projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4.1: fix session memory use calculation
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
363168b4 58#include <linux/sunrpc/clnt.h>
1da177e4
LT		 59
60#define NFSDDBG_FACILITY NFSDDBG_PROC
61
62/* Globals */
63static time_t lease_time = 90; /* default lease time */
d99a05ad 64static time_t user_lease_time = 90;
fd39ca9a 65static time_t boot_time;
1da177e4
LT		 66static u32 current_ownerid = 1;
67static u32 current_fileid = 1;
68static u32 current_delegid = 1;
69static u32 nfs4_init;
fd39ca9a
N		 70static stateid_t zerostateid; /* bits all 0 */
71static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 72static u64 current_sessionid = 1;
fd39ca9a
N		 73
74#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
75#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 76
1da177e4 77/* forward declarations */
fd39ca9a 78static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 79static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 80static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
81static void nfs4_set_recdir(char *recdir);
1da177e4 82
8b671b80
BF		 83/* Locking: */
84
85/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 86static DEFINE_MUTEX(client_mutex);
1da177e4 87
8b671b80
BF		 88/*
89 * Currently used for the del_recall_lru and file hash table. In an
90 * effort to decrease the scope of the client_mutex, this spinlock may
91 * eventually cover more:
92 */
93static DEFINE_SPINLOCK(recall_lock);
94
e18b890b
CL		 95static struct kmem_cache *stateowner_slab = NULL;
96static struct kmem_cache *file_slab = NULL;
97static struct kmem_cache *stateid_slab = NULL;
98static struct kmem_cache *deleg_slab = NULL;
e60d4398 99
1da177e4
LT		 100void
101nfs4_lock_state(void)
102{
353ab6e9 103 mutex_lock(&client_mutex);
1da177e4
LT		 104}
105
106void
107nfs4_unlock_state(void)
108{
353ab6e9 109 mutex_unlock(&client_mutex);
1da177e4
LT		 110}
111
112static inline u32
113opaque_hashval(const void *ptr, int nbytes)
114{
115 unsigned char *cptr = (unsigned char *) ptr;
116
117 u32 x = 0;
118 while (nbytes--) {
119 x *= 37;
120 x += *cptr++;
121 }
122 return x;
123}
124
1da177e4
LT		 125static struct list_head del_recall_lru;
126
13cd2184
N		 127static inline void
128put_nfs4_file(struct nfs4_file *fi)
129{
8b671b80
BF		 130 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
131 list_del(&fi->fi_hash);
132 spin_unlock(&recall_lock);
133 iput(fi->fi_inode);
134 kmem_cache_free(file_slab, fi);
135 }
13cd2184
N		 136}
137
138static inline void
139get_nfs4_file(struct nfs4_file *fi)
140{
8b671b80 141 atomic_inc(&fi->fi_ref);
13cd2184
N		 142}
143
ef0f3390 144static int num_delegations;
c2f1a551 145unsigned int max_delegations;
ef0f3390
N		 146
147/*
148 * Open owner state (share locks)
149 */
150
151/* hash tables for nfs4_stateowner */
152#define OWNER_HASH_BITS 8
153#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
154#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155
156#define ownerid_hashval(id) \
157 ((id) & OWNER_HASH_MASK)
158#define ownerstr_hashval(clientid, ownername) \
159 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160
161static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
162static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
163
164/* hash table for nfs4_file */
165#define FILE_HASH_BITS 8
166#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
167#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
168/* hash table for (open)nfs4_stateid */
169#define STATEID_HASH_BITS 10
170#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
171#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172
173#define file_hashval(x) \
174 hash_ptr(x, FILE_HASH_BITS)
175#define stateid_hashval(owner_id, file_id) \
176 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177
178static struct list_head file_hashtbl[FILE_HASH_SIZE];
179static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
180
1da177e4
LT		 181static struct nfs4_delegation *
182alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
183{
184 struct nfs4_delegation *dp;
185 struct nfs4_file *fp = stp->st_file;
c237dc03 186 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 187
188 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 189 if (fp->fi_had_conflict)
190 return NULL;
c2f1a551 191 if (num_delegations > max_delegations)
ef0f3390 192 return NULL;
5b2d21c1
N		 193 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
194 if (dp == NULL)
1da177e4 195 return dp;
ef0f3390 196 num_delegations++;
ea1da636
N		 197 INIT_LIST_HEAD(&dp->dl_perfile);
198 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 199 INIT_LIST_HEAD(&dp->dl_recall_lru);
200 dp->dl_client = clp;
13cd2184 201 get_nfs4_file(fp);
1da177e4
LT		 202 dp->dl_file = fp;
203 dp->dl_flock = NULL;
204 get_file(stp->st_vfs_file);
205 dp->dl_vfs_file = stp->st_vfs_file;
206 dp->dl_type = type;
b53d40c5 207 dp->dl_ident = cb->cb_ident;
78155ed7 208 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
6c02eaa1 212 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 213 dp->dl_time = 0;
214 atomic_set(&dp->dl_count, 1);
ea1da636
N		 215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 217 return dp;
218}
219
220void
221nfs4_put_delegation(struct nfs4_delegation *dp)
222{
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 225 put_nfs4_file(dp->dl_file);
5b2d21c1 226 kmem_cache_free(deleg_slab, dp);
ef0f3390 227 num_delegations--;
1da177e4
LT		 228 }
229}
230
231/* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
235 */
236static void
237nfs4_close_delegation(struct nfs4_delegation *dp)
238{
239 struct file *filp = dp->dl_vfs_file;
240
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
c907132d 245 if (dp->dl_flock)
a9933cea 246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 247 nfsd_close(filp);
1da177e4
LT		 248}
249
250/* Called under the state lock. */
251static void
252unhash_delegation(struct nfs4_delegation *dp)
253{
ea1da636
N		 254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
261}
262
263/*
264 * SETCLIENTID state
265 */
266
267/* Hash tables for nfs4_clientid state */
268#define CLIENT_HASH_BITS 4
269#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271
272#define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 274#define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 276/*
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
279 *
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
282 *
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
284 * setclientid info.
285 *
286 * client_lru holds client queue ordered by nfs4_client.cl_time
287 * for lease renewal.
288 *
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
291 */
292static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293static int reclaim_str_hashtbl_size = 0;
294static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head client_lru;
299static struct list_head close_lru;
300
2283963f
BF		 301static void unhash_generic_stateid(struct nfs4_stateid *stp)
302{
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
306}
307
308static void free_generic_stateid(struct nfs4_stateid *stp)
309{
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
312}
313
314static void release_lock_stateid(struct nfs4_stateid *stp)
315{
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
319}
320
f044ff83
BF		 321static void unhash_lockowner(struct nfs4_stateowner *sop)
322{
323 struct nfs4_stateid *stp;
324
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
332 }
333}
334
335static void release_lockowner(struct nfs4_stateowner *sop)
336{
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
339}
340
f1d110ca
BF		 341static void
342release_stateid_lockowners(struct nfs4_stateid *open_stp)
343{
344 struct nfs4_stateowner *lock_sop;
345
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 351 release_lockowner(lock_sop);
f1d110ca
BF		 352 }
353}
354
2283963f
BF		 355static void release_open_stateid(struct nfs4_stateid *stp)
356{
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
361}
362
f044ff83 363static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 364{
365 struct nfs4_stateid *stp;
366
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
f044ff83
BF		 369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 371 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
f1d110ca
BF		 375 }
376}
377
f044ff83 378static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 379{
f044ff83 380 unhash_openowner(sop);
f1d110ca
BF		 381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
383}
384
5282fd72
ME		 385static DEFINE_SPINLOCK(sessionid_lock);
386#define SESSION_HASH_SIZE 512
387static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
388
389static inline int
390hash_sessionid(struct nfs4_sessionid *sessionid)
391{
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
393
394 return sid->sequence % SESSION_HASH_SIZE;
395}
396
397static inline void
398dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
399{
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
402}
403
ec6b5d7b
AA		 404static void
405gen_sessionid(struct nfsd4_session *ses)
406{
407 struct nfs4_client *clp = ses->se_client;
408 struct nfsd4_sessionid *sid;
409
410 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
411 sid->clientid = clp->cl_clientid;
412 sid->sequence = current_sessionid++;
413 sid->reserved = 0;
414}
415
416/*
a649637c
AA		 417 * The protocol defines ca_maxresponssize_cached to include the size of
418 * the rpc header, but all we need to cache is the data starting after
419 * the end of the initial SEQUENCE operation--the rest we regenerate
420 * each time. Therefore we can advertise a ca_maxresponssize_cached
421 * value that is the number of bytes in our cache plus a few additional
422 * bytes. In order to stay on the safe side, and not promise more than
423 * we can cache, those additional bytes must be the minimum possible: 24
424 * bytes of rpc header (xid through accept state, with AUTH_NULL
425 * verifier), 12 for the compound header (with zero-length tag), and 44
426 * for the SEQUENCE op response:
427 */
428#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
429
430/*
431 * Give the client the number of ca_maxresponsesize_cached slots it
432 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
433 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
434 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 435 *
a649637c
AA		 436 * If we run out of reserved DRC memory we should (up to a point)
437 * re-negotiate active sessions and reduce their slot usage to make
438 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 439 */
a649637c 440static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 441{
a649637c 442 int mem, size = fchan->maxresp_cached;
ec6b5d7b 443
5d77ddfb
AA		 444 if (fchan->maxreqs < 1)
445 return nfserr_inval;
5d77ddfb 446
a649637c
AA		 447 if (size < NFSD_MIN_HDR_SEQ_SZ)
448 size = NFSD_MIN_HDR_SEQ_SZ;
449 size -= NFSD_MIN_HDR_SEQ_SZ;
450 if (size > NFSD_SLOT_CACHE_SIZE)
451 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 452
a649637c
AA		 453 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
454 mem = fchan->maxreqs * size;
455 if (mem > NFSD_MAX_MEM_PER_SESSION) {
456 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
457 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
458 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
459 mem = fchan->maxreqs * size;
460 }
ec6b5d7b 461
4bd9b0f4 462 spin_lock(&nfsd_drc_lock);
a649637c
AA		 463 /* bound the total session drc memory ussage */
464 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
465 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
466 mem = fchan->maxreqs * size;
467 }
0c193054 468 nfsd_drc_mem_used += mem;
4bd9b0f4 469 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 470
b101ebbc 471 if (fchan->maxreqs == 0)
bdac86e2 472 return nfserr_serverfault;
a649637c
AA		 473
474 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 475 return 0;
ec6b5d7b
AA		 476}
477
478/*
479 * fchan holds the client values on input, and the server values on output
ddc04fd4 480 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA		 481 */
482static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 483 struct nfsd4_channel_attrs *session_fchan,
484 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 485{
486 int status = 0;
ddc04fd4 487 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA		 488
489 /* headerpadsz set to zero in encode routine */
490
491 /* Use the client's max request and max response size if possible */
492 if (fchan->maxreq_sz > maxcount)
493 fchan->maxreq_sz = maxcount;
6c18ba9f 494 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 495
496 if (fchan->maxresp_sz > maxcount)
497 fchan->maxresp_sz = maxcount;
6c18ba9f 498 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 499
ec6b5d7b
AA		 500 /* Use the client's maxops if possible */
501 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
502 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 503 session_fchan->maxops = fchan->maxops;
ec6b5d7b 504
ec6b5d7b
AA		 505 /* FIXME: Error means no more DRC pages so the server should
506 * recover pages from existing sessions. For now fail session
507 * creation.
508 */
a649637c 509 status = set_forechannel_drc_size(fchan);
ec6b5d7b 510
a649637c 511 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 512 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 513
514 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 515 return status;
516}
517
557ce264
AA		 518static void
519free_session_slots(struct nfsd4_session *ses)
520{
521 int i;
522
523 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
524 kfree(ses->se_slots[i]);
525}
526
ec6b5d7b
AA		 527static int
528alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
529 struct nfsd4_create_session *cses)
530{
531 struct nfsd4_session *new, tmp;
557ce264
AA		 532 struct nfsd4_slot *sp;
533 int idx, slotsize, cachesize, i;
534 int status;
ec6b5d7b
AA		 535
536 memset(&tmp, 0, sizeof(tmp));
537
538 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 539 tmp.se_bchannel = cses->back_channel;
540 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
541 &cses->fore_channel);
ec6b5d7b
AA		 542 if (status)
543 goto out;
544
557ce264
AA		 545 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
546 + sizeof(struct nfsd4_session) > PAGE_SIZE);
547
548 status = nfserr_serverfault;
549 /* allocate struct nfsd4_session and slot table pointers in one piece */
550 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 551 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
552 if (!new)
553 goto out;
554
555 memcpy(new, &tmp, sizeof(*new));
556
557ce264
AA		 557 /* allocate each struct nfsd4_slot and data cache in one piece */
558 cachesize = new->se_fchannel.maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
559 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
560 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
561 if (!sp)
562 goto out_free;
563 new->se_slots[i] = sp;
564 }
565
ec6b5d7b
AA		 566 new->se_client = clp;
567 gen_sessionid(new);
568 idx = hash_sessionid(&new->se_sessionid);
569 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
570 NFS4_MAX_SESSIONID_LEN);
571
572 new->se_flags = cses->flags;
573 kref_init(&new->se_ref);
574 spin_lock(&sessionid_lock);
575 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
576 list_add(&new->se_perclnt, &clp->cl_sessions);
577 spin_unlock(&sessionid_lock);
578
579 status = nfs_ok;
580out:
581 return status;
557ce264
AA		 582out_free:
583 free_session_slots(new);
584 kfree(new);
585 goto out;
ec6b5d7b
AA		 586}
587
5282fd72
ME		 588/* caller must hold sessionid_lock */
589static struct nfsd4_session *
590find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
591{
592 struct nfsd4_session *elem;
593 int idx;
594
595 dump_sessionid(__func__, sessionid);
596 idx = hash_sessionid(sessionid);
597 dprintk("%s: idx is %d\n", __func__, idx);
598 /* Search in the appropriate list */
599 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
600 dump_sessionid("list traversal", &elem->se_sessionid);
601 if (!memcmp(elem->se_sessionid.data, sessionid->data,
602 NFS4_MAX_SESSIONID_LEN)) {
603 return elem;
604 }
605 }
606
607 dprintk("%s: session not found\n", __func__);
608 return NULL;
609}
610
611/* caller must hold sessionid_lock */
7116ed6b 612static void
5282fd72 613unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 614{
615 list_del(&ses->se_hash);
616 list_del(&ses->se_perclnt);
5282fd72
ME		 617}
618
619static void
620release_session(struct nfsd4_session *ses)
621{
622 spin_lock(&sessionid_lock);
623 unhash_session(ses);
624 spin_unlock(&sessionid_lock);
7116ed6b
AA		 625 nfsd4_put_session(ses);
626}
627
628void
629free_session(struct kref *kref)
630{
631 struct nfsd4_session *ses;
dd829c45 632 int mem;
7116ed6b
AA		 633
634 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 635 spin_lock(&nfsd_drc_lock);
dd829c45
BF		 636 mem = ses->se_fchannel.maxreqs
637 * (ses->se_fchannel.maxresp_cached - NFSD_MIN_HDR_SEQ_SZ);
638 nfsd_drc_mem_used -= mem;
be98d1bb 639 spin_unlock(&nfsd_drc_lock);
557ce264 640 free_session_slots(ses);
7116ed6b
AA		 641 kfree(ses);
642}
643
1da177e4
LT		 644static inline void
645renew_client(struct nfs4_client *clp)
646{
647 /*
648 * Move client to the end to the LRU list.
649 */
650 dprintk("renewing client (clientid %08x/%08x)\n",
651 clp->cl_clientid.cl_boot,
652 clp->cl_clientid.cl_id);
653 list_move_tail(&clp->cl_lru, &client_lru);
654 clp->cl_time = get_seconds();
655}
656
657/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
658static int
659STALE_CLIENTID(clientid_t *clid)
660{
661 if (clid->cl_boot == boot_time)
662 return 0;
60adfc50
AA		 663 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
664 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 665 return 1;
666}
667
668/*
669 * XXX Should we use a slab cache ?
670 * This type of memory management is somewhat inefficient, but we use it
671 * anyway since SETCLIENTID is not a common operation.
672 */
35bba9a3 673static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 674{
675 struct nfs4_client *clp;
676
35bba9a3
BF		 677 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
678 if (clp == NULL)
679 return NULL;
680 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
681 if (clp->cl_name.data == NULL) {
682 kfree(clp);
683 return NULL;
1da177e4 684 }
35bba9a3
BF		 685 memcpy(clp->cl_name.data, name.data, name.len);
686 clp->cl_name.len = name.len;
1da177e4
LT		 687 return clp;
688}
689
1b1a9b31
BF		 690static void
691shutdown_callback_client(struct nfs4_client *clp)
692{
c237dc03 693 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 694
1b1a9b31 695 if (clnt) {
404ec117
BF		 696 /*
697 * Callback threads take a reference on the client, so there
698 * should be no outstanding callbacks at this point.
699 */
c237dc03 700 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 701 rpc_shutdown_client(clnt);
702 }
703}
704
1da177e4
LT		 705static inline void
706free_client(struct nfs4_client *clp)
707{
1b1a9b31 708 shutdown_callback_client(clp);
38524ab3
AA		 709 if (clp->cl_cb_xprt)
710 svc_xprt_put(clp->cl_cb_xprt);
1da177e4
LT		 711 if (clp->cl_cred.cr_group_info)
712 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 713 kfree(clp->cl_principal);
1da177e4
LT		 714 kfree(clp->cl_name.data);
715 kfree(clp);
716}
717
718void
719put_nfs4_client(struct nfs4_client *clp)
720{
721 if (atomic_dec_and_test(&clp->cl_count))
722 free_client(clp);
723}
724
725static void
726expire_client(struct nfs4_client *clp)
727{
728 struct nfs4_stateowner *sop;
729 struct nfs4_delegation *dp;
1da177e4
LT		 730 struct list_head reaplist;
731
732 dprintk("NFSD: expire_client cl_count %d\n",
733 atomic_read(&clp->cl_count));
734
1da177e4
LT		 735 INIT_LIST_HEAD(&reaplist);
736 spin_lock(&recall_lock);
ea1da636
N		 737 while (!list_empty(&clp->cl_delegations)) {
738 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 739 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
740 dp->dl_flock);
ea1da636 741 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 742 list_move(&dp->dl_recall_lru, &reaplist);
743 }
744 spin_unlock(&recall_lock);
745 while (!list_empty(&reaplist)) {
746 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
747 list_del_init(&dp->dl_recall_lru);
748 unhash_delegation(dp);
749 }
750 list_del(&clp->cl_idhash);
751 list_del(&clp->cl_strhash);
752 list_del(&clp->cl_lru);
ea1da636
N		 753 while (!list_empty(&clp->cl_openowners)) {
754 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 755 release_openowner(sop);
1da177e4 756 }
c4bf7868
ME		 757 while (!list_empty(&clp->cl_sessions)) {
758 struct nfsd4_session *ses;
759 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
760 se_perclnt);
761 release_session(ses);
762 }
1da177e4
LT		 763 put_nfs4_client(clp);
764}
765
35bba9a3
BF		 766static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
767{
768 memcpy(target->cl_verifier.data, source->data,
769 sizeof(target->cl_verifier.data));
1da177e4
LT		 770}
771
35bba9a3
BF		 772static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
773{
1da177e4
LT		 774 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
775 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
776}
777
35bba9a3
BF		 778static void copy_cred(struct svc_cred *target, struct svc_cred *source)
779{
1da177e4
LT		 780 target->cr_uid = source->cr_uid;
781 target->cr_gid = source->cr_gid;
782 target->cr_group_info = source->cr_group_info;
783 get_group_info(target->cr_group_info);
784}
785
35bba9a3 786static int same_name(const char *n1, const char *n2)
599e0a22 787{
a55370a3 788 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 789}
790
791static int
599e0a22
BF		 792same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
793{
794 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 795}
796
797static int
599e0a22
BF		 798same_clid(clientid_t *cl1, clientid_t *cl2)
799{
800 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 801}
802
803/* XXX what about NGROUP */
804static int
599e0a22
BF		 805same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
806{
807 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 808}
809
5ec7b46c
BF		 810static void gen_clid(struct nfs4_client *clp)
811{
812 static u32 current_clientid = 1;
813
1da177e4
LT		 814 clp->cl_clientid.cl_boot = boot_time;
815 clp->cl_clientid.cl_id = current_clientid++;
816}
817
deda2faa
BF		 818static void gen_confirm(struct nfs4_client *clp)
819{
820 static u32 i;
821 u32 *p;
1da177e4 822
1da177e4 823 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 824 *p++ = get_seconds();
825 *p++ = i++;
1da177e4
LT		 826}
827
b09333c4
RL		 828static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
829 struct svc_rqst *rqstp, nfs4_verifier *verf)
830{
831 struct nfs4_client *clp;
832 struct sockaddr *sa = svc_addr(rqstp);
833 char *princ;
834
835 clp = alloc_client(name);
836 if (clp == NULL)
837 return NULL;
838
839 princ = svc_gss_principal(rqstp);
840 if (princ) {
841 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
842 if (clp->cl_principal == NULL) {
843 free_client(clp);
844 return NULL;
845 }
846 }
847
848 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
849 atomic_set(&clp->cl_count, 1);
850 atomic_set(&clp->cl_cb_conn.cb_set, 0);
851 INIT_LIST_HEAD(&clp->cl_idhash);
852 INIT_LIST_HEAD(&clp->cl_strhash);
853 INIT_LIST_HEAD(&clp->cl_openowners);
854 INIT_LIST_HEAD(&clp->cl_delegations);
855 INIT_LIST_HEAD(&clp->cl_sessions);
856 INIT_LIST_HEAD(&clp->cl_lru);
857 clear_bit(0, &clp->cl_cb_slot_busy);
858 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
859 copy_verf(clp, verf);
860 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
861 clp->cl_flavor = rqstp->rq_flavor;
862 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
863 gen_confirm(clp);
864
865 return clp;
866}
867
35bba9a3
BF		 868static int check_name(struct xdr_netobj name)
869{
1da177e4
LT		 870 if (name.len == 0)
871 return 0;
872 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 873 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 874 return 0;
875 }
876 return 1;
877}
878
fd39ca9a 879static void
1da177e4
LT		 880add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
881{
882 unsigned int idhashval;
883
884 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
885 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
886 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
887 list_add_tail(&clp->cl_lru, &client_lru);
888 clp->cl_time = get_seconds();
889}
890
fd39ca9a 891static void
1da177e4
LT		 892move_to_confirmed(struct nfs4_client *clp)
893{
894 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
895 unsigned int strhashval;
896
897 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
898 list_del_init(&clp->cl_strhash);
f116629d 899 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 900 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 901 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
902 renew_client(clp);
903}
904
905static struct nfs4_client *
906find_confirmed_client(clientid_t *clid)
907{
908 struct nfs4_client *clp;
909 unsigned int idhashval = clientid_hashval(clid->cl_id);
910
911 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 912 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 913 return clp;
914 }
915 return NULL;
916}
917
918static struct nfs4_client *
919find_unconfirmed_client(clientid_t *clid)
920{
921 struct nfs4_client *clp;
922 unsigned int idhashval = clientid_hashval(clid->cl_id);
923
924 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 925 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 926 return clp;
927 }
928 return NULL;
929}
930
a1bcecd2
AA		 931/*
932 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
933 * parameter. Matching is based on the fact the at least one of the
934 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
935 *
936 * FIXME: we need to unify the clientid namespaces for nfsv4.x
937 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
938 * and SET_CLIENTID{,_CONFIRM}
939 */
940static inline int
941match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
942{
943 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
944 return use_exchange_id == has_exchange_flags;
945}
946
28ce6054 947static struct nfs4_client *
a1bcecd2
AA		 948find_confirmed_client_by_str(const char *dname, unsigned int hashval,
949 bool use_exchange_id)
28ce6054
N		 950{
951 struct nfs4_client *clp;
952
953 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 954 if (same_name(clp->cl_recdir, dname) &&
955 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 956 return clp;
957 }
958 return NULL;
959}
960
961static struct nfs4_client *
a1bcecd2
AA		 962find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
963 bool use_exchange_id)
28ce6054
N		 964{
965 struct nfs4_client *clp;
966
967 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 968 if (same_name(clp->cl_recdir, dname) &&
969 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 970 return clp;
971 }
972 return NULL;
973}
974
fd39ca9a 975static void
fbf4665f 976gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 977{
c237dc03 978 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 979 unsigned short expected_family;
980
981 /* Currently, we only support tcp and tcp6 for the callback channel */
982 if (se->se_callback_netid_len == 3 &&
983 !memcmp(se->se_callback_netid_val, "tcp", 3))
984 expected_family = AF_INET;
985 else if (se->se_callback_netid_len == 4 &&
986 !memcmp(se->se_callback_netid_val, "tcp6", 4))
987 expected_family = AF_INET6;
988 else
1da177e4
LT		 989 goto out_err;
990
aa9a4ec7
JL		 991 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
992 se->se_callback_addr_len,
993 (struct sockaddr *) &cb->cb_addr,
994 sizeof(cb->cb_addr));
995
7077ecba 996 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 997 goto out_err;
aa9a4ec7 998
fbf4665f
JL		 999 if (cb->cb_addr.ss_family == AF_INET6)
1000 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
1001
ab52ae6d 1002 cb->cb_minorversion = 0;
1da177e4
LT		 1003 cb->cb_prog = se->se_callback_prog;
1004 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 1005 return;
1006out_err:
aa9a4ec7
JL		 1007 cb->cb_addr.ss_family = AF_UNSPEC;
1008 cb->cb_addrlen = 0;
849823c5 1009 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 1010 "will not receive delegations\n",
1011 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1012
1da177e4
LT		 1013 return;
1014}
1015
074fe897 1016/*
557ce264 1017 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 1018 */
074fe897
AA		 1019void
1020nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 1021{
557ce264
AA		 1022 struct nfsd4_slot *slot = resp->cstate.slot;
1023 unsigned int base;
074fe897 1024
557ce264 1025 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1026
557ce264
AA		 1027 slot->sl_opcnt = resp->opcnt;
1028 slot->sl_status = resp->cstate.status;
074fe897 1029
bf864a31 1030 if (nfsd4_not_cached(resp)) {
557ce264 1031 slot->sl_datalen = 0;
bf864a31 1032 return;
074fe897 1033 }
557ce264
AA		 1034 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1035 base = (char *)resp->cstate.datap -
1036 (char *)resp->xbuf->head[0].iov_base;
1037 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1038 slot->sl_datalen))
1039 WARN("%s: sessions DRC could not cache compound\n", __func__);
1040 return;
074fe897
AA		 1041}
1042
1043/*
abfabf8c
AA		 1044 * Encode the replay sequence operation from the slot values.
1045 * If cachethis is FALSE encode the uncached rep error on the next
1046 * operation which sets resp->p and increments resp->opcnt for
1047 * nfs4svc_encode_compoundres.
074fe897 1048 *
074fe897 1049 */
abfabf8c
AA		 1050static __be32
1051nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1052 struct nfsd4_compoundres *resp)
074fe897 1053{
abfabf8c
AA		 1054 struct nfsd4_op *op;
1055 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1056
abfabf8c 1057 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1058 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1059
abfabf8c
AA		 1060 /* Encode the replayed sequence operation */
1061 op = &args->ops[resp->opcnt - 1];
1062 nfsd4_encode_operation(resp, op);
bf864a31 1063
abfabf8c 1064 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1065 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1066 op = &args->ops[resp->opcnt++];
1067 op->status = nfserr_retry_uncached_rep;
1068 nfsd4_encode_operation(resp, op);
074fe897 1069 }
abfabf8c 1070 return op->status;
074fe897
AA		 1071}
1072
1073/*
557ce264
AA		 1074 * The sequence operation is not cached because we can use the slot and
1075 * session values.
074fe897
AA		 1076 */
1077__be32
bf864a31
AA		 1078nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1079 struct nfsd4_sequence *seq)
074fe897 1080{
557ce264 1081 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1082 __be32 status;
1083
557ce264 1084 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1085
abfabf8c
AA		 1086 /* Either returns 0 or nfserr_retry_uncached */
1087 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1088 if (status == nfserr_retry_uncached_rep)
1089 return status;
074fe897 1090
557ce264
AA		 1091 /* The sequence operation has been encoded, cstate->datap set. */
1092 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1093
557ce264
AA		 1094 resp->opcnt = slot->sl_opcnt;
1095 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1096 status = slot->sl_status;
074fe897
AA		 1097
1098 return status;
1099}
1100
0733d213
AA		 1101/*
1102 * Set the exchange_id flags returned by the server.
1103 */
1104static void
1105nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1106{
1107 /* pNFS is not supported */
1108 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1109
1110 /* Referrals are supported, Migration is not. */
1111 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1112
1113 /* set the wire flags to return to client. */
1114 clid->flags = new->cl_exchange_flags;
1115}
1116
069b6ad4
AA		 1117__be32
1118nfsd4_exchange_id(struct svc_rqst *rqstp,
1119 struct nfsd4_compound_state *cstate,
1120 struct nfsd4_exchange_id *exid)
1121{
0733d213
AA		 1122 struct nfs4_client *unconf, *conf, *new;
1123 int status;
1124 unsigned int strhashval;
1125 char dname[HEXDIR_LEN];
363168b4 1126 char addr_str[INET6_ADDRSTRLEN];
0733d213 1127 nfs4_verifier verf = exid->verifier;
363168b4 1128 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1129
363168b4 1130 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1131 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1132 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1133 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1134 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1135
1136 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1137 return nfserr_inval;
1138
1139 /* Currently only support SP4_NONE */
1140 switch (exid->spa_how) {
1141 case SP4_NONE:
1142 break;
1143 case SP4_SSV:
1144 return nfserr_encr_alg_unsupp;
1145 default:
1146 BUG(); /* checked by xdr code */
1147 case SP4_MACH_CRED:
1148 return nfserr_serverfault; /* no excuse :-/ */
1149 }
1150
1151 status = nfs4_make_rec_clidname(dname, &exid->clname);
1152
1153 if (status)
1154 goto error;
1155
1156 strhashval = clientstr_hashval(dname);
1157
1158 nfs4_lock_state();
1159 status = nfs_ok;
1160
a1bcecd2 1161 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1162 if (conf) {
1163 if (!same_verf(&verf, &conf->cl_verifier)) {
1164 /* 18.35.4 case 8 */
1165 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1166 status = nfserr_not_same;
1167 goto out;
1168 }
1169 /* Client reboot: destroy old state */
1170 expire_client(conf);
1171 goto out_new;
1172 }
1173 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1174 /* 18.35.4 case 9 */
1175 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1176 status = nfserr_perm;
1177 goto out;
1178 }
1179 expire_client(conf);
1180 goto out_new;
1181 }
0733d213
AA		 1182 /*
1183 * Set bit when the owner id and verifier map to an already
1184 * confirmed client id (18.35.3).
1185 */
1186 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1187
1188 /*
1189 * Falling into 18.35.4 case 2, possible router replay.
1190 * Leave confirmed record intact and return same result.
1191 */
1192 copy_verf(conf, &verf);
1193 new = conf;
1194 goto out_copy;
6ddbbbfe
MS		 1195 }
1196
1197 /* 18.35.4 case 7 */
1198 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1199 status = nfserr_noent;
1200 goto out;
0733d213
AA		 1201 }
1202
a1bcecd2 1203 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1204 if (unconf) {
1205 /*
1206 * Possible retry or client restart. Per 18.35.4 case 4,
1207 * a new unconfirmed record should be generated regardless
1208 * of whether any properties have changed.
1209 */
1210 expire_client(unconf);
1211 }
1212
1213out_new:
1214 /* Normal case */
b09333c4 1215 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1216 if (new == NULL) {
bdac86e2 1217 status = nfserr_serverfault;
0733d213
AA		 1218 goto out;
1219 }
1220
0733d213 1221 gen_clid(new);
0733d213
AA		 1222 add_to_unconfirmed(new, strhashval);
1223out_copy:
1224 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1225 exid->clientid.cl_id = new->cl_clientid.cl_id;
1226
38eb76a5 1227 exid->seqid = 1;
0733d213
AA		 1228 nfsd4_set_ex_flags(new, exid);
1229
1230 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1231 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1232 status = nfs_ok;
1233
1234out:
1235 nfs4_unlock_state();
1236error:
1237 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1238 return status;
069b6ad4
AA		 1239}
1240
b85d4c01 1241static int
88e588d5 1242check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1243{
88e588d5
AA		 1244 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1245 slot_seqid);
b85d4c01
BH		 1246
1247 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1248 if (slot_inuse) {
1249 if (seqid == slot_seqid)
b85d4c01
BH		 1250 return nfserr_jukebox;
1251 else
1252 return nfserr_seq_misordered;
1253 }
1254 /* Normal */
88e588d5 1255 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1256 return nfs_ok;
1257 /* Replay */
88e588d5 1258 if (seqid == slot_seqid)
b85d4c01
BH		 1259 return nfserr_replay_cache;
1260 /* Wraparound */
88e588d5 1261 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1262 return nfs_ok;
1263 /* Misordered replay or misordered new request */
1264 return nfserr_seq_misordered;
1265}
1266
49557cc7
AA		 1267/*
1268 * Cache the create session result into the create session single DRC
1269 * slot cache by saving the xdr structure. sl_seqid has been set.
1270 * Do this for solo or embedded create session operations.
1271 */
1272static void
1273nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1274 struct nfsd4_clid_slot *slot, int nfserr)
1275{
1276 slot->sl_status = nfserr;
1277 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1278}
1279
1280static __be32
1281nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1282 struct nfsd4_clid_slot *slot)
1283{
1284 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1285 return slot->sl_status;
1286}
1287
069b6ad4
AA		 1288__be32
1289nfsd4_create_session(struct svc_rqst *rqstp,
1290 struct nfsd4_compound_state *cstate,
1291 struct nfsd4_create_session *cr_ses)
1292{
363168b4 1293 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1294 struct nfs4_client *conf, *unconf;
49557cc7 1295 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1296 int status = 0;
1297
1298 nfs4_lock_state();
1299 unconf = find_unconfirmed_client(&cr_ses->clientid);
1300 conf = find_confirmed_client(&cr_ses->clientid);
1301
1302 if (conf) {
49557cc7
AA		 1303 cs_slot = &conf->cl_cs_slot;
1304 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1305 if (status == nfserr_replay_cache) {
ec6b5d7b 1306 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1307 cs_slot->sl_seqid);
38eb76a5 1308 /* Return the cached reply status */
49557cc7 1309 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1310 goto out;
49557cc7 1311 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1312 status = nfserr_seq_misordered;
1313 dprintk("Sequence misordered!\n");
1314 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1315 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1316 goto out;
1317 }
49557cc7 1318 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1319 } else if (unconf) {
1320 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1321 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1322 status = nfserr_clid_inuse;
1323 goto out;
1324 }
1325
49557cc7
AA		 1326 cs_slot = &unconf->cl_cs_slot;
1327 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1328 if (status) {
1329 /* an unconfirmed replay returns misordered */
ec6b5d7b 1330 status = nfserr_seq_misordered;
49557cc7 1331 goto out_cache;
ec6b5d7b
AA		 1332 }
1333
49557cc7 1334 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1335 move_to_confirmed(unconf);
1336
1337 /*
1338 * We do not support RDMA or persistent sessions
1339 */
1340 cr_ses->flags &= ~SESSION4_PERSIST;
1341 cr_ses->flags &= ~SESSION4_RDMA;
1342
38524ab3
AA		 1343 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1344 unconf->cl_cb_xprt = rqstp->rq_xprt;
1345 svc_xprt_get(unconf->cl_cb_xprt);
1346 rpc_copy_addr(
1347 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1348 sa);
1349 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1350 unconf->cl_cb_conn.cb_minorversion =
1351 cstate->minorversion;
1352 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1353 unconf->cl_cb_seq_nr = 1;
38524ab3
AA		 1354 nfsd4_probe_callback(unconf);
1355 }
ec6b5d7b
AA		 1356 conf = unconf;
1357 } else {
1358 status = nfserr_stale_clientid;
1359 goto out;
1360 }
1361
1362 status = alloc_init_session(rqstp, conf, cr_ses);
1363 if (status)
1364 goto out;
1365
ec6b5d7b
AA		 1366 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1367 NFS4_MAX_SESSIONID_LEN);
49557cc7 1368 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1369
49557cc7
AA		 1370out_cache:
1371 /* cache solo and embedded create sessions under the state lock */
1372 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1373out:
1374 nfs4_unlock_state();
1375 dprintk("%s returns %d\n", __func__, ntohl(status));
1376 return status;
069b6ad4
AA		 1377}
1378
1379__be32
1380nfsd4_destroy_session(struct svc_rqst *r,
1381 struct nfsd4_compound_state *cstate,
1382 struct nfsd4_destroy_session *sessionid)
1383{
e10e0cfc
BH		 1384 struct nfsd4_session *ses;
1385 u32 status = nfserr_badsession;
1386
1387 /* Notes:
1388 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1389 * - Should we return nfserr_back_chan_busy if waiting for
1390 * callbacks on to-be-destroyed session?
1391 * - Do we need to clear any callback info from previous session?
1392 */
1393
1394 dump_sessionid(__func__, &sessionid->sessionid);
1395 spin_lock(&sessionid_lock);
1396 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1397 if (!ses) {
1398 spin_unlock(&sessionid_lock);
1399 goto out;
1400 }
1401
1402 unhash_session(ses);
1403 spin_unlock(&sessionid_lock);
1404
1405 /* wait for callbacks */
1406 shutdown_callback_client(ses->se_client);
1407 nfsd4_put_session(ses);
1408 status = nfs_ok;
1409out:
1410 dprintk("%s returns %d\n", __func__, ntohl(status));
1411 return status;
069b6ad4
AA		 1412}
1413
1414__be32
b85d4c01 1415nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1416 struct nfsd4_compound_state *cstate,
1417 struct nfsd4_sequence *seq)
1418{
f9bb94c4 1419 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1420 struct nfsd4_session *session;
1421 struct nfsd4_slot *slot;
1422 int status;
1423
f9bb94c4
AA		 1424 if (resp->opcnt != 1)
1425 return nfserr_sequence_pos;
1426
b85d4c01
BH		 1427 spin_lock(&sessionid_lock);
1428 status = nfserr_badsession;
1429 session = find_in_sessionid_hashtbl(&seq->sessionid);
1430 if (!session)
1431 goto out;
1432
1433 status = nfserr_badslot;
6c18ba9f 1434 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1435 goto out;
1436
557ce264 1437 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1438 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1439
a8dfdaeb
AA		 1440 /* We do not negotiate the number of slots yet, so set the
1441 * maxslots to the session maxreqs which is used to encode
1442 * sr_highest_slotid and the sr_target_slot id to maxslots */
1443 seq->maxslots = session->se_fchannel.maxreqs;
1444
88e588d5 1445 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1446 if (status == nfserr_replay_cache) {
1447 cstate->slot = slot;
1448 cstate->session = session;
da3846a2 1449 /* Return the cached reply status and set cstate->status
557ce264 1450 * for nfsd4_proc_compound processing */
bf864a31 1451 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1452 cstate->status = nfserr_replay_cache;
aaf84eb9 1453 goto out;
b85d4c01
BH		 1454 }
1455 if (status)
1456 goto out;
1457
1458 /* Success! bump slot seqid */
1459 slot->sl_inuse = true;
1460 slot->sl_seqid = seq->seqid;
557ce264 1461 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1462
1463 cstate->slot = slot;
1464 cstate->session = session;
1465
aaf84eb9 1466 /* Hold a session reference until done processing the compound:
b85d4c01
BH		 1467 * nfsd4_put_session called only if the cstate slot is set.
1468 */
b85d4c01
BH		 1469 nfsd4_get_session(session);
1470out:
1471 spin_unlock(&sessionid_lock);
aaf84eb9
BH		 1472 /* Renew the clientid on success and on replay */
1473 if (cstate->session) {
1474 nfs4_lock_state();
1475 renew_client(session->se_client);
1476 nfs4_unlock_state();
1477 }
b85d4c01
BH		 1478 dprintk("%s: return %d\n", __func__, ntohl(status));
1479 return status;
069b6ad4
AA		 1480}
1481
b37ad28b 1482__be32
b591480b
BF		 1483nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1484 struct nfsd4_setclientid *setclid)
1da177e4 1485{
363168b4 1486 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1487 struct xdr_netobj clname = {
1488 .len = setclid->se_namelen,
1489 .data = setclid->se_name,
1490 };
1491 nfs4_verifier clverifier = setclid->se_verf;
1492 unsigned int strhashval;
28ce6054 1493 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1494 __be32 status;
a55370a3 1495 char dname[HEXDIR_LEN];
1da177e4 1496
1da177e4 1497 if (!check_name(clname))
73aea4ec 1498 return nfserr_inval;
1da177e4 1499
a55370a3
N		 1500 status = nfs4_make_rec_clidname(dname, &clname);
1501 if (status)
73aea4ec 1502 return status;
a55370a3 1503
1da177e4
LT		 1504 /*
1505 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1506 * We get here on a DRC miss.
1507 */
1508
a55370a3 1509 strhashval = clientstr_hashval(dname);
1da177e4 1510
1da177e4 1511 nfs4_lock_state();
a1bcecd2 1512 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1513 if (conf) {
a186e767 1514 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1515 status = nfserr_clid_inuse;
026722c2 1516 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1517 char addr_str[INET6_ADDRSTRLEN];
1518 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1519 sizeof(addr_str));
1520 dprintk("NFSD: setclientid: string in use by client "
1521 "at %s\n", addr_str);
1da177e4
LT		 1522 goto out;
1523 }
1da177e4 1524 }
a186e767
BF		 1525 /*
1526 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1527 * has a description of SETCLIENTID request processing consisting
1528 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1529 */
a1bcecd2 1530 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1531 status = nfserr_resource;
1532 if (!conf) {
a186e767
BF		 1533 /*
1534 * RFC 3530 14.2.33 CASE 4:
1535 * placed first, because it is the normal case
1da177e4
LT		 1536 */
1537 if (unconf)
1538 expire_client(unconf);
b09333c4 1539 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1540 if (new == NULL)
1da177e4 1541 goto out;
1da177e4 1542 gen_clid(new);
599e0a22 1543 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1544 /*
a186e767
BF		 1545 * RFC 3530 14.2.33 CASE 1:
1546 * probable callback update
1da177e4 1547 */
31f4a6c1
N		 1548 if (unconf) {
1549 /* Note this is removing unconfirmed {*x***},
1550 * which is stronger than RFC recommended {vxc**}.
1551 * This has the advantage that there is at most
1552 * one {*x***} in either list at any time.
1553 */
1554 expire_client(unconf);
1da177e4 1555 }
b09333c4 1556 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1557 if (new == NULL)
1da177e4 1558 goto out;
1da177e4 1559 copy_clid(new, conf);
1da177e4
LT		 1560 } else if (!unconf) {
1561 /*
a186e767
BF		 1562 * RFC 3530 14.2.33 CASE 2:
1563 * probable client reboot; state will be removed if
1564 * confirmed.
1da177e4 1565 */
b09333c4 1566 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1567 if (new == NULL)
1da177e4 1568 goto out;
1da177e4 1569 gen_clid(new);
49ba8781 1570 } else {
a186e767
BF		 1571 /*
1572 * RFC 3530 14.2.33 CASE 3:
1573 * probable client reboot; state will be removed if
1574 * confirmed.
1da177e4
LT		 1575 */
1576 expire_client(unconf);
b09333c4 1577 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1578 if (new == NULL)
1da177e4 1579 goto out;
1da177e4 1580 gen_clid(new);
1da177e4 1581 }
fbf4665f 1582 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1583 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1584 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1585 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1586 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1587 status = nfs_ok;
1588out:
1589 nfs4_unlock_state();
1590 return status;
1591}
1592
1593
1594/*
a186e767
BF		 1595 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1596 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1597 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1598 */
b37ad28b 1599__be32
b591480b
BF		 1600nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1601 struct nfsd4_compound_state *cstate,
1602 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1603{
363168b4 1604 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1605 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1606 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1607 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1608 __be32 status;
1da177e4
LT		 1609
1610 if (STALE_CLIENTID(clid))
1611 return nfserr_stale_clientid;
1612 /*
1613 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1614 * We get here on a DRC miss.
1615 */
1616
1617 nfs4_lock_state();
21ab45a4
N		 1618
1619 conf = find_confirmed_client(clid);
1620 unconf = find_unconfirmed_client(clid);
1621
1622 status = nfserr_clid_inuse;
363168b4 1623 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1624 goto out;
363168b4 1625 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1626 goto out;
1627
a186e767
BF		 1628 /*
1629 * section 14.2.34 of RFC 3530 has a description of
1630 * SETCLIENTID_CONFIRM request processing consisting
1631 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1632 */
366e0c1d 1633 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1634 /*
1635 * RFC 3530 14.2.34 CASE 1:
1636 * callback update
1637 */
599e0a22 1638 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1639 status = nfserr_clid_inuse;
1640 else {
1a69c179
N		 1641 /* XXX: We just turn off callbacks until we can handle
1642 * change request correctly. */
c237dc03 1643 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1644 expire_client(unconf);
1da177e4 1645 status = nfs_ok;
1a69c179 1646
1da177e4 1647 }
f3aba4e5 1648 } else if (conf && !unconf) {
a186e767
BF		 1649 /*
1650 * RFC 3530 14.2.34 CASE 2:
1651 * probable retransmitted request; play it safe and
1652 * do nothing.
7c79f737 1653 */
599e0a22 1654 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1655 status = nfserr_clid_inuse;
21ab45a4 1656 else
1da177e4 1657 status = nfs_ok;
7c79f737 1658 } else if (!conf && unconf
599e0a22 1659 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1660 /*
1661 * RFC 3530 14.2.34 CASE 3:
1662 * Normal case; new or rebooted client:
7c79f737 1663 */
599e0a22 1664 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1665 status = nfserr_clid_inuse;
1666 } else {
1a69c179
N		 1667 unsigned int hash =
1668 clientstr_hashval(unconf->cl_recdir);
1669 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1670 hash, false);
1a69c179 1671 if (conf) {
c7b9a459 1672 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1673 expire_client(conf);
1674 }
1da177e4 1675 move_to_confirmed(unconf);
21ab45a4 1676 conf = unconf;
46f8a64b 1677 nfsd4_probe_callback(conf);
1a69c179 1678 status = nfs_ok;
1da177e4 1679 }
599e0a22
BF		 1680 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1681 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1682 &confirm)))) {
a186e767
BF		 1683 /*
1684 * RFC 3530 14.2.34 CASE 4:
1685 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1686 */
1da177e4 1687 status = nfserr_stale_clientid;
7c79f737 1688 } else {
08e8987c
N		 1689 /* check that we have hit one of the cases...*/
1690 status = nfserr_clid_inuse;
1691 }
1da177e4 1692out:
1da177e4
LT		 1693 nfs4_unlock_state();
1694 return status;
1695}
1696
1da177e4
LT		 1697/* OPEN Share state helper functions */
1698static inline struct nfs4_file *
1699alloc_init_file(struct inode *ino)
1700{
1701 struct nfs4_file *fp;
1702 unsigned int hashval = file_hashval(ino);
1703
e60d4398
N		 1704 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1705 if (fp) {
8b671b80 1706 atomic_set(&fp->fi_ref, 1);
1da177e4 1707 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1708 INIT_LIST_HEAD(&fp->fi_stateids);
1709 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1710 spin_lock(&recall_lock);
1da177e4 1711 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1712 spin_unlock(&recall_lock);
1da177e4
LT		 1713 fp->fi_inode = igrab(ino);
1714 fp->fi_id = current_fileid++;
47f9940c 1715 fp->fi_had_conflict = false;
1da177e4
LT		 1716 return fp;
1717 }
1718 return NULL;
1719}
1720
e60d4398 1721static void
e18b890b 1722nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1723{
e60d4398
N		 1724 if (*slab == NULL)
1725 return;
1a1d92c1 1726 kmem_cache_destroy(*slab);
e60d4398 1727 *slab = NULL;
1da177e4
LT		 1728}
1729
e8ff2a84 1730void
1da177e4
LT		 1731nfsd4_free_slabs(void)
1732{
e60d4398
N		 1733 nfsd4_free_slab(&stateowner_slab);
1734 nfsd4_free_slab(&file_slab);
5ac049ac 1735 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1736 nfsd4_free_slab(&deleg_slab);
e60d4398 1737}
1da177e4 1738
e60d4398
N		 1739static int
1740nfsd4_init_slabs(void)
1741{
1742 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1743 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1744 if (stateowner_slab == NULL)
1745 goto out_nomem;
1746 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1747 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1748 if (file_slab == NULL)
1749 goto out_nomem;
5ac049ac 1750 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1751 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1752 if (stateid_slab == NULL)
1753 goto out_nomem;
5b2d21c1 1754 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1755 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1756 if (deleg_slab == NULL)
1757 goto out_nomem;
e60d4398
N		 1758 return 0;
1759out_nomem:
1760 nfsd4_free_slabs();
1761 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1762 return -ENOMEM;
1da177e4
LT		 1763}
1764
1765void
1766nfs4_free_stateowner(struct kref *kref)
1767{
1768 struct nfs4_stateowner *sop =
1769 container_of(kref, struct nfs4_stateowner, so_ref);
1770 kfree(sop->so_owner.data);
1771 kmem_cache_free(stateowner_slab, sop);
1772}
1773
1774static inline struct nfs4_stateowner *
1775alloc_stateowner(struct xdr_netobj *owner)
1776{
1777 struct nfs4_stateowner *sop;
1778
1779 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1780 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1781 memcpy(sop->so_owner.data, owner->data, owner->len);
1782 sop->so_owner.len = owner->len;
1783 kref_init(&sop->so_ref);
1784 return sop;
1785 }
1786 kmem_cache_free(stateowner_slab, sop);
1787 }
1788 return NULL;
1789}
1790
1791static struct nfs4_stateowner *
1792alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1793 struct nfs4_stateowner *sop;
1794 struct nfs4_replay *rp;
1795 unsigned int idhashval;
1796
1797 if (!(sop = alloc_stateowner(&open->op_owner)))
1798 return NULL;
1799 idhashval = ownerid_hashval(current_ownerid);
1800 INIT_LIST_HEAD(&sop->so_idhash);
1801 INIT_LIST_HEAD(&sop->so_strhash);
1802 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1803 INIT_LIST_HEAD(&sop->so_stateids);
1804 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1805 INIT_LIST_HEAD(&sop->so_close_lru);
1806 sop->so_time = 0;
1807 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1808 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1809 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1810 sop->so_is_open_owner = 1;
1811 sop->so_id = current_ownerid++;
1812 sop->so_client = clp;
1813 sop->so_seqid = open->op_seqid;
1814 sop->so_confirmed = 0;
1815 rp = &sop->so_replay;
de1ae286 1816 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1817 rp->rp_buflen = 0;
1818 rp->rp_buf = rp->rp_ibuf;
1819 return sop;
1820}
1821
1da177e4
LT		 1822static inline void
1823init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1824 struct nfs4_stateowner *sop = open->op_stateowner;
1825 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1826
1827 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1828 INIT_LIST_HEAD(&stp->st_perstateowner);
1829 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1830 INIT_LIST_HEAD(&stp->st_perfile);
1831 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1832 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1833 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1834 stp->st_stateowner = sop;
13cd2184 1835 get_nfs4_file(fp);
1da177e4 1836 stp->st_file = fp;
78155ed7 1837 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1838 stp->st_stateid.si_stateownerid = sop->so_id;
1839 stp->st_stateid.si_fileid = fp->fi_id;
1840 stp->st_stateid.si_generation = 0;
1841 stp->st_access_bmap = 0;
1842 stp->st_deny_bmap = 0;
84459a11
AA		 1843 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1844 &stp->st_access_bmap);
1da177e4 1845 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1846 stp->st_openstp = NULL;
1da177e4
LT		 1847}
1848
fd39ca9a 1849static void
1da177e4
LT		 1850move_to_close_lru(struct nfs4_stateowner *sop)
1851{
1852 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1853
358dd55a 1854 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1855 sop->so_time = get_seconds();
1856}
1857
1da177e4 1858static int
599e0a22
BF		 1859same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1860 clientid_t *clid)
1861{
1862 return (sop->so_owner.len == owner->len) &&
1863 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1864 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1865}
1866
1867static struct nfs4_stateowner *
1868find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1869{
1870 struct nfs4_stateowner *so = NULL;
1871
1872 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1873 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1874 return so;
1875 }
1876 return NULL;
1877}
1878
1879/* search file_hashtbl[] for file */
1880static struct nfs4_file *
1881find_file(struct inode *ino)
1882{
1883 unsigned int hashval = file_hashval(ino);
1884 struct nfs4_file *fp;
1885
8b671b80 1886 spin_lock(&recall_lock);
1da177e4 1887 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1888 if (fp->fi_inode == ino) {
1889 get_nfs4_file(fp);
8b671b80 1890 spin_unlock(&recall_lock);
1da177e4 1891 return fp;
13cd2184 1892 }
1da177e4 1893 }
8b671b80 1894 spin_unlock(&recall_lock);
1da177e4
LT		 1895 return NULL;
1896}
1897
d87a8ade 1898static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1899{
d87a8ade 1900 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1901 return 0;
d87a8ade
AA		 1902 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1903 return 0;
1904 x &= ~NFS4_SHARE_ACCESS_MASK;
1905 if (minorversion && x) {
1906 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1907 return 0;
1908 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1909 return 0;
1910 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1911 }
1912 if (x)
8838dc43
BF		 1913 return 0;
1914 return 1;
ba5a6a19
BF		 1915}
1916
8838dc43 1917static inline int deny_valid(u32 x)
ba5a6a19 1918{
8838dc43
BF		 1919 /* Note: unlike access bits, deny bits may be zero. */
1920 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1921}
1da177e4 1922
4f83aa30
BF		 1923/*
1924 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1925 * st_{access,deny}_bmap field of the stateid, in order to track not
1926 * only what share bits are currently in force, but also what
1927 * combinations of share bits previous opens have used. This allows us
1928 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1929 * return an error if the client attempt to downgrade to a combination
1930 * of share bits not explicable by closing some of its previous opens.
1931 *
1932 * XXX: This enforcement is actually incomplete, since we don't keep
1933 * track of access/deny bit combinations; so, e.g., we allow:
1934 *
1935 * OPEN allow read, deny write
1936 * OPEN allow both, deny none
1937 * DOWNGRADE allow read, deny none
1938 *
1939 * which we should reject.
1940 */
fd39ca9a 1941static void
1da177e4
LT		 1942set_access(unsigned int *access, unsigned long bmap) {
1943 int i;
1944
1945 *access = 0;
1946 for (i = 1; i < 4; i++) {
1947 if (test_bit(i, &bmap))
1948 *access |= i;
1949 }
1950}
1951
fd39ca9a 1952static void
1da177e4
LT		 1953set_deny(unsigned int *deny, unsigned long bmap) {
1954 int i;
1955
1956 *deny = 0;
1957 for (i = 0; i < 4; i++) {
1958 if (test_bit(i, &bmap))
1959 *deny |= i ;
1960 }
1961}
1962
1963static int
1964test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1965 unsigned int access, deny;
1966
1967 set_access(&access, stp->st_access_bmap);
1968 set_deny(&deny, stp->st_deny_bmap);
1969 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1970 return 0;
1971 return 1;
1972}
1973
1974/*
1975 * Called to check deny when READ with all zero stateid or
1976 * WRITE with all zero or all one stateid
1977 */
b37ad28b 1978static __be32
1da177e4
LT		 1979nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1980{
1981 struct inode *ino = current_fh->fh_dentry->d_inode;
1982 struct nfs4_file *fp;
1983 struct nfs4_stateid *stp;
b37ad28b 1984 __be32 ret;
1da177e4
LT		 1985
1986 dprintk("NFSD: nfs4_share_conflict\n");
1987
1988 fp = find_file(ino);
13cd2184
N		 1989 if (!fp)
1990 return nfs_ok;
b700949b 1991 ret = nfserr_locked;
1da177e4 1992 /* Search for conflicting share reservations */
13cd2184
N		 1993 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1994 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1995 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1996 goto out;
1da177e4 1997 }
13cd2184
N		 1998 ret = nfs_ok;
1999out:
2000 put_nfs4_file(fp);
2001 return ret;
1da177e4
LT		 2002}
2003
2004static inline void
2005nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2006{
2007 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2008 drop_file_write_access(filp);
1da177e4
LT		 2009 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2010 }
2011}
2012
1da177e4
LT		 2013/*
2014 * Spawn a thread to perform a recall on the delegation represented
2015 * by the lease (file_lock)
2016 *
2017 * Called from break_lease() with lock_kernel() held.
2018 * Note: we assume break_lease will only call this *once* for any given
2019 * lease.
2020 */
2021static
2022void nfsd_break_deleg_cb(struct file_lock *fl)
2023{
63e4863f 2024 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2025
2026 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2027 if (!dp)
2028 return;
2029
2030 /* We're assuming the state code never drops its reference
2031 * without first removing the lease. Since we're in this lease
2032 * callback (and since the lease code is serialized by the kernel
2033 * lock) we know the server hasn't removed the lease yet, we know
2034 * it's safe to take a reference: */
2035 atomic_inc(&dp->dl_count);
cfdcad4d 2036 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2037
2038 spin_lock(&recall_lock);
2039 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2040 spin_unlock(&recall_lock);
2041
2042 /* only place dl_time is set. protected by lock_kernel*/
2043 dp->dl_time = get_seconds();
2044
0272e1fd
BF		 2045 /*
2046 * We don't want the locks code to timeout the lease for us;
2047 * we'll remove it ourself if the delegation isn't returned
2048 * in time.
2049 */
2050 fl->fl_break_time = 0;
1da177e4 2051
63e4863f
BF		 2052 dp->dl_file->fi_had_conflict = true;
2053 nfsd4_cb_recall(dp);
1da177e4
LT		 2054}
2055
2056/*
2057 * The file_lock is being reapd.
2058 *
2059 * Called by locks_free_lock() with lock_kernel() held.
2060 */
2061static
2062void nfsd_release_deleg_cb(struct file_lock *fl)
2063{
2064 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2065
2066 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2067
2068 if (!(fl->fl_flags & FL_LEASE) || !dp)
2069 return;
2070 dp->dl_flock = NULL;
2071}
2072
2073/*
2074 * Set the delegation file_lock back pointer.
2075 *
a9933cea 2076 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2077 */
2078static
2079void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2080{
2081 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2082
2083 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2084 if (!dp)
2085 return;
2086 dp->dl_flock = new;
2087}
2088
2089/*
a9933cea 2090 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2091 */
2092static
2093int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2094{
2095 struct nfs4_delegation *onlistd =
2096 (struct nfs4_delegation *)onlist->fl_owner;
2097 struct nfs4_delegation *tryd =
2098 (struct nfs4_delegation *)try->fl_owner;
2099
2100 if (onlist->fl_lmops != try->fl_lmops)
2101 return 0;
2102
2103 return onlistd->dl_client == tryd->dl_client;
2104}
2105
2106
2107static
2108int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2109{
2110 if (arg & F_UNLCK)
2111 return lease_modify(onlist, arg);
2112 else
2113 return -EAGAIN;
2114}
2115
7b021967 2116static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2117 .fl_break = nfsd_break_deleg_cb,
2118 .fl_release_private = nfsd_release_deleg_cb,
2119 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2120 .fl_mylease = nfsd_same_client_deleg_cb,
2121 .fl_change = nfsd_change_deleg_cb,
2122};
2123
2124
b37ad28b 2125__be32
6668958f
AA		 2126nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2127 struct nfsd4_open *open)
1da177e4 2128{
1da177e4
LT		 2129 clientid_t *clientid = &open->op_clientid;
2130 struct nfs4_client *clp = NULL;
2131 unsigned int strhashval;
2132 struct nfs4_stateowner *sop = NULL;
2133
1da177e4 2134 if (!check_name(open->op_owner))
0f442aa2 2135 return nfserr_inval;
1da177e4
LT		 2136
2137 if (STALE_CLIENTID(&open->op_clientid))
2138 return nfserr_stale_clientid;
2139
2140 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2141 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2142 open->op_stateowner = sop;
2143 if (!sop) {
2144 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2145 clp = find_confirmed_client(clientid);
2146 if (clp == NULL)
0f442aa2
BF		 2147 return nfserr_expired;
2148 goto renew;
1da177e4 2149 }
6668958f
AA		 2150 /* When sessions are used, skip open sequenceid processing */
2151 if (nfsd4_has_session(cstate))
2152 goto renew;
0f442aa2
BF		 2153 if (!sop->so_confirmed) {
2154 /* Replace unconfirmed owners without checking for replay. */
2155 clp = sop->so_client;
f044ff83 2156 release_openowner(sop);
0f442aa2
BF		 2157 open->op_stateowner = NULL;
2158 goto renew;
2159 }
2160 if (open->op_seqid == sop->so_seqid - 1) {
2161 if (sop->so_replay.rp_buflen)
a90b061c 2162 return nfserr_replay_me;
0f442aa2
BF		 2163 /* The original OPEN failed so spectacularly
2164 * that we don't even have replay data saved!
2165 * Therefore, we have no choice but to continue
2166 * processing this OPEN; presumably, we'll
2167 * fail again for the same reason.
2168 */
2169 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2170 goto renew;
2171 }
2172 if (open->op_seqid != sop->so_seqid)
2173 return nfserr_bad_seqid;
1da177e4 2174renew:
0f442aa2
BF		 2175 if (open->op_stateowner == NULL) {
2176 sop = alloc_init_open_stateowner(strhashval, clp, open);
2177 if (sop == NULL)
2178 return nfserr_resource;
2179 open->op_stateowner = sop;
2180 }
2181 list_del_init(&sop->so_close_lru);
1da177e4 2182 renew_client(sop->so_client);
0f442aa2 2183 return nfs_ok;
1da177e4
LT		 2184}
2185
b37ad28b 2186static inline __be32
4a6e43e6
N		 2187nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2188{
2189 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2190 return nfserr_openmode;
2191 else
2192 return nfs_ok;
2193}
2194
52f4fb43
N		 2195static struct nfs4_delegation *
2196find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2197{
2198 struct nfs4_delegation *dp;
2199
ea1da636 2200 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2201 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2202 return dp;
2203 }
2204 return NULL;
2205}
2206
b37ad28b 2207static __be32
567d9829
N		 2208nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2209 struct nfs4_delegation **dp)
2210{
2211 int flags;
b37ad28b 2212 __be32 status = nfserr_bad_stateid;
567d9829
N		 2213
2214 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2215 if (*dp == NULL)
c44c5eeb 2216 goto out;
567d9829
N		 2217 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2218 RD_STATE : WR_STATE;
2219 status = nfs4_check_delegmode(*dp, flags);
2220 if (status)
2221 *dp = NULL;
c44c5eeb
N		 2222out:
2223 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2224 return nfs_ok;
2225 if (status)
2226 return status;
2227 open->op_stateowner->so_confirmed = 1;
2228 return nfs_ok;
567d9829
N		 2229}
2230
b37ad28b 2231static __be32
1da177e4
LT		 2232nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2233{
2234 struct nfs4_stateid *local;
b37ad28b 2235 __be32 status = nfserr_share_denied;
1da177e4
LT		 2236 struct nfs4_stateowner *sop = open->op_stateowner;
2237
8beefa24 2238 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2239 /* ignore lock owners */
2240 if (local->st_stateowner->so_is_open_owner == 0)
2241 continue;
2242 /* remember if we have seen this open owner */
2243 if (local->st_stateowner == sop)
2244 *stpp = local;
2245 /* check for conflicting share reservations */
2246 if (!test_share(local, open))
2247 goto out;
2248 }
2249 status = 0;
2250out:
2251 return status;
2252}
2253
5ac049ac
N		 2254static inline struct nfs4_stateid *
2255nfs4_alloc_stateid(void)
2256{
2257 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2258}
2259
b37ad28b 2260static __be32
1da177e4 2261nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2262 struct nfs4_delegation *dp,
1da177e4
LT		 2263 struct svc_fh *cur_fh, int flags)
2264{
2265 struct nfs4_stateid *stp;
1da177e4 2266
5ac049ac 2267 stp = nfs4_alloc_stateid();
1da177e4
LT		 2268 if (stp == NULL)
2269 return nfserr_resource;
2270
567d9829
N		 2271 if (dp) {
2272 get_file(dp->dl_vfs_file);
2273 stp->st_vfs_file = dp->dl_vfs_file;
2274 } else {
b37ad28b 2275 __be32 status;
567d9829
N		 2276 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2277 &stp->st_vfs_file);
2278 if (status) {
2279 if (status == nfserr_dropit)
2280 status = nfserr_jukebox;
5ac049ac 2281 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2282 return status;
2283 }
1da177e4 2284 }
1da177e4
LT		 2285 *stpp = stp;
2286 return 0;
2287}
2288
b37ad28b 2289static inline __be32
1da177e4
LT		 2290nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2291 struct nfsd4_open *open)
2292{
2293 struct iattr iattr = {
2294 .ia_valid = ATTR_SIZE,
2295 .ia_size = 0,
2296 };
2297 if (!open->op_truncate)
2298 return 0;
2299 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2300 return nfserr_inval;
1da177e4
LT		 2301 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2302}
2303
b37ad28b 2304static __be32
1da177e4
LT		 2305nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2306{
2307 struct file *filp = stp->st_vfs_file;
7eaa36e2 2308 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2309 unsigned int share_access, new_writer;
b37ad28b 2310 __be32 status;
1da177e4
LT		 2311
2312 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2313 new_writer = (~share_access) & open->op_share_access
2314 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2315
6c26d08f 2316 if (new_writer) {
b37ad28b
AV		 2317 int err = get_write_access(inode);
2318 if (err)
2319 return nfserrno(err);
e518f056
BH		 2320 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2321 if (err)
2322 return nfserrno(err);
2323 file_take_write(filp);
6c26d08f 2324 }
1da177e4
LT		 2325 status = nfsd4_truncate(rqstp, cur_fh, open);
2326 if (status) {
6c26d08f
BF		 2327 if (new_writer)
2328 put_write_access(inode);
1da177e4
LT		 2329 return status;
2330 }
2331 /* remember the open */
6c26d08f 2332 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2333 __set_bit(open->op_share_access, &stp->st_access_bmap);
2334 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2335
2336 return nfs_ok;
2337}
2338
2339
1da177e4 2340static void
37515177 2341nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2342{
37515177
N		 2343 open->op_stateowner->so_confirmed = 1;
2344 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2345}
2346
2347/*
2348 * Attempt to hand out a delegation.
2349 */
2350static void
2351nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2352{
2353 struct nfs4_delegation *dp;
2354 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2355 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2356 struct file_lock fl, *flp = &fl;
2357 int status, flag = 0;
2358
2359 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2360 open->op_recall = 0;
2361 switch (open->op_claim_type) {
2362 case NFS4_OPEN_CLAIM_PREVIOUS:
2363 if (!atomic_read(&cb->cb_set))
2364 open->op_recall = 1;
2365 flag = open->op_delegate_type;
2366 if (flag == NFS4_OPEN_DELEGATE_NONE)
2367 goto out;
2368 break;
2369 case NFS4_OPEN_CLAIM_NULL:
2370 /* Let's not give out any delegations till everyone's
2371 * had the chance to reclaim theirs.... */
af558e33 2372 if (locks_in_grace())
7b190fec
N		 2373 goto out;
2374 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2375 goto out;
2376 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2377 flag = NFS4_OPEN_DELEGATE_WRITE;
2378 else
2379 flag = NFS4_OPEN_DELEGATE_READ;
2380 break;
2381 default:
2382 goto out;
2383 }
1da177e4
LT		 2384
2385 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2386 if (dp == NULL) {
2387 flag = NFS4_OPEN_DELEGATE_NONE;
2388 goto out;
2389 }
2390 locks_init_lock(&fl);
2391 fl.fl_lmops = &nfsd_lease_mng_ops;
2392 fl.fl_flags = FL_LEASE;
9167f501 2393 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2394 fl.fl_end = OFFSET_MAX;
2395 fl.fl_owner = (fl_owner_t)dp;
2396 fl.fl_file = stp->st_vfs_file;
2397 fl.fl_pid = current->tgid;
2398
a9933cea 2399 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2400 * the lock_manager callbacks fl_mylease and fl_change are used
2401 */
9167f501 2402 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2403 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2404 unhash_delegation(dp);
1da177e4
LT		 2405 flag = NFS4_OPEN_DELEGATE_NONE;
2406 goto out;
2407 }
2408
2409 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2410
2411 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2412 dp->dl_stateid.si_boot,
2413 dp->dl_stateid.si_stateownerid,
2414 dp->dl_stateid.si_fileid,
2415 dp->dl_stateid.si_generation);
2416out:
7b190fec
N		 2417 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2418 && flag == NFS4_OPEN_DELEGATE_NONE
2419 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2420 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2421 open->op_delegate_type = flag;
2422}
2423
2424/*
2425 * called with nfs4_lock_state() held.
2426 */
b37ad28b 2427__be32
1da177e4
LT		 2428nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2429{
6668958f 2430 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2431 struct nfs4_file *fp = NULL;
2432 struct inode *ino = current_fh->fh_dentry->d_inode;
2433 struct nfs4_stateid *stp = NULL;
567d9829 2434 struct nfs4_delegation *dp = NULL;
b37ad28b 2435 __be32 status;
1da177e4
LT		 2436
2437 status = nfserr_inval;
d87a8ade 2438 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2439 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2440 goto out;
2441 /*
2442 * Lookup file; if found, lookup stateid and check open request,
2443 * and check for delegations in the process of being recalled.
2444 * If not found, create the nfs4_file struct
2445 */
2446 fp = find_file(ino);
2447 if (fp) {
2448 if ((status = nfs4_check_open(fp, open, &stp)))
2449 goto out;
c44c5eeb
N		 2450 status = nfs4_check_deleg(fp, open, &dp);
2451 if (status)
2452 goto out;
1da177e4 2453 } else {
c44c5eeb
N		 2454 status = nfserr_bad_stateid;
2455 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2456 goto out;
1da177e4
LT		 2457 status = nfserr_resource;
2458 fp = alloc_init_file(ino);
2459 if (fp == NULL)
2460 goto out;
2461 }
2462
2463 /*
2464 * OPEN the file, or upgrade an existing OPEN.
2465 * If truncate fails, the OPEN fails.
2466 */
2467 if (stp) {
2468 /* Stateid was found, this is an OPEN upgrade */
2469 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2470 if (status)
2471 goto out;
444c2c07 2472 update_stateid(&stp->st_stateid);
1da177e4
LT		 2473 } else {
2474 /* Stateid was not found, this is a new OPEN */
2475 int flags = 0;
9ecb6a08 2476 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2477 flags |= NFSD_MAY_READ;
1da177e4 2478 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2479 flags |= NFSD_MAY_WRITE;
567d9829
N		 2480 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2481 if (status)
1da177e4
LT		 2482 goto out;
2483 init_stateid(stp, fp, open);
2484 status = nfsd4_truncate(rqstp, current_fh, open);
2485 if (status) {
2283963f 2486 release_open_stateid(stp);
1da177e4
LT		 2487 goto out;
2488 }
6668958f
AA		 2489 if (nfsd4_has_session(&resp->cstate))
2490 update_stateid(&stp->st_stateid);
1da177e4
LT		 2491 }
2492 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2493
6668958f
AA		 2494 if (nfsd4_has_session(&resp->cstate))
2495 open->op_stateowner->so_confirmed = 1;
2496
1da177e4
LT		 2497 /*
2498 * Attempt to hand out a delegation. No error return, because the
2499 * OPEN succeeds even if we fail.
2500 */
2501 nfs4_open_delegation(current_fh, open, stp);
2502
2503 status = nfs_ok;
2504
2505 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2506 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2507 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2508out:
13cd2184
N		 2509 if (fp)
2510 put_nfs4_file(fp);
37515177
N		 2511 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2512 nfs4_set_claim_prev(open);
1da177e4
LT		 2513 /*
2514 * To finish the open response, we just need to set the rflags.
2515 */
2516 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2517 if (!open->op_stateowner->so_confirmed &&
2518 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2519 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2520
2521 return status;
2522}
2523
b37ad28b 2524__be32
b591480b
BF		 2525nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2526 clientid_t *clid)
1da177e4
LT		 2527{
2528 struct nfs4_client *clp;
b37ad28b 2529 __be32 status;
1da177e4
LT		 2530
2531 nfs4_lock_state();
2532 dprintk("process_renew(%08x/%08x): starting\n",
2533 clid->cl_boot, clid->cl_id);
2534 status = nfserr_stale_clientid;
2535 if (STALE_CLIENTID(clid))
2536 goto out;
2537 clp = find_confirmed_client(clid);
2538 status = nfserr_expired;
2539 if (clp == NULL) {
2540 /* We assume the client took too long to RENEW. */
2541 dprintk("nfsd4_renew: clientid not found!\n");
2542 goto out;
2543 }
2544 renew_client(clp);
2545 status = nfserr_cb_path_down;
ea1da636 2546 if (!list_empty(&clp->cl_delegations)
c237dc03 2547 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2548 goto out;
2549 status = nfs_ok;
2550out:
2551 nfs4_unlock_state();
2552 return status;
2553}
2554
af558e33
BF		 2555struct lock_manager nfsd4_manager = {
2556};
2557
a76b4319 2558static void
af558e33 2559nfsd4_end_grace(void)
a76b4319
N		 2560{
2561 dprintk("NFSD: end of grace period\n");
c7b9a459 2562 nfsd4_recdir_purge_old();
af558e33 2563 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2564}
2565
fd39ca9a 2566static time_t
1da177e4
LT		 2567nfs4_laundromat(void)
2568{
2569 struct nfs4_client *clp;
2570 struct nfs4_stateowner *sop;
2571 struct nfs4_delegation *dp;
2572 struct list_head *pos, *next, reaplist;
2573 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2574 time_t t, clientid_val = NFSD_LEASE_TIME;
2575 time_t u, test_val = NFSD_LEASE_TIME;
2576
2577 nfs4_lock_state();
2578
2579 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2580 if (locks_in_grace())
2581 nfsd4_end_grace();
1da177e4
LT		 2582 list_for_each_safe(pos, next, &client_lru) {
2583 clp = list_entry(pos, struct nfs4_client, cl_lru);
2584 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2585 t = clp->cl_time - cutoff;
2586 if (clientid_val > t)
2587 clientid_val = t;
2588 break;
2589 }
2590 dprintk("NFSD: purging unused client (clientid %08x)\n",
2591 clp->cl_clientid.cl_id);
c7b9a459 2592 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2593 expire_client(clp);
2594 }
2595 INIT_LIST_HEAD(&reaplist);
2596 spin_lock(&recall_lock);
2597 list_for_each_safe(pos, next, &del_recall_lru) {
2598 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2599 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2600 u = dp->dl_time - cutoff;
2601 if (test_val > u)
2602 test_val = u;
2603 break;
2604 }
2605 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2606 dp, dp->dl_flock);
2607 list_move(&dp->dl_recall_lru, &reaplist);
2608 }
2609 spin_unlock(&recall_lock);
2610 list_for_each_safe(pos, next, &reaplist) {
2611 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2612 list_del_init(&dp->dl_recall_lru);
2613 unhash_delegation(dp);
2614 }
2615 test_val = NFSD_LEASE_TIME;
2616 list_for_each_safe(pos, next, &close_lru) {
2617 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2618 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2619 u = sop->so_time - cutoff;
2620 if (test_val > u)
2621 test_val = u;
2622 break;
2623 }
2624 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2625 sop->so_id);
f044ff83 2626 release_openowner(sop);
1da177e4
LT		 2627 }
2628 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2629 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2630 nfs4_unlock_state();
2631 return clientid_val;
2632}
2633
a254b246
HH		 2634static struct workqueue_struct *laundry_wq;
2635static void laundromat_main(struct work_struct *);
2636static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2637
2638static void
c4028958 2639laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2640{
2641 time_t t;
2642
2643 t = nfs4_laundromat();
2644 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2645 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2646}
2647
fd39ca9a 2648static struct nfs4_stateowner *
f8816512
N		 2649search_close_lru(u32 st_id, int flags)
2650{
1da177e4
LT		 2651 struct nfs4_stateowner *local = NULL;
2652
1da177e4
LT		 2653 if (flags & CLOSE_STATE) {
2654 list_for_each_entry(local, &close_lru, so_close_lru) {
2655 if (local->so_id == st_id)
2656 return local;
2657 }
2658 }
2659 return NULL;
2660}
2661
2662static inline int
2663nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2664{
7eaa36e2 2665 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2666}
2667
2668static int
2669STALE_STATEID(stateid_t *stateid)
2670{
78155ed7
BN		 2671 if (time_after((unsigned long)boot_time,
2672 (unsigned long)stateid->si_boot)) {
2673 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2674 stateid->si_boot, stateid->si_stateownerid,
2675 stateid->si_fileid, stateid->si_generation);
2676 return 1;
2677 }
2678 return 0;
2679}
2680
2681static int
2682EXPIRED_STATEID(stateid_t *stateid)
2683{
2684 if (time_before((unsigned long)boot_time,
2685 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2686 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2687 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2688 stateid->si_boot, stateid->si_stateownerid,
2689 stateid->si_fileid, stateid->si_generation);
2690 return 1;
2691 }
2692 return 0;
2693}
2694
2695static __be32
2696stateid_error_map(stateid_t *stateid)
2697{
2698 if (STALE_STATEID(stateid))
2699 return nfserr_stale_stateid;
2700 if (EXPIRED_STATEID(stateid))
2701 return nfserr_expired;
2702
2703 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2704 stateid->si_boot, stateid->si_stateownerid,
2705 stateid->si_fileid, stateid->si_generation);
2706 return nfserr_bad_stateid;
1da177e4
LT		 2707}
2708
2709static inline int
2710access_permit_read(unsigned long access_bmap)
2711{
2712 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2713 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2714 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2715}
2716
2717static inline int
2718access_permit_write(unsigned long access_bmap)
2719{
2720 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2721 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2722}
2723
2724static
b37ad28b 2725__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2726{
b37ad28b 2727 __be32 status = nfserr_openmode;
1da177e4
LT		 2728
2729 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2730 goto out;
2731 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2732 goto out;
2733 status = nfs_ok;
2734out:
2735 return status;
2736}
2737
b37ad28b 2738static inline __be32
1da177e4
LT		 2739check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2740{
203a8c8e 2741 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2742 return nfs_ok;
af558e33 2743 else if (locks_in_grace()) {
1da177e4
LT		 2744 /* Answer in remaining cases depends on existance of
2745 * conflicting state; so we must wait out the grace period. */
2746 return nfserr_grace;
2747 } else if (flags & WR_STATE)
2748 return nfs4_share_conflict(current_fh,
2749 NFS4_SHARE_DENY_WRITE);
2750 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2751 return nfs4_share_conflict(current_fh,
2752 NFS4_SHARE_DENY_READ);
2753}
2754
2755/*
2756 * Allow READ/WRITE during grace period on recovered state only for files
2757 * that are not able to provide mandatory locking.
2758 */
2759static inline int
18f82731 2760grace_disallows_io(struct inode *inode)
1da177e4 2761{
203a8c8e 2762 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2763}
2764
6668958f 2765static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2766{
6668958f
AA		 2767 /*
2768 * When sessions are used the stateid generation number is ignored
2769 * when it is zero.
2770 */
2771 if ((flags & HAS_SESSION) && in->si_generation == 0)
2772 goto out;
2773
0836f587
BF		 2774 /* If the client sends us a stateid from the future, it's buggy: */
2775 if (in->si_generation > ref->si_generation)
2776 return nfserr_bad_stateid;
2777 /*
2778 * The following, however, can happen. For example, if the
2779 * client sends an open and some IO at the same time, the open
2780 * may bump si_generation while the IO is still in flight.
2781 * Thanks to hard links and renames, the client never knows what
2782 * file an open will affect. So it could avoid that situation
2783 * only by serializing all opens and IO from the same open
2784 * owner. To recover from the old_stateid error, the client
2785 * will just have to retry the IO:
2786 */
2787 if (in->si_generation < ref->si_generation)
2788 return nfserr_old_stateid;
6668958f 2789out:
0836f587
BF		 2790 return nfs_ok;
2791}
2792
3e633079
BF		 2793static int is_delegation_stateid(stateid_t *stateid)
2794{
2795 return stateid->si_fileid == 0;
2796}
2797
1da177e4
LT		 2798/*
2799* Checks for stateid operations
2800*/
b37ad28b 2801__be32
dd453dfd
BH		 2802nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2803 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2804{
2805 struct nfs4_stateid *stp = NULL;
2806 struct nfs4_delegation *dp = NULL;
dd453dfd 2807 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2808 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2809 __be32 status;
1da177e4 2810
1da177e4
LT		 2811 if (filpp)
2812 *filpp = NULL;
2813
18f82731 2814 if (grace_disallows_io(ino))
1da177e4
LT		 2815 return nfserr_grace;
2816
6668958f
AA		 2817 if (nfsd4_has_session(cstate))
2818 flags |= HAS_SESSION;
2819
1da177e4
LT		 2820 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2821 return check_special_stateids(current_fh, stateid, flags);
2822
1da177e4
LT		 2823 status = nfserr_stale_stateid;
2824 if (STALE_STATEID(stateid))
2825 goto out;
2826
1da177e4 2827 status = nfserr_bad_stateid;
3e633079 2828 if (is_delegation_stateid(stateid)) {
a4455be0 2829 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2830 if (!dp) {
2831 status = stateid_error_map(stateid);
1da177e4 2832 goto out;
78155ed7 2833 }
6668958f
AA		 2834 status = check_stateid_generation(stateid, &dp->dl_stateid,
2835 flags);
0c2a498f
BF		 2836 if (status)
2837 goto out;
dc9bf700
BF		 2838 status = nfs4_check_delegmode(dp, flags);
2839 if (status)
2840 goto out;
2841 renew_client(dp->dl_client);
dc9bf700
BF		 2842 if (filpp)
2843 *filpp = dp->dl_vfs_file;
1da177e4 2844 } else { /* open or lock stateid */
a4455be0 2845 stp = find_stateid(stateid, flags);
78155ed7
BN		 2846 if (!stp) {
2847 status = stateid_error_map(stateid);
1da177e4 2848 goto out;
78155ed7 2849 }
6150ef0d 2850 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2851 goto out;
2852 if (!stp->st_stateowner->so_confirmed)
2853 goto out;
6668958f
AA		 2854 status = check_stateid_generation(stateid, &stp->st_stateid,
2855 flags);
0c2a498f
BF		 2856 if (status)
2857 goto out;
a4455be0
BF		 2858 status = nfs4_check_openmode(stp, flags);
2859 if (status)
1da177e4
LT		 2860 goto out;
2861 renew_client(stp->st_stateowner->so_client);
2862 if (filpp)
2863 *filpp = stp->st_vfs_file;
1da177e4
LT		 2864 }
2865 status = nfs_ok;
2866out:
2867 return status;
2868}
2869
4c4cd222
N		 2870static inline int
2871setlkflg (int type)
2872{
2873 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2874 RD_STATE : WR_STATE;
2875}
1da177e4
LT		 2876
2877/*
2878 * Checks for sequence id mutating operations.
2879 */
b37ad28b 2880static __be32
dd453dfd
BH		 2881nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2882 stateid_t *stateid, int flags,
2883 struct nfs4_stateowner **sopp,
2884 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2885{
1da177e4
LT		 2886 struct nfs4_stateid *stp;
2887 struct nfs4_stateowner *sop;
dd453dfd 2888 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2889 __be32 status;
1da177e4
LT		 2890
2891 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2892 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2893 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2894 stateid->si_generation);
3a4f98bb 2895
1da177e4
LT		 2896 *stpp = NULL;
2897 *sopp = NULL;
2898
1da177e4 2899 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2900 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2901 return nfserr_bad_stateid;
1da177e4
LT		 2902 }
2903
1da177e4 2904 if (STALE_STATEID(stateid))
3a4f98bb 2905 return nfserr_stale_stateid;
6668958f
AA		 2906
2907 if (nfsd4_has_session(cstate))
2908 flags |= HAS_SESSION;
2909
1da177e4
LT		 2910 /*
2911 * We return BAD_STATEID if filehandle doesn't match stateid,
2912 * the confirmed flag is incorrecly set, or the generation
2913 * number is incorrect.
1da177e4 2914 */
f8816512
N		 2915 stp = find_stateid(stateid, flags);
2916 if (stp == NULL) {
2917 /*
2918 * Also, we should make sure this isn't just the result of
2919 * a replayed close:
2920 */
2921 sop = search_close_lru(stateid->si_stateownerid, flags);
2922 if (sop == NULL)
78155ed7 2923 return stateid_error_map(stateid);
f8816512
N		 2924 *sopp = sop;
2925 goto check_replay;
2926 }
1da177e4 2927
39325bd0
BF		 2928 *stpp = stp;
2929 *sopp = sop = stp->st_stateowner;
2930
4c4cd222 2931 if (lock) {
4c4cd222 2932 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2933 struct nfs4_client *clp = sop->so_client;
4c4cd222 2934 int lkflg = 0;
b37ad28b 2935 __be32 status;
4c4cd222
N		 2936
2937 lkflg = setlkflg(lock->lk_type);
2938
2939 if (lock->lk_is_new) {
599e0a22
BF		 2940 if (!sop->so_is_open_owner)
2941 return nfserr_bad_stateid;
60adfc50
AA		 2942 if (!(flags & HAS_SESSION) &&
2943 !same_clid(&clp->cl_clientid, lockclid))
2944 return nfserr_bad_stateid;
599e0a22
BF		 2945 /* stp is the open stateid */
2946 status = nfs4_check_openmode(stp, lkflg);
2947 if (status)
2948 return status;
2949 } else {
2950 /* stp is the lock stateid */
2951 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2952 if (status)
2953 return status;
4c4cd222 2954 }
1da177e4
LT		 2955 }
2956
f3362737 2957 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2958 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2959 return nfserr_bad_stateid;
1da177e4
LT		 2960 }
2961
1da177e4
LT		 2962 /*
2963 * We now validate the seqid and stateid generation numbers.
2964 * For the moment, we ignore the possibility of
2965 * generation number wraparound.
2966 */
6668958f 2967 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2968 goto check_replay;
2969
3a4f98bb 2970 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2971 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2972 " unconfirmed stateowner!\n");
2973 return nfserr_bad_stateid;
1da177e4 2974 }
3a4f98bb 2975 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2976 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2977 " confirmed yet!\n");
2978 return nfserr_bad_stateid;
1da177e4 2979 }
6668958f 2980 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 2981 if (status)
2982 return status;
52fd004e 2983 renew_client(sop->so_client);
3a4f98bb 2984 return nfs_ok;
1da177e4 2985
1da177e4 2986check_replay:
bd9aac52 2987 if (seqid == sop->so_seqid - 1) {
849823c5 2988 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2989 /* indicate replay to calling function */
a90b061c 2990 return nfserr_replay_me;
1da177e4 2991 }
2fdada03 2992 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2993 sop->so_seqid, seqid);
2994 *sopp = NULL;
2995 return nfserr_bad_seqid;
1da177e4
LT		 2996}
2997
b37ad28b 2998__be32
ca364317 2999nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3000 struct nfsd4_open_confirm *oc)
1da177e4 3001{
b37ad28b 3002 __be32 status;
1da177e4
LT		 3003 struct nfs4_stateowner *sop;
3004 struct nfs4_stateid *stp;
3005
3006 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3007 (int)cstate->current_fh.fh_dentry->d_name.len,
3008 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3009
ca364317 3010 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3011 if (status)
3012 return status;
1da177e4
LT		 3013
3014 nfs4_lock_state();
3015
dd453dfd 3016 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3017 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3018 CONFIRM | OPEN_STATE,
1da177e4
LT		 3019 &oc->oc_stateowner, &stp, NULL)))
3020 goto out;
3021
3022 sop = oc->oc_stateowner;
3023 sop->so_confirmed = 1;
3024 update_stateid(&stp->st_stateid);
3025 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3026 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3027 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3028 stp->st_stateid.si_boot,
3029 stp->st_stateid.si_stateownerid,
3030 stp->st_stateid.si_fileid,
3031 stp->st_stateid.si_generation);
c7b9a459
N		 3032
3033 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3034out:
f2327d9a 3035 if (oc->oc_stateowner) {
1da177e4 3036 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3037 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3038 }
1da177e4
LT		 3039 nfs4_unlock_state();
3040 return status;
3041}
3042
3043
3044/*
3045 * unset all bits in union bitmap (bmap) that
3046 * do not exist in share (from successful OPEN_DOWNGRADE)
3047 */
3048static void
3049reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3050{
3051 int i;
3052 for (i = 1; i < 4; i++) {
3053 if ((i & access) != i)
3054 __clear_bit(i, bmap);
3055 }
3056}
3057
3058static void
3059reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3060{
3061 int i;
3062 for (i = 0; i < 4; i++) {
3063 if ((i & deny) != i)
3064 __clear_bit(i, bmap);
3065 }
3066}
3067
b37ad28b 3068__be32
ca364317
BF		 3069nfsd4_open_downgrade(struct svc_rqst *rqstp,
3070 struct nfsd4_compound_state *cstate,
a4f1706a 3071 struct nfsd4_open_downgrade *od)
1da177e4 3072{
b37ad28b 3073 __be32 status;
1da177e4
LT		 3074 struct nfs4_stateid *stp;
3075 unsigned int share_access;
3076
3077 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3078 (int)cstate->current_fh.fh_dentry->d_name.len,
3079 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3080
d87a8ade 3081 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3082 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3083 return nfserr_inval;
3084
3085 nfs4_lock_state();
dd453dfd 3086 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3087 od->od_seqid,
1da177e4 3088 &od->od_stateid,
f3362737 3089 OPEN_STATE,
1da177e4
LT		 3090 &od->od_stateowner, &stp, NULL)))
3091 goto out;
3092
3093 status = nfserr_inval;
3094 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3095 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3096 stp->st_access_bmap, od->od_share_access);
3097 goto out;
3098 }
3099 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3100 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3101 stp->st_deny_bmap, od->od_share_deny);
3102 goto out;
3103 }
3104 set_access(&share_access, stp->st_access_bmap);
3105 nfs4_file_downgrade(stp->st_vfs_file,
3106 share_access & ~od->od_share_access);
3107
3108 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3109 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3110
3111 update_stateid(&stp->st_stateid);
3112 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3113 status = nfs_ok;
3114out:
f2327d9a 3115 if (od->od_stateowner) {
1da177e4 3116 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3117 cstate->replay_owner = od->od_stateowner;
f2327d9a 3118 }
1da177e4
LT		 3119 nfs4_unlock_state();
3120 return status;
3121}
3122
3123/*
3124 * nfs4_unlock_state() called after encode
3125 */
b37ad28b 3126__be32
ca364317 3127nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3128 struct nfsd4_close *close)
1da177e4 3129{
b37ad28b 3130 __be32 status;
1da177e4
LT		 3131 struct nfs4_stateid *stp;
3132
3133 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3134 (int)cstate->current_fh.fh_dentry->d_name.len,
3135 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3136
3137 nfs4_lock_state();
3138 /* check close_lru for replay */
dd453dfd 3139 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3140 close->cl_seqid,
1da177e4 3141 &close->cl_stateid,
f3362737 3142 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3143 &close->cl_stateowner, &stp, NULL)))
3144 goto out;
1da177e4
LT		 3145 status = nfs_ok;
3146 update_stateid(&stp->st_stateid);
3147 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3148
04ef5954 3149 /* release_stateid() calls nfsd_close() if needed */
2283963f 3150 release_open_stateid(stp);
04ef5954
BF		 3151
3152 /* place unused nfs4_stateowners on so_close_lru list to be
3153 * released by the laundromat service after the lease period
3154 * to enable us to handle CLOSE replay
3155 */
3156 if (list_empty(&close->cl_stateowner->so_stateids))
3157 move_to_close_lru(close->cl_stateowner);
1da177e4 3158out:
f2327d9a 3159 if (close->cl_stateowner) {
1da177e4 3160 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3161 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3162 }
1da177e4
LT		 3163 nfs4_unlock_state();
3164 return status;
3165}
3166
b37ad28b 3167__be32
ca364317
BF		 3168nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3169 struct nfsd4_delegreturn *dr)
1da177e4 3170{
203a8c8e
BF		 3171 struct nfs4_delegation *dp;
3172 stateid_t *stateid = &dr->dr_stateid;
3173 struct inode *inode;
b37ad28b 3174 __be32 status;
6668958f 3175 int flags = 0;
1da177e4 3176
ca364317 3177 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3178 return status;
3179 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3180
6668958f
AA		 3181 if (nfsd4_has_session(cstate))
3182 flags |= HAS_SESSION;
1da177e4 3183 nfs4_lock_state();
203a8c8e
BF		 3184 status = nfserr_bad_stateid;
3185 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3186 goto out;
3187 status = nfserr_stale_stateid;
3188 if (STALE_STATEID(stateid))
3189 goto out;
7e0f7cf5 3190 status = nfserr_bad_stateid;
203a8c8e
BF		 3191 if (!is_delegation_stateid(stateid))
3192 goto out;
203a8c8e 3193 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3194 if (!dp) {
3195 status = stateid_error_map(stateid);
203a8c8e 3196 goto out;
78155ed7 3197 }
6668958f 3198 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3199 if (status)
3200 goto out;
3201 renew_client(dp->dl_client);
3202
3203 unhash_delegation(dp);
1da177e4 3204out:
203a8c8e
BF		 3205 nfs4_unlock_state();
3206
1da177e4
LT		 3207 return status;
3208}
3209
3210
3211/*
3212 * Lock owner state (byte-range locks)
3213 */
3214#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3215#define LOCK_HASH_BITS 8
3216#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3217#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3218
87df4de8
BH		 3219static inline u64
3220end_offset(u64 start, u64 len)
3221{
3222 u64 end;
3223
3224 end = start + len;
3225 return end >= start ? end: NFS4_MAX_UINT64;
3226}
3227
3228/* last octet in a range */
3229static inline u64
3230last_byte_offset(u64 start, u64 len)
3231{
3232 u64 end;
3233
3234 BUG_ON(!len);
3235 end = start + len;
3236 return end > start ? end - 1: NFS4_MAX_UINT64;
3237}
3238
1da177e4
LT		 3239#define lockownerid_hashval(id) \
3240 ((id) & LOCK_HASH_MASK)
3241
3242static inline unsigned int
3243lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3244 struct xdr_netobj *ownername)
3245{
3246 return (file_hashval(inode) + cl_id
3247 + opaque_hashval(ownername->data, ownername->len))
3248 & LOCK_HASH_MASK;
3249}
3250
3251static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3252static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3253static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3254
fd39ca9a 3255static struct nfs4_stateid *
1da177e4
LT		 3256find_stateid(stateid_t *stid, int flags)
3257{
9346eff0 3258 struct nfs4_stateid *local;
1da177e4
LT		 3259 u32 st_id = stid->si_stateownerid;
3260 u32 f_id = stid->si_fileid;
3261 unsigned int hashval;
3262
3263 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3264 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3265 hashval = stateid_hashval(st_id, f_id);
3266 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3267 if ((local->st_stateid.si_stateownerid == st_id) &&
3268 (local->st_stateid.si_fileid == f_id))
3269 return local;
3270 }
3271 }
9346eff0
KK		 3272
3273 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3274 hashval = stateid_hashval(st_id, f_id);
3275 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3276 if ((local->st_stateid.si_stateownerid == st_id) &&
3277 (local->st_stateid.si_fileid == f_id))
3278 return local;
3279 }
849823c5 3280 }
1da177e4
LT		 3281 return NULL;
3282}
3283
3284static struct nfs4_delegation *
3285find_delegation_stateid(struct inode *ino, stateid_t *stid)
3286{
13cd2184
N		 3287 struct nfs4_file *fp;
3288 struct nfs4_delegation *dl;
1da177e4
LT		 3289
3290 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3291 stid->si_boot, stid->si_stateownerid,
3292 stid->si_fileid, stid->si_generation);
3293
1da177e4 3294 fp = find_file(ino);
13cd2184
N		 3295 if (!fp)
3296 return NULL;
3297 dl = find_delegation_file(fp, stid);
3298 put_nfs4_file(fp);
3299 return dl;
1da177e4
LT		 3300}
3301
3302/*
3303 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3304 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3305 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3306 * locking, this prevents us from being completely protocol-compliant. The
3307 * real solution to this problem is to start using unsigned file offsets in
3308 * the VFS, but this is a very deep change!
3309 */
3310static inline void
3311nfs4_transform_lock_offset(struct file_lock *lock)
3312{
3313 if (lock->fl_start < 0)
3314 lock->fl_start = OFFSET_MAX;
3315 if (lock->fl_end < 0)
3316 lock->fl_end = OFFSET_MAX;
3317}
3318
d5b9026a
N		 3319/* Hack!: For now, we're defining this just so we can use a pointer to it
3320 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3321static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3322};
1da177e4
LT		 3323
3324static inline void
3325nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3326{
d5b9026a
N		 3327 struct nfs4_stateowner *sop;
3328 unsigned int hval;
1da177e4 3329
d5b9026a
N		 3330 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3331 sop = (struct nfs4_stateowner *) fl->fl_owner;
3332 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3333 kref_get(&sop->so_ref);
3334 deny->ld_sop = sop;
3335 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3336 } else {
3337 deny->ld_sop = NULL;
3338 deny->ld_clientid.cl_boot = 0;
3339 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3340 }
3341 deny->ld_start = fl->fl_start;
87df4de8
BH		 3342 deny->ld_length = NFS4_MAX_UINT64;
3343 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3344 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3345 deny->ld_type = NFS4_READ_LT;
3346 if (fl->fl_type != F_RDLCK)
3347 deny->ld_type = NFS4_WRITE_LT;
3348}
3349
1da177e4
LT		 3350static struct nfs4_stateowner *
3351find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3352 struct xdr_netobj *owner)
3353{
3354 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3355 struct nfs4_stateowner *op;
3356
3357 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3358 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3359 return op;
3360 }
3361 return NULL;
3362}
3363
3364/*
3365 * Alloc a lock owner structure.
3366 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3367 * occured.
3368 *
3369 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3370 */
3371
3372static struct nfs4_stateowner *
3373alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3374 struct nfs4_stateowner *sop;
3375 struct nfs4_replay *rp;
3376 unsigned int idhashval;
3377
3378 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3379 return NULL;
3380 idhashval = lockownerid_hashval(current_ownerid);
3381 INIT_LIST_HEAD(&sop->so_idhash);
3382 INIT_LIST_HEAD(&sop->so_strhash);
3383 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3384 INIT_LIST_HEAD(&sop->so_stateids);
3385 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3386 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3387 sop->so_time = 0;
3388 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3389 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3390 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3391 sop->so_is_open_owner = 0;
3392 sop->so_id = current_ownerid++;
3393 sop->so_client = clp;
b59e3c0e
NB		 3394 /* It is the openowner seqid that will be incremented in encode in the
3395 * case of new lockowners; so increment the lock seqid manually: */
3396 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3397 sop->so_confirmed = 1;
3398 rp = &sop->so_replay;
de1ae286 3399 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3400 rp->rp_buflen = 0;
3401 rp->rp_buf = rp->rp_ibuf;
3402 return sop;
3403}
3404
fd39ca9a 3405static struct nfs4_stateid *
1da177e4
LT		 3406alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3407{
3408 struct nfs4_stateid *stp;
3409 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3410
5ac049ac
N		 3411 stp = nfs4_alloc_stateid();
3412 if (stp == NULL)
1da177e4
LT		 3413 goto out;
3414 INIT_LIST_HEAD(&stp->st_hash);
3415 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3416 INIT_LIST_HEAD(&stp->st_perstateowner);
3417 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3418 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3419 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3420 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3421 stp->st_stateowner = sop;
13cd2184 3422 get_nfs4_file(fp);
1da177e4 3423 stp->st_file = fp;
78155ed7 3424 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3425 stp->st_stateid.si_stateownerid = sop->so_id;
3426 stp->st_stateid.si_fileid = fp->fi_id;
3427 stp->st_stateid.si_generation = 0;
3428 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3429 stp->st_access_bmap = open_stp->st_access_bmap;
3430 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3431 stp->st_openstp = open_stp;
1da177e4
LT		 3432
3433out:
3434 return stp;
3435}
3436
fd39ca9a 3437static int
1da177e4
LT		 3438check_lock_length(u64 offset, u64 length)
3439{
87df4de8 3440 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3441 LOFF_OVERFLOW(offset, length)));
3442}
3443
3444/*
3445 * LOCK operation
3446 */
b37ad28b 3447__be32
ca364317 3448nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3449 struct nfsd4_lock *lock)
1da177e4 3450{
3e9e3dbe 3451 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3452 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3453 struct nfs4_stateid *lock_stp;
3454 struct file *filp;
3455 struct file_lock file_lock;
8dc7c311 3456 struct file_lock conflock;
b37ad28b 3457 __be32 status = 0;
1da177e4 3458 unsigned int strhashval;
150b3934 3459 unsigned int cmd;
b8dd7b9a 3460 int err;
1da177e4
LT		 3461
3462 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3463 (long long) lock->lk_offset,
3464 (long long) lock->lk_length);
3465
1da177e4
LT		 3466 if (check_lock_length(lock->lk_offset, lock->lk_length))
3467 return nfserr_inval;
3468
ca364317 3469 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3470 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3471 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3472 return status;
3473 }
3474
1da177e4
LT		 3475 nfs4_lock_state();
3476
3477 if (lock->lk_is_new) {
893f8770
N		 3478 /*
3479 * Client indicates that this is a new lockowner.
3480 * Use open owner and open stateid to create lock owner and
3481 * lock stateid.
3482 */
1da177e4
LT		 3483 struct nfs4_stateid *open_stp = NULL;
3484 struct nfs4_file *fp;
3485
3486 status = nfserr_stale_clientid;
60adfc50
AA		 3487 if (!nfsd4_has_session(cstate) &&
3488 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3489 goto out;
1da177e4 3490
1da177e4 3491 /* validate and update open stateid and open seqid */
dd453dfd 3492 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3493 lock->lk_new_open_seqid,
3494 &lock->lk_new_open_stateid,
f3362737 3495 OPEN_STATE,
3a65588a 3496 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3497 lock);
37515177 3498 if (status)
1da177e4 3499 goto out;
3a65588a 3500 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3501 /* create lockowner and lock stateid */
3502 fp = open_stp->st_file;
3503 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3504 open_sop->so_client->cl_clientid.cl_id,
3505 &lock->v.new.owner);
3e9e3dbe
N		 3506 /* XXX: Do we need to check for duplicate stateowners on
3507 * the same file, or should they just be allowed (and
3508 * create new stateids)? */
1da177e4 3509 status = nfserr_resource;
b59e3c0e
NB		 3510 lock_sop = alloc_init_lock_stateowner(strhashval,
3511 open_sop->so_client, open_stp, lock);
3512 if (lock_sop == NULL)
1da177e4 3513 goto out;
b59e3c0e 3514 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3515 if (lock_stp == NULL)
1da177e4 3516 goto out;
1da177e4
LT		 3517 } else {
3518 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3519 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3520 lock->lk_old_lock_seqid,
3521 &lock->lk_old_lock_stateid,
f3362737 3522 LOCK_STATE,
3a65588a 3523 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3524 if (status)
3525 goto out;
3a65588a 3526 lock_sop = lock->lk_replay_owner;
1da177e4 3527 }
3a65588a 3528 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3529 filp = lock_stp->st_vfs_file;
3530
0dd395dc 3531 status = nfserr_grace;
af558e33 3532 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3533 goto out;
3534 status = nfserr_no_grace;
af558e33 3535 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3536 goto out;
3537
1da177e4
LT		 3538 locks_init_lock(&file_lock);
3539 switch (lock->lk_type) {
3540 case NFS4_READ_LT:
3541 case NFS4_READW_LT:
3542 file_lock.fl_type = F_RDLCK;
150b3934 3543 cmd = F_SETLK;
1da177e4
LT		 3544 break;
3545 case NFS4_WRITE_LT:
3546 case NFS4_WRITEW_LT:
3547 file_lock.fl_type = F_WRLCK;
150b3934 3548 cmd = F_SETLK;
1da177e4
LT		 3549 break;
3550 default:
3551 status = nfserr_inval;
3552 goto out;
3553 }
b59e3c0e 3554 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3555 file_lock.fl_pid = current->tgid;
3556 file_lock.fl_file = filp;
3557 file_lock.fl_flags = FL_POSIX;
d5b9026a 3558 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3559
3560 file_lock.fl_start = lock->lk_offset;
87df4de8 3561 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3562 nfs4_transform_lock_offset(&file_lock);
3563
3564 /*
3565 * Try to lock the file in the VFS.
3566 * Note: locks.c uses the BKL to protect the inode's lock list.
3567 */
3568
fd85b817 3569 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3570 switch (-err) {
1da177e4
LT		 3571 case 0: /* success! */
3572 update_stateid(&lock_stp->st_stateid);
3573 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3574 sizeof(stateid_t));
b8dd7b9a 3575 status = 0;
eb76b3fd
AA		 3576 break;
3577 case (EAGAIN): /* conflock holds conflicting lock */
3578 status = nfserr_denied;
3579 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3580 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3581 break;
1da177e4
LT		 3582 case (EDEADLK):
3583 status = nfserr_deadlock;
eb76b3fd 3584 break;
1da177e4 3585 default:
fd85b817 3586 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3587 status = nfserr_resource;
3588 break;
1da177e4 3589 }
1da177e4 3590out:
8a280510 3591 if (status && lock->lk_is_new && lock_sop)
f044ff83 3592 release_lockowner(lock_sop);
3a65588a
BF		 3593 if (lock->lk_replay_owner) {
3594 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3595 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3596 }
1da177e4
LT		 3597 nfs4_unlock_state();
3598 return status;
3599}
3600
55ef1274
BF		 3601/*
3602 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3603 * so we do a temporary open here just to get an open file to pass to
3604 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3605 * inode operation.)
3606 */
3607static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3608{
3609 struct file *file;
3610 int err;
3611
3612 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3613 if (err)
3614 return err;
3615 err = vfs_test_lock(file, lock);
3616 nfsd_close(file);
3617 return err;
3618}
3619
1da177e4
LT		 3620/*
3621 * LOCKT operation
3622 */
b37ad28b 3623__be32
ca364317
BF		 3624nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3625 struct nfsd4_lockt *lockt)
1da177e4
LT		 3626{
3627 struct inode *inode;
1da177e4 3628 struct file_lock file_lock;
fd85b817 3629 int error;
b37ad28b 3630 __be32 status;
1da177e4 3631
af558e33 3632 if (locks_in_grace())
1da177e4
LT		 3633 return nfserr_grace;
3634
3635 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3636 return nfserr_inval;
3637
3638 lockt->lt_stateowner = NULL;
3639 nfs4_lock_state();
3640
3641 status = nfserr_stale_clientid;
60adfc50 3642 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3643 goto out;
1da177e4 3644
ca364317 3645 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3646 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3647 if (status == nfserr_symlink)
3648 status = nfserr_inval;
3649 goto out;
3650 }
3651
ca364317 3652 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3653 locks_init_lock(&file_lock);
3654 switch (lockt->lt_type) {
3655 case NFS4_READ_LT:
3656 case NFS4_READW_LT:
3657 file_lock.fl_type = F_RDLCK;
3658 break;
3659 case NFS4_WRITE_LT:
3660 case NFS4_WRITEW_LT:
3661 file_lock.fl_type = F_WRLCK;
3662 break;
3663 default:
2fdada03 3664 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3665 status = nfserr_inval;
3666 goto out;
3667 }
3668
3669 lockt->lt_stateowner = find_lockstateowner_str(inode,
3670 &lockt->lt_clientid, &lockt->lt_owner);
3671 if (lockt->lt_stateowner)
3672 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3673 file_lock.fl_pid = current->tgid;
3674 file_lock.fl_flags = FL_POSIX;
3675
3676 file_lock.fl_start = lockt->lt_offset;
87df4de8 3677 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3678
3679 nfs4_transform_lock_offset(&file_lock);
3680
1da177e4 3681 status = nfs_ok;
55ef1274 3682 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3683 if (error) {
3684 status = nfserrno(error);
3685 goto out;
3686 }
9d6a8c5c 3687 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3688 status = nfserr_denied;
9d6a8c5c 3689 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3690 }
3691out:
3692 nfs4_unlock_state();
3693 return status;
3694}
3695
b37ad28b 3696__be32
ca364317 3697nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3698 struct nfsd4_locku *locku)
1da177e4
LT		 3699{
3700 struct nfs4_stateid *stp;
3701 struct file *filp = NULL;
3702 struct file_lock file_lock;
b37ad28b 3703 __be32 status;
b8dd7b9a 3704 int err;
1da177e4
LT		 3705
3706 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3707 (long long) locku->lu_offset,
3708 (long long) locku->lu_length);
3709
3710 if (check_lock_length(locku->lu_offset, locku->lu_length))
3711 return nfserr_inval;
3712
3713 nfs4_lock_state();
3714
dd453dfd 3715 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3716 locku->lu_seqid,
3717 &locku->lu_stateid,
f3362737 3718 LOCK_STATE,
1da177e4
LT		 3719 &locku->lu_stateowner, &stp, NULL)))
3720 goto out;
3721
3722 filp = stp->st_vfs_file;
3723 BUG_ON(!filp);
3724 locks_init_lock(&file_lock);
3725 file_lock.fl_type = F_UNLCK;
3726 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3727 file_lock.fl_pid = current->tgid;
3728 file_lock.fl_file = filp;
3729 file_lock.fl_flags = FL_POSIX;
d5b9026a 3730 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3731 file_lock.fl_start = locku->lu_offset;
3732
87df4de8 3733 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3734 nfs4_transform_lock_offset(&file_lock);
3735
3736 /*
3737 * Try to unlock the file in the VFS.
3738 */
fd85b817 3739 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3740 if (err) {
fd85b817 3741 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3742 goto out_nfserr;
3743 }
3744 /*
3745 * OK, unlock succeeded; the only thing left to do is update the stateid.
3746 */
3747 update_stateid(&stp->st_stateid);
3748 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3749
3750out:
f2327d9a 3751 if (locku->lu_stateowner) {
1da177e4 3752 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3753 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3754 }
1da177e4
LT		 3755 nfs4_unlock_state();
3756 return status;
3757
3758out_nfserr:
b8dd7b9a 3759 status = nfserrno(err);
1da177e4
LT		 3760 goto out;
3761}
3762
3763/*
3764 * returns
3765 * 1: locks held by lockowner
3766 * 0: no locks held by lockowner
3767 */
3768static int
3769check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3770{
3771 struct file_lock **flpp;
7eaa36e2 3772 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3773 int status = 0;
3774
3775 lock_kernel();
3776 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3777 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3778 status = 1;
3779 goto out;
796dadfd 3780 }
1da177e4
LT		 3781 }
3782out:
3783 unlock_kernel();
3784 return status;
3785}
3786
b37ad28b 3787__be32
b591480b
BF		 3788nfsd4_release_lockowner(struct svc_rqst *rqstp,
3789 struct nfsd4_compound_state *cstate,
3790 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3791{
3792 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3793 struct nfs4_stateowner *sop;
3794 struct nfs4_stateid *stp;
1da177e4 3795 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3796 struct list_head matches;
3797 int i;
b37ad28b 3798 __be32 status;
1da177e4
LT		 3799
3800 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3801 clid->cl_boot, clid->cl_id);
3802
3803 /* XXX check for lease expiration */
3804
3805 status = nfserr_stale_clientid;
849823c5 3806 if (STALE_CLIENTID(clid))
1da177e4 3807 return status;
1da177e4
LT		 3808
3809 nfs4_lock_state();
3810
3e9e3dbe
N		 3811 status = nfserr_locks_held;
3812 /* XXX: we're doing a linear search through all the lockowners.
3813 * Yipes! For now we'll just hope clients aren't really using
3814 * release_lockowner much, but eventually we have to fix these
3815 * data structures. */
3816 INIT_LIST_HEAD(&matches);
3817 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3818 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3819 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3820 continue;
3821 list_for_each_entry(stp, &sop->so_stateids,
3822 st_perstateowner) {
3823 if (check_for_locks(stp->st_vfs_file, sop))
3824 goto out;
3825 /* Note: so_perclient unused for lockowners,
3826 * so it's OK to fool with here. */
3827 list_add(&sop->so_perclient, &matches);
3828 }
1da177e4 3829 }
3e9e3dbe
N		 3830 }
3831 /* Clients probably won't expect us to return with some (but not all)
3832 * of the lockowner state released; so don't release any until all
3833 * have been checked. */
3834 status = nfs_ok;
0fa822e4
N		 3835 while (!list_empty(&matches)) {
3836 sop = list_entry(matches.next, struct nfs4_stateowner,
3837 so_perclient);
3838 /* unhash_stateowner deletes so_perclient only
3839 * for openowners. */
3840 list_del(&sop->so_perclient);
f044ff83 3841 release_lockowner(sop);
1da177e4
LT		 3842 }
3843out:
3844 nfs4_unlock_state();
3845 return status;
3846}
3847
3848static inline struct nfs4_client_reclaim *
a55370a3 3849alloc_reclaim(void)
1da177e4 3850{
a55370a3 3851 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3852}
3853
c7b9a459 3854int
a1bcecd2 3855nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3856{
3857 unsigned int strhashval = clientstr_hashval(name);
3858 struct nfs4_client *clp;
3859
a1bcecd2 3860 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3861 return clp ? 1 : 0;
3862}
3863
1da177e4
LT		 3864/*
3865 * failure => all reset bets are off, nfserr_no_grace...
3866 */
190e4fbf
N		 3867int
3868nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3869{
3870 unsigned int strhashval;
3871 struct nfs4_client_reclaim *crp = NULL;
3872
a55370a3
N		 3873 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3874 crp = alloc_reclaim();
1da177e4
LT		 3875 if (!crp)
3876 return 0;
a55370a3 3877 strhashval = clientstr_hashval(name);
1da177e4
LT		 3878 INIT_LIST_HEAD(&crp->cr_strhash);
3879 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3880 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3881 reclaim_str_hashtbl_size++;
3882 return 1;
3883}
3884
3885static void
3886nfs4_release_reclaim(void)
3887{
3888 struct nfs4_client_reclaim *crp = NULL;
3889 int i;
3890
1da177e4
LT		 3891 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3892 while (!list_empty(&reclaim_str_hashtbl[i])) {
3893 crp = list_entry(reclaim_str_hashtbl[i].next,
3894 struct nfs4_client_reclaim, cr_strhash);
3895 list_del(&crp->cr_strhash);
1da177e4
LT		 3896 kfree(crp);
3897 reclaim_str_hashtbl_size--;
3898 }
3899 }
3900 BUG_ON(reclaim_str_hashtbl_size);
3901}
3902
3903/*
3904 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3905static struct nfs4_client_reclaim *
1da177e4
LT		 3906nfs4_find_reclaim_client(clientid_t *clid)
3907{
3908 unsigned int strhashval;
3909 struct nfs4_client *clp;
3910 struct nfs4_client_reclaim *crp = NULL;
3911
3912
3913 /* find clientid in conf_id_hashtbl */
3914 clp = find_confirmed_client(clid);
3915 if (clp == NULL)
3916 return NULL;
3917
a55370a3
N		 3918 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3919 clp->cl_name.len, clp->cl_name.data,
3920 clp->cl_recdir);
1da177e4
LT		 3921
3922 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3923 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3924 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3925 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3926 return crp;
3927 }
3928 }
3929 return NULL;
3930}
3931
3932/*
3933* Called from OPEN. Look for clientid in reclaim list.
3934*/
b37ad28b 3935__be32
1da177e4
LT		 3936nfs4_check_open_reclaim(clientid_t *clid)
3937{
dfc83565 3938 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3939}
3940
ac4d8ff2 3941/* initialization to perform at module load time: */
1da177e4 3942
e8ff2a84 3943int
ac4d8ff2 3944nfs4_state_init(void)
1da177e4 3945{
e8ff2a84 3946 int i, status;
1da177e4 3947
e8ff2a84
BF		 3948 status = nfsd4_init_slabs();
3949 if (status)
3950 return status;
1da177e4
LT		 3951 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3952 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3953 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3954 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3955 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3956 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3957 }
5282fd72
ME		 3958 for (i = 0; i < SESSION_HASH_SIZE; i++)
3959 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3960 for (i = 0; i < FILE_HASH_SIZE; i++) {
3961 INIT_LIST_HEAD(&file_hashtbl[i]);
3962 }
3963 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3964 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3965 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3966 }
3967 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3968 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3969 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3970 }
3971 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3972 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3973 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3974 }
1da177e4 3975 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3976 INIT_LIST_HEAD(&close_lru);
3977 INIT_LIST_HEAD(&client_lru);
3978 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3979 reclaim_str_hashtbl_size = 0;
e8ff2a84 3980 return 0;
ac4d8ff2
N		 3981}
3982
190e4fbf
N		 3983static void
3984nfsd4_load_reboot_recovery_data(void)
3985{
3986 int status;
3987
0964a3d3
N		 3988 nfs4_lock_state();
3989 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3990 status = nfsd4_recdir_load();
0964a3d3 3991 nfs4_unlock_state();
190e4fbf
N		 3992 if (status)
3993 printk("NFSD: Failure reading reboot recovery data\n");
3994}
3995
9a8db97e
ME		 3996unsigned long
3997get_nfs4_grace_period(void)
3998{
3999 return max(user_lease_time, lease_time) * HZ;
4000}
4001
c2f1a551
MS		 4002/*
4003 * Since the lifetime of a delegation isn't limited to that of an open, a
4004 * client may quite reasonably hang on to a delegation as long as it has
4005 * the inode cached. This becomes an obvious problem the first time a
4006 * client's inode cache approaches the size of the server's total memory.
4007 *
4008 * For now we avoid this problem by imposing a hard limit on the number
4009 * of delegations, which varies according to the server's memory size.
4010 */
4011static void
4012set_max_delegations(void)
4013{
4014 /*
4015 * Allow at most 4 delegations per megabyte of RAM. Quick
4016 * estimates suggest that in the worst case (where every delegation
4017 * is for a different inode), a delegation could take about 1.5K,
4018 * giving a worst case usage of about 6% of memory.
4019 */
4020 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4021}
4022
ac4d8ff2
N		 4023/* initialization to perform when the nfsd service is started: */
4024
29ab23cc 4025static int
ac4d8ff2
N		 4026__nfs4_state_start(void)
4027{
9a8db97e 4028 unsigned long grace_time;
ac4d8ff2 4029
1da177e4 4030 boot_time = get_seconds();
af558e33 4031 grace_time = get_nfs4_grace_period();
d99a05ad 4032 lease_time = user_lease_time;
af558e33 4033 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4034 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4035 grace_time/HZ);
58da282b 4036 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 4037 if (laundry_wq == NULL)
4038 return -ENOMEM;
9a8db97e 4039 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4040 set_max_delegations();
80fc015b 4041 return set_callback_cred();
1da177e4
LT		 4042}
4043
29ab23cc 4044int
76a3550e 4045nfs4_state_start(void)
1da177e4 4046{
29ab23cc
BF		 4047 int ret;
4048
1da177e4 4049 if (nfs4_init)
29ab23cc 4050 return 0;
190e4fbf 4051 nfsd4_load_reboot_recovery_data();
29ab23cc
BF		 4052 ret = __nfs4_state_start();
4053 if (ret)
4054 return ret;
1da177e4 4055 nfs4_init = 1;
29ab23cc 4056 return 0;
1da177e4
LT		 4057}
4058
1da177e4
LT		 4059time_t
4060nfs4_lease_time(void)
4061{
4062 return lease_time;
4063}
4064
4065static void
4066__nfs4_state_shutdown(void)
4067{
4068 int i;
4069 struct nfs4_client *clp = NULL;
4070 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4071 struct list_head *pos, *next, reaplist;
4072
1da177e4
LT		 4073 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4074 while (!list_empty(&conf_id_hashtbl[i])) {
4075 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4076 expire_client(clp);
4077 }
4078 while (!list_empty(&unconf_str_hashtbl[i])) {
4079 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4080 expire_client(clp);
4081 }
4082 }
4083 INIT_LIST_HEAD(&reaplist);
4084 spin_lock(&recall_lock);
4085 list_for_each_safe(pos, next, &del_recall_lru) {
4086 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4087 list_move(&dp->dl_recall_lru, &reaplist);
4088 }
4089 spin_unlock(&recall_lock);
4090 list_for_each_safe(pos, next, &reaplist) {
4091 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4092 list_del_init(&dp->dl_recall_lru);
4093 unhash_delegation(dp);
4094 }
4095
190e4fbf 4096 nfsd4_shutdown_recdir();
1da177e4 4097 nfs4_init = 0;
1da177e4
LT		 4098}
4099
4100void
4101nfs4_state_shutdown(void)
4102{
5e8d5c29
N		 4103 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4104 destroy_workqueue(laundry_wq);
2c5e7615 4105 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4106 nfs4_lock_state();
4107 nfs4_release_reclaim();
4108 __nfs4_state_shutdown();
1da177e4
LT		 4109 nfs4_unlock_state();
4110}
4111
3dd98a3b
JL		 4112/*
4113 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4114 * accessed when nfsd is starting.
4115 */
0964a3d3
N		 4116static void
4117nfs4_set_recdir(char *recdir)
4118{
0964a3d3 4119 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4120}
4121
4122/*
4123 * Change the NFSv4 recovery directory to recdir.
4124 */
4125int
4126nfs4_reset_recoverydir(char *recdir)
4127{
4128 int status;
a63bb996 4129 struct path path;
0964a3d3 4130
a63bb996 4131 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4132 if (status)
4133 return status;
4134 status = -ENOTDIR;
a63bb996 4135 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4136 nfs4_set_recdir(recdir);
4137 status = 0;
4138 }
a63bb996 4139 path_put(&path);
0964a3d3
N		 4140 return status;
4141}
4142
3dd98a3b
JL		 4143char *
4144nfs4_recoverydir(void)
4145{
4146 return user_recovery_dirname;
4147}
4148
1da177e4
LT		 4149/*
4150 * Called when leasetime is changed.
4151 *
d99a05ad
N		 4152 * The only way the protocol gives us to handle on-the-fly lease changes is to
4153 * simulate a reboot. Instead of doing that, we just wait till the next time
4154 * we start to register any changes in lease time. If the administrator
4155 * really wants to change the lease time *now*, they can go ahead and bring
4156 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4157 *
4158 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4159 * when nfsd is starting
1da177e4
LT		 4160 */
4161void
4162nfs4_reset_lease(time_t leasetime)
4163{
d99a05ad 4164 user_lease_time = leasetime;
1da177e4 4165}
Linux 6.x block layer and io_uring tree(s)