]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/nfsd/nfs4state.c
projects / mirror_ubuntu-zesty-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: modify nfsd4.1 backchannel to use new xprt class
[mirror_ubuntu-zesty-kernel.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;
452
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 }
b101ebbc 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
480 */
481static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 482 struct nfsd4_channel_attrs *session_fchan,
483 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 484{
485 int status = 0;
486 __u32 maxcount = svc_max_payload(rqstp);
487
488 /* headerpadsz set to zero in encode routine */
489
490 /* Use the client's max request and max response size if possible */
491 if (fchan->maxreq_sz > maxcount)
492 fchan->maxreq_sz = maxcount;
6c18ba9f 493 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 494
495 if (fchan->maxresp_sz > maxcount)
496 fchan->maxresp_sz = maxcount;
6c18ba9f 497 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 498
ec6b5d7b
AA		 499 /* Use the client's maxops if possible */
500 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
501 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 502 session_fchan->maxops = fchan->maxops;
ec6b5d7b 503
ec6b5d7b
AA		 504 /* FIXME: Error means no more DRC pages so the server should
505 * recover pages from existing sessions. For now fail session
506 * creation.
507 */
a649637c 508 status = set_forechannel_drc_size(fchan);
ec6b5d7b 509
a649637c 510 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 511 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 512
513 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 514 return status;
515}
516
557ce264
AA		 517static void
518free_session_slots(struct nfsd4_session *ses)
519{
520 int i;
521
522 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
523 kfree(ses->se_slots[i]);
524}
525
ec6b5d7b
AA		 526static int
527alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
528 struct nfsd4_create_session *cses)
529{
530 struct nfsd4_session *new, tmp;
557ce264
AA		 531 struct nfsd4_slot *sp;
532 int idx, slotsize, cachesize, i;
533 int status;
ec6b5d7b
AA		 534
535 memset(&tmp, 0, sizeof(tmp));
536
537 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 538 tmp.se_bchannel = cses->back_channel;
539 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
540 &cses->fore_channel);
ec6b5d7b
AA		 541 if (status)
542 goto out;
543
557ce264
AA		 544 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
545 + sizeof(struct nfsd4_session) > PAGE_SIZE);
546
547 status = nfserr_serverfault;
548 /* allocate struct nfsd4_session and slot table pointers in one piece */
549 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 550 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
551 if (!new)
552 goto out;
553
554 memcpy(new, &tmp, sizeof(*new));
555
557ce264
AA		 556 /* allocate each struct nfsd4_slot and data cache in one piece */
557 cachesize = new->se_fchannel.maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
558 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
559 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
560 if (!sp)
561 goto out_free;
562 new->se_slots[i] = sp;
563 }
564
ec6b5d7b
AA		 565 new->se_client = clp;
566 gen_sessionid(new);
567 idx = hash_sessionid(&new->se_sessionid);
568 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
569 NFS4_MAX_SESSIONID_LEN);
570
571 new->se_flags = cses->flags;
572 kref_init(&new->se_ref);
573 spin_lock(&sessionid_lock);
574 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
575 list_add(&new->se_perclnt, &clp->cl_sessions);
576 spin_unlock(&sessionid_lock);
577
578 status = nfs_ok;
579out:
580 return status;
557ce264
AA		 581out_free:
582 free_session_slots(new);
583 kfree(new);
584 goto out;
ec6b5d7b
AA		 585}
586
5282fd72
ME		 587/* caller must hold sessionid_lock */
588static struct nfsd4_session *
589find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
590{
591 struct nfsd4_session *elem;
592 int idx;
593
594 dump_sessionid(__func__, sessionid);
595 idx = hash_sessionid(sessionid);
596 dprintk("%s: idx is %d\n", __func__, idx);
597 /* Search in the appropriate list */
598 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
599 dump_sessionid("list traversal", &elem->se_sessionid);
600 if (!memcmp(elem->se_sessionid.data, sessionid->data,
601 NFS4_MAX_SESSIONID_LEN)) {
602 return elem;
603 }
604 }
605
606 dprintk("%s: session not found\n", __func__);
607 return NULL;
608}
609
610/* caller must hold sessionid_lock */
7116ed6b 611static void
5282fd72 612unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 613{
614 list_del(&ses->se_hash);
615 list_del(&ses->se_perclnt);
5282fd72
ME		 616}
617
618static void
619release_session(struct nfsd4_session *ses)
620{
621 spin_lock(&sessionid_lock);
622 unhash_session(ses);
623 spin_unlock(&sessionid_lock);
7116ed6b
AA		 624 nfsd4_put_session(ses);
625}
626
627void
628free_session(struct kref *kref)
629{
630 struct nfsd4_session *ses;
631
632 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 633 spin_lock(&nfsd_drc_lock);
0c193054 634 nfsd_drc_mem_used -= ses->se_fchannel.maxreqs * NFSD_SLOT_CACHE_SIZE;
be98d1bb 635 spin_unlock(&nfsd_drc_lock);
557ce264 636 free_session_slots(ses);
7116ed6b
AA		 637 kfree(ses);
638}
639
1da177e4
LT		 640static inline void
641renew_client(struct nfs4_client *clp)
642{
643 /*
644 * Move client to the end to the LRU list.
645 */
646 dprintk("renewing client (clientid %08x/%08x)\n",
647 clp->cl_clientid.cl_boot,
648 clp->cl_clientid.cl_id);
649 list_move_tail(&clp->cl_lru, &client_lru);
650 clp->cl_time = get_seconds();
651}
652
653/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
654static int
655STALE_CLIENTID(clientid_t *clid)
656{
657 if (clid->cl_boot == boot_time)
658 return 0;
60adfc50
AA		 659 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
660 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 661 return 1;
662}
663
664/*
665 * XXX Should we use a slab cache ?
666 * This type of memory management is somewhat inefficient, but we use it
667 * anyway since SETCLIENTID is not a common operation.
668 */
35bba9a3 669static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 670{
671 struct nfs4_client *clp;
672
35bba9a3
BF		 673 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
674 if (clp == NULL)
675 return NULL;
676 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
677 if (clp->cl_name.data == NULL) {
678 kfree(clp);
679 return NULL;
1da177e4 680 }
35bba9a3
BF		 681 memcpy(clp->cl_name.data, name.data, name.len);
682 clp->cl_name.len = name.len;
1da177e4
LT		 683 return clp;
684}
685
1b1a9b31
BF		 686static void
687shutdown_callback_client(struct nfs4_client *clp)
688{
c237dc03 689 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 690
1b1a9b31 691 if (clnt) {
404ec117
BF		 692 /*
693 * Callback threads take a reference on the client, so there
694 * should be no outstanding callbacks at this point.
695 */
c237dc03 696 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 697 rpc_shutdown_client(clnt);
698 }
699}
700
1da177e4
LT		 701static inline void
702free_client(struct nfs4_client *clp)
703{
1b1a9b31 704 shutdown_callback_client(clp);
38524ab3
AA		 705 if (clp->cl_cb_xprt)
706 svc_xprt_put(clp->cl_cb_xprt);
1da177e4
LT		 707 if (clp->cl_cred.cr_group_info)
708 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 709 kfree(clp->cl_principal);
1da177e4
LT		 710 kfree(clp->cl_name.data);
711 kfree(clp);
712}
713
714void
715put_nfs4_client(struct nfs4_client *clp)
716{
717 if (atomic_dec_and_test(&clp->cl_count))
718 free_client(clp);
719}
720
721static void
722expire_client(struct nfs4_client *clp)
723{
724 struct nfs4_stateowner *sop;
725 struct nfs4_delegation *dp;
1da177e4
LT		 726 struct list_head reaplist;
727
728 dprintk("NFSD: expire_client cl_count %d\n",
729 atomic_read(&clp->cl_count));
730
1da177e4
LT		 731 INIT_LIST_HEAD(&reaplist);
732 spin_lock(&recall_lock);
ea1da636
N		 733 while (!list_empty(&clp->cl_delegations)) {
734 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 735 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
736 dp->dl_flock);
ea1da636 737 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 738 list_move(&dp->dl_recall_lru, &reaplist);
739 }
740 spin_unlock(&recall_lock);
741 while (!list_empty(&reaplist)) {
742 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
743 list_del_init(&dp->dl_recall_lru);
744 unhash_delegation(dp);
745 }
746 list_del(&clp->cl_idhash);
747 list_del(&clp->cl_strhash);
748 list_del(&clp->cl_lru);
ea1da636
N		 749 while (!list_empty(&clp->cl_openowners)) {
750 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 751 release_openowner(sop);
1da177e4 752 }
c4bf7868
ME		 753 while (!list_empty(&clp->cl_sessions)) {
754 struct nfsd4_session *ses;
755 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
756 se_perclnt);
757 release_session(ses);
758 }
1da177e4
LT		 759 put_nfs4_client(clp);
760}
761
35bba9a3
BF		 762static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
763{
1da177e4
LT		 764 struct nfs4_client *clp;
765
35bba9a3
BF		 766 clp = alloc_client(name);
767 if (clp == NULL)
768 return NULL;
a55370a3 769 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 770 atomic_set(&clp->cl_count, 1);
c237dc03 771 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 772 INIT_LIST_HEAD(&clp->cl_idhash);
773 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 774 INIT_LIST_HEAD(&clp->cl_openowners);
775 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 776 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 777 INIT_LIST_HEAD(&clp->cl_lru);
199ff35e
RL		 778 clear_bit(0, &clp->cl_cb_slot_busy);
779 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1da177e4
LT		 780 return clp;
781}
782
35bba9a3
BF		 783static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
784{
785 memcpy(target->cl_verifier.data, source->data,
786 sizeof(target->cl_verifier.data));
1da177e4
LT		 787}
788
35bba9a3
BF		 789static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
790{
1da177e4
LT		 791 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
792 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
793}
794
35bba9a3
BF		 795static void copy_cred(struct svc_cred *target, struct svc_cred *source)
796{
1da177e4
LT		 797 target->cr_uid = source->cr_uid;
798 target->cr_gid = source->cr_gid;
799 target->cr_group_info = source->cr_group_info;
800 get_group_info(target->cr_group_info);
801}
802
35bba9a3 803static int same_name(const char *n1, const char *n2)
599e0a22 804{
a55370a3 805 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 806}
807
808static int
599e0a22
BF		 809same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
810{
811 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 812}
813
814static int
599e0a22
BF		 815same_clid(clientid_t *cl1, clientid_t *cl2)
816{
817 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 818}
819
820/* XXX what about NGROUP */
821static int
599e0a22
BF		 822same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
823{
824 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 825}
826
5ec7b46c
BF		 827static void gen_clid(struct nfs4_client *clp)
828{
829 static u32 current_clientid = 1;
830
1da177e4
LT		 831 clp->cl_clientid.cl_boot = boot_time;
832 clp->cl_clientid.cl_id = current_clientid++;
833}
834
deda2faa
BF		 835static void gen_confirm(struct nfs4_client *clp)
836{
837 static u32 i;
838 u32 *p;
1da177e4 839
1da177e4 840 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 841 *p++ = get_seconds();
842 *p++ = i++;
1da177e4
LT		 843}
844
35bba9a3
BF		 845static int check_name(struct xdr_netobj name)
846{
1da177e4
LT		 847 if (name.len == 0)
848 return 0;
849 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 850 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 851 return 0;
852 }
853 return 1;
854}
855
fd39ca9a 856static void
1da177e4
LT		 857add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
858{
859 unsigned int idhashval;
860
861 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
862 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
863 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
864 list_add_tail(&clp->cl_lru, &client_lru);
865 clp->cl_time = get_seconds();
866}
867
fd39ca9a 868static void
1da177e4
LT		 869move_to_confirmed(struct nfs4_client *clp)
870{
871 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
872 unsigned int strhashval;
873
874 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
875 list_del_init(&clp->cl_strhash);
f116629d 876 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 877 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 878 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
879 renew_client(clp);
880}
881
882static struct nfs4_client *
883find_confirmed_client(clientid_t *clid)
884{
885 struct nfs4_client *clp;
886 unsigned int idhashval = clientid_hashval(clid->cl_id);
887
888 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 889 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 890 return clp;
891 }
892 return NULL;
893}
894
895static struct nfs4_client *
896find_unconfirmed_client(clientid_t *clid)
897{
898 struct nfs4_client *clp;
899 unsigned int idhashval = clientid_hashval(clid->cl_id);
900
901 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 902 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 903 return clp;
904 }
905 return NULL;
906}
907
a1bcecd2
AA		 908/*
909 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
910 * parameter. Matching is based on the fact the at least one of the
911 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
912 *
913 * FIXME: we need to unify the clientid namespaces for nfsv4.x
914 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
915 * and SET_CLIENTID{,_CONFIRM}
916 */
917static inline int
918match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
919{
920 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
921 return use_exchange_id == has_exchange_flags;
922}
923
28ce6054 924static struct nfs4_client *
a1bcecd2
AA		 925find_confirmed_client_by_str(const char *dname, unsigned int hashval,
926 bool use_exchange_id)
28ce6054
N		 927{
928 struct nfs4_client *clp;
929
930 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 931 if (same_name(clp->cl_recdir, dname) &&
932 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 933 return clp;
934 }
935 return NULL;
936}
937
938static struct nfs4_client *
a1bcecd2
AA		 939find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
940 bool use_exchange_id)
28ce6054
N		 941{
942 struct nfs4_client *clp;
943
944 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 945 if (same_name(clp->cl_recdir, dname) &&
946 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 947 return clp;
948 }
949 return NULL;
950}
951
fd39ca9a 952static void
fbf4665f 953gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 954{
c237dc03 955 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 956 unsigned short expected_family;
957
958 /* Currently, we only support tcp and tcp6 for the callback channel */
959 if (se->se_callback_netid_len == 3 &&
960 !memcmp(se->se_callback_netid_val, "tcp", 3))
961 expected_family = AF_INET;
962 else if (se->se_callback_netid_len == 4 &&
963 !memcmp(se->se_callback_netid_val, "tcp6", 4))
964 expected_family = AF_INET6;
965 else
1da177e4
LT		 966 goto out_err;
967
aa9a4ec7
JL		 968 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
969 se->se_callback_addr_len,
970 (struct sockaddr *) &cb->cb_addr,
971 sizeof(cb->cb_addr));
972
7077ecba 973 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 974 goto out_err;
aa9a4ec7 975
fbf4665f
JL		 976 if (cb->cb_addr.ss_family == AF_INET6)
977 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
978
ab52ae6d 979 cb->cb_minorversion = 0;
1da177e4
LT		 980 cb->cb_prog = se->se_callback_prog;
981 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 982 return;
983out_err:
aa9a4ec7
JL		 984 cb->cb_addr.ss_family = AF_UNSPEC;
985 cb->cb_addrlen = 0;
849823c5 986 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 987 "will not receive delegations\n",
988 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
989
1da177e4
LT		 990 return;
991}
992
074fe897 993/*
557ce264 994 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897
AA		 995 */
996void
997nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
998{
557ce264
AA		 999 struct nfsd4_slot *slot = resp->cstate.slot;
1000 unsigned int base;
074fe897 1001
557ce264 1002 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1003
557ce264
AA		 1004 slot->sl_opcnt = resp->opcnt;
1005 slot->sl_status = resp->cstate.status;
bf864a31
AA		 1006
1007 if (nfsd4_not_cached(resp)) {
557ce264 1008 slot->sl_datalen = 0;
bf864a31
AA		 1009 return;
1010 }
557ce264
AA		 1011 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1012 base = (char *)resp->cstate.datap -
1013 (char *)resp->xbuf->head[0].iov_base;
1014 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1015 slot->sl_datalen))
1016 WARN("%s: sessions DRC could not cache compound\n", __func__);
1017 return;
074fe897
AA		 1018}
1019
abfabf8c
AA		 1020/*
1021 * Encode the replay sequence operation from the slot values.
1022 * If cachethis is FALSE encode the uncached rep error on the next
1023 * operation which sets resp->p and increments resp->opcnt for
1024 * nfs4svc_encode_compoundres.
1025 *
1026 */
1027static __be32
1028nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1029 struct nfsd4_compoundres *resp)
1030{
1031 struct nfsd4_op *op;
1032 struct nfsd4_slot *slot = resp->cstate.slot;
1033
1034 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1035 resp->opcnt, resp->cstate.slot->sl_cachethis);
abfabf8c
AA		 1036
1037 /* Encode the replayed sequence operation */
1038 op = &args->ops[resp->opcnt - 1];
1039 nfsd4_encode_operation(resp, op);
1040
1041 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1042 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1043 op = &args->ops[resp->opcnt++];
1044 op->status = nfserr_retry_uncached_rep;
1045 nfsd4_encode_operation(resp, op);
1046 }
1047 return op->status;
1048}
1049
074fe897 1050/*
557ce264
AA		 1051 * The sequence operation is not cached because we can use the slot and
1052 * session values.
074fe897
AA		 1053 */
1054__be32
bf864a31
AA		 1055nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1056 struct nfsd4_sequence *seq)
074fe897 1057{
557ce264 1058 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1059 __be32 status;
1060
557ce264 1061 dprintk("--> %s slot %p\n", __func__, slot);
abfabf8c
AA		 1062
1063 /* Either returns 0 or nfserr_retry_uncached */
1064 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1065 if (status == nfserr_retry_uncached_rep)
1066 return status;
074fe897 1067
557ce264
AA		 1068 /* The sequence operation has been encoded, cstate->datap set. */
1069 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1070
557ce264
AA		 1071 resp->opcnt = slot->sl_opcnt;
1072 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1073 status = slot->sl_status;
074fe897
AA		 1074
1075 return status;
1076}
1077
0733d213
AA		 1078/*
1079 * Set the exchange_id flags returned by the server.
1080 */
1081static void
1082nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1083{
1084 /* pNFS is not supported */
1085 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1086
1087 /* Referrals are supported, Migration is not. */
1088 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1089
1090 /* set the wire flags to return to client. */
1091 clid->flags = new->cl_exchange_flags;
1092}
1093
069b6ad4
AA		 1094__be32
1095nfsd4_exchange_id(struct svc_rqst *rqstp,
1096 struct nfsd4_compound_state *cstate,
1097 struct nfsd4_exchange_id *exid)
1098{
0733d213
AA		 1099 struct nfs4_client *unconf, *conf, *new;
1100 int status;
1101 unsigned int strhashval;
1102 char dname[HEXDIR_LEN];
363168b4 1103 char addr_str[INET6_ADDRSTRLEN];
0733d213 1104 nfs4_verifier verf = exid->verifier;
363168b4 1105 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1106
363168b4 1107 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1108 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1109 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1110 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1111 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1112
1113 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1114 return nfserr_inval;
1115
1116 /* Currently only support SP4_NONE */
1117 switch (exid->spa_how) {
1118 case SP4_NONE:
1119 break;
1120 case SP4_SSV:
1121 return nfserr_encr_alg_unsupp;
1122 default:
1123 BUG(); /* checked by xdr code */
1124 case SP4_MACH_CRED:
1125 return nfserr_serverfault; /* no excuse :-/ */
1126 }
1127
1128 status = nfs4_make_rec_clidname(dname, &exid->clname);
1129
1130 if (status)
1131 goto error;
1132
1133 strhashval = clientstr_hashval(dname);
1134
1135 nfs4_lock_state();
1136 status = nfs_ok;
1137
a1bcecd2 1138 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1139 if (conf) {
1140 if (!same_verf(&verf, &conf->cl_verifier)) {
1141 /* 18.35.4 case 8 */
1142 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1143 status = nfserr_not_same;
1144 goto out;
1145 }
1146 /* Client reboot: destroy old state */
1147 expire_client(conf);
1148 goto out_new;
1149 }
1150 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1151 /* 18.35.4 case 9 */
1152 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1153 status = nfserr_perm;
1154 goto out;
1155 }
1156 expire_client(conf);
1157 goto out_new;
1158 }
0733d213
AA		 1159 /*
1160 * Set bit when the owner id and verifier map to an already
1161 * confirmed client id (18.35.3).
1162 */
1163 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1164
1165 /*
1166 * Falling into 18.35.4 case 2, possible router replay.
1167 * Leave confirmed record intact and return same result.
1168 */
1169 copy_verf(conf, &verf);
1170 new = conf;
1171 goto out_copy;
6ddbbbfe
MS		 1172 }
1173
1174 /* 18.35.4 case 7 */
1175 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1176 status = nfserr_noent;
1177 goto out;
0733d213
AA		 1178 }
1179
a1bcecd2 1180 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1181 if (unconf) {
1182 /*
1183 * Possible retry or client restart. Per 18.35.4 case 4,
1184 * a new unconfirmed record should be generated regardless
1185 * of whether any properties have changed.
1186 */
1187 expire_client(unconf);
1188 }
1189
1190out_new:
1191 /* Normal case */
1192 new = create_client(exid->clname, dname);
1193 if (new == NULL) {
bdac86e2 1194 status = nfserr_serverfault;
0733d213
AA		 1195 goto out;
1196 }
1197
1198 copy_verf(new, &verf);
1199 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1200 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1201 gen_clid(new);
1202 gen_confirm(new);
1203 add_to_unconfirmed(new, strhashval);
1204out_copy:
1205 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1206 exid->clientid.cl_id = new->cl_clientid.cl_id;
1207
38eb76a5 1208 exid->seqid = 1;
0733d213
AA		 1209 nfsd4_set_ex_flags(new, exid);
1210
1211 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1212 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1213 status = nfs_ok;
1214
1215out:
1216 nfs4_unlock_state();
1217error:
1218 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1219 return status;
069b6ad4
AA		 1220}
1221
b85d4c01 1222static int
88e588d5 1223check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1224{
88e588d5
AA		 1225 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1226 slot_seqid);
b85d4c01
BH		 1227
1228 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1229 if (slot_inuse) {
1230 if (seqid == slot_seqid)
b85d4c01
BH		 1231 return nfserr_jukebox;
1232 else
1233 return nfserr_seq_misordered;
1234 }
1235 /* Normal */
88e588d5 1236 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1237 return nfs_ok;
1238 /* Replay */
88e588d5 1239 if (seqid == slot_seqid)
b85d4c01
BH		 1240 return nfserr_replay_cache;
1241 /* Wraparound */
88e588d5 1242 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1243 return nfs_ok;
1244 /* Misordered replay or misordered new request */
1245 return nfserr_seq_misordered;
1246}
1247
49557cc7
AA		 1248/*
1249 * Cache the create session result into the create session single DRC
1250 * slot cache by saving the xdr structure. sl_seqid has been set.
1251 * Do this for solo or embedded create session operations.
1252 */
1253static void
1254nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1255 struct nfsd4_clid_slot *slot, int nfserr)
1256{
1257 slot->sl_status = nfserr;
1258 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1259}
1260
1261static __be32
1262nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1263 struct nfsd4_clid_slot *slot)
1264{
1265 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1266 return slot->sl_status;
1267}
1268
069b6ad4
AA		 1269__be32
1270nfsd4_create_session(struct svc_rqst *rqstp,
1271 struct nfsd4_compound_state *cstate,
1272 struct nfsd4_create_session *cr_ses)
1273{
363168b4 1274 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1275 struct nfs4_client *conf, *unconf;
49557cc7 1276 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1277 int status = 0;
1278
1279 nfs4_lock_state();
1280 unconf = find_unconfirmed_client(&cr_ses->clientid);
1281 conf = find_confirmed_client(&cr_ses->clientid);
1282
1283 if (conf) {
49557cc7
AA		 1284 cs_slot = &conf->cl_cs_slot;
1285 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1286 if (status == nfserr_replay_cache) {
ec6b5d7b 1287 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1288 cs_slot->sl_seqid);
38eb76a5 1289 /* Return the cached reply status */
49557cc7 1290 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1291 goto out;
49557cc7 1292 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1293 status = nfserr_seq_misordered;
1294 dprintk("Sequence misordered!\n");
1295 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1296 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1297 goto out;
1298 }
49557cc7 1299 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1300 } else if (unconf) {
1301 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1302 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1303 status = nfserr_clid_inuse;
1304 goto out;
1305 }
1306
49557cc7
AA		 1307 cs_slot = &unconf->cl_cs_slot;
1308 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1309 if (status) {
1310 /* an unconfirmed replay returns misordered */
ec6b5d7b 1311 status = nfserr_seq_misordered;
49557cc7 1312 goto out_cache;
ec6b5d7b
AA		 1313 }
1314
49557cc7 1315 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1316 move_to_confirmed(unconf);
1317
1318 /*
1319 * We do not support RDMA or persistent sessions
1320 */
1321 cr_ses->flags &= ~SESSION4_PERSIST;
1322 cr_ses->flags &= ~SESSION4_RDMA;
1323
38524ab3
AA		 1324 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1325 unconf->cl_cb_xprt = rqstp->rq_xprt;
1326 svc_xprt_get(unconf->cl_cb_xprt);
1327 rpc_copy_addr(
1328 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1329 sa);
1330 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1331 unconf->cl_cb_conn.cb_minorversion =
1332 cstate->minorversion;
1333 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1334 unconf->cl_cb_seq_nr = 1;
38524ab3
AA		 1335 nfsd4_probe_callback(unconf);
1336 }
ec6b5d7b
AA		 1337 conf = unconf;
1338 } else {
1339 status = nfserr_stale_clientid;
1340 goto out;
1341 }
1342
1343 status = alloc_init_session(rqstp, conf, cr_ses);
1344 if (status)
1345 goto out;
1346
ec6b5d7b
AA		 1347 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1348 NFS4_MAX_SESSIONID_LEN);
49557cc7 1349 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1350
49557cc7
AA		 1351out_cache:
1352 /* cache solo and embedded create sessions under the state lock */
1353 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1354out:
1355 nfs4_unlock_state();
1356 dprintk("%s returns %d\n", __func__, ntohl(status));
1357 return status;
069b6ad4
AA		 1358}
1359
1360__be32
1361nfsd4_destroy_session(struct svc_rqst *r,
1362 struct nfsd4_compound_state *cstate,
1363 struct nfsd4_destroy_session *sessionid)
1364{
e10e0cfc
BH		 1365 struct nfsd4_session *ses;
1366 u32 status = nfserr_badsession;
1367
1368 /* Notes:
1369 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1370 * - Should we return nfserr_back_chan_busy if waiting for
1371 * callbacks on to-be-destroyed session?
1372 * - Do we need to clear any callback info from previous session?
1373 */
1374
1375 dump_sessionid(__func__, &sessionid->sessionid);
1376 spin_lock(&sessionid_lock);
1377 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1378 if (!ses) {
1379 spin_unlock(&sessionid_lock);
1380 goto out;
1381 }
1382
1383 unhash_session(ses);
1384 spin_unlock(&sessionid_lock);
1385
1386 /* wait for callbacks */
1387 shutdown_callback_client(ses->se_client);
1388 nfsd4_put_session(ses);
1389 status = nfs_ok;
1390out:
1391 dprintk("%s returns %d\n", __func__, ntohl(status));
1392 return status;
069b6ad4
AA		 1393}
1394
1395__be32
b85d4c01 1396nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1397 struct nfsd4_compound_state *cstate,
1398 struct nfsd4_sequence *seq)
1399{
f9bb94c4 1400 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1401 struct nfsd4_session *session;
1402 struct nfsd4_slot *slot;
1403 int status;
1404
f9bb94c4
AA		 1405 if (resp->opcnt != 1)
1406 return nfserr_sequence_pos;
1407
b85d4c01
BH		 1408 spin_lock(&sessionid_lock);
1409 status = nfserr_badsession;
1410 session = find_in_sessionid_hashtbl(&seq->sessionid);
1411 if (!session)
1412 goto out;
1413
1414 status = nfserr_badslot;
6c18ba9f 1415 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1416 goto out;
1417
557ce264 1418 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1419 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1420
a8dfdaeb
AA		 1421 /* We do not negotiate the number of slots yet, so set the
1422 * maxslots to the session maxreqs which is used to encode
1423 * sr_highest_slotid and the sr_target_slot id to maxslots */
1424 seq->maxslots = session->se_fchannel.maxreqs;
1425
88e588d5 1426 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1427 if (status == nfserr_replay_cache) {
1428 cstate->slot = slot;
1429 cstate->session = session;
da3846a2 1430 /* Return the cached reply status and set cstate->status
557ce264 1431 * for nfsd4_proc_compound processing */
bf864a31 1432 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1433 cstate->status = nfserr_replay_cache;
aaf84eb9 1434 goto out;
b85d4c01
BH		 1435 }
1436 if (status)
1437 goto out;
1438
1439 /* Success! bump slot seqid */
1440 slot->sl_inuse = true;
1441 slot->sl_seqid = seq->seqid;
557ce264 1442 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1443
1444 cstate->slot = slot;
1445 cstate->session = session;
1446
aaf84eb9 1447 /* Hold a session reference until done processing the compound:
b85d4c01
BH		 1448 * nfsd4_put_session called only if the cstate slot is set.
1449 */
b85d4c01
BH		 1450 nfsd4_get_session(session);
1451out:
1452 spin_unlock(&sessionid_lock);
aaf84eb9
BH		 1453 /* Renew the clientid on success and on replay */
1454 if (cstate->session) {
1455 nfs4_lock_state();
1456 renew_client(session->se_client);
1457 nfs4_unlock_state();
1458 }
b85d4c01
BH		 1459 dprintk("%s: return %d\n", __func__, ntohl(status));
1460 return status;
069b6ad4
AA		 1461}
1462
b37ad28b 1463__be32
b591480b
BF		 1464nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1465 struct nfsd4_setclientid *setclid)
1da177e4 1466{
363168b4 1467 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1468 struct xdr_netobj clname = {
1469 .len = setclid->se_namelen,
1470 .data = setclid->se_name,
1471 };
1472 nfs4_verifier clverifier = setclid->se_verf;
1473 unsigned int strhashval;
28ce6054 1474 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1475 __be32 status;
68e76ad0 1476 char *princ;
a55370a3 1477 char dname[HEXDIR_LEN];
1da177e4 1478
1da177e4 1479 if (!check_name(clname))
73aea4ec 1480 return nfserr_inval;
1da177e4 1481
a55370a3
N		 1482 status = nfs4_make_rec_clidname(dname, &clname);
1483 if (status)
73aea4ec 1484 return status;
a55370a3 1485
1da177e4
LT		 1486 /*
1487 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1488 * We get here on a DRC miss.
1489 */
1490
a55370a3 1491 strhashval = clientstr_hashval(dname);
1da177e4 1492
1da177e4 1493 nfs4_lock_state();
a1bcecd2 1494 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1495 if (conf) {
a186e767 1496 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1497 status = nfserr_clid_inuse;
026722c2 1498 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1499 char addr_str[INET6_ADDRSTRLEN];
1500 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1501 sizeof(addr_str));
1502 dprintk("NFSD: setclientid: string in use by client "
1503 "at %s\n", addr_str);
1da177e4
LT		 1504 goto out;
1505 }
1da177e4 1506 }
a186e767
BF		 1507 /*
1508 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1509 * has a description of SETCLIENTID request processing consisting
1510 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1511 */
a1bcecd2 1512 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1513 status = nfserr_resource;
1514 if (!conf) {
a186e767
BF		 1515 /*
1516 * RFC 3530 14.2.33 CASE 4:
1517 * placed first, because it is the normal case
1da177e4
LT		 1518 */
1519 if (unconf)
1520 expire_client(unconf);
a55370a3
N		 1521 new = create_client(clname, dname);
1522 if (new == NULL)
1da177e4 1523 goto out;
1da177e4 1524 gen_clid(new);
599e0a22 1525 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1526 /*
a186e767
BF		 1527 * RFC 3530 14.2.33 CASE 1:
1528 * probable callback update
1da177e4 1529 */
31f4a6c1
N		 1530 if (unconf) {
1531 /* Note this is removing unconfirmed {*x***},
1532 * which is stronger than RFC recommended {vxc**}.
1533 * This has the advantage that there is at most
1534 * one {*x***} in either list at any time.
1535 */
1536 expire_client(unconf);
1da177e4 1537 }
a55370a3
N		 1538 new = create_client(clname, dname);
1539 if (new == NULL)
1da177e4 1540 goto out;
1da177e4 1541 copy_clid(new, conf);
1da177e4
LT		 1542 } else if (!unconf) {
1543 /*
a186e767
BF		 1544 * RFC 3530 14.2.33 CASE 2:
1545 * probable client reboot; state will be removed if
1546 * confirmed.
1da177e4 1547 */
a55370a3
N		 1548 new = create_client(clname, dname);
1549 if (new == NULL)
1da177e4 1550 goto out;
1da177e4 1551 gen_clid(new);
49ba8781 1552 } else {
a186e767
BF		 1553 /*
1554 * RFC 3530 14.2.33 CASE 3:
1555 * probable client reboot; state will be removed if
1556 * confirmed.
1da177e4
LT		 1557 */
1558 expire_client(unconf);
a55370a3
N		 1559 new = create_client(clname, dname);
1560 if (new == NULL)
1da177e4 1561 goto out;
1da177e4 1562 gen_clid(new);
1da177e4 1563 }
c175b83c 1564 copy_verf(new, &clverifier);
363168b4 1565 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1566 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1567 princ = svc_gss_principal(rqstp);
1568 if (princ) {
1569 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1570 if (new->cl_principal == NULL) {
1571 free_client(new);
1572 goto out;
1573 }
1574 }
c175b83c
BF		 1575 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1576 gen_confirm(new);
fbf4665f 1577 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1578 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1579 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1580 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1581 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1582 status = nfs_ok;
1583out:
1584 nfs4_unlock_state();
1585 return status;
1586}
1587
1588
1589/*
a186e767
BF		 1590 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1591 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1592 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1593 */
b37ad28b 1594__be32
b591480b
BF		 1595nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1596 struct nfsd4_compound_state *cstate,
1597 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1598{
363168b4 1599 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1600 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1601 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1602 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1603 __be32 status;
1da177e4
LT		 1604
1605 if (STALE_CLIENTID(clid))
1606 return nfserr_stale_clientid;
1607 /*
1608 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1609 * We get here on a DRC miss.
1610 */
1611
1612 nfs4_lock_state();
21ab45a4
N		 1613
1614 conf = find_confirmed_client(clid);
1615 unconf = find_unconfirmed_client(clid);
1616
1617 status = nfserr_clid_inuse;
363168b4 1618 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1619 goto out;
363168b4 1620 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1621 goto out;
1622
a186e767
BF		 1623 /*
1624 * section 14.2.34 of RFC 3530 has a description of
1625 * SETCLIENTID_CONFIRM request processing consisting
1626 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1627 */
366e0c1d 1628 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1629 /*
1630 * RFC 3530 14.2.34 CASE 1:
1631 * callback update
1632 */
599e0a22 1633 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1634 status = nfserr_clid_inuse;
1635 else {
1a69c179
N		 1636 /* XXX: We just turn off callbacks until we can handle
1637 * change request correctly. */
c237dc03 1638 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1639 expire_client(unconf);
1da177e4 1640 status = nfs_ok;
1a69c179 1641
1da177e4 1642 }
f3aba4e5 1643 } else if (conf && !unconf) {
a186e767
BF		 1644 /*
1645 * RFC 3530 14.2.34 CASE 2:
1646 * probable retransmitted request; play it safe and
1647 * do nothing.
7c79f737 1648 */
599e0a22 1649 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1650 status = nfserr_clid_inuse;
21ab45a4 1651 else
1da177e4 1652 status = nfs_ok;
7c79f737 1653 } else if (!conf && unconf
599e0a22 1654 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1655 /*
1656 * RFC 3530 14.2.34 CASE 3:
1657 * Normal case; new or rebooted client:
7c79f737 1658 */
599e0a22 1659 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1660 status = nfserr_clid_inuse;
1661 } else {
1a69c179
N		 1662 unsigned int hash =
1663 clientstr_hashval(unconf->cl_recdir);
1664 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1665 hash, false);
1a69c179 1666 if (conf) {
c7b9a459 1667 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1668 expire_client(conf);
1669 }
1da177e4 1670 move_to_confirmed(unconf);
21ab45a4 1671 conf = unconf;
46f8a64b 1672 nfsd4_probe_callback(conf);
1a69c179 1673 status = nfs_ok;
1da177e4 1674 }
599e0a22
BF		 1675 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1676 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1677 &confirm)))) {
a186e767
BF		 1678 /*
1679 * RFC 3530 14.2.34 CASE 4:
1680 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1681 */
1da177e4 1682 status = nfserr_stale_clientid;
7c79f737 1683 } else {
08e8987c
N		 1684 /* check that we have hit one of the cases...*/
1685 status = nfserr_clid_inuse;
1686 }
1da177e4 1687out:
1da177e4
LT		 1688 nfs4_unlock_state();
1689 return status;
1690}
1691
1da177e4
LT		 1692/* OPEN Share state helper functions */
1693static inline struct nfs4_file *
1694alloc_init_file(struct inode *ino)
1695{
1696 struct nfs4_file *fp;
1697 unsigned int hashval = file_hashval(ino);
1698
e60d4398
N		 1699 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1700 if (fp) {
8b671b80 1701 atomic_set(&fp->fi_ref, 1);
1da177e4 1702 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1703 INIT_LIST_HEAD(&fp->fi_stateids);
1704 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1705 spin_lock(&recall_lock);
1da177e4 1706 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1707 spin_unlock(&recall_lock);
1da177e4
LT		 1708 fp->fi_inode = igrab(ino);
1709 fp->fi_id = current_fileid++;
47f9940c 1710 fp->fi_had_conflict = false;
1da177e4
LT		 1711 return fp;
1712 }
1713 return NULL;
1714}
1715
e60d4398 1716static void
e18b890b 1717nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1718{
e60d4398
N		 1719 if (*slab == NULL)
1720 return;
1a1d92c1 1721 kmem_cache_destroy(*slab);
e60d4398 1722 *slab = NULL;
1da177e4
LT		 1723}
1724
e8ff2a84 1725void
1da177e4
LT		 1726nfsd4_free_slabs(void)
1727{
e60d4398
N		 1728 nfsd4_free_slab(&stateowner_slab);
1729 nfsd4_free_slab(&file_slab);
5ac049ac 1730 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1731 nfsd4_free_slab(&deleg_slab);
e60d4398 1732}
1da177e4 1733
e60d4398
N		 1734static int
1735nfsd4_init_slabs(void)
1736{
1737 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1738 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1739 if (stateowner_slab == NULL)
1740 goto out_nomem;
1741 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1742 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1743 if (file_slab == NULL)
1744 goto out_nomem;
5ac049ac 1745 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1746 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1747 if (stateid_slab == NULL)
1748 goto out_nomem;
5b2d21c1 1749 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1750 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1751 if (deleg_slab == NULL)
1752 goto out_nomem;
e60d4398
N		 1753 return 0;
1754out_nomem:
1755 nfsd4_free_slabs();
1756 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1757 return -ENOMEM;
1da177e4
LT		 1758}
1759
1760void
1761nfs4_free_stateowner(struct kref *kref)
1762{
1763 struct nfs4_stateowner *sop =
1764 container_of(kref, struct nfs4_stateowner, so_ref);
1765 kfree(sop->so_owner.data);
1766 kmem_cache_free(stateowner_slab, sop);
1767}
1768
1769static inline struct nfs4_stateowner *
1770alloc_stateowner(struct xdr_netobj *owner)
1771{
1772 struct nfs4_stateowner *sop;
1773
1774 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1775 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1776 memcpy(sop->so_owner.data, owner->data, owner->len);
1777 sop->so_owner.len = owner->len;
1778 kref_init(&sop->so_ref);
1779 return sop;
1780 }
1781 kmem_cache_free(stateowner_slab, sop);
1782 }
1783 return NULL;
1784}
1785
1786static struct nfs4_stateowner *
1787alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1788 struct nfs4_stateowner *sop;
1789 struct nfs4_replay *rp;
1790 unsigned int idhashval;
1791
1792 if (!(sop = alloc_stateowner(&open->op_owner)))
1793 return NULL;
1794 idhashval = ownerid_hashval(current_ownerid);
1795 INIT_LIST_HEAD(&sop->so_idhash);
1796 INIT_LIST_HEAD(&sop->so_strhash);
1797 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1798 INIT_LIST_HEAD(&sop->so_stateids);
1799 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1800 INIT_LIST_HEAD(&sop->so_close_lru);
1801 sop->so_time = 0;
1802 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1803 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1804 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1805 sop->so_is_open_owner = 1;
1806 sop->so_id = current_ownerid++;
1807 sop->so_client = clp;
1808 sop->so_seqid = open->op_seqid;
1809 sop->so_confirmed = 0;
1810 rp = &sop->so_replay;
de1ae286 1811 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1812 rp->rp_buflen = 0;
1813 rp->rp_buf = rp->rp_ibuf;
1814 return sop;
1815}
1816
1da177e4
LT		 1817static inline void
1818init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1819 struct nfs4_stateowner *sop = open->op_stateowner;
1820 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1821
1822 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1823 INIT_LIST_HEAD(&stp->st_perstateowner);
1824 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1825 INIT_LIST_HEAD(&stp->st_perfile);
1826 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1827 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1828 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1829 stp->st_stateowner = sop;
13cd2184 1830 get_nfs4_file(fp);
1da177e4 1831 stp->st_file = fp;
78155ed7 1832 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1833 stp->st_stateid.si_stateownerid = sop->so_id;
1834 stp->st_stateid.si_fileid = fp->fi_id;
1835 stp->st_stateid.si_generation = 0;
1836 stp->st_access_bmap = 0;
1837 stp->st_deny_bmap = 0;
84459a11
AA		 1838 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1839 &stp->st_access_bmap);
1da177e4 1840 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1841 stp->st_openstp = NULL;
1da177e4
LT		 1842}
1843
fd39ca9a 1844static void
1da177e4
LT		 1845move_to_close_lru(struct nfs4_stateowner *sop)
1846{
1847 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1848
358dd55a 1849 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1850 sop->so_time = get_seconds();
1851}
1852
1da177e4 1853static int
599e0a22
BF		 1854same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1855 clientid_t *clid)
1856{
1857 return (sop->so_owner.len == owner->len) &&
1858 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1859 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1860}
1861
1862static struct nfs4_stateowner *
1863find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1864{
1865 struct nfs4_stateowner *so = NULL;
1866
1867 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1868 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1869 return so;
1870 }
1871 return NULL;
1872}
1873
1874/* search file_hashtbl[] for file */
1875static struct nfs4_file *
1876find_file(struct inode *ino)
1877{
1878 unsigned int hashval = file_hashval(ino);
1879 struct nfs4_file *fp;
1880
8b671b80 1881 spin_lock(&recall_lock);
1da177e4 1882 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1883 if (fp->fi_inode == ino) {
1884 get_nfs4_file(fp);
8b671b80 1885 spin_unlock(&recall_lock);
1da177e4 1886 return fp;
13cd2184 1887 }
1da177e4 1888 }
8b671b80 1889 spin_unlock(&recall_lock);
1da177e4
LT		 1890 return NULL;
1891}
1892
d87a8ade 1893static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1894{
d87a8ade 1895 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1896 return 0;
d87a8ade
AA		 1897 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1898 return 0;
1899 x &= ~NFS4_SHARE_ACCESS_MASK;
1900 if (minorversion && x) {
1901 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1902 return 0;
1903 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1904 return 0;
1905 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1906 }
1907 if (x)
8838dc43
BF		 1908 return 0;
1909 return 1;
ba5a6a19
BF		 1910}
1911
8838dc43 1912static inline int deny_valid(u32 x)
ba5a6a19 1913{
8838dc43
BF		 1914 /* Note: unlike access bits, deny bits may be zero. */
1915 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1916}
1da177e4 1917
4f83aa30
BF		 1918/*
1919 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1920 * st_{access,deny}_bmap field of the stateid, in order to track not
1921 * only what share bits are currently in force, but also what
1922 * combinations of share bits previous opens have used. This allows us
1923 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1924 * return an error if the client attempt to downgrade to a combination
1925 * of share bits not explicable by closing some of its previous opens.
1926 *
1927 * XXX: This enforcement is actually incomplete, since we don't keep
1928 * track of access/deny bit combinations; so, e.g., we allow:
1929 *
1930 * OPEN allow read, deny write
1931 * OPEN allow both, deny none
1932 * DOWNGRADE allow read, deny none
1933 *
1934 * which we should reject.
1935 */
fd39ca9a 1936static void
1da177e4
LT		 1937set_access(unsigned int *access, unsigned long bmap) {
1938 int i;
1939
1940 *access = 0;
1941 for (i = 1; i < 4; i++) {
1942 if (test_bit(i, &bmap))
1943 *access |= i;
1944 }
1945}
1946
fd39ca9a 1947static void
1da177e4
LT		 1948set_deny(unsigned int *deny, unsigned long bmap) {
1949 int i;
1950
1951 *deny = 0;
1952 for (i = 0; i < 4; i++) {
1953 if (test_bit(i, &bmap))
1954 *deny |= i ;
1955 }
1956}
1957
1958static int
1959test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1960 unsigned int access, deny;
1961
1962 set_access(&access, stp->st_access_bmap);
1963 set_deny(&deny, stp->st_deny_bmap);
1964 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1965 return 0;
1966 return 1;
1967}
1968
1969/*
1970 * Called to check deny when READ with all zero stateid or
1971 * WRITE with all zero or all one stateid
1972 */
b37ad28b 1973static __be32
1da177e4
LT		 1974nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1975{
1976 struct inode *ino = current_fh->fh_dentry->d_inode;
1977 struct nfs4_file *fp;
1978 struct nfs4_stateid *stp;
b37ad28b 1979 __be32 ret;
1da177e4
LT		 1980
1981 dprintk("NFSD: nfs4_share_conflict\n");
1982
1983 fp = find_file(ino);
13cd2184
N		 1984 if (!fp)
1985 return nfs_ok;
b700949b 1986 ret = nfserr_locked;
1da177e4 1987 /* Search for conflicting share reservations */
13cd2184
N		 1988 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1989 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1990 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1991 goto out;
1da177e4 1992 }
13cd2184
N		 1993 ret = nfs_ok;
1994out:
1995 put_nfs4_file(fp);
1996 return ret;
1da177e4
LT		 1997}
1998
1999static inline void
2000nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2001{
2002 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2003 drop_file_write_access(filp);
1da177e4
LT		 2004 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2005 }
2006}
2007
1da177e4
LT		 2008/*
2009 * Spawn a thread to perform a recall on the delegation represented
2010 * by the lease (file_lock)
2011 *
2012 * Called from break_lease() with lock_kernel() held.
2013 * Note: we assume break_lease will only call this *once* for any given
2014 * lease.
2015 */
2016static
2017void nfsd_break_deleg_cb(struct file_lock *fl)
2018{
63e4863f 2019 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2020
2021 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2022 if (!dp)
2023 return;
2024
2025 /* We're assuming the state code never drops its reference
2026 * without first removing the lease. Since we're in this lease
2027 * callback (and since the lease code is serialized by the kernel
2028 * lock) we know the server hasn't removed the lease yet, we know
2029 * it's safe to take a reference: */
2030 atomic_inc(&dp->dl_count);
cfdcad4d 2031 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2032
2033 spin_lock(&recall_lock);
2034 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2035 spin_unlock(&recall_lock);
2036
2037 /* only place dl_time is set. protected by lock_kernel*/
2038 dp->dl_time = get_seconds();
2039
0272e1fd
BF		 2040 /*
2041 * We don't want the locks code to timeout the lease for us;
2042 * we'll remove it ourself if the delegation isn't returned
2043 * in time.
2044 */
2045 fl->fl_break_time = 0;
1da177e4 2046
63e4863f
BF		 2047 dp->dl_file->fi_had_conflict = true;
2048 nfsd4_cb_recall(dp);
1da177e4
LT		 2049}
2050
2051/*
2052 * The file_lock is being reapd.
2053 *
2054 * Called by locks_free_lock() with lock_kernel() held.
2055 */
2056static
2057void nfsd_release_deleg_cb(struct file_lock *fl)
2058{
2059 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2060
2061 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2062
2063 if (!(fl->fl_flags & FL_LEASE) || !dp)
2064 return;
2065 dp->dl_flock = NULL;
2066}
2067
2068/*
2069 * Set the delegation file_lock back pointer.
2070 *
a9933cea 2071 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2072 */
2073static
2074void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2075{
2076 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2077
2078 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2079 if (!dp)
2080 return;
2081 dp->dl_flock = new;
2082}
2083
2084/*
a9933cea 2085 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2086 */
2087static
2088int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2089{
2090 struct nfs4_delegation *onlistd =
2091 (struct nfs4_delegation *)onlist->fl_owner;
2092 struct nfs4_delegation *tryd =
2093 (struct nfs4_delegation *)try->fl_owner;
2094
2095 if (onlist->fl_lmops != try->fl_lmops)
2096 return 0;
2097
2098 return onlistd->dl_client == tryd->dl_client;
2099}
2100
2101
2102static
2103int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2104{
2105 if (arg & F_UNLCK)
2106 return lease_modify(onlist, arg);
2107 else
2108 return -EAGAIN;
2109}
2110
fd39ca9a 2111static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2112 .fl_break = nfsd_break_deleg_cb,
2113 .fl_release_private = nfsd_release_deleg_cb,
2114 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2115 .fl_mylease = nfsd_same_client_deleg_cb,
2116 .fl_change = nfsd_change_deleg_cb,
2117};
2118
2119
b37ad28b 2120__be32
6668958f
AA		 2121nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2122 struct nfsd4_open *open)
1da177e4 2123{
1da177e4
LT		 2124 clientid_t *clientid = &open->op_clientid;
2125 struct nfs4_client *clp = NULL;
2126 unsigned int strhashval;
2127 struct nfs4_stateowner *sop = NULL;
2128
1da177e4 2129 if (!check_name(open->op_owner))
0f442aa2 2130 return nfserr_inval;
1da177e4
LT		 2131
2132 if (STALE_CLIENTID(&open->op_clientid))
2133 return nfserr_stale_clientid;
2134
2135 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2136 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2137 open->op_stateowner = sop;
2138 if (!sop) {
2139 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2140 clp = find_confirmed_client(clientid);
2141 if (clp == NULL)
0f442aa2
BF		 2142 return nfserr_expired;
2143 goto renew;
1da177e4 2144 }
6668958f
AA		 2145 /* When sessions are used, skip open sequenceid processing */
2146 if (nfsd4_has_session(cstate))
2147 goto renew;
0f442aa2
BF		 2148 if (!sop->so_confirmed) {
2149 /* Replace unconfirmed owners without checking for replay. */
2150 clp = sop->so_client;
f044ff83 2151 release_openowner(sop);
0f442aa2
BF		 2152 open->op_stateowner = NULL;
2153 goto renew;
2154 }
2155 if (open->op_seqid == sop->so_seqid - 1) {
2156 if (sop->so_replay.rp_buflen)
a90b061c 2157 return nfserr_replay_me;
0f442aa2
BF		 2158 /* The original OPEN failed so spectacularly
2159 * that we don't even have replay data saved!
2160 * Therefore, we have no choice but to continue
2161 * processing this OPEN; presumably, we'll
2162 * fail again for the same reason.
2163 */
2164 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2165 goto renew;
2166 }
2167 if (open->op_seqid != sop->so_seqid)
2168 return nfserr_bad_seqid;
1da177e4 2169renew:
0f442aa2
BF		 2170 if (open->op_stateowner == NULL) {
2171 sop = alloc_init_open_stateowner(strhashval, clp, open);
2172 if (sop == NULL)
2173 return nfserr_resource;
2174 open->op_stateowner = sop;
2175 }
2176 list_del_init(&sop->so_close_lru);
1da177e4 2177 renew_client(sop->so_client);
0f442aa2 2178 return nfs_ok;
1da177e4
LT		 2179}
2180
b37ad28b 2181static inline __be32
4a6e43e6
N		 2182nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2183{
2184 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2185 return nfserr_openmode;
2186 else
2187 return nfs_ok;
2188}
2189
52f4fb43
N		 2190static struct nfs4_delegation *
2191find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2192{
2193 struct nfs4_delegation *dp;
2194
ea1da636 2195 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2196 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2197 return dp;
2198 }
2199 return NULL;
2200}
2201
b37ad28b 2202static __be32
567d9829
N		 2203nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2204 struct nfs4_delegation **dp)
2205{
2206 int flags;
b37ad28b 2207 __be32 status = nfserr_bad_stateid;
567d9829
N		 2208
2209 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2210 if (*dp == NULL)
c44c5eeb 2211 goto out;
567d9829
N		 2212 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2213 RD_STATE : WR_STATE;
2214 status = nfs4_check_delegmode(*dp, flags);
2215 if (status)
2216 *dp = NULL;
c44c5eeb
N		 2217out:
2218 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2219 return nfs_ok;
2220 if (status)
2221 return status;
2222 open->op_stateowner->so_confirmed = 1;
2223 return nfs_ok;
567d9829
N		 2224}
2225
b37ad28b 2226static __be32
1da177e4
LT		 2227nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2228{
2229 struct nfs4_stateid *local;
b37ad28b 2230 __be32 status = nfserr_share_denied;
1da177e4
LT		 2231 struct nfs4_stateowner *sop = open->op_stateowner;
2232
8beefa24 2233 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2234 /* ignore lock owners */
2235 if (local->st_stateowner->so_is_open_owner == 0)
2236 continue;
2237 /* remember if we have seen this open owner */
2238 if (local->st_stateowner == sop)
2239 *stpp = local;
2240 /* check for conflicting share reservations */
2241 if (!test_share(local, open))
2242 goto out;
2243 }
2244 status = 0;
2245out:
2246 return status;
2247}
2248
5ac049ac
N		 2249static inline struct nfs4_stateid *
2250nfs4_alloc_stateid(void)
2251{
2252 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2253}
2254
b37ad28b 2255static __be32
1da177e4 2256nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2257 struct nfs4_delegation *dp,
1da177e4
LT		 2258 struct svc_fh *cur_fh, int flags)
2259{
2260 struct nfs4_stateid *stp;
1da177e4 2261
5ac049ac 2262 stp = nfs4_alloc_stateid();
1da177e4
LT		 2263 if (stp == NULL)
2264 return nfserr_resource;
2265
567d9829
N		 2266 if (dp) {
2267 get_file(dp->dl_vfs_file);
2268 stp->st_vfs_file = dp->dl_vfs_file;
2269 } else {
b37ad28b 2270 __be32 status;
567d9829
N		 2271 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2272 &stp->st_vfs_file);
2273 if (status) {
2274 if (status == nfserr_dropit)
2275 status = nfserr_jukebox;
5ac049ac 2276 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2277 return status;
2278 }
1da177e4 2279 }
1da177e4
LT		 2280 *stpp = stp;
2281 return 0;
2282}
2283
b37ad28b 2284static inline __be32
1da177e4
LT		 2285nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2286 struct nfsd4_open *open)
2287{
2288 struct iattr iattr = {
2289 .ia_valid = ATTR_SIZE,
2290 .ia_size = 0,
2291 };
2292 if (!open->op_truncate)
2293 return 0;
2294 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2295 return nfserr_inval;
1da177e4
LT		 2296 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2297}
2298
b37ad28b 2299static __be32
1da177e4
LT		 2300nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2301{
2302 struct file *filp = stp->st_vfs_file;
7eaa36e2 2303 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2304 unsigned int share_access, new_writer;
b37ad28b 2305 __be32 status;
1da177e4
LT		 2306
2307 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2308 new_writer = (~share_access) & open->op_share_access
2309 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2310
6c26d08f 2311 if (new_writer) {
b37ad28b
AV		 2312 int err = get_write_access(inode);
2313 if (err)
2314 return nfserrno(err);
e518f056
BH		 2315 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2316 if (err)
2317 return nfserrno(err);
2318 file_take_write(filp);
6c26d08f 2319 }
1da177e4
LT		 2320 status = nfsd4_truncate(rqstp, cur_fh, open);
2321 if (status) {
6c26d08f
BF		 2322 if (new_writer)
2323 put_write_access(inode);
1da177e4
LT		 2324 return status;
2325 }
2326 /* remember the open */
6c26d08f 2327 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2328 __set_bit(open->op_share_access, &stp->st_access_bmap);
2329 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2330
2331 return nfs_ok;
2332}
2333
2334
1da177e4 2335static void
37515177 2336nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2337{
37515177
N		 2338 open->op_stateowner->so_confirmed = 1;
2339 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2340}
2341
2342/*
2343 * Attempt to hand out a delegation.
2344 */
2345static void
2346nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2347{
2348 struct nfs4_delegation *dp;
2349 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2350 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2351 struct file_lock fl, *flp = &fl;
2352 int status, flag = 0;
2353
2354 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2355 open->op_recall = 0;
2356 switch (open->op_claim_type) {
2357 case NFS4_OPEN_CLAIM_PREVIOUS:
2358 if (!atomic_read(&cb->cb_set))
2359 open->op_recall = 1;
2360 flag = open->op_delegate_type;
2361 if (flag == NFS4_OPEN_DELEGATE_NONE)
2362 goto out;
2363 break;
2364 case NFS4_OPEN_CLAIM_NULL:
2365 /* Let's not give out any delegations till everyone's
2366 * had the chance to reclaim theirs.... */
af558e33 2367 if (locks_in_grace())
7b190fec
N		 2368 goto out;
2369 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2370 goto out;
2371 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2372 flag = NFS4_OPEN_DELEGATE_WRITE;
2373 else
2374 flag = NFS4_OPEN_DELEGATE_READ;
2375 break;
2376 default:
2377 goto out;
2378 }
1da177e4
LT		 2379
2380 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2381 if (dp == NULL) {
2382 flag = NFS4_OPEN_DELEGATE_NONE;
2383 goto out;
2384 }
2385 locks_init_lock(&fl);
2386 fl.fl_lmops = &nfsd_lease_mng_ops;
2387 fl.fl_flags = FL_LEASE;
9167f501 2388 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2389 fl.fl_end = OFFSET_MAX;
2390 fl.fl_owner = (fl_owner_t)dp;
2391 fl.fl_file = stp->st_vfs_file;
2392 fl.fl_pid = current->tgid;
2393
a9933cea 2394 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2395 * the lock_manager callbacks fl_mylease and fl_change are used
2396 */
9167f501 2397 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2398 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2399 unhash_delegation(dp);
1da177e4
LT		 2400 flag = NFS4_OPEN_DELEGATE_NONE;
2401 goto out;
2402 }
2403
2404 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2405
2406 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2407 dp->dl_stateid.si_boot,
2408 dp->dl_stateid.si_stateownerid,
2409 dp->dl_stateid.si_fileid,
2410 dp->dl_stateid.si_generation);
2411out:
7b190fec
N		 2412 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2413 && flag == NFS4_OPEN_DELEGATE_NONE
2414 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2415 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2416 open->op_delegate_type = flag;
2417}
2418
2419/*
2420 * called with nfs4_lock_state() held.
2421 */
b37ad28b 2422__be32
1da177e4
LT		 2423nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2424{
6668958f 2425 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2426 struct nfs4_file *fp = NULL;
2427 struct inode *ino = current_fh->fh_dentry->d_inode;
2428 struct nfs4_stateid *stp = NULL;
567d9829 2429 struct nfs4_delegation *dp = NULL;
b37ad28b 2430 __be32 status;
1da177e4
LT		 2431
2432 status = nfserr_inval;
d87a8ade 2433 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2434 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2435 goto out;
2436 /*
2437 * Lookup file; if found, lookup stateid and check open request,
2438 * and check for delegations in the process of being recalled.
2439 * If not found, create the nfs4_file struct
2440 */
2441 fp = find_file(ino);
2442 if (fp) {
2443 if ((status = nfs4_check_open(fp, open, &stp)))
2444 goto out;
c44c5eeb
N		 2445 status = nfs4_check_deleg(fp, open, &dp);
2446 if (status)
2447 goto out;
1da177e4 2448 } else {
c44c5eeb
N		 2449 status = nfserr_bad_stateid;
2450 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2451 goto out;
1da177e4
LT		 2452 status = nfserr_resource;
2453 fp = alloc_init_file(ino);
2454 if (fp == NULL)
2455 goto out;
2456 }
2457
2458 /*
2459 * OPEN the file, or upgrade an existing OPEN.
2460 * If truncate fails, the OPEN fails.
2461 */
2462 if (stp) {
2463 /* Stateid was found, this is an OPEN upgrade */
2464 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2465 if (status)
2466 goto out;
444c2c07 2467 update_stateid(&stp->st_stateid);
1da177e4
LT		 2468 } else {
2469 /* Stateid was not found, this is a new OPEN */
2470 int flags = 0;
9ecb6a08 2471 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2472 flags |= NFSD_MAY_READ;
1da177e4 2473 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2474 flags |= NFSD_MAY_WRITE;
567d9829
N		 2475 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2476 if (status)
1da177e4
LT		 2477 goto out;
2478 init_stateid(stp, fp, open);
2479 status = nfsd4_truncate(rqstp, current_fh, open);
2480 if (status) {
2283963f 2481 release_open_stateid(stp);
1da177e4
LT		 2482 goto out;
2483 }
6668958f
AA		 2484 if (nfsd4_has_session(&resp->cstate))
2485 update_stateid(&stp->st_stateid);
1da177e4
LT		 2486 }
2487 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2488
6668958f
AA		 2489 if (nfsd4_has_session(&resp->cstate))
2490 open->op_stateowner->so_confirmed = 1;
2491
1da177e4
LT		 2492 /*
2493 * Attempt to hand out a delegation. No error return, because the
2494 * OPEN succeeds even if we fail.
2495 */
2496 nfs4_open_delegation(current_fh, open, stp);
2497
2498 status = nfs_ok;
2499
2500 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2501 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2502 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2503out:
13cd2184
N		 2504 if (fp)
2505 put_nfs4_file(fp);
37515177
N		 2506 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2507 nfs4_set_claim_prev(open);
1da177e4
LT		 2508 /*
2509 * To finish the open response, we just need to set the rflags.
2510 */
2511 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2512 if (!open->op_stateowner->so_confirmed &&
2513 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2514 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2515
2516 return status;
2517}
2518
b37ad28b 2519__be32
b591480b
BF		 2520nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2521 clientid_t *clid)
1da177e4
LT		 2522{
2523 struct nfs4_client *clp;
b37ad28b 2524 __be32 status;
1da177e4
LT		 2525
2526 nfs4_lock_state();
2527 dprintk("process_renew(%08x/%08x): starting\n",
2528 clid->cl_boot, clid->cl_id);
2529 status = nfserr_stale_clientid;
2530 if (STALE_CLIENTID(clid))
2531 goto out;
2532 clp = find_confirmed_client(clid);
2533 status = nfserr_expired;
2534 if (clp == NULL) {
2535 /* We assume the client took too long to RENEW. */
2536 dprintk("nfsd4_renew: clientid not found!\n");
2537 goto out;
2538 }
2539 renew_client(clp);
2540 status = nfserr_cb_path_down;
ea1da636 2541 if (!list_empty(&clp->cl_delegations)
c237dc03 2542 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2543 goto out;
2544 status = nfs_ok;
2545out:
2546 nfs4_unlock_state();
2547 return status;
2548}
2549
af558e33
BF		 2550struct lock_manager nfsd4_manager = {
2551};
2552
a76b4319 2553static void
af558e33 2554nfsd4_end_grace(void)
a76b4319
N		 2555{
2556 dprintk("NFSD: end of grace period\n");
c7b9a459 2557 nfsd4_recdir_purge_old();
af558e33 2558 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2559}
2560
fd39ca9a 2561static time_t
1da177e4
LT		 2562nfs4_laundromat(void)
2563{
2564 struct nfs4_client *clp;
2565 struct nfs4_stateowner *sop;
2566 struct nfs4_delegation *dp;
2567 struct list_head *pos, *next, reaplist;
2568 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2569 time_t t, clientid_val = NFSD_LEASE_TIME;
2570 time_t u, test_val = NFSD_LEASE_TIME;
2571
2572 nfs4_lock_state();
2573
2574 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2575 if (locks_in_grace())
2576 nfsd4_end_grace();
1da177e4
LT		 2577 list_for_each_safe(pos, next, &client_lru) {
2578 clp = list_entry(pos, struct nfs4_client, cl_lru);
2579 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2580 t = clp->cl_time - cutoff;
2581 if (clientid_val > t)
2582 clientid_val = t;
2583 break;
2584 }
2585 dprintk("NFSD: purging unused client (clientid %08x)\n",
2586 clp->cl_clientid.cl_id);
c7b9a459 2587 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2588 expire_client(clp);
2589 }
2590 INIT_LIST_HEAD(&reaplist);
2591 spin_lock(&recall_lock);
2592 list_for_each_safe(pos, next, &del_recall_lru) {
2593 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2594 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2595 u = dp->dl_time - cutoff;
2596 if (test_val > u)
2597 test_val = u;
2598 break;
2599 }
2600 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2601 dp, dp->dl_flock);
2602 list_move(&dp->dl_recall_lru, &reaplist);
2603 }
2604 spin_unlock(&recall_lock);
2605 list_for_each_safe(pos, next, &reaplist) {
2606 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2607 list_del_init(&dp->dl_recall_lru);
2608 unhash_delegation(dp);
2609 }
2610 test_val = NFSD_LEASE_TIME;
2611 list_for_each_safe(pos, next, &close_lru) {
2612 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2613 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2614 u = sop->so_time - cutoff;
2615 if (test_val > u)
2616 test_val = u;
2617 break;
2618 }
2619 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2620 sop->so_id);
f044ff83 2621 release_openowner(sop);
1da177e4
LT		 2622 }
2623 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2624 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2625 nfs4_unlock_state();
2626 return clientid_val;
2627}
2628
a254b246
HH		 2629static struct workqueue_struct *laundry_wq;
2630static void laundromat_main(struct work_struct *);
2631static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2632
2633static void
c4028958 2634laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2635{
2636 time_t t;
2637
2638 t = nfs4_laundromat();
2639 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2640 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2641}
2642
fd39ca9a 2643static struct nfs4_stateowner *
f8816512
N		 2644search_close_lru(u32 st_id, int flags)
2645{
1da177e4
LT		 2646 struct nfs4_stateowner *local = NULL;
2647
1da177e4
LT		 2648 if (flags & CLOSE_STATE) {
2649 list_for_each_entry(local, &close_lru, so_close_lru) {
2650 if (local->so_id == st_id)
2651 return local;
2652 }
2653 }
2654 return NULL;
2655}
2656
2657static inline int
2658nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2659{
7eaa36e2 2660 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2661}
2662
2663static int
2664STALE_STATEID(stateid_t *stateid)
2665{
78155ed7
BN		 2666 if (time_after((unsigned long)boot_time,
2667 (unsigned long)stateid->si_boot)) {
2668 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2669 stateid->si_boot, stateid->si_stateownerid,
2670 stateid->si_fileid, stateid->si_generation);
2671 return 1;
2672 }
2673 return 0;
2674}
2675
2676static int
2677EXPIRED_STATEID(stateid_t *stateid)
2678{
2679 if (time_before((unsigned long)boot_time,
2680 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2681 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2682 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2683 stateid->si_boot, stateid->si_stateownerid,
2684 stateid->si_fileid, stateid->si_generation);
2685 return 1;
2686 }
2687 return 0;
2688}
2689
2690static __be32
2691stateid_error_map(stateid_t *stateid)
2692{
2693 if (STALE_STATEID(stateid))
2694 return nfserr_stale_stateid;
2695 if (EXPIRED_STATEID(stateid))
2696 return nfserr_expired;
2697
2698 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2699 stateid->si_boot, stateid->si_stateownerid,
2700 stateid->si_fileid, stateid->si_generation);
2701 return nfserr_bad_stateid;
1da177e4
LT		 2702}
2703
2704static inline int
2705access_permit_read(unsigned long access_bmap)
2706{
2707 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2708 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2709 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2710}
2711
2712static inline int
2713access_permit_write(unsigned long access_bmap)
2714{
2715 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2716 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2717}
2718
2719static
b37ad28b 2720__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2721{
b37ad28b 2722 __be32 status = nfserr_openmode;
1da177e4
LT		 2723
2724 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2725 goto out;
2726 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2727 goto out;
2728 status = nfs_ok;
2729out:
2730 return status;
2731}
2732
b37ad28b 2733static inline __be32
1da177e4
LT		 2734check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2735{
203a8c8e 2736 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2737 return nfs_ok;
af558e33 2738 else if (locks_in_grace()) {
1da177e4
LT		 2739 /* Answer in remaining cases depends on existance of
2740 * conflicting state; so we must wait out the grace period. */
2741 return nfserr_grace;
2742 } else if (flags & WR_STATE)
2743 return nfs4_share_conflict(current_fh,
2744 NFS4_SHARE_DENY_WRITE);
2745 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2746 return nfs4_share_conflict(current_fh,
2747 NFS4_SHARE_DENY_READ);
2748}
2749
2750/*
2751 * Allow READ/WRITE during grace period on recovered state only for files
2752 * that are not able to provide mandatory locking.
2753 */
2754static inline int
18f82731 2755grace_disallows_io(struct inode *inode)
1da177e4 2756{
203a8c8e 2757 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2758}
2759
6668958f 2760static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2761{
6668958f
AA		 2762 /*
2763 * When sessions are used the stateid generation number is ignored
2764 * when it is zero.
2765 */
2766 if ((flags & HAS_SESSION) && in->si_generation == 0)
2767 goto out;
2768
0836f587
BF		 2769 /* If the client sends us a stateid from the future, it's buggy: */
2770 if (in->si_generation > ref->si_generation)
2771 return nfserr_bad_stateid;
2772 /*
2773 * The following, however, can happen. For example, if the
2774 * client sends an open and some IO at the same time, the open
2775 * may bump si_generation while the IO is still in flight.
2776 * Thanks to hard links and renames, the client never knows what
2777 * file an open will affect. So it could avoid that situation
2778 * only by serializing all opens and IO from the same open
2779 * owner. To recover from the old_stateid error, the client
2780 * will just have to retry the IO:
2781 */
2782 if (in->si_generation < ref->si_generation)
2783 return nfserr_old_stateid;
6668958f 2784out:
0836f587
BF		 2785 return nfs_ok;
2786}
2787
3e633079
BF		 2788static int is_delegation_stateid(stateid_t *stateid)
2789{
2790 return stateid->si_fileid == 0;
2791}
2792
1da177e4
LT		 2793/*
2794* Checks for stateid operations
2795*/
b37ad28b 2796__be32
dd453dfd
BH		 2797nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2798 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2799{
2800 struct nfs4_stateid *stp = NULL;
2801 struct nfs4_delegation *dp = NULL;
dd453dfd 2802 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2803 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2804 __be32 status;
1da177e4 2805
1da177e4
LT		 2806 if (filpp)
2807 *filpp = NULL;
2808
18f82731 2809 if (grace_disallows_io(ino))
1da177e4
LT		 2810 return nfserr_grace;
2811
6668958f
AA		 2812 if (nfsd4_has_session(cstate))
2813 flags |= HAS_SESSION;
2814
1da177e4
LT		 2815 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2816 return check_special_stateids(current_fh, stateid, flags);
2817
1da177e4
LT		 2818 status = nfserr_stale_stateid;
2819 if (STALE_STATEID(stateid))
2820 goto out;
2821
1da177e4 2822 status = nfserr_bad_stateid;
3e633079 2823 if (is_delegation_stateid(stateid)) {
a4455be0 2824 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2825 if (!dp) {
2826 status = stateid_error_map(stateid);
1da177e4 2827 goto out;
78155ed7 2828 }
6668958f
AA		 2829 status = check_stateid_generation(stateid, &dp->dl_stateid,
2830 flags);
0c2a498f
BF		 2831 if (status)
2832 goto out;
dc9bf700
BF		 2833 status = nfs4_check_delegmode(dp, flags);
2834 if (status)
2835 goto out;
2836 renew_client(dp->dl_client);
dc9bf700
BF		 2837 if (filpp)
2838 *filpp = dp->dl_vfs_file;
1da177e4 2839 } else { /* open or lock stateid */
a4455be0 2840 stp = find_stateid(stateid, flags);
78155ed7
BN		 2841 if (!stp) {
2842 status = stateid_error_map(stateid);
1da177e4 2843 goto out;
78155ed7 2844 }
6150ef0d 2845 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2846 goto out;
2847 if (!stp->st_stateowner->so_confirmed)
2848 goto out;
6668958f
AA		 2849 status = check_stateid_generation(stateid, &stp->st_stateid,
2850 flags);
0c2a498f
BF		 2851 if (status)
2852 goto out;
a4455be0
BF		 2853 status = nfs4_check_openmode(stp, flags);
2854 if (status)
1da177e4
LT		 2855 goto out;
2856 renew_client(stp->st_stateowner->so_client);
2857 if (filpp)
2858 *filpp = stp->st_vfs_file;
1da177e4
LT		 2859 }
2860 status = nfs_ok;
2861out:
2862 return status;
2863}
2864
4c4cd222
N		 2865static inline int
2866setlkflg (int type)
2867{
2868 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2869 RD_STATE : WR_STATE;
2870}
1da177e4
LT		 2871
2872/*
2873 * Checks for sequence id mutating operations.
2874 */
b37ad28b 2875static __be32
dd453dfd
BH		 2876nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2877 stateid_t *stateid, int flags,
2878 struct nfs4_stateowner **sopp,
2879 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2880{
1da177e4
LT		 2881 struct nfs4_stateid *stp;
2882 struct nfs4_stateowner *sop;
dd453dfd 2883 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2884 __be32 status;
1da177e4
LT		 2885
2886 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2887 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2888 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2889 stateid->si_generation);
3a4f98bb 2890
1da177e4
LT		 2891 *stpp = NULL;
2892 *sopp = NULL;
2893
1da177e4 2894 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2895 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2896 return nfserr_bad_stateid;
1da177e4
LT		 2897 }
2898
1da177e4 2899 if (STALE_STATEID(stateid))
3a4f98bb 2900 return nfserr_stale_stateid;
6668958f
AA		 2901
2902 if (nfsd4_has_session(cstate))
2903 flags |= HAS_SESSION;
2904
1da177e4
LT		 2905 /*
2906 * We return BAD_STATEID if filehandle doesn't match stateid,
2907 * the confirmed flag is incorrecly set, or the generation
2908 * number is incorrect.
1da177e4 2909 */
f8816512
N		 2910 stp = find_stateid(stateid, flags);
2911 if (stp == NULL) {
2912 /*
2913 * Also, we should make sure this isn't just the result of
2914 * a replayed close:
2915 */
2916 sop = search_close_lru(stateid->si_stateownerid, flags);
2917 if (sop == NULL)
78155ed7 2918 return stateid_error_map(stateid);
f8816512
N		 2919 *sopp = sop;
2920 goto check_replay;
2921 }
1da177e4 2922
39325bd0
BF		 2923 *stpp = stp;
2924 *sopp = sop = stp->st_stateowner;
2925
4c4cd222 2926 if (lock) {
4c4cd222 2927 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2928 struct nfs4_client *clp = sop->so_client;
4c4cd222 2929 int lkflg = 0;
b37ad28b 2930 __be32 status;
4c4cd222
N		 2931
2932 lkflg = setlkflg(lock->lk_type);
2933
2934 if (lock->lk_is_new) {
599e0a22
BF		 2935 if (!sop->so_is_open_owner)
2936 return nfserr_bad_stateid;
60adfc50
AA		 2937 if (!(flags & HAS_SESSION) &&
2938 !same_clid(&clp->cl_clientid, lockclid))
2939 return nfserr_bad_stateid;
599e0a22
BF		 2940 /* stp is the open stateid */
2941 status = nfs4_check_openmode(stp, lkflg);
2942 if (status)
2943 return status;
2944 } else {
2945 /* stp is the lock stateid */
2946 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2947 if (status)
2948 return status;
4c4cd222 2949 }
1da177e4
LT		 2950 }
2951
f3362737 2952 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2953 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2954 return nfserr_bad_stateid;
1da177e4
LT		 2955 }
2956
1da177e4
LT		 2957 /*
2958 * We now validate the seqid and stateid generation numbers.
2959 * For the moment, we ignore the possibility of
2960 * generation number wraparound.
2961 */
6668958f 2962 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2963 goto check_replay;
2964
3a4f98bb 2965 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2966 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2967 " unconfirmed stateowner!\n");
2968 return nfserr_bad_stateid;
1da177e4 2969 }
3a4f98bb 2970 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2971 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2972 " confirmed yet!\n");
2973 return nfserr_bad_stateid;
1da177e4 2974 }
6668958f 2975 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 2976 if (status)
2977 return status;
52fd004e 2978 renew_client(sop->so_client);
3a4f98bb 2979 return nfs_ok;
1da177e4 2980
1da177e4 2981check_replay:
bd9aac52 2982 if (seqid == sop->so_seqid - 1) {
849823c5 2983 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2984 /* indicate replay to calling function */
a90b061c 2985 return nfserr_replay_me;
1da177e4 2986 }
2fdada03 2987 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2988 sop->so_seqid, seqid);
2989 *sopp = NULL;
2990 return nfserr_bad_seqid;
1da177e4
LT		 2991}
2992
b37ad28b 2993__be32
ca364317 2994nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2995 struct nfsd4_open_confirm *oc)
1da177e4 2996{
b37ad28b 2997 __be32 status;
1da177e4
LT		 2998 struct nfs4_stateowner *sop;
2999 struct nfs4_stateid *stp;
3000
3001 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3002 (int)cstate->current_fh.fh_dentry->d_name.len,
3003 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3004
ca364317 3005 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3006 if (status)
3007 return status;
1da177e4
LT		 3008
3009 nfs4_lock_state();
3010
dd453dfd 3011 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3012 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3013 CONFIRM | OPEN_STATE,
1da177e4
LT		 3014 &oc->oc_stateowner, &stp, NULL)))
3015 goto out;
3016
3017 sop = oc->oc_stateowner;
3018 sop->so_confirmed = 1;
3019 update_stateid(&stp->st_stateid);
3020 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3021 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3022 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3023 stp->st_stateid.si_boot,
3024 stp->st_stateid.si_stateownerid,
3025 stp->st_stateid.si_fileid,
3026 stp->st_stateid.si_generation);
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
LT		 3284
3285 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3286 stid->si_boot, stid->si_stateownerid,
3287 stid->si_fileid, stid->si_generation);
3288
1da177e4 3289 fp = find_file(ino);
13cd2184
N		 3290 if (!fp)
3291 return NULL;
3292 dl = find_delegation_file(fp, stid);
3293 put_nfs4_file(fp);
3294 return dl;
1da177e4
LT		 3295}
3296
3297/*
3298 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3299 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3300 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3301 * locking, this prevents us from being completely protocol-compliant. The
3302 * real solution to this problem is to start using unsigned file offsets in
3303 * the VFS, but this is a very deep change!
3304 */
3305static inline void
3306nfs4_transform_lock_offset(struct file_lock *lock)
3307{
3308 if (lock->fl_start < 0)
3309 lock->fl_start = OFFSET_MAX;
3310 if (lock->fl_end < 0)
3311 lock->fl_end = OFFSET_MAX;
3312}
3313
d5b9026a
N		 3314/* Hack!: For now, we're defining this just so we can use a pointer to it
3315 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3316static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3317};
1da177e4
LT		 3318
3319static inline void
3320nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3321{
d5b9026a
N		 3322 struct nfs4_stateowner *sop;
3323 unsigned int hval;
1da177e4 3324
d5b9026a
N		 3325 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3326 sop = (struct nfs4_stateowner *) fl->fl_owner;
3327 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3328 kref_get(&sop->so_ref);
3329 deny->ld_sop = sop;
3330 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3331 } else {
3332 deny->ld_sop = NULL;
3333 deny->ld_clientid.cl_boot = 0;
3334 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3335 }
3336 deny->ld_start = fl->fl_start;
87df4de8
BH		 3337 deny->ld_length = NFS4_MAX_UINT64;
3338 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3339 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3340 deny->ld_type = NFS4_READ_LT;
3341 if (fl->fl_type != F_RDLCK)
3342 deny->ld_type = NFS4_WRITE_LT;
3343}
3344
1da177e4
LT		 3345static struct nfs4_stateowner *
3346find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3347 struct xdr_netobj *owner)
3348{
3349 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3350 struct nfs4_stateowner *op;
3351
3352 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3353 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3354 return op;
3355 }
3356 return NULL;
3357}
3358
3359/*
3360 * Alloc a lock owner structure.
3361 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3362 * occured.
3363 *
3364 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3365 */
3366
3367static struct nfs4_stateowner *
3368alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3369 struct nfs4_stateowner *sop;
3370 struct nfs4_replay *rp;
3371 unsigned int idhashval;
3372
3373 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3374 return NULL;
3375 idhashval = lockownerid_hashval(current_ownerid);
3376 INIT_LIST_HEAD(&sop->so_idhash);
3377 INIT_LIST_HEAD(&sop->so_strhash);
3378 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3379 INIT_LIST_HEAD(&sop->so_stateids);
3380 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3381 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3382 sop->so_time = 0;
3383 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3384 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3385 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3386 sop->so_is_open_owner = 0;
3387 sop->so_id = current_ownerid++;
3388 sop->so_client = clp;
b59e3c0e
NB		 3389 /* It is the openowner seqid that will be incremented in encode in the
3390 * case of new lockowners; so increment the lock seqid manually: */
3391 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3392 sop->so_confirmed = 1;
3393 rp = &sop->so_replay;
de1ae286 3394 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3395 rp->rp_buflen = 0;
3396 rp->rp_buf = rp->rp_ibuf;
3397 return sop;
3398}
3399
fd39ca9a 3400static struct nfs4_stateid *
1da177e4
LT		 3401alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3402{
3403 struct nfs4_stateid *stp;
3404 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3405
5ac049ac
N		 3406 stp = nfs4_alloc_stateid();
3407 if (stp == NULL)
1da177e4
LT		 3408 goto out;
3409 INIT_LIST_HEAD(&stp->st_hash);
3410 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3411 INIT_LIST_HEAD(&stp->st_perstateowner);
3412 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3413 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3414 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3415 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3416 stp->st_stateowner = sop;
13cd2184 3417 get_nfs4_file(fp);
1da177e4 3418 stp->st_file = fp;
78155ed7 3419 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3420 stp->st_stateid.si_stateownerid = sop->so_id;
3421 stp->st_stateid.si_fileid = fp->fi_id;
3422 stp->st_stateid.si_generation = 0;
3423 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3424 stp->st_access_bmap = open_stp->st_access_bmap;
3425 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3426 stp->st_openstp = open_stp;
1da177e4
LT		 3427
3428out:
3429 return stp;
3430}
3431
fd39ca9a 3432static int
1da177e4
LT		 3433check_lock_length(u64 offset, u64 length)
3434{
87df4de8 3435 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3436 LOFF_OVERFLOW(offset, length)));
3437}
3438
3439/*
3440 * LOCK operation
3441 */
b37ad28b 3442__be32
ca364317 3443nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3444 struct nfsd4_lock *lock)
1da177e4 3445{
3e9e3dbe 3446 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3447 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3448 struct nfs4_stateid *lock_stp;
3449 struct file *filp;
3450 struct file_lock file_lock;
8dc7c311 3451 struct file_lock conflock;
b37ad28b 3452 __be32 status = 0;
1da177e4 3453 unsigned int strhashval;
150b3934 3454 unsigned int cmd;
b8dd7b9a 3455 int err;
1da177e4
LT		 3456
3457 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3458 (long long) lock->lk_offset,
3459 (long long) lock->lk_length);
3460
1da177e4
LT		 3461 if (check_lock_length(lock->lk_offset, lock->lk_length))
3462 return nfserr_inval;
3463
ca364317 3464 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3465 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3466 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3467 return status;
3468 }
3469
1da177e4
LT		 3470 nfs4_lock_state();
3471
3472 if (lock->lk_is_new) {
893f8770
N		 3473 /*
3474 * Client indicates that this is a new lockowner.
3475 * Use open owner and open stateid to create lock owner and
3476 * lock stateid.
3477 */
1da177e4
LT		 3478 struct nfs4_stateid *open_stp = NULL;
3479 struct nfs4_file *fp;
3480
3481 status = nfserr_stale_clientid;
60adfc50
AA		 3482 if (!nfsd4_has_session(cstate) &&
3483 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3484 goto out;
1da177e4 3485
1da177e4 3486 /* validate and update open stateid and open seqid */
dd453dfd 3487 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3488 lock->lk_new_open_seqid,
3489 &lock->lk_new_open_stateid,
f3362737 3490 OPEN_STATE,
3a65588a 3491 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3492 lock);
37515177 3493 if (status)
1da177e4 3494 goto out;
3a65588a 3495 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3496 /* create lockowner and lock stateid */
3497 fp = open_stp->st_file;
3498 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3499 open_sop->so_client->cl_clientid.cl_id,
3500 &lock->v.new.owner);
3e9e3dbe
N		 3501 /* XXX: Do we need to check for duplicate stateowners on
3502 * the same file, or should they just be allowed (and
3503 * create new stateids)? */
1da177e4 3504 status = nfserr_resource;
b59e3c0e
NB		 3505 lock_sop = alloc_init_lock_stateowner(strhashval,
3506 open_sop->so_client, open_stp, lock);
3507 if (lock_sop == NULL)
1da177e4 3508 goto out;
b59e3c0e 3509 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3510 if (lock_stp == NULL)
1da177e4 3511 goto out;
1da177e4
LT		 3512 } else {
3513 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3514 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3515 lock->lk_old_lock_seqid,
3516 &lock->lk_old_lock_stateid,
f3362737 3517 LOCK_STATE,
3a65588a 3518 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3519 if (status)
3520 goto out;
3a65588a 3521 lock_sop = lock->lk_replay_owner;
1da177e4 3522 }
3a65588a 3523 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3524 filp = lock_stp->st_vfs_file;
3525
0dd395dc 3526 status = nfserr_grace;
af558e33 3527 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3528 goto out;
3529 status = nfserr_no_grace;
af558e33 3530 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3531 goto out;
3532
1da177e4
LT		 3533 locks_init_lock(&file_lock);
3534 switch (lock->lk_type) {
3535 case NFS4_READ_LT:
3536 case NFS4_READW_LT:
3537 file_lock.fl_type = F_RDLCK;
150b3934 3538 cmd = F_SETLK;
1da177e4
LT		 3539 break;
3540 case NFS4_WRITE_LT:
3541 case NFS4_WRITEW_LT:
3542 file_lock.fl_type = F_WRLCK;
150b3934 3543 cmd = F_SETLK;
1da177e4
LT		 3544 break;
3545 default:
3546 status = nfserr_inval;
3547 goto out;
3548 }
b59e3c0e 3549 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3550 file_lock.fl_pid = current->tgid;
3551 file_lock.fl_file = filp;
3552 file_lock.fl_flags = FL_POSIX;
d5b9026a 3553 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3554
3555 file_lock.fl_start = lock->lk_offset;
87df4de8 3556 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3557 nfs4_transform_lock_offset(&file_lock);
3558
3559 /*
3560 * Try to lock the file in the VFS.
3561 * Note: locks.c uses the BKL to protect the inode's lock list.
3562 */
3563
fd85b817 3564 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3565 switch (-err) {
1da177e4
LT		 3566 case 0: /* success! */
3567 update_stateid(&lock_stp->st_stateid);
3568 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3569 sizeof(stateid_t));
b8dd7b9a 3570 status = 0;
eb76b3fd
AA		 3571 break;
3572 case (EAGAIN): /* conflock holds conflicting lock */
3573 status = nfserr_denied;
3574 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3575 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3576 break;
1da177e4
LT		 3577 case (EDEADLK):
3578 status = nfserr_deadlock;
eb76b3fd 3579 break;
1da177e4 3580 default:
fd85b817 3581 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3582 status = nfserr_resource;
3583 break;
1da177e4 3584 }
1da177e4 3585out:
8a280510 3586 if (status && lock->lk_is_new && lock_sop)
f044ff83 3587 release_lockowner(lock_sop);
3a65588a
BF		 3588 if (lock->lk_replay_owner) {
3589 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3590 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3591 }
1da177e4
LT		 3592 nfs4_unlock_state();
3593 return status;
3594}
3595
55ef1274
BF		 3596/*
3597 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3598 * so we do a temporary open here just to get an open file to pass to
3599 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3600 * inode operation.)
3601 */
3602static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3603{
3604 struct file *file;
3605 int err;
3606
3607 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3608 if (err)
3609 return err;
3610 err = vfs_test_lock(file, lock);
3611 nfsd_close(file);
3612 return err;
3613}
3614
1da177e4
LT		 3615/*
3616 * LOCKT operation
3617 */
b37ad28b 3618__be32
ca364317
BF		 3619nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3620 struct nfsd4_lockt *lockt)
1da177e4
LT		 3621{
3622 struct inode *inode;
1da177e4 3623 struct file_lock file_lock;
fd85b817 3624 int error;
b37ad28b 3625 __be32 status;
1da177e4 3626
af558e33 3627 if (locks_in_grace())
1da177e4
LT		 3628 return nfserr_grace;
3629
3630 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3631 return nfserr_inval;
3632
3633 lockt->lt_stateowner = NULL;
3634 nfs4_lock_state();
3635
3636 status = nfserr_stale_clientid;
60adfc50 3637 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3638 goto out;
1da177e4 3639
ca364317 3640 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3641 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3642 if (status == nfserr_symlink)
3643 status = nfserr_inval;
3644 goto out;
3645 }
3646
ca364317 3647 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3648 locks_init_lock(&file_lock);
3649 switch (lockt->lt_type) {
3650 case NFS4_READ_LT:
3651 case NFS4_READW_LT:
3652 file_lock.fl_type = F_RDLCK;
3653 break;
3654 case NFS4_WRITE_LT:
3655 case NFS4_WRITEW_LT:
3656 file_lock.fl_type = F_WRLCK;
3657 break;
3658 default:
2fdada03 3659 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3660 status = nfserr_inval;
3661 goto out;
3662 }
3663
3664 lockt->lt_stateowner = find_lockstateowner_str(inode,
3665 &lockt->lt_clientid, &lockt->lt_owner);
3666 if (lockt->lt_stateowner)
3667 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3668 file_lock.fl_pid = current->tgid;
3669 file_lock.fl_flags = FL_POSIX;
3670
3671 file_lock.fl_start = lockt->lt_offset;
87df4de8 3672 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3673
3674 nfs4_transform_lock_offset(&file_lock);
3675
1da177e4 3676 status = nfs_ok;
55ef1274 3677 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3678 if (error) {
3679 status = nfserrno(error);
3680 goto out;
3681 }
9d6a8c5c 3682 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3683 status = nfserr_denied;
9d6a8c5c 3684 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3685 }
3686out:
3687 nfs4_unlock_state();
3688 return status;
3689}
3690
b37ad28b 3691__be32
ca364317 3692nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3693 struct nfsd4_locku *locku)
1da177e4
LT		 3694{
3695 struct nfs4_stateid *stp;
3696 struct file *filp = NULL;
3697 struct file_lock file_lock;
b37ad28b 3698 __be32 status;
b8dd7b9a 3699 int err;
1da177e4
LT		 3700
3701 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3702 (long long) locku->lu_offset,
3703 (long long) locku->lu_length);
3704
3705 if (check_lock_length(locku->lu_offset, locku->lu_length))
3706 return nfserr_inval;
3707
3708 nfs4_lock_state();
3709
dd453dfd 3710 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3711 locku->lu_seqid,
3712 &locku->lu_stateid,
f3362737 3713 LOCK_STATE,
1da177e4
LT		 3714 &locku->lu_stateowner, &stp, NULL)))
3715 goto out;
3716
3717 filp = stp->st_vfs_file;
3718 BUG_ON(!filp);
3719 locks_init_lock(&file_lock);
3720 file_lock.fl_type = F_UNLCK;
3721 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3722 file_lock.fl_pid = current->tgid;
3723 file_lock.fl_file = filp;
3724 file_lock.fl_flags = FL_POSIX;
d5b9026a 3725 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3726 file_lock.fl_start = locku->lu_offset;
3727
87df4de8 3728 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3729 nfs4_transform_lock_offset(&file_lock);
3730
3731 /*
3732 * Try to unlock the file in the VFS.
3733 */
fd85b817 3734 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3735 if (err) {
fd85b817 3736 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3737 goto out_nfserr;
3738 }
3739 /*
3740 * OK, unlock succeeded; the only thing left to do is update the stateid.
3741 */
3742 update_stateid(&stp->st_stateid);
3743 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3744
3745out:
f2327d9a 3746 if (locku->lu_stateowner) {
1da177e4 3747 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3748 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3749 }
1da177e4
LT		 3750 nfs4_unlock_state();
3751 return status;
3752
3753out_nfserr:
b8dd7b9a 3754 status = nfserrno(err);
1da177e4
LT		 3755 goto out;
3756}
3757
3758/*
3759 * returns
3760 * 1: locks held by lockowner
3761 * 0: no locks held by lockowner
3762 */
3763static int
3764check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3765{
3766 struct file_lock **flpp;
7eaa36e2 3767 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3768 int status = 0;
3769
3770 lock_kernel();
3771 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3772 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3773 status = 1;
3774 goto out;
796dadfd 3775 }
1da177e4
LT		 3776 }
3777out:
3778 unlock_kernel();
3779 return status;
3780}
3781
b37ad28b 3782__be32
b591480b
BF		 3783nfsd4_release_lockowner(struct svc_rqst *rqstp,
3784 struct nfsd4_compound_state *cstate,
3785 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3786{
3787 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3788 struct nfs4_stateowner *sop;
3789 struct nfs4_stateid *stp;
1da177e4 3790 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3791 struct list_head matches;
3792 int i;
b37ad28b 3793 __be32 status;
1da177e4
LT		 3794
3795 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3796 clid->cl_boot, clid->cl_id);
3797
3798 /* XXX check for lease expiration */
3799
3800 status = nfserr_stale_clientid;
849823c5 3801 if (STALE_CLIENTID(clid))
1da177e4 3802 return status;
1da177e4
LT		 3803
3804 nfs4_lock_state();
3805
3e9e3dbe
N		 3806 status = nfserr_locks_held;
3807 /* XXX: we're doing a linear search through all the lockowners.
3808 * Yipes! For now we'll just hope clients aren't really using
3809 * release_lockowner much, but eventually we have to fix these
3810 * data structures. */
3811 INIT_LIST_HEAD(&matches);
3812 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3813 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3814 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3815 continue;
3816 list_for_each_entry(stp, &sop->so_stateids,
3817 st_perstateowner) {
3818 if (check_for_locks(stp->st_vfs_file, sop))
3819 goto out;
3820 /* Note: so_perclient unused for lockowners,
3821 * so it's OK to fool with here. */
3822 list_add(&sop->so_perclient, &matches);
3823 }
1da177e4 3824 }
3e9e3dbe
N		 3825 }
3826 /* Clients probably won't expect us to return with some (but not all)
3827 * of the lockowner state released; so don't release any until all
3828 * have been checked. */
3829 status = nfs_ok;
0fa822e4
N		 3830 while (!list_empty(&matches)) {
3831 sop = list_entry(matches.next, struct nfs4_stateowner,
3832 so_perclient);
3833 /* unhash_stateowner deletes so_perclient only
3834 * for openowners. */
3835 list_del(&sop->so_perclient);
f044ff83 3836 release_lockowner(sop);
1da177e4
LT		 3837 }
3838out:
3839 nfs4_unlock_state();
3840 return status;
3841}
3842
3843static inline struct nfs4_client_reclaim *
a55370a3 3844alloc_reclaim(void)
1da177e4 3845{
a55370a3 3846 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3847}
3848
c7b9a459 3849int
a1bcecd2 3850nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3851{
3852 unsigned int strhashval = clientstr_hashval(name);
3853 struct nfs4_client *clp;
3854
a1bcecd2 3855 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3856 return clp ? 1 : 0;
3857}
3858
1da177e4
LT		 3859/*
3860 * failure => all reset bets are off, nfserr_no_grace...
3861 */
190e4fbf
N		 3862int
3863nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3864{
3865 unsigned int strhashval;
3866 struct nfs4_client_reclaim *crp = NULL;
3867
a55370a3
N		 3868 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3869 crp = alloc_reclaim();
1da177e4
LT		 3870 if (!crp)
3871 return 0;
a55370a3 3872 strhashval = clientstr_hashval(name);
1da177e4
LT		 3873 INIT_LIST_HEAD(&crp->cr_strhash);
3874 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3875 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3876 reclaim_str_hashtbl_size++;
3877 return 1;
3878}
3879
3880static void
3881nfs4_release_reclaim(void)
3882{
3883 struct nfs4_client_reclaim *crp = NULL;
3884 int i;
3885
1da177e4
LT		 3886 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3887 while (!list_empty(&reclaim_str_hashtbl[i])) {
3888 crp = list_entry(reclaim_str_hashtbl[i].next,
3889 struct nfs4_client_reclaim, cr_strhash);
3890 list_del(&crp->cr_strhash);
1da177e4
LT		 3891 kfree(crp);
3892 reclaim_str_hashtbl_size--;
3893 }
3894 }
3895 BUG_ON(reclaim_str_hashtbl_size);
3896}
3897
3898/*
3899 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3900static struct nfs4_client_reclaim *
1da177e4
LT		 3901nfs4_find_reclaim_client(clientid_t *clid)
3902{
3903 unsigned int strhashval;
3904 struct nfs4_client *clp;
3905 struct nfs4_client_reclaim *crp = NULL;
3906
3907
3908 /* find clientid in conf_id_hashtbl */
3909 clp = find_confirmed_client(clid);
3910 if (clp == NULL)
3911 return NULL;
3912
a55370a3
N		 3913 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3914 clp->cl_name.len, clp->cl_name.data,
3915 clp->cl_recdir);
1da177e4
LT		 3916
3917 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3918 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3919 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3920 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3921 return crp;
3922 }
3923 }
3924 return NULL;
3925}
3926
3927/*
3928* Called from OPEN. Look for clientid in reclaim list.
3929*/
b37ad28b 3930__be32
1da177e4
LT		 3931nfs4_check_open_reclaim(clientid_t *clid)
3932{
dfc83565 3933 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3934}
3935
ac4d8ff2 3936/* initialization to perform at module load time: */
1da177e4 3937
e8ff2a84 3938int
ac4d8ff2 3939nfs4_state_init(void)
1da177e4 3940{
e8ff2a84 3941 int i, status;
1da177e4 3942
e8ff2a84
BF		 3943 status = nfsd4_init_slabs();
3944 if (status)
3945 return status;
1da177e4
LT		 3946 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3947 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3948 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3949 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3950 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3951 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3952 }
5282fd72
ME		 3953 for (i = 0; i < SESSION_HASH_SIZE; i++)
3954 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3955 for (i = 0; i < FILE_HASH_SIZE; i++) {
3956 INIT_LIST_HEAD(&file_hashtbl[i]);
3957 }
3958 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3959 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3960 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3961 }
3962 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3963 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3964 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3965 }
3966 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3967 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3968 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3969 }
1da177e4 3970 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3971 INIT_LIST_HEAD(&close_lru);
3972 INIT_LIST_HEAD(&client_lru);
3973 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3974 reclaim_str_hashtbl_size = 0;
e8ff2a84 3975 return 0;
ac4d8ff2
N		 3976}
3977
190e4fbf
N		 3978static void
3979nfsd4_load_reboot_recovery_data(void)
3980{
3981 int status;
3982
0964a3d3
N		 3983 nfs4_lock_state();
3984 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3985 status = nfsd4_recdir_load();
0964a3d3 3986 nfs4_unlock_state();
190e4fbf
N		 3987 if (status)
3988 printk("NFSD: Failure reading reboot recovery data\n");
3989}
3990
9a8db97e
ME		 3991unsigned long
3992get_nfs4_grace_period(void)
3993{
3994 return max(user_lease_time, lease_time) * HZ;
3995}
3996
c2f1a551
MS		 3997/*
3998 * Since the lifetime of a delegation isn't limited to that of an open, a
3999 * client may quite reasonably hang on to a delegation as long as it has
4000 * the inode cached. This becomes an obvious problem the first time a
4001 * client's inode cache approaches the size of the server's total memory.
4002 *
4003 * For now we avoid this problem by imposing a hard limit on the number
4004 * of delegations, which varies according to the server's memory size.
4005 */
4006static void
4007set_max_delegations(void)
4008{
4009 /*
4010 * Allow at most 4 delegations per megabyte of RAM. Quick
4011 * estimates suggest that in the worst case (where every delegation
4012 * is for a different inode), a delegation could take about 1.5K,
4013 * giving a worst case usage of about 6% of memory.
4014 */
4015 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4016}
4017
ac4d8ff2
N		 4018/* initialization to perform when the nfsd service is started: */
4019
29ab23cc 4020static int
ac4d8ff2
N		 4021__nfs4_state_start(void)
4022{
9a8db97e 4023 unsigned long grace_time;
ac4d8ff2 4024
1da177e4 4025 boot_time = get_seconds();
af558e33 4026 grace_time = get_nfs4_grace_period();
d99a05ad 4027 lease_time = user_lease_time;
af558e33 4028 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4029 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4030 grace_time/HZ);
58da282b 4031 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 4032 if (laundry_wq == NULL)
4033 return -ENOMEM;
9a8db97e 4034 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4035 set_max_delegations();
80fc015b 4036 return set_callback_cred();
1da177e4
LT		 4037}
4038
29ab23cc 4039int
76a3550e 4040nfs4_state_start(void)
1da177e4 4041{
29ab23cc
BF		 4042 int ret;
4043
1da177e4 4044 if (nfs4_init)
29ab23cc 4045 return 0;
190e4fbf 4046 nfsd4_load_reboot_recovery_data();
29ab23cc
BF		 4047 ret = __nfs4_state_start();
4048 if (ret)
4049 return ret;
1da177e4 4050 nfs4_init = 1;
29ab23cc 4051 return 0;
1da177e4
LT		 4052}
4053
1da177e4
LT		 4054time_t
4055nfs4_lease_time(void)
4056{
4057 return lease_time;
4058}
4059
4060static void
4061__nfs4_state_shutdown(void)
4062{
4063 int i;
4064 struct nfs4_client *clp = NULL;
4065 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4066 struct list_head *pos, *next, reaplist;
4067
1da177e4
LT		 4068 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4069 while (!list_empty(&conf_id_hashtbl[i])) {
4070 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4071 expire_client(clp);
4072 }
4073 while (!list_empty(&unconf_str_hashtbl[i])) {
4074 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4075 expire_client(clp);
4076 }
4077 }
4078 INIT_LIST_HEAD(&reaplist);
4079 spin_lock(&recall_lock);
4080 list_for_each_safe(pos, next, &del_recall_lru) {
4081 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4082 list_move(&dp->dl_recall_lru, &reaplist);
4083 }
4084 spin_unlock(&recall_lock);
4085 list_for_each_safe(pos, next, &reaplist) {
4086 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4087 list_del_init(&dp->dl_recall_lru);
4088 unhash_delegation(dp);
4089 }
4090
190e4fbf 4091 nfsd4_shutdown_recdir();
1da177e4 4092 nfs4_init = 0;
1da177e4
LT		 4093}
4094
4095void
4096nfs4_state_shutdown(void)
4097{
5e8d5c29
N		 4098 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4099 destroy_workqueue(laundry_wq);
2c5e7615 4100 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4101 nfs4_lock_state();
4102 nfs4_release_reclaim();
4103 __nfs4_state_shutdown();
1da177e4
LT		 4104 nfs4_unlock_state();
4105}
4106
3dd98a3b
JL		 4107/*
4108 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4109 * accessed when nfsd is starting.
4110 */
0964a3d3
N		 4111static void
4112nfs4_set_recdir(char *recdir)
4113{
0964a3d3 4114 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4115}
4116
4117/*
4118 * Change the NFSv4 recovery directory to recdir.
4119 */
4120int
4121nfs4_reset_recoverydir(char *recdir)
4122{
4123 int status;
a63bb996 4124 struct path path;
0964a3d3 4125
a63bb996 4126 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4127 if (status)
4128 return status;
4129 status = -ENOTDIR;
a63bb996 4130 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4131 nfs4_set_recdir(recdir);
4132 status = 0;
4133 }
a63bb996 4134 path_put(&path);
0964a3d3
N		 4135 return status;
4136}
4137
3dd98a3b
JL		 4138char *
4139nfs4_recoverydir(void)
4140{
4141 return user_recovery_dirname;
4142}
4143
1da177e4
LT		 4144/*
4145 * Called when leasetime is changed.
4146 *
d99a05ad
N		 4147 * The only way the protocol gives us to handle on-the-fly lease changes is to
4148 * simulate a reboot. Instead of doing that, we just wait till the next time
4149 * we start to register any changes in lease time. If the administrator
4150 * really wants to change the lease time *now*, they can go ahead and bring
4151 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4152 *
4153 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4154 * when nfsd is starting
1da177e4
LT		 4155 */
4156void
4157nfs4_reset_lease(time_t leasetime)
4158{
d99a05ad 4159 user_lease_time = leasetime;
1da177e4 4160}
ubuntu-zesty-kernel mirror