837 struct svc_rqst *rqstp, nfs4_verifier *verf)
838{
839 struct nfs4_client *clp;
840 struct sockaddr *sa = svc_addr(rqstp);
841 char *princ;
842
843 clp = alloc_client(name);
844 if (clp == NULL)
845 return NULL;
846
847 princ = svc_gss_principal(rqstp);
848 if (princ) {
849 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
850 if (clp->cl_principal == NULL) {
851 free_client(clp);
852 return NULL;
853 }
854 }
855
856 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
857 atomic_set(&clp->cl_count, 1);
858 atomic_set(&clp->cl_cb_conn.cb_set, 0);
859 INIT_LIST_HEAD(&clp->cl_idhash);
860 INIT_LIST_HEAD(&clp->cl_strhash);
861 INIT_LIST_HEAD(&clp->cl_openowners);
862 INIT_LIST_HEAD(&clp->cl_delegations);
863 INIT_LIST_HEAD(&clp->cl_sessions);
864 INIT_LIST_HEAD(&clp->cl_lru);
865 clear_bit(0, &clp->cl_cb_slot_busy);
866 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
867 copy_verf(clp, verf);
868 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
869 clp->cl_flavor = rqstp->rq_flavor;
870 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
871 gen_confirm(clp);
872
873 return clp;
874}
875
35bba9a3
BF
876static int check_name(struct xdr_netobj name)
877{
1da177e4
LT
878 if (name.len == 0)
879 return 0;
880 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 881 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT
882 return 0;
883 }
884 return 1;
885}
886
fd39ca9a 887static void
1da177e4
LT
888add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
889{
890 unsigned int idhashval;
891
892 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
893 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
894 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
895 list_add_tail(&clp->cl_lru, &client_lru);
896 clp->cl_time = get_seconds();
897}
898
fd39ca9a 899static void
1da177e4
LT
900move_to_confirmed(struct nfs4_client *clp)
901{
902 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
903 unsigned int strhashval;
904
905 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
906 list_del_init(&clp->cl_strhash);
f116629d 907 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 908 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT
909 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
910 renew_client(clp);
911}
912
913static struct nfs4_client *
914find_confirmed_client(clientid_t *clid)
915{
916 struct nfs4_client *clp;
917 unsigned int idhashval = clientid_hashval(clid->cl_id);
918
919 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 920 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
921 return clp;
922 }
923 return NULL;
924}
925
926static struct nfs4_client *
927find_unconfirmed_client(clientid_t *clid)
928{
929 struct nfs4_client *clp;
930 unsigned int idhashval = clientid_hashval(clid->cl_id);
931
932 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 933 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
934 return clp;
935 }
936 return NULL;
937}
938
a1bcecd2
AA
939/*
940 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
941 * parameter. Matching is based on the fact the at least one of the
942 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
943 *
944 * FIXME: we need to unify the clientid namespaces for nfsv4.x
945 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
946 * and SET_CLIENTID{,_CONFIRM}
947 */
948static inline int
949match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
950{
951 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
952 return use_exchange_id == has_exchange_flags;
953}
954
28ce6054 955static struct nfs4_client *
a1bcecd2
AA
956find_confirmed_client_by_str(const char *dname, unsigned int hashval,
957 bool use_exchange_id)
28ce6054
N
958{
959 struct nfs4_client *clp;
960
961 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
962 if (same_name(clp->cl_recdir, dname) &&
963 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
964 return clp;
965 }
966 return NULL;
967}
968
969static struct nfs4_client *
a1bcecd2
AA
970find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
971 bool use_exchange_id)
28ce6054
N
972{
973 struct nfs4_client *clp;
974
975 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
976 if (same_name(clp->cl_recdir, dname) &&
977 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
978 return clp;
979 }
980 return NULL;
981}
982
fd39ca9a 983static void
fbf4665f 984gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 985{
c237dc03 986 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL
987 unsigned short expected_family;
988
989 /* Currently, we only support tcp and tcp6 for the callback channel */
990 if (se->se_callback_netid_len == 3 &&
991 !memcmp(se->se_callback_netid_val, "tcp", 3))
992 expected_family = AF_INET;
993 else if (se->se_callback_netid_len == 4 &&
994 !memcmp(se->se_callback_netid_val, "tcp6", 4))
995 expected_family = AF_INET6;
996 else
1da177e4
LT
997 goto out_err;
998
aa9a4ec7
JL
999 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1000 se->se_callback_addr_len,
1001 (struct sockaddr *) &cb->cb_addr,
1002 sizeof(cb->cb_addr));
1003
7077ecba 1004 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 1005 goto out_err;
aa9a4ec7 1006
fbf4665f
JL
1007 if (cb->cb_addr.ss_family == AF_INET6)
1008 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
1009
ab52ae6d 1010 cb->cb_minorversion = 0;
1da177e4
LT
1011 cb->cb_prog = se->se_callback_prog;
1012 cb->cb_ident = se->se_callback_ident;
1da177e4
LT
1013 return;
1014out_err:
aa9a4ec7
JL
1015 cb->cb_addr.ss_family = AF_UNSPEC;
1016 cb->cb_addrlen = 0;
849823c5 1017 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT
1018 "will not receive delegations\n",
1019 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1020
1da177e4
LT
1021 return;
1022}
1023
074fe897 1024/*
557ce264 1025 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 1026 */
074fe897
AA
1027void
1028nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 1029{
557ce264
AA
1030 struct nfsd4_slot *slot = resp->cstate.slot;
1031 unsigned int base;
074fe897 1032
557ce264 1033 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1034
557ce264
AA
1035 slot->sl_opcnt = resp->opcnt;
1036 slot->sl_status = resp->cstate.status;
074fe897 1037
bf864a31 1038 if (nfsd4_not_cached(resp)) {
557ce264 1039 slot->sl_datalen = 0;
bf864a31 1040 return;
074fe897 1041 }
557ce264
AA
1042 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1043 base = (char *)resp->cstate.datap -
1044 (char *)resp->xbuf->head[0].iov_base;
1045 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1046 slot->sl_datalen))
1047 WARN("%s: sessions DRC could not cache compound\n", __func__);
1048 return;
074fe897
AA
1049}
1050
1051/*
abfabf8c
AA
1052 * Encode the replay sequence operation from the slot values.
1053 * If cachethis is FALSE encode the uncached rep error on the next
1054 * operation which sets resp->p and increments resp->opcnt for
1055 * nfs4svc_encode_compoundres.
074fe897 1056 *
074fe897 1057 */
abfabf8c
AA
1058static __be32
1059nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1060 struct nfsd4_compoundres *resp)
074fe897 1061{
abfabf8c
AA
1062 struct nfsd4_op *op;
1063 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1064
abfabf8c 1065 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1066 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1067
abfabf8c
AA
1068 /* Encode the replayed sequence operation */
1069 op = &args->ops[resp->opcnt - 1];
1070 nfsd4_encode_operation(resp, op);
bf864a31 1071
abfabf8c 1072 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1073 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA
1074 op = &args->ops[resp->opcnt++];
1075 op->status = nfserr_retry_uncached_rep;
1076 nfsd4_encode_operation(resp, op);
074fe897 1077 }
abfabf8c 1078 return op->status;
074fe897
AA
1079}
1080
1081/*
557ce264
AA
1082 * The sequence operation is not cached because we can use the slot and
1083 * session values.
074fe897
AA
1084 */
1085__be32
bf864a31
AA
1086nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1087 struct nfsd4_sequence *seq)
074fe897 1088{
557ce264 1089 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA
1090 __be32 status;
1091
557ce264 1092 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1093
abfabf8c
AA
1094 /* Either returns 0 or nfserr_retry_uncached */
1095 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1096 if (status == nfserr_retry_uncached_rep)
1097 return status;
074fe897 1098
557ce264
AA
1099 /* The sequence operation has been encoded, cstate->datap set. */
1100 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1101
557ce264
AA
1102 resp->opcnt = slot->sl_opcnt;
1103 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1104 status = slot->sl_status;
074fe897
AA
1105
1106 return status;
1107}
1108
0733d213
AA
1109/*
1110 * Set the exchange_id flags returned by the server.
1111 */
1112static void
1113nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1114{
1115 /* pNFS is not supported */
1116 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1117
1118 /* Referrals are supported, Migration is not. */
1119 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1120
1121 /* set the wire flags to return to client. */
1122 clid->flags = new->cl_exchange_flags;
1123}
1124
069b6ad4
AA
1125__be32
1126nfsd4_exchange_id(struct svc_rqst *rqstp,
1127 struct nfsd4_compound_state *cstate,
1128 struct nfsd4_exchange_id *exid)
1129{
0733d213
AA
1130 struct nfs4_client *unconf, *conf, *new;
1131 int status;
1132 unsigned int strhashval;
1133 char dname[HEXDIR_LEN];
363168b4 1134 char addr_str[INET6_ADDRSTRLEN];
0733d213 1135 nfs4_verifier verf = exid->verifier;
363168b4 1136 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1137
363168b4 1138 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1139 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1140 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1141 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1142 addr_str, exid->flags, exid->spa_how);
0733d213
AA
1143
1144 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1145 return nfserr_inval;
1146
1147 /* Currently only support SP4_NONE */
1148 switch (exid->spa_how) {
1149 case SP4_NONE:
1150 break;
1151 case SP4_SSV:
1152 return nfserr_encr_alg_unsupp;
1153 default:
1154 BUG(); /* checked by xdr code */
1155 case SP4_MACH_CRED:
1156 return nfserr_serverfault; /* no excuse :-/ */
1157 }
1158
1159 status = nfs4_make_rec_clidname(dname, &exid->clname);
1160
1161 if (status)
1162 goto error;
1163
1164 strhashval = clientstr_hashval(dname);
1165
1166 nfs4_lock_state();
1167 status = nfs_ok;
1168
a1bcecd2 1169 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1170 if (conf) {
1171 if (!same_verf(&verf, &conf->cl_verifier)) {
1172 /* 18.35.4 case 8 */
1173 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1174 status = nfserr_not_same;
1175 goto out;
1176 }
1177 /* Client reboot: destroy old state */
1178 expire_client(conf);
1179 goto out_new;
1180 }
1181 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1182 /* 18.35.4 case 9 */
1183 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1184 status = nfserr_perm;
1185 goto out;
1186 }
1187 expire_client(conf);
1188 goto out_new;
1189 }
0733d213
AA
1190 /*
1191 * Set bit when the owner id and verifier map to an already
1192 * confirmed client id (18.35.3).
1193 */
1194 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1195
1196 /*
1197 * Falling into 18.35.4 case 2, possible router replay.
1198 * Leave confirmed record intact and return same result.
1199 */
1200 copy_verf(conf, &verf);
1201 new = conf;
1202 goto out_copy;
6ddbbbfe
MS
1203 }
1204
1205 /* 18.35.4 case 7 */
1206 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1207 status = nfserr_noent;
1208 goto out;
0733d213
AA
1209 }
1210
a1bcecd2 1211 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1212 if (unconf) {
1213 /*
1214 * Possible retry or client restart. Per 18.35.4 case 4,
1215 * a new unconfirmed record should be generated regardless
1216 * of whether any properties have changed.
1217 */
1218 expire_client(unconf);
1219 }
1220
1221out_new:
1222 /* Normal case */
b09333c4 1223 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1224 if (new == NULL) {
bdac86e2 1225 status = nfserr_serverfault;
0733d213
AA
1226 goto out;
1227 }
1228
0733d213 1229 gen_clid(new);
0733d213
AA
1230 add_to_unconfirmed(new, strhashval);
1231out_copy:
1232 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1233 exid->clientid.cl_id = new->cl_clientid.cl_id;
1234
38eb76a5 1235 exid->seqid = 1;
0733d213
AA
1236 nfsd4_set_ex_flags(new, exid);
1237
1238 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1239 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA
1240 status = nfs_ok;
1241
1242out:
1243 nfs4_unlock_state();
1244error:
1245 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1246 return status;
069b6ad4
AA
1247}
1248
b85d4c01 1249static int
88e588d5 1250check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1251{
88e588d5
AA
1252 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1253 slot_seqid);
b85d4c01
BH
1254
1255 /* The slot is in use, and no response has been sent. */
88e588d5
AA
1256 if (slot_inuse) {
1257 if (seqid == slot_seqid)
b85d4c01
BH
1258 return nfserr_jukebox;
1259 else
1260 return nfserr_seq_misordered;
1261 }
1262 /* Normal */
88e588d5 1263 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH
1264 return nfs_ok;
1265 /* Replay */
88e588d5 1266 if (seqid == slot_seqid)
b85d4c01
BH
1267 return nfserr_replay_cache;
1268 /* Wraparound */
88e588d5 1269 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH
1270 return nfs_ok;
1271 /* Misordered replay or misordered new request */
1272 return nfserr_seq_misordered;
1273}
1274
49557cc7
AA
1275/*
1276 * Cache the create session result into the create session single DRC
1277 * slot cache by saving the xdr structure. sl_seqid has been set.
1278 * Do this for solo or embedded create session operations.
1279 */
1280static void
1281nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1282 struct nfsd4_clid_slot *slot, int nfserr)
1283{
1284 slot->sl_status = nfserr;
1285 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1286}
1287
1288static __be32
1289nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1290 struct nfsd4_clid_slot *slot)
1291{
1292 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1293 return slot->sl_status;
1294}
1295
069b6ad4
AA
1296__be32
1297nfsd4_create_session(struct svc_rqst *rqstp,
1298 struct nfsd4_compound_state *cstate,
1299 struct nfsd4_create_session *cr_ses)
1300{
363168b4 1301 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1302 struct nfs4_client *conf, *unconf;
49557cc7 1303 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA
1304 int status = 0;
1305
1306 nfs4_lock_state();
1307 unconf = find_unconfirmed_client(&cr_ses->clientid);
1308 conf = find_confirmed_client(&cr_ses->clientid);
1309
1310 if (conf) {
49557cc7
AA
1311 cs_slot = &conf->cl_cs_slot;
1312 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1313 if (status == nfserr_replay_cache) {
ec6b5d7b 1314 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1315 cs_slot->sl_seqid);
38eb76a5 1316 /* Return the cached reply status */
49557cc7 1317 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1318 goto out;
49557cc7 1319 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA
1320 status = nfserr_seq_misordered;
1321 dprintk("Sequence misordered!\n");
1322 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1323 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA
1324 goto out;
1325 }
49557cc7 1326 cs_slot->sl_seqid++;
ec6b5d7b
AA
1327 } else if (unconf) {
1328 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1329 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA
1330 status = nfserr_clid_inuse;
1331 goto out;
1332 }
1333
49557cc7
AA
1334 cs_slot = &unconf->cl_cs_slot;
1335 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA
1336 if (status) {
1337 /* an unconfirmed replay returns misordered */
ec6b5d7b 1338 status = nfserr_seq_misordered;
49557cc7 1339 goto out_cache;
ec6b5d7b
AA
1340 }
1341
49557cc7 1342 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA
1343 move_to_confirmed(unconf);
1344
1345 /*
1346 * We do not support RDMA or persistent sessions
1347 */
1348 cr_ses->flags &= ~SESSION4_PERSIST;
1349 cr_ses->flags &= ~SESSION4_RDMA;
1350
38524ab3
AA
1351 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1352 unconf->cl_cb_xprt = rqstp->rq_xprt;
1353 svc_xprt_get(unconf->cl_cb_xprt);
1354 rpc_copy_addr(
1355 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1356 sa);
1357 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1358 unconf->cl_cb_conn.cb_minorversion =
1359 cstate->minorversion;
1360 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1361 unconf->cl_cb_seq_nr = 1;
38524ab3
AA
1362 nfsd4_probe_callback(unconf);
1363 }
ec6b5d7b
AA
1364 conf = unconf;
1365 } else {
1366 status = nfserr_stale_clientid;
1367 goto out;
1368 }
1369
1370 status = alloc_init_session(rqstp, conf, cr_ses);
1371 if (status)
1372 goto out;
1373
ec6b5d7b
AA
1374 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1375 NFS4_MAX_SESSIONID_LEN);
49557cc7 1376 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1377
49557cc7
AA
1378out_cache:
1379 /* cache solo and embedded create sessions under the state lock */
1380 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA
1381out:
1382 nfs4_unlock_state();
1383 dprintk("%s returns %d\n", __func__, ntohl(status));
1384 return status;
069b6ad4
AA
1385}
1386
1387__be32
1388nfsd4_destroy_session(struct svc_rqst *r,
1389 struct nfsd4_compound_state *cstate,
1390 struct nfsd4_destroy_session *sessionid)
1391{
e10e0cfc
BH
1392 struct nfsd4_session *ses;
1393 u32 status = nfserr_badsession;
1394
1395 /* Notes:
1396 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1397 * - Should we return nfserr_back_chan_busy if waiting for
1398 * callbacks on to-be-destroyed session?
1399 * - Do we need to clear any callback info from previous session?
1400 */
1401
1402 dump_sessionid(__func__, &sessionid->sessionid);
1403 spin_lock(&sessionid_lock);
1404 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1405 if (!ses) {
1406 spin_unlock(&sessionid_lock);
1407 goto out;
1408 }
1409
1410 unhash_session(ses);
1411 spin_unlock(&sessionid_lock);
1412
1413 /* wait for callbacks */
1414 shutdown_callback_client(ses->se_client);
1415 nfsd4_put_session(ses);
1416 status = nfs_ok;
1417out:
1418 dprintk("%s returns %d\n", __func__, ntohl(status));
1419 return status;
069b6ad4
AA
1420}
1421
1422__be32
b85d4c01 1423nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA
1424 struct nfsd4_compound_state *cstate,
1425 struct nfsd4_sequence *seq)
1426{
f9bb94c4 1427 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH
1428 struct nfsd4_session *session;
1429 struct nfsd4_slot *slot;
1430 int status;
1431
f9bb94c4
AA
1432 if (resp->opcnt != 1)
1433 return nfserr_sequence_pos;
1434
b85d4c01
BH
1435 spin_lock(&sessionid_lock);
1436 status = nfserr_badsession;
1437 session = find_in_sessionid_hashtbl(&seq->sessionid);
1438 if (!session)
1439 goto out;
1440
1441 status = nfserr_badslot;
6c18ba9f 1442 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH
1443 goto out;
1444
557ce264 1445 slot = session->se_slots[seq->slotid];
b85d4c01
BH
1446 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1447
a8dfdaeb
AA
1448 /* We do not negotiate the number of slots yet, so set the
1449 * maxslots to the session maxreqs which is used to encode
1450 * sr_highest_slotid and the sr_target_slot id to maxslots */
1451 seq->maxslots = session->se_fchannel.maxreqs;
1452
88e588d5 1453 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH
1454 if (status == nfserr_replay_cache) {
1455 cstate->slot = slot;
1456 cstate->session = session;
da3846a2 1457 /* Return the cached reply status and set cstate->status
557ce264 1458 * for nfsd4_proc_compound processing */
bf864a31 1459 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1460 cstate->status = nfserr_replay_cache;
aaf84eb9 1461 goto out;
b85d4c01
BH
1462 }
1463 if (status)
1464 goto out;
1465
1466 /* Success! bump slot seqid */
1467 slot->sl_inuse = true;
1468 slot->sl_seqid = seq->seqid;
557ce264 1469 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH
1470
1471 cstate->slot = slot;
1472 cstate->session = session;
1473
aaf84eb9 1474 /* Hold a session reference until done processing the compound:
b85d4c01
BH
1475 * nfsd4_put_session called only if the cstate slot is set.
1476 */
b85d4c01
BH
1477 nfsd4_get_session(session);
1478out:
1479 spin_unlock(&sessionid_lock);
aaf84eb9
BH
1480 /* Renew the clientid on success and on replay */
1481 if (cstate->session) {
1482 nfs4_lock_state();
1483 renew_client(session->se_client);
1484 nfs4_unlock_state();
1485 }
b85d4c01
BH
1486 dprintk("%s: return %d\n", __func__, ntohl(status));
1487 return status;
069b6ad4
AA
1488}
1489
b37ad28b 1490__be32
b591480b
BF
1491nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1492 struct nfsd4_setclientid *setclid)
1da177e4 1493{
363168b4 1494 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT
1495 struct xdr_netobj clname = {
1496 .len = setclid->se_namelen,
1497 .data = setclid->se_name,
1498 };
1499 nfs4_verifier clverifier = setclid->se_verf;
1500 unsigned int strhashval;
28ce6054 1501 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1502 __be32 status;
a55370a3 1503 char dname[HEXDIR_LEN];
1da177e4 1504
1da177e4 1505 if (!check_name(clname))
73aea4ec 1506 return nfserr_inval;
1da177e4 1507
a55370a3
N
1508 status = nfs4_make_rec_clidname(dname, &clname);
1509 if (status)
73aea4ec 1510 return status;
a55370a3 1511
1da177e4
LT
1512 /*
1513 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1514 * We get here on a DRC miss.
1515 */
1516
a55370a3 1517 strhashval = clientstr_hashval(dname);
1da177e4 1518
1da177e4 1519 nfs4_lock_state();
a1bcecd2 1520 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1521 if (conf) {
a186e767 1522 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1523 status = nfserr_clid_inuse;
026722c2 1524 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL
1525 char addr_str[INET6_ADDRSTRLEN];
1526 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1527 sizeof(addr_str));
1528 dprintk("NFSD: setclientid: string in use by client "
1529 "at %s\n", addr_str);
1da177e4
LT
1530 goto out;
1531 }
1da177e4 1532 }
a186e767
BF
1533 /*
1534 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1535 * has a description of SETCLIENTID request processing consisting
1536 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1537 */
a1bcecd2 1538 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT
1539 status = nfserr_resource;
1540 if (!conf) {
a186e767
BF
1541 /*
1542 * RFC 3530 14.2.33 CASE 4:
1543 * placed first, because it is the normal case
1da177e4
LT
1544 */
1545 if (unconf)
1546 expire_client(unconf);
b09333c4 1547 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1548 if (new == NULL)
1da177e4 1549 goto out;
1da177e4 1550 gen_clid(new);
599e0a22 1551 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1552 /*
a186e767
BF
1553 * RFC 3530 14.2.33 CASE 1:
1554 * probable callback update
1da177e4 1555 */
31f4a6c1
N
1556 if (unconf) {
1557 /* Note this is removing unconfirmed {*x***},
1558 * which is stronger than RFC recommended {vxc**}.
1559 * This has the advantage that there is at most
1560 * one {*x***} in either list at any time.
1561 */
1562 expire_client(unconf);
1da177e4 1563 }
b09333c4 1564 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1565 if (new == NULL)
1da177e4 1566 goto out;
1da177e4 1567 copy_clid(new, conf);
1da177e4
LT
1568 } else if (!unconf) {
1569 /*
a186e767
BF
1570 * RFC 3530 14.2.33 CASE 2:
1571 * probable client reboot; state will be removed if
1572 * confirmed.
1da177e4 1573 */
b09333c4 1574 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1575 if (new == NULL)
1da177e4 1576 goto out;
1da177e4 1577 gen_clid(new);
49ba8781 1578 } else {
a186e767
BF
1579 /*
1580 * RFC 3530 14.2.33 CASE 3:
1581 * probable client reboot; state will be removed if
1582 * confirmed.
1da177e4
LT
1583 */
1584 expire_client(unconf);
b09333c4 1585 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1586 if (new == NULL)
1da177e4 1587 goto out;
1da177e4 1588 gen_clid(new);
1da177e4 1589 }
fbf4665f 1590 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1591 add_to_unconfirmed(new, strhashval);
1da177e4
LT
1592 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1593 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1594 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1595 status = nfs_ok;
1596out:
1597 nfs4_unlock_state();
1598 return status;
1599}
1600
1601
1602/*
a186e767
BF
1603 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1604 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1605 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1606 */
b37ad28b 1607__be32
b591480b
BF
1608nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1609 struct nfsd4_compound_state *cstate,
1610 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1611{
363168b4 1612 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1613 struct nfs4_client *conf, *unconf;
1da177e4
LT
1614 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1615 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1616 __be32 status;
1da177e4
LT
1617
1618 if (STALE_CLIENTID(clid))
1619 return nfserr_stale_clientid;
1620 /*
1621 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1622 * We get here on a DRC miss.
1623 */
1624
1625 nfs4_lock_state();
21ab45a4
N
1626
1627 conf = find_confirmed_client(clid);
1628 unconf = find_unconfirmed_client(clid);
1629
1630 status = nfserr_clid_inuse;
363168b4 1631 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1632 goto out;
363168b4 1633 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N
1634 goto out;
1635
a186e767
BF
1636 /*
1637 * section 14.2.34 of RFC 3530 has a description of
1638 * SETCLIENTID_CONFIRM request processing consisting
1639 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1640 */
366e0c1d 1641 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF
1642 /*
1643 * RFC 3530 14.2.34 CASE 1:
1644 * callback update
1645 */
599e0a22 1646 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT
1647 status = nfserr_clid_inuse;
1648 else {
1a69c179
N
1649 /* XXX: We just turn off callbacks until we can handle
1650 * change request correctly. */
c237dc03 1651 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1652 expire_client(unconf);
1da177e4 1653 status = nfs_ok;
1a69c179 1654
1da177e4 1655 }
f3aba4e5 1656 } else if (conf && !unconf) {
a186e767
BF
1657 /*
1658 * RFC 3530 14.2.34 CASE 2:
1659 * probable retransmitted request; play it safe and
1660 * do nothing.
7c79f737 1661 */
599e0a22 1662 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1663 status = nfserr_clid_inuse;
21ab45a4 1664 else
1da177e4 1665 status = nfs_ok;
7c79f737 1666 } else if (!conf && unconf
599e0a22 1667 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF
1668 /*
1669 * RFC 3530 14.2.34 CASE 3:
1670 * Normal case; new or rebooted client:
7c79f737 1671 */
599e0a22 1672 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT
1673 status = nfserr_clid_inuse;
1674 } else {
1a69c179
N
1675 unsigned int hash =
1676 clientstr_hashval(unconf->cl_recdir);
1677 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1678 hash, false);
1a69c179 1679 if (conf) {
c7b9a459 1680 nfsd4_remove_clid_dir(conf);
1a69c179
N
1681 expire_client(conf);
1682 }
1da177e4 1683 move_to_confirmed(unconf);
21ab45a4 1684 conf = unconf;
46f8a64b 1685 nfsd4_probe_callback(conf);
1a69c179 1686 status = nfs_ok;
1da177e4 1687 }
599e0a22
BF
1688 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1689 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1690 &confirm)))) {
a186e767
BF
1691 /*
1692 * RFC 3530 14.2.34 CASE 4:
1693 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1694 */
1da177e4 1695 status = nfserr_stale_clientid;
7c79f737 1696 } else {
08e8987c
N
1697 /* check that we have hit one of the cases...*/
1698 status = nfserr_clid_inuse;
1699 }
1da177e4 1700out:
1da177e4
LT
1701 nfs4_unlock_state();
1702 return status;
1703}
1704
1da177e4
LT
1705/* OPEN Share state helper functions */
1706static inline struct nfs4_file *
1707alloc_init_file(struct inode *ino)
1708{
1709 struct nfs4_file *fp;
1710 unsigned int hashval = file_hashval(ino);
1711
e60d4398
N
1712 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1713 if (fp) {
8b671b80 1714 atomic_set(&fp->fi_ref, 1);
1da177e4 1715 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N
1716 INIT_LIST_HEAD(&fp->fi_stateids);
1717 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1718 spin_lock(&recall_lock);
1da177e4 1719 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1720 spin_unlock(&recall_lock);
1da177e4
LT
1721 fp->fi_inode = igrab(ino);
1722 fp->fi_id = current_fileid++;
47f9940c 1723 fp->fi_had_conflict = false;
1da177e4
LT
1724 return fp;
1725 }
1726 return NULL;
1727}
1728
e60d4398 1729static void
e18b890b 1730nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1731{
e60d4398
N
1732 if (*slab == NULL)
1733 return;
1a1d92c1 1734 kmem_cache_destroy(*slab);
e60d4398 1735 *slab = NULL;
1da177e4
LT
1736}
1737
e8ff2a84 1738void
1da177e4
LT
1739nfsd4_free_slabs(void)
1740{
e60d4398
N
1741 nfsd4_free_slab(&stateowner_slab);
1742 nfsd4_free_slab(&file_slab);
5ac049ac 1743 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1744 nfsd4_free_slab(&deleg_slab);
e60d4398 1745}
1da177e4 1746
e60d4398
N
1747static int
1748nfsd4_init_slabs(void)
1749{
1750 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1751 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N
1752 if (stateowner_slab == NULL)
1753 goto out_nomem;
1754 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1755 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N
1756 if (file_slab == NULL)
1757 goto out_nomem;
5ac049ac 1758 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1759 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N
1760 if (stateid_slab == NULL)
1761 goto out_nomem;
5b2d21c1 1762 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1763 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N
1764 if (deleg_slab == NULL)
1765 goto out_nomem;
e60d4398
N
1766 return 0;
1767out_nomem:
1768 nfsd4_free_slabs();
1769 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1770 return -ENOMEM;
1da177e4
LT
1771}
1772
1773void
1774nfs4_free_stateowner(struct kref *kref)
1775{
1776 struct nfs4_stateowner *sop =
1777 container_of(kref, struct nfs4_stateowner, so_ref);
1778 kfree(sop->so_owner.data);
1779 kmem_cache_free(stateowner_slab, sop);
1780}
1781
1782static inline struct nfs4_stateowner *
1783alloc_stateowner(struct xdr_netobj *owner)
1784{
1785 struct nfs4_stateowner *sop;
1786
1787 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1788 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1789 memcpy(sop->so_owner.data, owner->data, owner->len);
1790 sop->so_owner.len = owner->len;
1791 kref_init(&sop->so_ref);
1792 return sop;
1793 }
1794 kmem_cache_free(stateowner_slab, sop);
1795 }
1796 return NULL;
1797}
1798
1799static struct nfs4_stateowner *
1800alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1801 struct nfs4_stateowner *sop;
1802 struct nfs4_replay *rp;
1803 unsigned int idhashval;
1804
1805 if (!(sop = alloc_stateowner(&open->op_owner)))
1806 return NULL;
1807 idhashval = ownerid_hashval(current_ownerid);
1808 INIT_LIST_HEAD(&sop->so_idhash);
1809 INIT_LIST_HEAD(&sop->so_strhash);
1810 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
1811 INIT_LIST_HEAD(&sop->so_stateids);
1812 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT
1813 INIT_LIST_HEAD(&sop->so_close_lru);
1814 sop->so_time = 0;
1815 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1816 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1817 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT
1818 sop->so_is_open_owner = 1;
1819 sop->so_id = current_ownerid++;
1820 sop->so_client = clp;
1821 sop->so_seqid = open->op_seqid;
1822 sop->so_confirmed = 0;
1823 rp = &sop->so_replay;
de1ae286 1824 rp->rp_status = nfserr_serverfault;
1da177e4
LT
1825 rp->rp_buflen = 0;
1826 rp->rp_buf = rp->rp_ibuf;
1827 return sop;
1828}
1829
1da177e4
LT
1830static inline void
1831init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1832 struct nfs4_stateowner *sop = open->op_stateowner;
1833 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1834
1835 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N
1836 INIT_LIST_HEAD(&stp->st_perstateowner);
1837 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT
1838 INIT_LIST_HEAD(&stp->st_perfile);
1839 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1840 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1841 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1842 stp->st_stateowner = sop;
13cd2184 1843 get_nfs4_file(fp);
1da177e4 1844 stp->st_file = fp;
78155ed7 1845 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
1846 stp->st_stateid.si_stateownerid = sop->so_id;
1847 stp->st_stateid.si_fileid = fp->fi_id;
1848 stp->st_stateid.si_generation = 0;
1849 stp->st_access_bmap = 0;
1850 stp->st_deny_bmap = 0;
84459a11
AA
1851 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1852 &stp->st_access_bmap);
1da177e4 1853 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1854 stp->st_openstp = NULL;
1da177e4
LT
1855}
1856
fd39ca9a 1857static void
1da177e4
LT
1858move_to_close_lru(struct nfs4_stateowner *sop)
1859{
1860 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1861
358dd55a 1862 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT
1863 sop->so_time = get_seconds();
1864}
1865
1da177e4 1866static int
599e0a22
BF
1867same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1868 clientid_t *clid)
1869{
1870 return (sop->so_owner.len == owner->len) &&
1871 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1872 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT
1873}
1874
1875static struct nfs4_stateowner *
1876find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1877{
1878 struct nfs4_stateowner *so = NULL;
1879
1880 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1881 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT
1882 return so;
1883 }
1884 return NULL;
1885}
1886
1887/* search file_hashtbl[] for file */
1888static struct nfs4_file *
1889find_file(struct inode *ino)
1890{
1891 unsigned int hashval = file_hashval(ino);
1892 struct nfs4_file *fp;
1893
8b671b80 1894 spin_lock(&recall_lock);
1da177e4 1895 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N
1896 if (fp->fi_inode == ino) {
1897 get_nfs4_file(fp);
8b671b80 1898 spin_unlock(&recall_lock);
1da177e4 1899 return fp;
13cd2184 1900 }
1da177e4 1901 }
8b671b80 1902 spin_unlock(&recall_lock);
1da177e4
LT
1903 return NULL;
1904}
1905
d87a8ade 1906static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1907{
d87a8ade 1908 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1909 return 0;
d87a8ade
AA
1910 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1911 return 0;
1912 x &= ~NFS4_SHARE_ACCESS_MASK;
1913 if (minorversion && x) {
1914 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1915 return 0;
1916 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1917 return 0;
1918 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1919 }
1920 if (x)
8838dc43
BF
1921 return 0;
1922 return 1;
ba5a6a19
BF
1923}
1924
8838dc43 1925static inline int deny_valid(u32 x)
ba5a6a19 1926{
8838dc43
BF
1927 /* Note: unlike access bits, deny bits may be zero. */
1928 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1929}
1da177e4 1930
4f83aa30
BF
1931/*
1932 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1933 * st_{access,deny}_bmap field of the stateid, in order to track not
1934 * only what share bits are currently in force, but also what
1935 * combinations of share bits previous opens have used. This allows us
1936 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1937 * return an error if the client attempt to downgrade to a combination
1938 * of share bits not explicable by closing some of its previous opens.
1939 *
1940 * XXX: This enforcement is actually incomplete, since we don't keep
1941 * track of access/deny bit combinations; so, e.g., we allow:
1942 *
1943 * OPEN allow read, deny write
1944 * OPEN allow both, deny none
1945 * DOWNGRADE allow read, deny none
1946 *
1947 * which we should reject.
1948 */
fd39ca9a 1949static void
1da177e4
LT
1950set_access(unsigned int *access, unsigned long bmap) {
1951 int i;
1952
1953 *access = 0;
1954 for (i = 1; i < 4; i++) {
1955 if (test_bit(i, &bmap))
1956 *access |= i;
1957 }
1958}
1959
fd39ca9a 1960static void
1da177e4
LT
1961set_deny(unsigned int *deny, unsigned long bmap) {
1962 int i;
1963
1964 *deny = 0;
1965 for (i = 0; i < 4; i++) {
1966 if (test_bit(i, &bmap))
1967 *deny |= i ;
1968 }
1969}
1970
1971static int
1972test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1973 unsigned int access, deny;
1974
1975 set_access(&access, stp->st_access_bmap);
1976 set_deny(&deny, stp->st_deny_bmap);
1977 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1978 return 0;
1979 return 1;
1980}
1981
1982/*
1983 * Called to check deny when READ with all zero stateid or
1984 * WRITE with all zero or all one stateid
1985 */
b37ad28b 1986static __be32
1da177e4
LT
1987nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1988{
1989 struct inode *ino = current_fh->fh_dentry->d_inode;
1990 struct nfs4_file *fp;
1991 struct nfs4_stateid *stp;
b37ad28b 1992 __be32 ret;
1da177e4
LT
1993
1994 dprintk("NFSD: nfs4_share_conflict\n");
1995
1996 fp = find_file(ino);
13cd2184
N
1997 if (!fp)
1998 return nfs_ok;
b700949b 1999 ret = nfserr_locked;
1da177e4 2000 /* Search for conflicting share reservations */
13cd2184
N
2001 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2002 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2003 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2004 goto out;
1da177e4 2005 }
13cd2184
N
2006 ret = nfs_ok;
2007out:
2008 put_nfs4_file(fp);
2009 return ret;
1da177e4
LT
2010}
2011
2012static inline void
2013nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2014{
2015 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2016 drop_file_write_access(filp);
1da177e4
LT
2017 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2018 }
2019}
2020
1da177e4
LT
2021/*
2022 * Spawn a thread to perform a recall on the delegation represented
2023 * by the lease (file_lock)
2024 *
2025 * Called from break_lease() with lock_kernel() held.
2026 * Note: we assume break_lease will only call this *once* for any given
2027 * lease.
2028 */
2029static
2030void nfsd_break_deleg_cb(struct file_lock *fl)
2031{
63e4863f 2032 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT
2033
2034 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2035 if (!dp)
2036 return;
2037
2038 /* We're assuming the state code never drops its reference
2039 * without first removing the lease. Since we're in this lease
2040 * callback (and since the lease code is serialized by the kernel
2041 * lock) we know the server hasn't removed the lease yet, we know
2042 * it's safe to take a reference: */
2043 atomic_inc(&dp->dl_count);
cfdcad4d 2044 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT
2045
2046 spin_lock(&recall_lock);
2047 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2048 spin_unlock(&recall_lock);
2049
2050 /* only place dl_time is set. protected by lock_kernel*/
2051 dp->dl_time = get_seconds();
2052
0272e1fd
BF
2053 /*
2054 * We don't want the locks code to timeout the lease for us;
2055 * we'll remove it ourself if the delegation isn't returned
2056 * in time.
2057 */
2058 fl->fl_break_time = 0;
1da177e4 2059
63e4863f
BF
2060 dp->dl_file->fi_had_conflict = true;
2061 nfsd4_cb_recall(dp);
1da177e4
LT
2062}
2063
2064/*
2065 * The file_lock is being reapd.
2066 *
2067 * Called by locks_free_lock() with lock_kernel() held.
2068 */
2069static
2070void nfsd_release_deleg_cb(struct file_lock *fl)
2071{
2072 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2073
2074 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2075
2076 if (!(fl->fl_flags & FL_LEASE) || !dp)
2077 return;
2078 dp->dl_flock = NULL;
2079}
2080
2081/*
2082 * Set the delegation file_lock back pointer.
2083 *
a9933cea 2084 * Called from setlease() with lock_kernel() held.
1da177e4
LT
2085 */
2086static
2087void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2088{
2089 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2090
2091 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2092 if (!dp)
2093 return;
2094 dp->dl_flock = new;
2095}
2096
2097/*
a9933cea 2098 * Called from setlease() with lock_kernel() held
1da177e4
LT
2099 */
2100static
2101int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2102{
2103 struct nfs4_delegation *onlistd =
2104 (struct nfs4_delegation *)onlist->fl_owner;
2105 struct nfs4_delegation *tryd =
2106 (struct nfs4_delegation *)try->fl_owner;
2107
2108 if (onlist->fl_lmops != try->fl_lmops)
2109 return 0;
2110
2111 return onlistd->dl_client == tryd->dl_client;
2112}
2113
2114
2115static
2116int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2117{
2118 if (arg & F_UNLCK)
2119 return lease_modify(onlist, arg);
2120 else
2121 return -EAGAIN;
2122}
2123
7b021967 2124static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT
2125 .fl_break = nfsd_break_deleg_cb,
2126 .fl_release_private = nfsd_release_deleg_cb,
2127 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2128 .fl_mylease = nfsd_same_client_deleg_cb,
2129 .fl_change = nfsd_change_deleg_cb,
2130};
2131
2132
b37ad28b 2133__be32
6668958f
AA
2134nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2135 struct nfsd4_open *open)
1da177e4 2136{
1da177e4
LT
2137 clientid_t *clientid = &open->op_clientid;
2138 struct nfs4_client *clp = NULL;
2139 unsigned int strhashval;
2140 struct nfs4_stateowner *sop = NULL;
2141
1da177e4 2142 if (!check_name(open->op_owner))
0f442aa2 2143 return nfserr_inval;
1da177e4
LT
2144
2145 if (STALE_CLIENTID(&open->op_clientid))
2146 return nfserr_stale_clientid;
2147
2148 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2149 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF
2150 open->op_stateowner = sop;
2151 if (!sop) {
2152 /* Make sure the client's lease hasn't expired. */
1da177e4
LT
2153 clp = find_confirmed_client(clientid);
2154 if (clp == NULL)
0f442aa2
BF
2155 return nfserr_expired;
2156 goto renew;
1da177e4 2157 }
6668958f
AA
2158 /* When sessions are used, skip open sequenceid processing */
2159 if (nfsd4_has_session(cstate))
2160 goto renew;
0f442aa2
BF
2161 if (!sop->so_confirmed) {
2162 /* Replace unconfirmed owners without checking for replay. */
2163 clp = sop->so_client;
f044ff83 2164 release_openowner(sop);
0f442aa2
BF
2165 open->op_stateowner = NULL;
2166 goto renew;
2167 }
2168 if (open->op_seqid == sop->so_seqid - 1) {
2169 if (sop->so_replay.rp_buflen)
a90b061c 2170 return nfserr_replay_me;
0f442aa2
BF
2171 /* The original OPEN failed so spectacularly
2172 * that we don't even have replay data saved!
2173 * Therefore, we have no choice but to continue
2174 * processing this OPEN; presumably, we'll
2175 * fail again for the same reason.
2176 */
2177 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2178 goto renew;
2179 }
2180 if (open->op_seqid != sop->so_seqid)
2181 return nfserr_bad_seqid;
1da177e4 2182renew:
0f442aa2
BF
2183 if (open->op_stateowner == NULL) {
2184 sop = alloc_init_open_stateowner(strhashval, clp, open);
2185 if (sop == NULL)
2186 return nfserr_resource;
2187 open->op_stateowner = sop;
2188 }
2189 list_del_init(&sop->so_close_lru);
1da177e4 2190 renew_client(sop->so_client);
0f442aa2 2191 return nfs_ok;
1da177e4
LT
2192}
2193
b37ad28b 2194static inline __be32
4a6e43e6
N
2195nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2196{
2197 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2198 return nfserr_openmode;
2199 else
2200 return nfs_ok;
2201}
2202
52f4fb43
N
2203static struct nfs4_delegation *
2204find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2205{
2206 struct nfs4_delegation *dp;
2207
ea1da636 2208 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N
2209 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2210 return dp;
2211 }
2212 return NULL;
2213}
2214
b37ad28b 2215static __be32
567d9829
N
2216nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2217 struct nfs4_delegation **dp)
2218{
2219 int flags;
b37ad28b 2220 __be32 status = nfserr_bad_stateid;
567d9829
N
2221
2222 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2223 if (*dp == NULL)
c44c5eeb 2224 goto out;
567d9829
N
2225 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2226 RD_STATE : WR_STATE;
2227 status = nfs4_check_delegmode(*dp, flags);
2228 if (status)
2229 *dp = NULL;
c44c5eeb
N
2230out:
2231 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2232 return nfs_ok;
2233 if (status)
2234 return status;
2235 open->op_stateowner->so_confirmed = 1;
2236 return nfs_ok;
567d9829
N
2237}
2238
b37ad28b 2239static __be32
1da177e4
LT
2240nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2241{
2242 struct nfs4_stateid *local;
b37ad28b 2243 __be32 status = nfserr_share_denied;
1da177e4
LT
2244 struct nfs4_stateowner *sop = open->op_stateowner;
2245
8beefa24 2246 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT
2247 /* ignore lock owners */
2248 if (local->st_stateowner->so_is_open_owner == 0)
2249 continue;
2250 /* remember if we have seen this open owner */
2251 if (local->st_stateowner == sop)
2252 *stpp = local;
2253 /* check for conflicting share reservations */
2254 if (!test_share(local, open))
2255 goto out;
2256 }
2257 status = 0;
2258out:
2259 return status;
2260}
2261
5ac049ac
N
2262static inline struct nfs4_stateid *
2263nfs4_alloc_stateid(void)
2264{
2265 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2266}
2267
b37ad28b 2268static __be32
1da177e4 2269nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2270 struct nfs4_delegation *dp,
1da177e4
LT
2271 struct svc_fh *cur_fh, int flags)
2272{
2273 struct nfs4_stateid *stp;
1da177e4 2274
5ac049ac 2275 stp = nfs4_alloc_stateid();
1da177e4
LT
2276 if (stp == NULL)
2277 return nfserr_resource;
2278
567d9829
N
2279 if (dp) {
2280 get_file(dp->dl_vfs_file);
2281 stp->st_vfs_file = dp->dl_vfs_file;
2282 } else {
b37ad28b 2283 __be32 status;
567d9829
N
2284 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2285 &stp->st_vfs_file);
2286 if (status) {
2287 if (status == nfserr_dropit)
2288 status = nfserr_jukebox;
5ac049ac 2289 kmem_cache_free(stateid_slab, stp);
567d9829
N
2290 return status;
2291 }
1da177e4 2292 }
1da177e4
LT
2293 *stpp = stp;
2294 return 0;
2295}
2296
b37ad28b 2297static inline __be32
1da177e4
LT
2298nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2299 struct nfsd4_open *open)
2300{
2301 struct iattr iattr = {
2302 .ia_valid = ATTR_SIZE,
2303 .ia_size = 0,
2304 };
2305 if (!open->op_truncate)
2306 return 0;
2307 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2308 return nfserr_inval;
1da177e4
LT
2309 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2310}
2311
b37ad28b 2312static __be32
1da177e4
LT
2313nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2314{
2315 struct file *filp = stp->st_vfs_file;
7eaa36e2 2316 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2317 unsigned int share_access, new_writer;
b37ad28b 2318 __be32 status;
1da177e4
LT
2319
2320 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF
2321 new_writer = (~share_access) & open->op_share_access
2322 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2323
6c26d08f 2324 if (new_writer) {
b37ad28b
AV
2325 int err = get_write_access(inode);
2326 if (err)
2327 return nfserrno(err);
e518f056
BH
2328 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2329 if (err)
2330 return nfserrno(err);
2331 file_take_write(filp);
6c26d08f 2332 }
1da177e4
LT
2333 status = nfsd4_truncate(rqstp, cur_fh, open);
2334 if (status) {
6c26d08f
BF
2335 if (new_writer)
2336 put_write_access(inode);
1da177e4
LT
2337 return status;
2338 }
2339 /* remember the open */
6c26d08f 2340 filp->f_mode |= open->op_share_access;
b55e0ba1
BF
2341 __set_bit(open->op_share_access, &stp->st_access_bmap);
2342 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT
2343
2344 return nfs_ok;
2345}
2346
2347
1da177e4 2348static void
37515177 2349nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2350{
37515177
N
2351 open->op_stateowner->so_confirmed = 1;
2352 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT
2353}
2354
2355/*
2356 * Attempt to hand out a delegation.
2357 */
2358static void
2359nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2360{
2361 struct nfs4_delegation *dp;
2362 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2363 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT
2364 struct file_lock fl, *flp = &fl;
2365 int status, flag = 0;
2366
2367 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N
2368 open->op_recall = 0;
2369 switch (open->op_claim_type) {
2370 case NFS4_OPEN_CLAIM_PREVIOUS:
2371 if (!atomic_read(&cb->cb_set))
2372 open->op_recall = 1;
2373 flag = open->op_delegate_type;
2374 if (flag == NFS4_OPEN_DELEGATE_NONE)
2375 goto out;
2376 break;
2377 case NFS4_OPEN_CLAIM_NULL:
2378 /* Let's not give out any delegations till everyone's
2379 * had the chance to reclaim theirs.... */
af558e33 2380 if (locks_in_grace())
7b190fec
N
2381 goto out;
2382 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2383 goto out;
2384 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2385 flag = NFS4_OPEN_DELEGATE_WRITE;
2386 else
2387 flag = NFS4_OPEN_DELEGATE_READ;
2388 break;
2389 default:
2390 goto out;
2391 }
1da177e4
LT
2392
2393 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2394 if (dp == NULL) {
2395 flag = NFS4_OPEN_DELEGATE_NONE;
2396 goto out;
2397 }
2398 locks_init_lock(&fl);
2399 fl.fl_lmops = &nfsd_lease_mng_ops;
2400 fl.fl_flags = FL_LEASE;
9167f501 2401 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT
2402 fl.fl_end = OFFSET_MAX;
2403 fl.fl_owner = (fl_owner_t)dp;
2404 fl.fl_file = stp->st_vfs_file;
2405 fl.fl_pid = current->tgid;
2406
a9933cea 2407 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT
2408 * the lock_manager callbacks fl_mylease and fl_change are used
2409 */
9167f501 2410 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2411 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2412 unhash_delegation(dp);
1da177e4
LT
2413 flag = NFS4_OPEN_DELEGATE_NONE;
2414 goto out;
2415 }
2416
2417 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2418
8c10cbdb
BH
2419 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2420 STATEID_VAL(&dp->dl_stateid));
1da177e4 2421out:
7b190fec
N
2422 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2423 && flag == NFS4_OPEN_DELEGATE_NONE
2424 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2425 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT
2426 open->op_delegate_type = flag;
2427}
2428
2429/*
2430 * called with nfs4_lock_state() held.
2431 */
b37ad28b 2432__be32
1da177e4
LT
2433nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2434{
6668958f 2435 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT
2436 struct nfs4_file *fp = NULL;
2437 struct inode *ino = current_fh->fh_dentry->d_inode;
2438 struct nfs4_stateid *stp = NULL;
567d9829 2439 struct nfs4_delegation *dp = NULL;
b37ad28b 2440 __be32 status;
1da177e4
LT
2441
2442 status = nfserr_inval;
d87a8ade 2443 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2444 || !deny_valid(open->op_share_deny))
1da177e4
LT
2445 goto out;
2446 /*
2447 * Lookup file; if found, lookup stateid and check open request,
2448 * and check for delegations in the process of being recalled.
2449 * If not found, create the nfs4_file struct
2450 */
2451 fp = find_file(ino);
2452 if (fp) {
2453 if ((status = nfs4_check_open(fp, open, &stp)))
2454 goto out;
c44c5eeb
N
2455 status = nfs4_check_deleg(fp, open, &dp);
2456 if (status)
2457 goto out;
1da177e4 2458 } else {
c44c5eeb
N
2459 status = nfserr_bad_stateid;
2460 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2461 goto out;
1da177e4
LT
2462 status = nfserr_resource;
2463 fp = alloc_init_file(ino);
2464 if (fp == NULL)
2465 goto out;
2466 }
2467
2468 /*
2469 * OPEN the file, or upgrade an existing OPEN.
2470 * If truncate fails, the OPEN fails.
2471 */
2472 if (stp) {
2473 /* Stateid was found, this is an OPEN upgrade */
2474 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2475 if (status)
2476 goto out;
444c2c07 2477 update_stateid(&stp->st_stateid);
1da177e4
LT
2478 } else {
2479 /* Stateid was not found, this is a new OPEN */
2480 int flags = 0;
9ecb6a08 2481 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2482 flags |= NFSD_MAY_READ;
1da177e4 2483 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2484 flags |= NFSD_MAY_WRITE;
567d9829
N
2485 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2486 if (status)
1da177e4
LT
2487 goto out;
2488 init_stateid(stp, fp, open);
2489 status = nfsd4_truncate(rqstp, current_fh, open);
2490 if (status) {
2283963f 2491 release_open_stateid(stp);
1da177e4
LT
2492 goto out;
2493 }
6668958f
AA
2494 if (nfsd4_has_session(&resp->cstate))
2495 update_stateid(&stp->st_stateid);
1da177e4
LT
2496 }
2497 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2498
6668958f
AA
2499 if (nfsd4_has_session(&resp->cstate))
2500 open->op_stateowner->so_confirmed = 1;
2501
1da177e4
LT
2502 /*
2503 * Attempt to hand out a delegation. No error return, because the
2504 * OPEN succeeds even if we fail.
2505 */
2506 nfs4_open_delegation(current_fh, open, stp);
2507
2508 status = nfs_ok;
2509
8c10cbdb
BH
2510 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2511 STATEID_VAL(&stp->st_stateid));
1da177e4 2512out:
13cd2184
N
2513 if (fp)
2514 put_nfs4_file(fp);
37515177
N
2515 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2516 nfs4_set_claim_prev(open);
1da177e4
LT
2517 /*
2518 * To finish the open response, we just need to set the rflags.
2519 */
2520 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA
2521 if (!open->op_stateowner->so_confirmed &&
2522 !nfsd4_has_session(&resp->cstate))
1da177e4
LT
2523 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2524
2525 return status;
2526}
2527
b37ad28b 2528__be32
b591480b
BF
2529nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2530 clientid_t *clid)
1da177e4
LT
2531{
2532 struct nfs4_client *clp;
b37ad28b 2533 __be32 status;
1da177e4
LT
2534
2535 nfs4_lock_state();
2536 dprintk("process_renew(%08x/%08x): starting\n",
2537 clid->cl_boot, clid->cl_id);
2538 status = nfserr_stale_clientid;
2539 if (STALE_CLIENTID(clid))
2540 goto out;
2541 clp = find_confirmed_client(clid);
2542 status = nfserr_expired;
2543 if (clp == NULL) {
2544 /* We assume the client took too long to RENEW. */
2545 dprintk("nfsd4_renew: clientid not found!\n");
2546 goto out;
2547 }
2548 renew_client(clp);
2549 status = nfserr_cb_path_down;
ea1da636 2550 if (!list_empty(&clp->cl_delegations)
c237dc03 2551 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT
2552 goto out;
2553 status = nfs_ok;
2554out:
2555 nfs4_unlock_state();
2556 return status;
2557}
2558
af558e33
BF
2559struct lock_manager nfsd4_manager = {
2560};
2561
a76b4319 2562static void
af558e33 2563nfsd4_end_grace(void)
a76b4319
N
2564{
2565 dprintk("NFSD: end of grace period\n");
c7b9a459 2566 nfsd4_recdir_purge_old();
af558e33 2567 locks_end_grace(&nfsd4_manager);
a76b4319
N
2568}
2569
fd39ca9a 2570static time_t
1da177e4
LT
2571nfs4_laundromat(void)
2572{
2573 struct nfs4_client *clp;
2574 struct nfs4_stateowner *sop;
2575 struct nfs4_delegation *dp;
2576 struct list_head *pos, *next, reaplist;
2577 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2578 time_t t, clientid_val = NFSD_LEASE_TIME;
2579 time_t u, test_val = NFSD_LEASE_TIME;
2580
2581 nfs4_lock_state();
2582
2583 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF
2584 if (locks_in_grace())
2585 nfsd4_end_grace();
1da177e4
LT
2586 list_for_each_safe(pos, next, &client_lru) {
2587 clp = list_entry(pos, struct nfs4_client, cl_lru);
2588 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2589 t = clp->cl_time - cutoff;
2590 if (clientid_val > t)
2591 clientid_val = t;
2592 break;
2593 }
2594 dprintk("NFSD: purging unused client (clientid %08x)\n",
2595 clp->cl_clientid.cl_id);
c7b9a459 2596 nfsd4_remove_clid_dir(clp);
1da177e4
LT
2597 expire_client(clp);
2598 }
2599 INIT_LIST_HEAD(&reaplist);
2600 spin_lock(&recall_lock);
2601 list_for_each_safe(pos, next, &del_recall_lru) {
2602 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2603 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2604 u = dp->dl_time - cutoff;
2605 if (test_val > u)
2606 test_val = u;
2607 break;
2608 }
2609 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2610 dp, dp->dl_flock);
2611 list_move(&dp->dl_recall_lru, &reaplist);
2612 }
2613 spin_unlock(&recall_lock);
2614 list_for_each_safe(pos, next, &reaplist) {
2615 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2616 list_del_init(&dp->dl_recall_lru);
2617 unhash_delegation(dp);
2618 }
2619 test_val = NFSD_LEASE_TIME;
2620 list_for_each_safe(pos, next, &close_lru) {
2621 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2622 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2623 u = sop->so_time - cutoff;
2624 if (test_val > u)
2625 test_val = u;
2626 break;
2627 }
2628 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2629 sop->so_id);
f044ff83 2630 release_openowner(sop);
1da177e4
LT
2631 }
2632 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2633 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2634 nfs4_unlock_state();
2635 return clientid_val;
2636}
2637
a254b246
HH
2638static struct workqueue_struct *laundry_wq;
2639static void laundromat_main(struct work_struct *);
2640static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2641
2642static void
c4028958 2643laundromat_main(struct work_struct *not_used)
1da177e4
LT
2644{
2645 time_t t;
2646
2647 t = nfs4_laundromat();
2648 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2649 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT
2650}
2651
fd39ca9a 2652static struct nfs4_stateowner *
f8816512
N
2653search_close_lru(u32 st_id, int flags)
2654{
1da177e4
LT
2655 struct nfs4_stateowner *local = NULL;
2656
1da177e4
LT
2657 if (flags & CLOSE_STATE) {
2658 list_for_each_entry(local, &close_lru, so_close_lru) {
2659 if (local->so_id == st_id)
2660 return local;
2661 }
2662 }
2663 return NULL;
2664}
2665
2666static inline int
2667nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2668{
7eaa36e2 2669 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT
2670}
2671
2672static int
2673STALE_STATEID(stateid_t *stateid)
2674{
78155ed7
BN
2675 if (time_after((unsigned long)boot_time,
2676 (unsigned long)stateid->si_boot)) {
8c10cbdb
BH
2677 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2678 STATEID_VAL(stateid));
78155ed7
BN
2679 return 1;
2680 }
2681 return 0;
2682}
2683
2684static int
2685EXPIRED_STATEID(stateid_t *stateid)
2686{
2687 if (time_before((unsigned long)boot_time,
2688 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2689 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
8c10cbdb
BH
2690 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2691 STATEID_VAL(stateid));
78155ed7
BN
2692 return 1;
2693 }
2694 return 0;
2695}
2696
2697static __be32
2698stateid_error_map(stateid_t *stateid)
2699{
2700 if (STALE_STATEID(stateid))
2701 return nfserr_stale_stateid;
2702 if (EXPIRED_STATEID(stateid))
2703 return nfserr_expired;
2704
8c10cbdb
BH
2705 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2706 STATEID_VAL(stateid));
78155ed7 2707 return nfserr_bad_stateid;
1da177e4
LT
2708}
2709
2710static inline int
2711access_permit_read(unsigned long access_bmap)
2712{
2713 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2714 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2715 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2716}
2717
2718static inline int
2719access_permit_write(unsigned long access_bmap)
2720{
2721 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2722 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2723}
2724
2725static
b37ad28b 2726__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2727{
b37ad28b 2728 __be32 status = nfserr_openmode;
1da177e4
LT
2729
2730 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2731 goto out;
2732 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2733 goto out;
2734 status = nfs_ok;
2735out:
2736 return status;
2737}
2738
b37ad28b 2739static inline __be32
1da177e4
LT
2740check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2741{
203a8c8e 2742 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2743 return nfs_ok;
af558e33 2744 else if (locks_in_grace()) {
1da177e4
LT
2745 /* Answer in remaining cases depends on existance of
2746 * conflicting state; so we must wait out the grace period. */
2747 return nfserr_grace;
2748 } else if (flags & WR_STATE)
2749 return nfs4_share_conflict(current_fh,
2750 NFS4_SHARE_DENY_WRITE);
2751 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2752 return nfs4_share_conflict(current_fh,
2753 NFS4_SHARE_DENY_READ);
2754}
2755
2756/*
2757 * Allow READ/WRITE during grace period on recovered state only for files
2758 * that are not able to provide mandatory locking.
2759 */
2760static inline int
18f82731 2761grace_disallows_io(struct inode *inode)
1da177e4 2762{
203a8c8e 2763 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT
2764}
2765
6668958f 2766static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2767{
6668958f
AA
2768 /*
2769 * When sessions are used the stateid generation number is ignored
2770 * when it is zero.
2771 */
2772 if ((flags & HAS_SESSION) && in->si_generation == 0)
2773 goto out;
2774
0836f587
BF
2775 /* If the client sends us a stateid from the future, it's buggy: */
2776 if (in->si_generation > ref->si_generation)
2777 return nfserr_bad_stateid;
2778 /*
2779 * The following, however, can happen. For example, if the
2780 * client sends an open and some IO at the same time, the open
2781 * may bump si_generation while the IO is still in flight.
2782 * Thanks to hard links and renames, the client never knows what
2783 * file an open will affect. So it could avoid that situation
2784 * only by serializing all opens and IO from the same open
2785 * owner. To recover from the old_stateid error, the client
2786 * will just have to retry the IO:
2787 */
2788 if (in->si_generation < ref->si_generation)
2789 return nfserr_old_stateid;
6668958f 2790out:
0836f587
BF
2791 return nfs_ok;
2792}
2793
3e633079
BF
2794static int is_delegation_stateid(stateid_t *stateid)
2795{
2796 return stateid->si_fileid == 0;
2797}
2798
1da177e4
LT
2799/*
2800* Checks for stateid operations
2801*/
b37ad28b 2802__be32
dd453dfd
BH
2803nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2804 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT
2805{
2806 struct nfs4_stateid *stp = NULL;
2807 struct nfs4_delegation *dp = NULL;
dd453dfd 2808 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2809 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2810 __be32 status;
1da177e4 2811
1da177e4
LT
2812 if (filpp)
2813 *filpp = NULL;
2814
18f82731 2815 if (grace_disallows_io(ino))
1da177e4
LT
2816 return nfserr_grace;
2817
6668958f
AA
2818 if (nfsd4_has_session(cstate))
2819 flags |= HAS_SESSION;
2820
1da177e4
LT
2821 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2822 return check_special_stateids(current_fh, stateid, flags);
2823
1da177e4
LT
2824 status = nfserr_stale_stateid;
2825 if (STALE_STATEID(stateid))
2826 goto out;
2827
1da177e4 2828 status = nfserr_bad_stateid;
3e633079 2829 if (is_delegation_stateid(stateid)) {
a4455be0 2830 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN
2831 if (!dp) {
2832 status = stateid_error_map(stateid);
1da177e4 2833 goto out;
78155ed7 2834 }
6668958f
AA
2835 status = check_stateid_generation(stateid, &dp->dl_stateid,
2836 flags);
0c2a498f
BF
2837 if (status)
2838 goto out;
dc9bf700
BF
2839 status = nfs4_check_delegmode(dp, flags);
2840 if (status)
2841 goto out;
2842 renew_client(dp->dl_client);
dc9bf700
BF
2843 if (filpp)
2844 *filpp = dp->dl_vfs_file;
1da177e4 2845 } else { /* open or lock stateid */
a4455be0 2846 stp = find_stateid(stateid, flags);
78155ed7
BN
2847 if (!stp) {
2848 status = stateid_error_map(stateid);
1da177e4 2849 goto out;
78155ed7 2850 }
6150ef0d 2851 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT
2852 goto out;
2853 if (!stp->st_stateowner->so_confirmed)
2854 goto out;
6668958f
AA
2855 status = check_stateid_generation(stateid, &stp->st_stateid,
2856 flags);
0c2a498f
BF
2857 if (status)
2858 goto out;
a4455be0
BF
2859 status = nfs4_check_openmode(stp, flags);
2860 if (status)
1da177e4
LT
2861 goto out;
2862 renew_client(stp->st_stateowner->so_client);
2863 if (filpp)
2864 *filpp = stp->st_vfs_file;
1da177e4
LT
2865 }
2866 status = nfs_ok;
2867out:
2868 return status;
2869}
2870
4c4cd222
N
2871static inline int
2872setlkflg (int type)
2873{
2874 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2875 RD_STATE : WR_STATE;
2876}
1da177e4
LT
2877
2878/*
2879 * Checks for sequence id mutating operations.
2880 */
b37ad28b 2881static __be32
dd453dfd
BH
2882nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2883 stateid_t *stateid, int flags,
2884 struct nfs4_stateowner **sopp,
2885 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2886{
1da177e4
LT
2887 struct nfs4_stateid *stp;
2888 struct nfs4_stateowner *sop;
dd453dfd 2889 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2890 __be32 status;
1da177e4 2891
8c10cbdb
BH
2892 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2893 seqid, STATEID_VAL(stateid));
3a4f98bb 2894
1da177e4
LT
2895 *stpp = NULL;
2896 *sopp = NULL;
2897
1da177e4 2898 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2899 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2900 return nfserr_bad_stateid;
1da177e4
LT
2901 }
2902
1da177e4 2903 if (STALE_STATEID(stateid))
3a4f98bb 2904 return nfserr_stale_stateid;
6668958f
AA
2905
2906 if (nfsd4_has_session(cstate))
2907 flags |= HAS_SESSION;
2908
1da177e4
LT
2909 /*
2910 * We return BAD_STATEID if filehandle doesn't match stateid,
2911 * the confirmed flag is incorrecly set, or the generation
2912 * number is incorrect.
1da177e4 2913 */
f8816512
N
2914 stp = find_stateid(stateid, flags);
2915 if (stp == NULL) {
2916 /*
2917 * Also, we should make sure this isn't just the result of
2918 * a replayed close:
2919 */
2920 sop = search_close_lru(stateid->si_stateownerid, flags);
2921 if (sop == NULL)
78155ed7 2922 return stateid_error_map(stateid);
f8816512
N
2923 *sopp = sop;
2924 goto check_replay;
2925 }
1da177e4 2926
39325bd0
BF
2927 *stpp = stp;
2928 *sopp = sop = stp->st_stateowner;
2929
4c4cd222 2930 if (lock) {
4c4cd222 2931 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2932 struct nfs4_client *clp = sop->so_client;
4c4cd222 2933 int lkflg = 0;
b37ad28b 2934 __be32 status;
4c4cd222
N
2935
2936 lkflg = setlkflg(lock->lk_type);
2937
2938 if (lock->lk_is_new) {
599e0a22
BF
2939 if (!sop->so_is_open_owner)
2940 return nfserr_bad_stateid;
60adfc50
AA
2941 if (!(flags & HAS_SESSION) &&
2942 !same_clid(&clp->cl_clientid, lockclid))
2943 return nfserr_bad_stateid;
599e0a22
BF
2944 /* stp is the open stateid */
2945 status = nfs4_check_openmode(stp, lkflg);
2946 if (status)
2947 return status;
2948 } else {
2949 /* stp is the lock stateid */
2950 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2951 if (status)
2952 return status;
4c4cd222 2953 }
1da177e4
LT
2954 }
2955
f3362737 2956 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2957 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2958 return nfserr_bad_stateid;
1da177e4
LT
2959 }
2960
1da177e4
LT
2961 /*
2962 * We now validate the seqid and stateid generation numbers.
2963 * For the moment, we ignore the possibility of
2964 * generation number wraparound.
2965 */
6668958f 2966 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT
2967 goto check_replay;
2968
3a4f98bb 2969 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2970 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N
2971 " unconfirmed stateowner!\n");
2972 return nfserr_bad_stateid;
1da177e4 2973 }
3a4f98bb 2974 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2975 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N
2976 " confirmed yet!\n");
2977 return nfserr_bad_stateid;
1da177e4 2978 }
6668958f 2979 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF
2980 if (status)
2981 return status;
52fd004e 2982 renew_client(sop->so_client);
3a4f98bb 2983 return nfs_ok;
1da177e4 2984
1da177e4 2985check_replay:
bd9aac52 2986 if (seqid == sop->so_seqid - 1) {
849823c5 2987 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2988 /* indicate replay to calling function */
a90b061c 2989 return nfserr_replay_me;
1da177e4 2990 }
2fdada03 2991 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N
2992 sop->so_seqid, seqid);
2993 *sopp = NULL;
2994 return nfserr_bad_seqid;
1da177e4
LT
2995}
2996
b37ad28b 2997__be32
ca364317 2998nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2999 struct nfsd4_open_confirm *oc)
1da177e4 3000{
b37ad28b 3001 __be32 status;
1da177e4
LT
3002 struct nfs4_stateowner *sop;
3003 struct nfs4_stateid *stp;
3004
3005 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF
3006 (int)cstate->current_fh.fh_dentry->d_name.len,
3007 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3008
ca364317 3009 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF
3010 if (status)
3011 return status;
1da177e4
LT
3012
3013 nfs4_lock_state();
3014
dd453dfd 3015 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3016 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3017 CONFIRM | OPEN_STATE,
1da177e4
LT
3018 &oc->oc_stateowner, &stp, NULL)))
3019 goto out;
3020
3021 sop = oc->oc_stateowner;
3022 sop->so_confirmed = 1;
3023 update_stateid(&stp->st_stateid);
3024 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH
3025 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3026 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N
3027
3028 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3029out:
f2327d9a 3030 if (oc->oc_stateowner) {
1da177e4 3031 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3032 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3033 }
1da177e4
LT
3034 nfs4_unlock_state();
3035 return status;
3036}
3037
3038
3039/*
3040 * unset all bits in union bitmap (bmap) that
3041 * do not exist in share (from successful OPEN_DOWNGRADE)
3042 */
3043static void
3044reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3045{
3046 int i;
3047 for (i = 1; i < 4; i++) {
3048 if ((i & access) != i)
3049 __clear_bit(i, bmap);
3050 }
3051}
3052
3053static void
3054reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3055{
3056 int i;
3057 for (i = 0; i < 4; i++) {
3058 if ((i & deny) != i)
3059 __clear_bit(i, bmap);
3060 }
3061}
3062
b37ad28b 3063__be32
ca364317
BF
3064nfsd4_open_downgrade(struct svc_rqst *rqstp,
3065 struct nfsd4_compound_state *cstate,
a4f1706a 3066 struct nfsd4_open_downgrade *od)
1da177e4 3067{
b37ad28b 3068 __be32 status;
1da177e4
LT
3069 struct nfs4_stateid *stp;
3070 unsigned int share_access;
3071
3072 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF
3073 (int)cstate->current_fh.fh_dentry->d_name.len,
3074 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3075
d87a8ade 3076 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3077 || !deny_valid(od->od_share_deny))
1da177e4
LT
3078 return nfserr_inval;
3079
3080 nfs4_lock_state();
dd453dfd 3081 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3082 od->od_seqid,
1da177e4 3083 &od->od_stateid,
f3362737 3084 OPEN_STATE,
1da177e4
LT
3085 &od->od_stateowner, &stp, NULL)))
3086 goto out;
3087
3088 status = nfserr_inval;
3089 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3090 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3091 stp->st_access_bmap, od->od_share_access);
3092 goto out;
3093 }
3094 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3095 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3096 stp->st_deny_bmap, od->od_share_deny);
3097 goto out;
3098 }
3099 set_access(&share_access, stp->st_access_bmap);
3100 nfs4_file_downgrade(stp->st_vfs_file,
3101 share_access & ~od->od_share_access);
3102
3103 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3104 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3105
3106 update_stateid(&stp->st_stateid);
3107 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3108 status = nfs_ok;
3109out:
f2327d9a 3110 if (od->od_stateowner) {
1da177e4 3111 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3112 cstate->replay_owner = od->od_stateowner;
f2327d9a 3113 }
1da177e4
LT
3114 nfs4_unlock_state();
3115 return status;
3116}
3117
3118/*
3119 * nfs4_unlock_state() called after encode
3120 */
b37ad28b 3121__be32
ca364317 3122nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3123 struct nfsd4_close *close)
1da177e4 3124{
b37ad28b 3125 __be32 status;
1da177e4
LT
3126 struct nfs4_stateid *stp;
3127
3128 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF
3129 (int)cstate->current_fh.fh_dentry->d_name.len,
3130 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT
3131
3132 nfs4_lock_state();
3133 /* check close_lru for replay */
dd453dfd 3134 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3135 close->cl_seqid,
1da177e4 3136 &close->cl_stateid,
f3362737 3137 OPEN_STATE | CLOSE_STATE,
1da177e4
LT
3138 &close->cl_stateowner, &stp, NULL)))
3139 goto out;
1da177e4
LT
3140 status = nfs_ok;
3141 update_stateid(&stp->st_stateid);
3142 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3143
04ef5954 3144 /* release_stateid() calls nfsd_close() if needed */
2283963f 3145 release_open_stateid(stp);
04ef5954
BF
3146
3147 /* place unused nfs4_stateowners on so_close_lru list to be
3148 * released by the laundromat service after the lease period
3149 * to enable us to handle CLOSE replay
3150 */
3151 if (list_empty(&close->cl_stateowner->so_stateids))
3152 move_to_close_lru(close->cl_stateowner);
1da177e4 3153out:
f2327d9a 3154 if (close->cl_stateowner) {
1da177e4 3155 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3156 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3157 }
1da177e4
LT
3158 nfs4_unlock_state();
3159 return status;
3160}
3161
b37ad28b 3162__be32
ca364317
BF
3163nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3164 struct nfsd4_delegreturn *dr)
1da177e4 3165{
203a8c8e
BF
3166 struct nfs4_delegation *dp;
3167 stateid_t *stateid = &dr->dr_stateid;
3168 struct inode *inode;
b37ad28b 3169 __be32 status;
6668958f 3170 int flags = 0;
1da177e4 3171
ca364317 3172 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF
3173 return status;
3174 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3175
6668958f
AA
3176 if (nfsd4_has_session(cstate))
3177 flags |= HAS_SESSION;
1da177e4 3178 nfs4_lock_state();
203a8c8e
BF
3179 status = nfserr_bad_stateid;
3180 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3181 goto out;
3182 status = nfserr_stale_stateid;
3183 if (STALE_STATEID(stateid))
3184 goto out;
7e0f7cf5 3185 status = nfserr_bad_stateid;
203a8c8e
BF
3186 if (!is_delegation_stateid(stateid))
3187 goto out;
203a8c8e 3188 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN
3189 if (!dp) {
3190 status = stateid_error_map(stateid);
203a8c8e 3191 goto out;
78155ed7 3192 }
6668958f 3193 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF
3194 if (status)
3195 goto out;
3196 renew_client(dp->dl_client);
3197
3198 unhash_delegation(dp);
1da177e4 3199out:
203a8c8e
BF
3200 nfs4_unlock_state();
3201
1da177e4
LT
3202 return status;
3203}
3204
3205
3206/*
3207 * Lock owner state (byte-range locks)
3208 */
3209#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3210#define LOCK_HASH_BITS 8
3211#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3212#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3213
87df4de8
BH
3214static inline u64
3215end_offset(u64 start, u64 len)
3216{
3217 u64 end;
3218
3219 end = start + len;
3220 return end >= start ? end: NFS4_MAX_UINT64;
3221}
3222
3223/* last octet in a range */
3224static inline u64
3225last_byte_offset(u64 start, u64 len)
3226{
3227 u64 end;
3228
3229 BUG_ON(!len);
3230 end = start + len;
3231 return end > start ? end - 1: NFS4_MAX_UINT64;
3232}
3233
1da177e4
LT
3234#define lockownerid_hashval(id) \
3235 ((id) & LOCK_HASH_MASK)
3236
3237static inline unsigned int
3238lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3239 struct xdr_netobj *ownername)
3240{
3241 return (file_hashval(inode) + cl_id
3242 + opaque_hashval(ownername->data, ownername->len))
3243 & LOCK_HASH_MASK;
3244}
3245
3246static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3247static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3248static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3249
fd39ca9a 3250static struct nfs4_stateid *
1da177e4
LT
3251find_stateid(stateid_t *stid, int flags)
3252{
9346eff0 3253 struct nfs4_stateid *local;
1da177e4
LT
3254 u32 st_id = stid->si_stateownerid;
3255 u32 f_id = stid->si_fileid;
3256 unsigned int hashval;
3257
3258 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3259 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3260 hashval = stateid_hashval(st_id, f_id);
3261 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3262 if ((local->st_stateid.si_stateownerid == st_id) &&
3263 (local->st_stateid.si_fileid == f_id))
3264 return local;
3265 }
3266 }
9346eff0
KK
3267
3268 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3269 hashval = stateid_hashval(st_id, f_id);
3270 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3271 if ((local->st_stateid.si_stateownerid == st_id) &&
3272 (local->st_stateid.si_fileid == f_id))
3273 return local;
3274 }
849823c5 3275 }
1da177e4
LT
3276 return NULL;
3277}
3278
3279static struct nfs4_delegation *
3280find_delegation_stateid(struct inode *ino, stateid_t *stid)
3281{
13cd2184
N
3282 struct nfs4_file *fp;
3283 struct nfs4_delegation *dl;
1da177e4 3284
8c10cbdb
BH
3285 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3286 STATEID_VAL(stid));
1da177e4 3287
1da177e4 3288 fp = find_file(ino);
13cd2184
N
3289 if (!fp)
3290 return NULL;
3291 dl = find_delegation_file(fp, stid);
3292 put_nfs4_file(fp);
3293 return dl;
1da177e4
LT
3294}
3295
3296/*
3297 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3298 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3299 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3300 * locking, this prevents us from being completely protocol-compliant. The
3301 * real solution to this problem is to start using unsigned file offsets in
3302 * the VFS, but this is a very deep change!
3303 */
3304static inline void
3305nfs4_transform_lock_offset(struct file_lock *lock)
3306{
3307 if (lock->fl_start < 0)
3308 lock->fl_start = OFFSET_MAX;
3309 if (lock->fl_end < 0)
3310 lock->fl_end = OFFSET_MAX;
3311}
3312
d5b9026a
N
3313/* Hack!: For now, we're defining this just so we can use a pointer to it
3314 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3315static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3316};
1da177e4
LT
3317
3318static inline void
3319nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3320{
d5b9026a
N
3321 struct nfs4_stateowner *sop;
3322 unsigned int hval;
1da177e4 3323
d5b9026a
N
3324 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3325 sop = (struct nfs4_stateowner *) fl->fl_owner;
3326 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT
3327 kref_get(&sop->so_ref);
3328 deny->ld_sop = sop;
3329 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N
3330 } else {
3331 deny->ld_sop = NULL;
3332 deny->ld_clientid.cl_boot = 0;
3333 deny->ld_clientid.cl_id = 0;
1da177e4
LT
3334 }
3335 deny->ld_start = fl->fl_start;
87df4de8
BH
3336 deny->ld_length = NFS4_MAX_UINT64;
3337 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT
3338 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3339 deny->ld_type = NFS4_READ_LT;
3340 if (fl->fl_type != F_RDLCK)
3341 deny->ld_type = NFS4_WRITE_LT;
3342}
3343
1da177e4
LT
3344static struct nfs4_stateowner *
3345find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3346 struct xdr_netobj *owner)
3347{
3348 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3349 struct nfs4_stateowner *op;
3350
3351 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3352 if (same_owner_str(op, owner, clid))
1da177e4
LT
3353 return op;
3354 }
3355 return NULL;
3356}
3357
3358/*
3359 * Alloc a lock owner structure.
3360 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3361 * occured.
3362 *
3363 * strhashval = lock_ownerstr_hashval
1da177e4
LT
3364 */
3365
3366static struct nfs4_stateowner *
3367alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3368 struct nfs4_stateowner *sop;
3369 struct nfs4_replay *rp;
3370 unsigned int idhashval;
3371
3372 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3373 return NULL;
3374 idhashval = lockownerid_hashval(current_ownerid);
3375 INIT_LIST_HEAD(&sop->so_idhash);
3376 INIT_LIST_HEAD(&sop->so_strhash);
3377 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
3378 INIT_LIST_HEAD(&sop->so_stateids);
3379 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT
3380 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3381 sop->so_time = 0;
3382 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3383 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3384 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT
3385 sop->so_is_open_owner = 0;
3386 sop->so_id = current_ownerid++;
3387 sop->so_client = clp;
b59e3c0e
NB
3388 /* It is the openowner seqid that will be incremented in encode in the
3389 * case of new lockowners; so increment the lock seqid manually: */
3390 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT
3391 sop->so_confirmed = 1;
3392 rp = &sop->so_replay;
de1ae286 3393 rp->rp_status = nfserr_serverfault;
1da177e4
LT
3394 rp->rp_buflen = 0;
3395 rp->rp_buf = rp->rp_ibuf;
3396 return sop;
3397}
3398
fd39ca9a 3399static struct nfs4_stateid *
1da177e4
LT
3400alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3401{
3402 struct nfs4_stateid *stp;
3403 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3404
5ac049ac
N
3405 stp = nfs4_alloc_stateid();
3406 if (stp == NULL)
1da177e4
LT
3407 goto out;
3408 INIT_LIST_HEAD(&stp->st_hash);
3409 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N
3410 INIT_LIST_HEAD(&stp->st_perstateowner);
3411 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3412 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3413 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3414 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3415 stp->st_stateowner = sop;
13cd2184 3416 get_nfs4_file(fp);
1da177e4 3417 stp->st_file = fp;
78155ed7 3418 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
3419 stp->st_stateid.si_stateownerid = sop->so_id;
3420 stp->st_stateid.si_fileid = fp->fi_id;
3421 stp->st_stateid.si_generation = 0;
3422 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3423 stp->st_access_bmap = open_stp->st_access_bmap;
3424 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3425 stp->st_openstp = open_stp;
1da177e4
LT
3426
3427out:
3428 return stp;
3429}
3430
fd39ca9a 3431static int
1da177e4
LT
3432check_lock_length(u64 offset, u64 length)
3433{
87df4de8 3434 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT
3435 LOFF_OVERFLOW(offset, length)));
3436}
3437
3438/*
3439 * LOCK operation
3440 */
b37ad28b 3441__be32
ca364317 3442nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3443 struct nfsd4_lock *lock)
1da177e4 3444{
3e9e3dbe 3445 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3446 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT
3447 struct nfs4_stateid *lock_stp;
3448 struct file *filp;
3449 struct file_lock file_lock;
8dc7c311 3450 struct file_lock conflock;
b37ad28b 3451 __be32 status = 0;
1da177e4 3452 unsigned int strhashval;
150b3934 3453 unsigned int cmd;
b8dd7b9a 3454 int err;
1da177e4
LT
3455
3456 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3457 (long long) lock->lk_offset,
3458 (long long) lock->lk_length);
3459
1da177e4
LT
3460 if (check_lock_length(lock->lk_offset, lock->lk_length))
3461 return nfserr_inval;
3462
ca364317 3463 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3464 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA
3465 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3466 return status;
3467 }
3468
1da177e4
LT
3469 nfs4_lock_state();
3470
3471 if (lock->lk_is_new) {
893f8770
N
3472 /*
3473 * Client indicates that this is a new lockowner.
3474 * Use open owner and open stateid to create lock owner and
3475 * lock stateid.
3476 */
1da177e4
LT
3477 struct nfs4_stateid *open_stp = NULL;
3478 struct nfs4_file *fp;
3479
3480 status = nfserr_stale_clientid;
60adfc50
AA
3481 if (!nfsd4_has_session(cstate) &&
3482 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3483 goto out;
1da177e4 3484
1da177e4 3485 /* validate and update open stateid and open seqid */
dd453dfd 3486 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3487 lock->lk_new_open_seqid,
3488 &lock->lk_new_open_stateid,
f3362737 3489 OPEN_STATE,
3a65588a 3490 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3491 lock);
37515177 3492 if (status)
1da177e4 3493 goto out;
3a65588a 3494 open_sop = lock->lk_replay_owner;
1da177e4
LT
3495 /* create lockowner and lock stateid */
3496 fp = open_stp->st_file;
3497 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3498 open_sop->so_client->cl_clientid.cl_id,
3499 &lock->v.new.owner);
3e9e3dbe
N
3500 /* XXX: Do we need to check for duplicate stateowners on
3501 * the same file, or should they just be allowed (and
3502 * create new stateids)? */
1da177e4 3503 status = nfserr_resource;
b59e3c0e
NB
3504 lock_sop = alloc_init_lock_stateowner(strhashval,
3505 open_sop->so_client, open_stp, lock);
3506 if (lock_sop == NULL)
1da177e4 3507 goto out;
b59e3c0e 3508 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3509 if (lock_stp == NULL)
1da177e4 3510 goto out;
1da177e4
LT
3511 } else {
3512 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3513 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3514 lock->lk_old_lock_seqid,
3515 &lock->lk_old_lock_stateid,
f3362737 3516 LOCK_STATE,
3a65588a 3517 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT
3518 if (status)
3519 goto out;
3a65588a 3520 lock_sop = lock->lk_replay_owner;
1da177e4 3521 }
3a65588a 3522 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT
3523 filp = lock_stp->st_vfs_file;
3524
0dd395dc 3525 status = nfserr_grace;
af558e33 3526 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N
3527 goto out;
3528 status = nfserr_no_grace;
af558e33 3529 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N
3530 goto out;
3531
1da177e4
LT
3532 locks_init_lock(&file_lock);
3533 switch (lock->lk_type) {
3534 case NFS4_READ_LT:
3535 case NFS4_READW_LT:
3536 file_lock.fl_type = F_RDLCK;
150b3934 3537 cmd = F_SETLK;
1da177e4
LT
3538 break;
3539 case NFS4_WRITE_LT:
3540 case NFS4_WRITEW_LT:
3541 file_lock.fl_type = F_WRLCK;
150b3934 3542 cmd = F_SETLK;
1da177e4
LT
3543 break;
3544 default:
3545 status = nfserr_inval;
3546 goto out;
3547 }
b59e3c0e 3548 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT
3549 file_lock.fl_pid = current->tgid;
3550 file_lock.fl_file = filp;
3551 file_lock.fl_flags = FL_POSIX;
d5b9026a 3552 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3553
3554 file_lock.fl_start = lock->lk_offset;
87df4de8 3555 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT
3556 nfs4_transform_lock_offset(&file_lock);
3557
3558 /*
3559 * Try to lock the file in the VFS.
3560 * Note: locks.c uses the BKL to protect the inode's lock list.
3561 */
3562
fd85b817 3563 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3564 switch (-err) {
1da177e4
LT
3565 case 0: /* success! */
3566 update_stateid(&lock_stp->st_stateid);
3567 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3568 sizeof(stateid_t));
b8dd7b9a 3569 status = 0;
eb76b3fd
AA
3570 break;
3571 case (EAGAIN): /* conflock holds conflicting lock */
3572 status = nfserr_denied;
3573 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3574 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3575 break;
1da177e4
LT
3576 case (EDEADLK):
3577 status = nfserr_deadlock;
eb76b3fd 3578 break;
1da177e4 3579 default:
fd85b817 3580 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA
3581 status = nfserr_resource;
3582 break;
1da177e4 3583 }
1da177e4 3584out:
8a280510 3585 if (status && lock->lk_is_new && lock_sop)
f044ff83 3586 release_lockowner(lock_sop);
3a65588a
BF
3587 if (lock->lk_replay_owner) {
3588 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3589 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3590 }
1da177e4
LT
3591 nfs4_unlock_state();
3592 return status;
3593}
3594
55ef1274
BF
3595/*
3596 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3597 * so we do a temporary open here just to get an open file to pass to
3598 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3599 * inode operation.)
3600 */
3601static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3602{
3603 struct file *file;
3604 int err;
3605
3606 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3607 if (err)
3608 return err;
3609 err = vfs_test_lock(file, lock);
3610 nfsd_close(file);
3611 return err;
3612}
3613
1da177e4
LT
3614/*
3615 * LOCKT operation
3616 */
b37ad28b 3617__be32
ca364317
BF
3618nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3619 struct nfsd4_lockt *lockt)
1da177e4
LT
3620{
3621 struct inode *inode;
1da177e4 3622 struct file_lock file_lock;
fd85b817 3623 int error;
b37ad28b 3624 __be32 status;
1da177e4 3625
af558e33 3626 if (locks_in_grace())
1da177e4
LT
3627 return nfserr_grace;
3628
3629 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3630 return nfserr_inval;
3631
3632 lockt->lt_stateowner = NULL;
3633 nfs4_lock_state();
3634
3635 status = nfserr_stale_clientid;
60adfc50 3636 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3637 goto out;
1da177e4 3638
ca364317 3639 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3640 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT
3641 if (status == nfserr_symlink)
3642 status = nfserr_inval;
3643 goto out;
3644 }
3645
ca364317 3646 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
3647 locks_init_lock(&file_lock);
3648 switch (lockt->lt_type) {
3649 case NFS4_READ_LT:
3650 case NFS4_READW_LT:
3651 file_lock.fl_type = F_RDLCK;
3652 break;
3653 case NFS4_WRITE_LT:
3654 case NFS4_WRITEW_LT:
3655 file_lock.fl_type = F_WRLCK;
3656 break;
3657 default:
2fdada03 3658 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT
3659 status = nfserr_inval;
3660 goto out;
3661 }
3662
3663 lockt->lt_stateowner = find_lockstateowner_str(inode,
3664 &lockt->lt_clientid, &lockt->lt_owner);
3665 if (lockt->lt_stateowner)
3666 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3667 file_lock.fl_pid = current->tgid;
3668 file_lock.fl_flags = FL_POSIX;
3669
3670 file_lock.fl_start = lockt->lt_offset;
87df4de8 3671 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT
3672
3673 nfs4_transform_lock_offset(&file_lock);
3674
1da177e4 3675 status = nfs_ok;
55ef1274 3676 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME
3677 if (error) {
3678 status = nfserrno(error);
3679 goto out;
3680 }
9d6a8c5c 3681 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3682 status = nfserr_denied;
9d6a8c5c 3683 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT
3684 }
3685out:
3686 nfs4_unlock_state();
3687 return status;
3688}
3689
b37ad28b 3690__be32
ca364317 3691nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3692 struct nfsd4_locku *locku)
1da177e4
LT
3693{
3694 struct nfs4_stateid *stp;
3695 struct file *filp = NULL;
3696 struct file_lock file_lock;
b37ad28b 3697 __be32 status;
b8dd7b9a 3698 int err;
1da177e4
LT
3699
3700 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3701 (long long) locku->lu_offset,
3702 (long long) locku->lu_length);
3703
3704 if (check_lock_length(locku->lu_offset, locku->lu_length))
3705 return nfserr_inval;
3706
3707 nfs4_lock_state();
3708
dd453dfd 3709 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3710 locku->lu_seqid,
3711 &locku->lu_stateid,
f3362737 3712 LOCK_STATE,
1da177e4
LT
3713 &locku->lu_stateowner, &stp, NULL)))
3714 goto out;
3715
3716 filp = stp->st_vfs_file;
3717 BUG_ON(!filp);
3718 locks_init_lock(&file_lock);
3719 file_lock.fl_type = F_UNLCK;
3720 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3721 file_lock.fl_pid = current->tgid;
3722 file_lock.fl_file = filp;
3723 file_lock.fl_flags = FL_POSIX;
d5b9026a 3724 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3725 file_lock.fl_start = locku->lu_offset;
3726
87df4de8 3727 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT
3728 nfs4_transform_lock_offset(&file_lock);
3729
3730 /*
3731 * Try to unlock the file in the VFS.
3732 */
fd85b817 3733 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3734 if (err) {
fd85b817 3735 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT
3736 goto out_nfserr;
3737 }
3738 /*
3739 * OK, unlock succeeded; the only thing left to do is update the stateid.
3740 */
3741 update_stateid(&stp->st_stateid);
3742 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3743
3744out:
f2327d9a 3745 if (locku->lu_stateowner) {
1da177e4 3746 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3747 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3748 }
1da177e4
LT
3749 nfs4_unlock_state();
3750 return status;
3751
3752out_nfserr:
b8dd7b9a 3753 status = nfserrno(err);
1da177e4
LT
3754 goto out;
3755}
3756
3757/*
3758 * returns
3759 * 1: locks held by lockowner
3760 * 0: no locks held by lockowner
3761 */
3762static int
3763check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3764{
3765 struct file_lock **flpp;
7eaa36e2 3766 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT
3767 int status = 0;
3768
3769 lock_kernel();
3770 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3771 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT
3772 status = 1;
3773 goto out;
796dadfd 3774 }
1da177e4
LT
3775 }
3776out:
3777 unlock_kernel();
3778 return status;
3779}
3780
b37ad28b 3781__be32
b591480b
BF
3782nfsd4_release_lockowner(struct svc_rqst *rqstp,
3783 struct nfsd4_compound_state *cstate,
3784 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT
3785{
3786 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N
3787 struct nfs4_stateowner *sop;
3788 struct nfs4_stateid *stp;
1da177e4 3789 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N
3790 struct list_head matches;
3791 int i;
b37ad28b 3792 __be32 status;
1da177e4
LT
3793
3794 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3795 clid->cl_boot, clid->cl_id);
3796
3797 /* XXX check for lease expiration */
3798
3799 status = nfserr_stale_clientid;
849823c5 3800 if (STALE_CLIENTID(clid))
1da177e4 3801 return status;
1da177e4
LT
3802
3803 nfs4_lock_state();
3804
3e9e3dbe
N
3805 status = nfserr_locks_held;
3806 /* XXX: we're doing a linear search through all the lockowners.
3807 * Yipes! For now we'll just hope clients aren't really using
3808 * release_lockowner much, but eventually we have to fix these
3809 * data structures. */
3810 INIT_LIST_HEAD(&matches);
3811 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3812 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3813 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N
3814 continue;
3815 list_for_each_entry(stp, &sop->so_stateids,
3816 st_perstateowner) {
3817 if (check_for_locks(stp->st_vfs_file, sop))
3818 goto out;
3819 /* Note: so_perclient unused for lockowners,
3820 * so it's OK to fool with here. */
3821 list_add(&sop->so_perclient, &matches);
3822 }
1da177e4 3823 }
3e9e3dbe
N
3824 }
3825 /* Clients probably won't expect us to return with some (but not all)
3826 * of the lockowner state released; so don't release any until all
3827 * have been checked. */
3828 status = nfs_ok;
0fa822e4
N
3829 while (!list_empty(&matches)) {
3830 sop = list_entry(matches.next, struct nfs4_stateowner,
3831 so_perclient);
3832 /* unhash_stateowner deletes so_perclient only
3833 * for openowners. */
3834 list_del(&sop->so_perclient);
f044ff83 3835 release_lockowner(sop);
1da177e4
LT
3836 }
3837out:
3838 nfs4_unlock_state();
3839 return status;
3840}
3841
3842static inline struct nfs4_client_reclaim *
a55370a3 3843alloc_reclaim(void)
1da177e4 3844{
a55370a3 3845 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT
3846}
3847
c7b9a459 3848int
a1bcecd2 3849nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N
3850{
3851 unsigned int strhashval = clientstr_hashval(name);
3852 struct nfs4_client *clp;
3853
a1bcecd2 3854 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N
3855 return clp ? 1 : 0;
3856}
3857
1da177e4
LT
3858/*
3859 * failure => all reset bets are off, nfserr_no_grace...
3860 */
190e4fbf
N
3861int
3862nfs4_client_to_reclaim(const char *name)
1da177e4
LT
3863{
3864 unsigned int strhashval;
3865 struct nfs4_client_reclaim *crp = NULL;
3866
a55370a3
N
3867 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3868 crp = alloc_reclaim();
1da177e4
LT
3869 if (!crp)
3870 return 0;
a55370a3 3871 strhashval = clientstr_hashval(name);
1da177e4
LT
3872 INIT_LIST_HEAD(&crp->cr_strhash);
3873 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3874 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT
3875 reclaim_str_hashtbl_size++;
3876 return 1;
3877}
3878
3879static void
3880nfs4_release_reclaim(void)
3881{
3882 struct nfs4_client_reclaim *crp = NULL;
3883 int i;
3884
1da177e4
LT
3885 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3886 while (!list_empty(&reclaim_str_hashtbl[i])) {
3887 crp = list_entry(reclaim_str_hashtbl[i].next,
3888 struct nfs4_client_reclaim, cr_strhash);
3889 list_del(&crp->cr_strhash);
1da177e4
LT
3890 kfree(crp);
3891 reclaim_str_hashtbl_size--;
3892 }
3893 }
3894 BUG_ON(reclaim_str_hashtbl_size);
3895}
3896
3897/*
3898 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3899static struct nfs4_client_reclaim *
1da177e4
LT
3900nfs4_find_reclaim_client(clientid_t *clid)
3901{
3902 unsigned int strhashval;
3903 struct nfs4_client *clp;
3904 struct nfs4_client_reclaim *crp = NULL;
3905
3906
3907 /* find clientid in conf_id_hashtbl */
3908 clp = find_confirmed_client(clid);
3909 if (clp == NULL)
3910 return NULL;
3911
a55370a3
N
3912 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3913 clp->cl_name.len, clp->cl_name.data,
3914 clp->cl_recdir);
1da177e4
LT
3915
3916 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3917 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3918 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3919 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT
3920 return crp;
3921 }
3922 }
3923 return NULL;
3924}
3925
3926/*
3927* Called from OPEN. Look for clientid in reclaim list.
3928*/
b37ad28b 3929__be32
1da177e4
LT
3930nfs4_check_open_reclaim(clientid_t *clid)
3931{
dfc83565 3932 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT
3933}
3934
ac4d8ff2 3935/* initialization to perform at module load time: */
1da177e4 3936
e8ff2a84 3937int
ac4d8ff2 3938nfs4_state_init(void)
1da177e4 3939{
e8ff2a84 3940 int i, status;
1da177e4 3941
e8ff2a84
BF
3942 status = nfsd4_init_slabs();
3943 if (status)
3944 return status;
1da177e4
LT
3945 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3946 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3947 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3948 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3949 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3950 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3951 }
5282fd72
ME
3952 for (i = 0; i < SESSION_HASH_SIZE; i++)
3953 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT
3954 for (i = 0; i < FILE_HASH_SIZE; i++) {
3955 INIT_LIST_HEAD(&file_hashtbl[i]);
3956 }
3957 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3958 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3959 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3960 }
3961 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3962 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3963 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3964 }
3965 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3966 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3967 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3968 }
1da177e4 3969 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT
3970 INIT_LIST_HEAD(&close_lru);
3971 INIT_LIST_HEAD(&client_lru);
3972 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3973 reclaim_str_hashtbl_size = 0;
e8ff2a84 3974 return 0;
ac4d8ff2
N
3975}
3976
190e4fbf
N
3977static void
3978nfsd4_load_reboot_recovery_data(void)
3979{
3980 int status;
3981
0964a3d3
N
3982 nfs4_lock_state();
3983 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3984 status = nfsd4_recdir_load();
0964a3d3 3985 nfs4_unlock_state();
190e4fbf
N
3986 if (status)
3987 printk("NFSD: Failure reading reboot recovery data\n");
3988}
3989
9a8db97e
ME
3990unsigned long
3991get_nfs4_grace_period(void)
3992{
3993 return max(user_lease_time, lease_time) * HZ;
3994}
3995
c2f1a551
MS
3996/*
3997 * Since the lifetime of a delegation isn't limited to that of an open, a
3998 * client may quite reasonably hang on to a delegation as long as it has
3999 * the inode cached. This becomes an obvious problem the first time a
4000 * client's inode cache approaches the size of the server's total memory.
4001 *
4002 * For now we avoid this problem by imposing a hard limit on the number
4003 * of delegations, which varies according to the server's memory size.
4004 */
4005static void
4006set_max_delegations(void)
4007{
4008 /*
4009 * Allow at most 4 delegations per megabyte of RAM. Quick
4010 * estimates suggest that in the worst case (where every delegation
4011 * is for a different inode), a delegation could take about 1.5K,
4012 * giving a worst case usage of about 6% of memory.
4013 */
4014 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4015}
4016
ac4d8ff2
N
4017/* initialization to perform when the nfsd service is started: */
4018
29ab23cc 4019static int
ac4d8ff2
N
4020__nfs4_state_start(void)
4021{
9a8db97e 4022 unsigned long grace_time;
ac4d8ff2 4023
1da177e4 4024 boot_time = get_seconds();
af558e33 4025 grace_time = get_nfs4_grace_period();
d99a05ad 4026 lease_time = user_lease_time;
af558e33 4027 locks_start_grace(&nfsd4_manager);
9a8db97e
ME
4028 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4029 grace_time/HZ);
58da282b 4030 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF
4031 if (laundry_wq == NULL)
4032 return -ENOMEM;
9a8db97e 4033 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4034 set_max_delegations();
80fc015b 4035 return set_callback_cred();
1da177e4
LT
4036}
4037
29ab23cc 4038int
76a3550e 4039nfs4_state_start(void)
1da177e4 4040{
29ab23cc
BF
4041 int ret;
4042
1da177e4 4043 if (nfs4_init)
29ab23cc 4044 return 0;
190e4fbf 4045 nfsd4_load_reboot_recovery_data();
29ab23cc
BF
4046 ret = __nfs4_state_start();
4047 if (ret)
4048 return ret;
1da177e4 4049 nfs4_init = 1;
29ab23cc 4050 return 0;
1da177e4
LT
4051}
4052
1da177e4
LT
4053time_t
4054nfs4_lease_time(void)
4055{
4056 return lease_time;
4057}
4058
4059static void
4060__nfs4_state_shutdown(void)
4061{
4062 int i;
4063 struct nfs4_client *clp = NULL;
4064 struct nfs4_delegation *dp = NULL;
1da177e4
LT
4065 struct list_head *pos, *next, reaplist;
4066
1da177e4
LT
4067 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4068 while (!list_empty(&conf_id_hashtbl[i])) {
4069 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4070 expire_client(clp);
4071 }
4072 while (!list_empty(&unconf_str_hashtbl[i])) {
4073 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4074 expire_client(clp);
4075 }
4076 }
4077 INIT_LIST_HEAD(&reaplist);
4078 spin_lock(&recall_lock);
4079 list_for_each_safe(pos, next, &del_recall_lru) {
4080 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4081 list_move(&dp->dl_recall_lru, &reaplist);
4082 }
4083 spin_unlock(&recall_lock);
4084 list_for_each_safe(pos, next, &reaplist) {
4085 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4086 list_del_init(&dp->dl_recall_lru);
4087 unhash_delegation(dp);
4088 }
4089
190e4fbf 4090 nfsd4_shutdown_recdir();
1da177e4 4091 nfs4_init = 0;
1da177e4
LT
4092}
4093
4094void
4095nfs4_state_shutdown(void)
4096{
5e8d5c29
N
4097 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4098 destroy_workqueue(laundry_wq);
2c5e7615 4099 locks_end_grace(&nfsd4_manager);
1da177e4
LT
4100 nfs4_lock_state();
4101 nfs4_release_reclaim();
4102 __nfs4_state_shutdown();
1da177e4
LT
4103 nfs4_unlock_state();
4104}
4105
3dd98a3b
JL
4106/*
4107 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4108 * accessed when nfsd is starting.
4109 */
0964a3d3
N
4110static void
4111nfs4_set_recdir(char *recdir)
4112{
0964a3d3 4113 strcpy(user_recovery_dirname, recdir);
0964a3d3
N
4114}
4115
4116/*
4117 * Change the NFSv4 recovery directory to recdir.
4118 */
4119int
4120nfs4_reset_recoverydir(char *recdir)
4121{
4122 int status;
a63bb996 4123 struct path path;
0964a3d3 4124
a63bb996 4125 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N
4126 if (status)
4127 return status;
4128 status = -ENOTDIR;
a63bb996 4129 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N
4130 nfs4_set_recdir(recdir);
4131 status = 0;
4132 }
a63bb996 4133 path_put(&path);
0964a3d3
N
4134 return status;
4135}
4136
3dd98a3b
JL
4137char *
4138nfs4_recoverydir(void)
4139{
4140 return user_recovery_dirname;
4141}
4142
1da177e4
LT
4143/*
4144 * Called when leasetime is changed.
4145 *
d99a05ad
N
4146 * The only way the protocol gives us to handle on-the-fly lease changes is to
4147 * simulate a reboot. Instead of doing that, we just wait till the next time
4148 * we start to register any changes in lease time. If the administrator
4149 * really wants to change the lease time *now*, they can go ahead and bring
4150 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL
4151 *
4152 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4153 * when nfsd is starting
1da177e4
LT
4154 */
4155void
4156nfs4_reset_lease(time_t leasetime)
4157{
d99a05ad 4158 user_lease_time = leasetime;
1da177e4 4159}
This page took 0.743453 seconds and 5 git commands to generate.