projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: miscellaneous process_open2 cleanup
[deliverable/linux.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4 1/*
1da177e4
LT		 2* Copyright (c) 2001 The Regents of the University of Michigan.
3* All rights reserved.
4*
5* Kendrick Smith <kmsmith@umich.edu>
6* Andy Adamson <kandros@umich.edu>
7*
8* Redistribution and use in source and binary forms, with or without
9* modification, are permitted provided that the following conditions
10* are met:
11*
12* 1. Redistributions of source code must retain the above copyright
13* notice, this list of conditions and the following disclaimer.
14* 2. Redistributions in binary form must reproduce the above copyright
15* notice, this list of conditions and the following disclaimer in the
16* documentation and/or other materials provided with the distribution.
17* 3. Neither the name of the University nor the names of its
18* contributors may be used to endorse or promote products derived
19* from this software without specific prior written permission.
20*
21* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32*
33*/
34
aceaf78d 35#include <linux/file.h>
1da177e4 36#include <linux/smp_lock.h>
5a0e3ad6 37#include <linux/slab.h>
0964a3d3 38#include <linux/namei.h>
c2f1a551 39#include <linux/swap.h>
68e76ad0 40#include <linux/sunrpc/svcauth_gss.h>
363168b4 41#include <linux/sunrpc/clnt.h>
9a74af21 42#include "xdr4.h"
0a3adade 43#include "vfs.h"
1da177e4
LT		 44
45#define NFSDDBG_FACILITY NFSDDBG_PROC
46
47/* Globals */
cf07d2ea 48time_t nfsd4_lease = 90; /* default lease time */
efc4bb4f 49time_t nfsd4_grace = 90;
fd39ca9a 50static time_t boot_time;
1da177e4
LT		 51static u32 current_ownerid = 1;
52static u32 current_fileid = 1;
53static u32 current_delegid = 1;
fd39ca9a
N		 54static stateid_t zerostateid; /* bits all 0 */
55static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 56static u64 current_sessionid = 1;
fd39ca9a
N		 57
58#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 60
1da177e4 61/* forward declarations */
fd39ca9a 62static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 63static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 64static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
65static void nfs4_set_recdir(char *recdir);
1da177e4 66
8b671b80
BF		 67/* Locking: */
68
69/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 70static DEFINE_MUTEX(client_mutex);
1da177e4 71
8b671b80
BF		 72/*
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
76 */
77static DEFINE_SPINLOCK(recall_lock);
78
e18b890b
CL		 79static struct kmem_cache *stateowner_slab = NULL;
80static struct kmem_cache *file_slab = NULL;
81static struct kmem_cache *stateid_slab = NULL;
82static struct kmem_cache *deleg_slab = NULL;
e60d4398 83
1da177e4
LT		 84void
85nfs4_lock_state(void)
86{
353ab6e9 87 mutex_lock(&client_mutex);
1da177e4
LT		 88}
89
90void
91nfs4_unlock_state(void)
92{
353ab6e9 93 mutex_unlock(&client_mutex);
1da177e4
LT		 94}
95
96static inline u32
97opaque_hashval(const void *ptr, int nbytes)
98{
99 unsigned char *cptr = (unsigned char *) ptr;
100
101 u32 x = 0;
102 while (nbytes--) {
103 x *= 37;
104 x += *cptr++;
105 }
106 return x;
107}
108
1da177e4
LT		 109static struct list_head del_recall_lru;
110
13cd2184
N		 111static inline void
112put_nfs4_file(struct nfs4_file *fi)
113{
8b671b80
BF		 114 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
115 list_del(&fi->fi_hash);
116 spin_unlock(&recall_lock);
117 iput(fi->fi_inode);
118 kmem_cache_free(file_slab, fi);
119 }
13cd2184
N		 120}
121
122static inline void
123get_nfs4_file(struct nfs4_file *fi)
124{
8b671b80 125 atomic_inc(&fi->fi_ref);
13cd2184
N		 126}
127
ef0f3390 128static int num_delegations;
c2f1a551 129unsigned int max_delegations;
ef0f3390
N		 130
131/*
132 * Open owner state (share locks)
133 */
134
135/* hash tables for nfs4_stateowner */
136#define OWNER_HASH_BITS 8
137#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
139
140#define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142#define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
144
145static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
146static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
147
148/* hash table for nfs4_file */
149#define FILE_HASH_BITS 8
150#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
152/* hash table for (open)nfs4_stateid */
153#define STATEID_HASH_BITS 10
154#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
156
157#define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159#define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
161
162static struct list_head file_hashtbl[FILE_HASH_SIZE];
163static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
164
1da177e4
LT		 165static struct nfs4_delegation *
166alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
167{
168 struct nfs4_delegation *dp;
169 struct nfs4_file *fp = stp->st_file;
c237dc03 170 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 171
172 dprintk("NFSD alloc_init_deleg\n");
c3e48080
BF		 173 /*
174 * Major work on the lease subsystem (for example, to support
175 * calbacks on stat) will be required before we can support
176 * write delegations properly.
177 */
178 if (type != NFS4_OPEN_DELEGATE_READ)
179 return NULL;
47f9940c
MS		 180 if (fp->fi_had_conflict)
181 return NULL;
c2f1a551 182 if (num_delegations > max_delegations)
ef0f3390 183 return NULL;
5b2d21c1
N		 184 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
185 if (dp == NULL)
1da177e4 186 return dp;
ef0f3390 187 num_delegations++;
ea1da636
N		 188 INIT_LIST_HEAD(&dp->dl_perfile);
189 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 190 INIT_LIST_HEAD(&dp->dl_recall_lru);
191 dp->dl_client = clp;
13cd2184 192 get_nfs4_file(fp);
1da177e4
LT		 193 dp->dl_file = fp;
194 dp->dl_flock = NULL;
195 get_file(stp->st_vfs_file);
196 dp->dl_vfs_file = stp->st_vfs_file;
197 dp->dl_type = type;
b53d40c5 198 dp->dl_ident = cb->cb_ident;
e4e83ea4 199 dp->dl_stateid.si_boot = boot_time;
1da177e4
LT		 200 dp->dl_stateid.si_stateownerid = current_delegid++;
201 dp->dl_stateid.si_fileid = 0;
202 dp->dl_stateid.si_generation = 0;
6c02eaa1 203 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 204 dp->dl_time = 0;
205 atomic_set(&dp->dl_count, 1);
ea1da636
N		 206 list_add(&dp->dl_perfile, &fp->fi_delegations);
207 list_add(&dp->dl_perclnt, &clp->cl_delegations);
b5a1a81e 208 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
1da177e4
LT		 209 return dp;
210}
211
212void
213nfs4_put_delegation(struct nfs4_delegation *dp)
214{
215 if (atomic_dec_and_test(&dp->dl_count)) {
216 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 217 put_nfs4_file(dp->dl_file);
5b2d21c1 218 kmem_cache_free(deleg_slab, dp);
ef0f3390 219 num_delegations--;
1da177e4
LT		 220 }
221}
222
223/* Remove the associated file_lock first, then remove the delegation.
224 * lease_modify() is called to remove the FS_LEASE file_lock from
225 * the i_flock list, eventually calling nfsd's lock_manager
226 * fl_release_callback.
227 */
228static void
229nfs4_close_delegation(struct nfs4_delegation *dp)
230{
231 struct file *filp = dp->dl_vfs_file;
232
233 dprintk("NFSD: close_delegation dp %p\n",dp);
234 dp->dl_vfs_file = NULL;
235 /* The following nfsd_close may not actually close the file,
236 * but we want to remove the lease in any case. */
c907132d 237 if (dp->dl_flock)
a9933cea 238 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 239 nfsd_close(filp);
1da177e4
LT		 240}
241
242/* Called under the state lock. */
243static void
244unhash_delegation(struct nfs4_delegation *dp)
245{
ea1da636
N		 246 list_del_init(&dp->dl_perfile);
247 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 248 spin_lock(&recall_lock);
249 list_del_init(&dp->dl_recall_lru);
250 spin_unlock(&recall_lock);
251 nfs4_close_delegation(dp);
252 nfs4_put_delegation(dp);
253}
254
255/*
256 * SETCLIENTID state
257 */
258
36acb66b 259/* client_lock protects the client lru list and session hash table */
9089f1b4
BH		 260static DEFINE_SPINLOCK(client_lock);
261
1da177e4
LT		 262/* Hash tables for nfs4_clientid state */
263#define CLIENT_HASH_BITS 4
264#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
265#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
266
267#define clientid_hashval(id) \
268 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 269#define clientstr_hashval(name) \
270 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 271/*
272 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
273 * used in reboot/reset lease grace period processing
274 *
275 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
276 * setclientid_confirmed info.
277 *
278 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
279 * setclientid info.
280 *
281 * client_lru holds client queue ordered by nfs4_client.cl_time
282 * for lease renewal.
283 *
284 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
285 * for last close replay.
286 */
287static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
288static int reclaim_str_hashtbl_size = 0;
289static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
290static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
291static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
292static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
293static struct list_head client_lru;
294static struct list_head close_lru;
295
2283963f
BF		 296static void unhash_generic_stateid(struct nfs4_stateid *stp)
297{
298 list_del(&stp->st_hash);
299 list_del(&stp->st_perfile);
300 list_del(&stp->st_perstateowner);
301}
302
303static void free_generic_stateid(struct nfs4_stateid *stp)
304{
305 put_nfs4_file(stp->st_file);
306 kmem_cache_free(stateid_slab, stp);
307}
308
309static void release_lock_stateid(struct nfs4_stateid *stp)
310{
311 unhash_generic_stateid(stp);
312 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
313 free_generic_stateid(stp);
314}
315
f044ff83
BF		 316static void unhash_lockowner(struct nfs4_stateowner *sop)
317{
318 struct nfs4_stateid *stp;
319
320 list_del(&sop->so_idhash);
321 list_del(&sop->so_strhash);
322 list_del(&sop->so_perstateid);
323 while (!list_empty(&sop->so_stateids)) {
324 stp = list_first_entry(&sop->so_stateids,
325 struct nfs4_stateid, st_perstateowner);
326 release_lock_stateid(stp);
327 }
328}
329
330static void release_lockowner(struct nfs4_stateowner *sop)
331{
332 unhash_lockowner(sop);
333 nfs4_put_stateowner(sop);
334}
335
f1d110ca
BF		 336static void
337release_stateid_lockowners(struct nfs4_stateid *open_stp)
338{
339 struct nfs4_stateowner *lock_sop;
340
341 while (!list_empty(&open_stp->st_lockowners)) {
342 lock_sop = list_entry(open_stp->st_lockowners.next,
343 struct nfs4_stateowner, so_perstateid);
344 /* list_del(&open_stp->st_lockowners); */
345 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 346 release_lockowner(lock_sop);
f1d110ca
BF		 347 }
348}
349
2283963f
BF		 350static void release_open_stateid(struct nfs4_stateid *stp)
351{
352 unhash_generic_stateid(stp);
353 release_stateid_lockowners(stp);
354 nfsd_close(stp->st_vfs_file);
355 free_generic_stateid(stp);
356}
357
f044ff83 358static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 359{
360 struct nfs4_stateid *stp;
361
362 list_del(&sop->so_idhash);
363 list_del(&sop->so_strhash);
f044ff83
BF		 364 list_del(&sop->so_perclient);
365 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 366 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 367 stp = list_first_entry(&sop->so_stateids,
368 struct nfs4_stateid, st_perstateowner);
369 release_open_stateid(stp);
f1d110ca
BF		 370 }
371}
372
f044ff83 373static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 374{
f044ff83 375 unhash_openowner(sop);
f1d110ca
BF		 376 list_del(&sop->so_close_lru);
377 nfs4_put_stateowner(sop);
378}
379
5282fd72
ME		 380#define SESSION_HASH_SIZE 512
381static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
382
383static inline int
384hash_sessionid(struct nfs4_sessionid *sessionid)
385{
386 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
387
388 return sid->sequence % SESSION_HASH_SIZE;
389}
390
391static inline void
392dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
393{
394 u32 *ptr = (u32 *)(&sessionid->data[0]);
395 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
396}
397
ec6b5d7b
AA		 398static void
399gen_sessionid(struct nfsd4_session *ses)
400{
401 struct nfs4_client *clp = ses->se_client;
402 struct nfsd4_sessionid *sid;
403
404 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
405 sid->clientid = clp->cl_clientid;
406 sid->sequence = current_sessionid++;
407 sid->reserved = 0;
408}
409
410/*
a649637c
AA		 411 * The protocol defines ca_maxresponssize_cached to include the size of
412 * the rpc header, but all we need to cache is the data starting after
413 * the end of the initial SEQUENCE operation--the rest we regenerate
414 * each time. Therefore we can advertise a ca_maxresponssize_cached
415 * value that is the number of bytes in our cache plus a few additional
416 * bytes. In order to stay on the safe side, and not promise more than
417 * we can cache, those additional bytes must be the minimum possible: 24
418 * bytes of rpc header (xid through accept state, with AUTH_NULL
419 * verifier), 12 for the compound header (with zero-length tag), and 44
420 * for the SEQUENCE op response:
421 */
422#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
423
424/*
425 * Give the client the number of ca_maxresponsesize_cached slots it
426 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
427 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
428 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 429 *
a649637c
AA		 430 * If we run out of reserved DRC memory we should (up to a point)
431 * re-negotiate active sessions and reduce their slot usage to make
432 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 433 */
a649637c 434static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 435{
a649637c 436 int mem, size = fchan->maxresp_cached;
ec6b5d7b 437
5d77ddfb
AA		 438 if (fchan->maxreqs < 1)
439 return nfserr_inval;
5d77ddfb 440
a649637c
AA		 441 if (size < NFSD_MIN_HDR_SEQ_SZ)
442 size = NFSD_MIN_HDR_SEQ_SZ;
443 size -= NFSD_MIN_HDR_SEQ_SZ;
444 if (size > NFSD_SLOT_CACHE_SIZE)
445 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 446
a649637c
AA		 447 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
448 mem = fchan->maxreqs * size;
449 if (mem > NFSD_MAX_MEM_PER_SESSION) {
450 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
451 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
452 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
453 mem = fchan->maxreqs * size;
454 }
ec6b5d7b 455
4bd9b0f4 456 spin_lock(&nfsd_drc_lock);
a649637c
AA		 457 /* bound the total session drc memory ussage */
458 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
459 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
460 mem = fchan->maxreqs * size;
461 }
0c193054 462 nfsd_drc_mem_used += mem;
4bd9b0f4 463 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 464
b101ebbc 465 if (fchan->maxreqs == 0)
4731030d 466 return nfserr_jukebox;
a649637c
AA		 467
468 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 469 return 0;
ec6b5d7b
AA		 470}
471
472/*
473 * fchan holds the client values on input, and the server values on output
ddc04fd4 474 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA		 475 */
476static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 477 struct nfsd4_channel_attrs *session_fchan,
478 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 479{
480 int status = 0;
ddc04fd4 481 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA		 482
483 /* headerpadsz set to zero in encode routine */
484
485 /* Use the client's max request and max response size if possible */
486 if (fchan->maxreq_sz > maxcount)
487 fchan->maxreq_sz = maxcount;
6c18ba9f 488 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 489
490 if (fchan->maxresp_sz > maxcount)
491 fchan->maxresp_sz = maxcount;
6c18ba9f 492 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 493
ec6b5d7b
AA		 494 /* Use the client's maxops if possible */
495 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
496 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 497 session_fchan->maxops = fchan->maxops;
ec6b5d7b 498
ec6b5d7b
AA		 499 /* FIXME: Error means no more DRC pages so the server should
500 * recover pages from existing sessions. For now fail session
501 * creation.
502 */
a649637c 503 status = set_forechannel_drc_size(fchan);
ec6b5d7b 504
a649637c 505 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 506 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 507
508 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 509 return status;
510}
511
557ce264
AA		 512static void
513free_session_slots(struct nfsd4_session *ses)
514{
515 int i;
516
517 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
518 kfree(ses->se_slots[i]);
519}
520
efe0cb6d
BF		 521/*
522 * We don't actually need to cache the rpc and session headers, so we
523 * can allocate a little less for each slot:
524 */
525static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
526{
527 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
528}
529
ec6b5d7b
AA		 530static int
531alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
532 struct nfsd4_create_session *cses)
533{
534 struct nfsd4_session *new, tmp;
557ce264
AA		 535 struct nfsd4_slot *sp;
536 int idx, slotsize, cachesize, i;
537 int status;
ec6b5d7b
AA		 538
539 memset(&tmp, 0, sizeof(tmp));
540
541 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 542 tmp.se_bchannel = cses->back_channel;
543 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
544 &cses->fore_channel);
ec6b5d7b
AA		 545 if (status)
546 goto out;
547
557ce264
AA		 548 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
549 + sizeof(struct nfsd4_session) > PAGE_SIZE);
550
4731030d 551 status = nfserr_jukebox;
557ce264
AA		 552 /* allocate struct nfsd4_session and slot table pointers in one piece */
553 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 554 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
555 if (!new)
556 goto out;
557
558 memcpy(new, &tmp, sizeof(*new));
559
557ce264 560 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 561 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA		 562 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
563 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
564 if (!sp)
565 goto out_free;
566 new->se_slots[i] = sp;
567 }
568
ec6b5d7b
AA		 569 new->se_client = clp;
570 gen_sessionid(new);
571 idx = hash_sessionid(&new->se_sessionid);
572 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
573 NFS4_MAX_SESSIONID_LEN);
574
575 new->se_flags = cses->flags;
576 kref_init(&new->se_ref);
9089f1b4 577 spin_lock(&client_lock);
ec6b5d7b
AA		 578 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
579 list_add(&new->se_perclnt, &clp->cl_sessions);
9089f1b4 580 spin_unlock(&client_lock);
ec6b5d7b
AA		 581
582 status = nfs_ok;
583out:
584 return status;
557ce264
AA		 585out_free:
586 free_session_slots(new);
587 kfree(new);
588 goto out;
ec6b5d7b
AA		 589}
590
9089f1b4 591/* caller must hold client_lock */
5282fd72
ME		 592static struct nfsd4_session *
593find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
594{
595 struct nfsd4_session *elem;
596 int idx;
597
598 dump_sessionid(__func__, sessionid);
599 idx = hash_sessionid(sessionid);
5282fd72
ME		 600 /* Search in the appropriate list */
601 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
5282fd72
ME		 602 if (!memcmp(elem->se_sessionid.data, sessionid->data,
603 NFS4_MAX_SESSIONID_LEN)) {
604 return elem;
605 }
606 }
607
608 dprintk("%s: session not found\n", __func__);
609 return NULL;
610}
611
9089f1b4 612/* caller must hold client_lock */
7116ed6b 613static void
5282fd72 614unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 615{
616 list_del(&ses->se_hash);
617 list_del(&ses->se_perclnt);
5282fd72
ME		 618}
619
7116ed6b
AA		 620void
621free_session(struct kref *kref)
622{
623 struct nfsd4_session *ses;
dd829c45 624 int mem;
7116ed6b
AA		 625
626 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 627 spin_lock(&nfsd_drc_lock);
efe0cb6d 628 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 629 nfsd_drc_mem_used -= mem;
be98d1bb 630 spin_unlock(&nfsd_drc_lock);
557ce264 631 free_session_slots(ses);
7116ed6b
AA		 632 kfree(ses);
633}
634
36acb66b 635/* must be called under the client_lock */
1da177e4 636static inline void
36acb66b 637renew_client_locked(struct nfs4_client *clp)
1da177e4 638{
07cd4909
BH		 639 if (is_client_expired(clp)) {
640 dprintk("%s: client (clientid %08x/%08x) already expired\n",
641 __func__,
642 clp->cl_clientid.cl_boot,
643 clp->cl_clientid.cl_id);
644 return;
645 }
646
1da177e4
LT		 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
36acb66b
BH		 657static inline void
658renew_client(struct nfs4_client *clp)
659{
660 spin_lock(&client_lock);
661 renew_client_locked(clp);
662 spin_unlock(&client_lock);
663}
664
1da177e4
LT		 665/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
666static int
667STALE_CLIENTID(clientid_t *clid)
668{
669 if (clid->cl_boot == boot_time)
670 return 0;
60adfc50
AA		 671 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
672 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 673 return 1;
674}
675
676/*
677 * XXX Should we use a slab cache ?
678 * This type of memory management is somewhat inefficient, but we use it
679 * anyway since SETCLIENTID is not a common operation.
680 */
35bba9a3 681static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 682{
683 struct nfs4_client *clp;
684
35bba9a3
BF		 685 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
686 if (clp == NULL)
687 return NULL;
688 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
689 if (clp->cl_name.data == NULL) {
690 kfree(clp);
691 return NULL;
1da177e4 692 }
35bba9a3
BF		 693 memcpy(clp->cl_name.data, name.data, name.len);
694 clp->cl_name.len = name.len;
1da177e4
LT		 695 return clp;
696}
697
698static inline void
699free_client(struct nfs4_client *clp)
700{
701 if (clp->cl_cred.cr_group_info)
702 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 703 kfree(clp->cl_principal);
1da177e4
LT		 704 kfree(clp->cl_name.data);
705 kfree(clp);
706}
707
d7682988
BH		 708void
709release_session_client(struct nfsd4_session *session)
710{
711 struct nfs4_client *clp = session->se_client;
712
713 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
714 return;
715 if (is_client_expired(clp)) {
716 free_client(clp);
717 session->se_client = NULL;
718 } else
719 renew_client_locked(clp);
720 spin_unlock(&client_lock);
d7682988
BH		 721}
722
84d38ac9
BH		 723/* must be called under the client_lock */
724static inline void
725unhash_client_locked(struct nfs4_client *clp)
726{
07cd4909 727 mark_client_expired(clp);
84d38ac9
BH		 728 list_del(&clp->cl_lru);
729 while (!list_empty(&clp->cl_sessions)) {
730 struct nfsd4_session *ses;
731 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
732 se_perclnt);
733 unhash_session(ses);
734 nfsd4_put_session(ses);
735 }
736}
737
1da177e4
LT		 738static void
739expire_client(struct nfs4_client *clp)
740{
741 struct nfs4_stateowner *sop;
742 struct nfs4_delegation *dp;
1da177e4
LT		 743 struct list_head reaplist;
744
1da177e4
LT		 745 INIT_LIST_HEAD(&reaplist);
746 spin_lock(&recall_lock);
ea1da636
N		 747 while (!list_empty(&clp->cl_delegations)) {
748 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 749 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
750 dp->dl_flock);
ea1da636 751 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 752 list_move(&dp->dl_recall_lru, &reaplist);
753 }
754 spin_unlock(&recall_lock);
755 while (!list_empty(&reaplist)) {
756 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
757 list_del_init(&dp->dl_recall_lru);
758 unhash_delegation(dp);
759 }
ea1da636
N		 760 while (!list_empty(&clp->cl_openowners)) {
761 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 762 release_openowner(sop);
1da177e4 763 }
84d38ac9
BH		 764 nfsd4_set_callback_client(clp, NULL);
765 if (clp->cl_cb_conn.cb_xprt)
766 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
36acb66b
BH		 767 list_del(&clp->cl_idhash);
768 list_del(&clp->cl_strhash);
be1fdf6c 769 spin_lock(&client_lock);
84d38ac9 770 unhash_client_locked(clp);
46583e25
BH		 771 if (atomic_read(&clp->cl_refcount) == 0)
772 free_client(clp);
be1fdf6c 773 spin_unlock(&client_lock);
1da177e4
LT		 774}
775
35bba9a3
BF		 776static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
777{
778 memcpy(target->cl_verifier.data, source->data,
779 sizeof(target->cl_verifier.data));
1da177e4
LT		 780}
781
35bba9a3
BF		 782static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
783{
1da177e4
LT		 784 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
785 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
786}
787
35bba9a3
BF		 788static void copy_cred(struct svc_cred *target, struct svc_cred *source)
789{
1da177e4
LT		 790 target->cr_uid = source->cr_uid;
791 target->cr_gid = source->cr_gid;
792 target->cr_group_info = source->cr_group_info;
793 get_group_info(target->cr_group_info);
794}
795
35bba9a3 796static int same_name(const char *n1, const char *n2)
599e0a22 797{
a55370a3 798 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 799}
800
801static int
599e0a22
BF		 802same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
803{
804 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 805}
806
807static int
599e0a22
BF		 808same_clid(clientid_t *cl1, clientid_t *cl2)
809{
810 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 811}
812
813/* XXX what about NGROUP */
814static int
599e0a22
BF		 815same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
816{
817 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 818}
819
5ec7b46c
BF		 820static void gen_clid(struct nfs4_client *clp)
821{
822 static u32 current_clientid = 1;
823
1da177e4
LT		 824 clp->cl_clientid.cl_boot = boot_time;
825 clp->cl_clientid.cl_id = current_clientid++;
826}
827
deda2faa
BF		 828static void gen_confirm(struct nfs4_client *clp)
829{
830 static u32 i;
831 u32 *p;
1da177e4 832
1da177e4 833 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 834 *p++ = get_seconds();
835 *p++ = i++;
1da177e4
LT		 836}
837
b09333c4
RL		 838static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
839 struct svc_rqst *rqstp, nfs4_verifier *verf)
840{
841 struct nfs4_client *clp;
842 struct sockaddr *sa = svc_addr(rqstp);
843 char *princ;
844
845 clp = alloc_client(name);
846 if (clp == NULL)
847 return NULL;
848
849 princ = svc_gss_principal(rqstp);
850 if (princ) {
851 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
852 if (clp->cl_principal == NULL) {
853 free_client(clp);
854 return NULL;
855 }
856 }
857
858 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
46583e25 859 atomic_set(&clp->cl_refcount, 0);
2bf23875 860 atomic_set(&clp->cl_cb_set, 0);
b09333c4
RL		 861 INIT_LIST_HEAD(&clp->cl_idhash);
862 INIT_LIST_HEAD(&clp->cl_strhash);
863 INIT_LIST_HEAD(&clp->cl_openowners);
864 INIT_LIST_HEAD(&clp->cl_delegations);
865 INIT_LIST_HEAD(&clp->cl_sessions);
866 INIT_LIST_HEAD(&clp->cl_lru);
07cd4909 867 clp->cl_time = get_seconds();
b09333c4
RL		 868 clear_bit(0, &clp->cl_cb_slot_busy);
869 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
870 copy_verf(clp, verf);
871 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
872 clp->cl_flavor = rqstp->rq_flavor;
873 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
874 gen_confirm(clp);
875
876 return clp;
877}
878
35bba9a3
BF		 879static int check_name(struct xdr_netobj name)
880{
1da177e4
LT		 881 if (name.len == 0)
882 return 0;
883 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 884 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 885 return 0;
886 }
887 return 1;
888}
889
fd39ca9a 890static void
1da177e4
LT		 891add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
892{
893 unsigned int idhashval;
894
895 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
896 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
897 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
36acb66b 898 renew_client(clp);
1da177e4
LT		 899}
900
fd39ca9a 901static void
1da177e4
LT		 902move_to_confirmed(struct nfs4_client *clp)
903{
904 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
905 unsigned int strhashval;
906
907 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
f116629d 908 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 909 strhashval = clientstr_hashval(clp->cl_recdir);
328efbab 910 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1da177e4
LT		 911 renew_client(clp);
912}
913
914static struct nfs4_client *
915find_confirmed_client(clientid_t *clid)
916{
917 struct nfs4_client *clp;
918 unsigned int idhashval = clientid_hashval(clid->cl_id);
919
920 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 921 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 922 return clp;
923 }
924 return NULL;
925}
926
927static struct nfs4_client *
928find_unconfirmed_client(clientid_t *clid)
929{
930 struct nfs4_client *clp;
931 unsigned int idhashval = clientid_hashval(clid->cl_id);
932
933 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 934 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 935 return clp;
936 }
937 return NULL;
938}
939
a1bcecd2
AA		 940/*
941 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
942 * parameter. Matching is based on the fact the at least one of the
943 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
944 *
945 * FIXME: we need to unify the clientid namespaces for nfsv4.x
946 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
947 * and SET_CLIENTID{,_CONFIRM}
948 */
949static inline int
950match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
951{
952 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
953 return use_exchange_id == has_exchange_flags;
954}
955
28ce6054 956static struct nfs4_client *
a1bcecd2
AA		 957find_confirmed_client_by_str(const char *dname, unsigned int hashval,
958 bool use_exchange_id)
28ce6054
N		 959{
960 struct nfs4_client *clp;
961
962 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 963 if (same_name(clp->cl_recdir, dname) &&
964 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 965 return clp;
966 }
967 return NULL;
968}
969
970static struct nfs4_client *
a1bcecd2
AA		 971find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
972 bool use_exchange_id)
28ce6054
N		 973{
974 struct nfs4_client *clp;
975
976 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 977 if (same_name(clp->cl_recdir, dname) &&
978 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 979 return clp;
980 }
981 return NULL;
982}
983
fd39ca9a 984static void
fbf4665f 985gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 986{
c237dc03 987 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 988 unsigned short expected_family;
989
990 /* Currently, we only support tcp and tcp6 for the callback channel */
991 if (se->se_callback_netid_len == 3 &&
992 !memcmp(se->se_callback_netid_val, "tcp", 3))
993 expected_family = AF_INET;
994 else if (se->se_callback_netid_len == 4 &&
995 !memcmp(se->se_callback_netid_val, "tcp6", 4))
996 expected_family = AF_INET6;
997 else
1da177e4
LT		 998 goto out_err;
999
aa9a4ec7
JL		 1000 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1001 se->se_callback_addr_len,
1002 (struct sockaddr *) &cb->cb_addr,
1003 sizeof(cb->cb_addr));
1004
7077ecba 1005 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 1006 goto out_err;
aa9a4ec7 1007
fbf4665f
JL		 1008 if (cb->cb_addr.ss_family == AF_INET6)
1009 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
1010
ab52ae6d 1011 cb->cb_minorversion = 0;
1da177e4
LT		 1012 cb->cb_prog = se->se_callback_prog;
1013 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 1014 return;
1015out_err:
aa9a4ec7
JL		 1016 cb->cb_addr.ss_family = AF_UNSPEC;
1017 cb->cb_addrlen = 0;
849823c5 1018 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 1019 "will not receive delegations\n",
1020 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1021
1da177e4
LT		 1022 return;
1023}
1024
074fe897 1025/*
557ce264 1026 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 1027 */
074fe897
AA		 1028void
1029nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 1030{
557ce264
AA		 1031 struct nfsd4_slot *slot = resp->cstate.slot;
1032 unsigned int base;
074fe897 1033
557ce264 1034 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1035
557ce264
AA		 1036 slot->sl_opcnt = resp->opcnt;
1037 slot->sl_status = resp->cstate.status;
074fe897 1038
bf864a31 1039 if (nfsd4_not_cached(resp)) {
557ce264 1040 slot->sl_datalen = 0;
bf864a31 1041 return;
074fe897 1042 }
557ce264
AA		 1043 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1044 base = (char *)resp->cstate.datap -
1045 (char *)resp->xbuf->head[0].iov_base;
1046 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1047 slot->sl_datalen))
1048 WARN("%s: sessions DRC could not cache compound\n", __func__);
1049 return;
074fe897
AA		 1050}
1051
1052/*
abfabf8c
AA		 1053 * Encode the replay sequence operation from the slot values.
1054 * If cachethis is FALSE encode the uncached rep error on the next
1055 * operation which sets resp->p and increments resp->opcnt for
1056 * nfs4svc_encode_compoundres.
074fe897 1057 *
074fe897 1058 */
abfabf8c
AA		 1059static __be32
1060nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1061 struct nfsd4_compoundres *resp)
074fe897 1062{
abfabf8c
AA		 1063 struct nfsd4_op *op;
1064 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1065
abfabf8c 1066 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1067 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1068
abfabf8c
AA		 1069 /* Encode the replayed sequence operation */
1070 op = &args->ops[resp->opcnt - 1];
1071 nfsd4_encode_operation(resp, op);
bf864a31 1072
abfabf8c 1073 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1074 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1075 op = &args->ops[resp->opcnt++];
1076 op->status = nfserr_retry_uncached_rep;
1077 nfsd4_encode_operation(resp, op);
074fe897 1078 }
abfabf8c 1079 return op->status;
074fe897
AA		 1080}
1081
1082/*
557ce264
AA		 1083 * The sequence operation is not cached because we can use the slot and
1084 * session values.
074fe897
AA		 1085 */
1086__be32
bf864a31
AA		 1087nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1088 struct nfsd4_sequence *seq)
074fe897 1089{
557ce264 1090 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1091 __be32 status;
1092
557ce264 1093 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1094
abfabf8c
AA		 1095 /* Either returns 0 or nfserr_retry_uncached */
1096 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1097 if (status == nfserr_retry_uncached_rep)
1098 return status;
074fe897 1099
557ce264
AA		 1100 /* The sequence operation has been encoded, cstate->datap set. */
1101 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1102
557ce264
AA		 1103 resp->opcnt = slot->sl_opcnt;
1104 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1105 status = slot->sl_status;
074fe897
AA		 1106
1107 return status;
1108}
1109
0733d213
AA		 1110/*
1111 * Set the exchange_id flags returned by the server.
1112 */
1113static void
1114nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1115{
1116 /* pNFS is not supported */
1117 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1118
1119 /* Referrals are supported, Migration is not. */
1120 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1121
1122 /* set the wire flags to return to client. */
1123 clid->flags = new->cl_exchange_flags;
1124}
1125
069b6ad4
AA		 1126__be32
1127nfsd4_exchange_id(struct svc_rqst *rqstp,
1128 struct nfsd4_compound_state *cstate,
1129 struct nfsd4_exchange_id *exid)
1130{
0733d213
AA		 1131 struct nfs4_client *unconf, *conf, *new;
1132 int status;
1133 unsigned int strhashval;
1134 char dname[HEXDIR_LEN];
363168b4 1135 char addr_str[INET6_ADDRSTRLEN];
0733d213 1136 nfs4_verifier verf = exid->verifier;
363168b4 1137 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1138
363168b4 1139 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1140 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1141 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1142 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1143 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1144
1145 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1146 return nfserr_inval;
1147
1148 /* Currently only support SP4_NONE */
1149 switch (exid->spa_how) {
1150 case SP4_NONE:
1151 break;
1152 case SP4_SSV:
1153 return nfserr_encr_alg_unsupp;
1154 default:
1155 BUG(); /* checked by xdr code */
1156 case SP4_MACH_CRED:
1157 return nfserr_serverfault; /* no excuse :-/ */
1158 }
1159
1160 status = nfs4_make_rec_clidname(dname, &exid->clname);
1161
1162 if (status)
1163 goto error;
1164
1165 strhashval = clientstr_hashval(dname);
1166
1167 nfs4_lock_state();
1168 status = nfs_ok;
1169
a1bcecd2 1170 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1171 if (conf) {
1172 if (!same_verf(&verf, &conf->cl_verifier)) {
1173 /* 18.35.4 case 8 */
1174 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1175 status = nfserr_not_same;
1176 goto out;
1177 }
1178 /* Client reboot: destroy old state */
1179 expire_client(conf);
1180 goto out_new;
1181 }
1182 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1183 /* 18.35.4 case 9 */
1184 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1185 status = nfserr_perm;
1186 goto out;
1187 }
1188 expire_client(conf);
1189 goto out_new;
1190 }
0733d213
AA		 1191 /*
1192 * Set bit when the owner id and verifier map to an already
1193 * confirmed client id (18.35.3).
1194 */
1195 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1196
1197 /*
1198 * Falling into 18.35.4 case 2, possible router replay.
1199 * Leave confirmed record intact and return same result.
1200 */
1201 copy_verf(conf, &verf);
1202 new = conf;
1203 goto out_copy;
6ddbbbfe
MS		 1204 }
1205
1206 /* 18.35.4 case 7 */
1207 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1208 status = nfserr_noent;
1209 goto out;
0733d213
AA		 1210 }
1211
a1bcecd2 1212 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1213 if (unconf) {
1214 /*
1215 * Possible retry or client restart. Per 18.35.4 case 4,
1216 * a new unconfirmed record should be generated regardless
1217 * of whether any properties have changed.
1218 */
1219 expire_client(unconf);
1220 }
1221
1222out_new:
1223 /* Normal case */
b09333c4 1224 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1225 if (new == NULL) {
4731030d 1226 status = nfserr_jukebox;
0733d213
AA		 1227 goto out;
1228 }
1229
0733d213 1230 gen_clid(new);
0733d213
AA		 1231 add_to_unconfirmed(new, strhashval);
1232out_copy:
1233 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1234 exid->clientid.cl_id = new->cl_clientid.cl_id;
1235
38eb76a5 1236 exid->seqid = 1;
0733d213
AA		 1237 nfsd4_set_ex_flags(new, exid);
1238
1239 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1240 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1241 status = nfs_ok;
1242
1243out:
1244 nfs4_unlock_state();
1245error:
1246 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1247 return status;
069b6ad4
AA		 1248}
1249
b85d4c01 1250static int
88e588d5 1251check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1252{
88e588d5
AA		 1253 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1254 slot_seqid);
b85d4c01
BH		 1255
1256 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1257 if (slot_inuse) {
1258 if (seqid == slot_seqid)
b85d4c01
BH		 1259 return nfserr_jukebox;
1260 else
1261 return nfserr_seq_misordered;
1262 }
1263 /* Normal */
88e588d5 1264 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1265 return nfs_ok;
1266 /* Replay */
88e588d5 1267 if (seqid == slot_seqid)
b85d4c01
BH		 1268 return nfserr_replay_cache;
1269 /* Wraparound */
88e588d5 1270 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1271 return nfs_ok;
1272 /* Misordered replay or misordered new request */
1273 return nfserr_seq_misordered;
1274}
1275
49557cc7
AA		 1276/*
1277 * Cache the create session result into the create session single DRC
1278 * slot cache by saving the xdr structure. sl_seqid has been set.
1279 * Do this for solo or embedded create session operations.
1280 */
1281static void
1282nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1283 struct nfsd4_clid_slot *slot, int nfserr)
1284{
1285 slot->sl_status = nfserr;
1286 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1287}
1288
1289static __be32
1290nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1291 struct nfsd4_clid_slot *slot)
1292{
1293 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1294 return slot->sl_status;
1295}
1296
069b6ad4
AA		 1297__be32
1298nfsd4_create_session(struct svc_rqst *rqstp,
1299 struct nfsd4_compound_state *cstate,
1300 struct nfsd4_create_session *cr_ses)
1301{
363168b4 1302 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1303 struct nfs4_client *conf, *unconf;
49557cc7 1304 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1305 int status = 0;
1306
1307 nfs4_lock_state();
1308 unconf = find_unconfirmed_client(&cr_ses->clientid);
1309 conf = find_confirmed_client(&cr_ses->clientid);
1310
1311 if (conf) {
49557cc7
AA		 1312 cs_slot = &conf->cl_cs_slot;
1313 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1314 if (status == nfserr_replay_cache) {
ec6b5d7b 1315 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1316 cs_slot->sl_seqid);
38eb76a5 1317 /* Return the cached reply status */
49557cc7 1318 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1319 goto out;
49557cc7 1320 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1321 status = nfserr_seq_misordered;
1322 dprintk("Sequence misordered!\n");
1323 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1324 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1325 goto out;
1326 }
49557cc7 1327 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1328 } else if (unconf) {
1329 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1330 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1331 status = nfserr_clid_inuse;
1332 goto out;
1333 }
1334
49557cc7
AA		 1335 cs_slot = &unconf->cl_cs_slot;
1336 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1337 if (status) {
1338 /* an unconfirmed replay returns misordered */
ec6b5d7b 1339 status = nfserr_seq_misordered;
49557cc7 1340 goto out_cache;
ec6b5d7b
AA		 1341 }
1342
49557cc7 1343 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1344 move_to_confirmed(unconf);
1345
38524ab3 1346 if (cr_ses->flags & SESSION4_BACK_CHAN) {
2bf23875
BF		 1347 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1348 svc_xprt_get(rqstp->rq_xprt);
38524ab3
AA		 1349 rpc_copy_addr(
1350 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1351 sa);
1352 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1353 unconf->cl_cb_conn.cb_minorversion =
1354 cstate->minorversion;
1355 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1356 unconf->cl_cb_seq_nr = 1;
4b21d0de 1357 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
38524ab3 1358 }
ec6b5d7b
AA		 1359 conf = unconf;
1360 } else {
1361 status = nfserr_stale_clientid;
1362 goto out;
1363 }
1364
408b79bc
BF		 1365 /*
1366 * We do not support RDMA or persistent sessions
1367 */
1368 cr_ses->flags &= ~SESSION4_PERSIST;
1369 cr_ses->flags &= ~SESSION4_RDMA;
1370
ec6b5d7b
AA		 1371 status = alloc_init_session(rqstp, conf, cr_ses);
1372 if (status)
1373 goto out;
1374
ec6b5d7b
AA		 1375 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1376 NFS4_MAX_SESSIONID_LEN);
49557cc7 1377 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1378
49557cc7
AA		 1379out_cache:
1380 /* cache solo and embedded create sessions under the state lock */
1381 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1382out:
1383 nfs4_unlock_state();
1384 dprintk("%s returns %d\n", __func__, ntohl(status));
1385 return status;
069b6ad4
AA		 1386}
1387
57716355
BF		 1388static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1389{
1390 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1391 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1392
1393 return argp->opcnt == resp->opcnt;
1394}
1395
5d4cec2f
BF		 1396static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1397{
1398 if (!session)
1399 return 0;
1400 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1401}
1402
069b6ad4
AA		 1403__be32
1404nfsd4_destroy_session(struct svc_rqst *r,
1405 struct nfsd4_compound_state *cstate,
1406 struct nfsd4_destroy_session *sessionid)
1407{
e10e0cfc
BH		 1408 struct nfsd4_session *ses;
1409 u32 status = nfserr_badsession;
1410
1411 /* Notes:
1412 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1413 * - Should we return nfserr_back_chan_busy if waiting for
1414 * callbacks on to-be-destroyed session?
1415 * - Do we need to clear any callback info from previous session?
1416 */
1417
5d4cec2f 1418 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
57716355
BF		 1419 if (!nfsd4_last_compound_op(r))
1420 return nfserr_not_only_op;
1421 }
e10e0cfc 1422 dump_sessionid(__func__, &sessionid->sessionid);
9089f1b4 1423 spin_lock(&client_lock);
e10e0cfc
BH		 1424 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1425 if (!ses) {
9089f1b4 1426 spin_unlock(&client_lock);
e10e0cfc
BH		 1427 goto out;
1428 }
1429
1430 unhash_session(ses);
9089f1b4 1431 spin_unlock(&client_lock);
e10e0cfc 1432
ab707e15 1433 nfs4_lock_state();
e10e0cfc 1434 /* wait for callbacks */
b5a1a81e 1435 nfsd4_set_callback_client(ses->se_client, NULL);
ab707e15 1436 nfs4_unlock_state();
e10e0cfc
BH		 1437 nfsd4_put_session(ses);
1438 status = nfs_ok;
1439out:
1440 dprintk("%s returns %d\n", __func__, ntohl(status));
1441 return status;
069b6ad4
AA		 1442}
1443
1444__be32
b85d4c01 1445nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1446 struct nfsd4_compound_state *cstate,
1447 struct nfsd4_sequence *seq)
1448{
f9bb94c4 1449 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1450 struct nfsd4_session *session;
1451 struct nfsd4_slot *slot;
1452 int status;
1453
f9bb94c4
AA		 1454 if (resp->opcnt != 1)
1455 return nfserr_sequence_pos;
1456
9089f1b4 1457 spin_lock(&client_lock);
b85d4c01
BH		 1458 status = nfserr_badsession;
1459 session = find_in_sessionid_hashtbl(&seq->sessionid);
1460 if (!session)
1461 goto out;
1462
1463 status = nfserr_badslot;
6c18ba9f 1464 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1465 goto out;
1466
557ce264 1467 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1468 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1469
a8dfdaeb
AA		 1470 /* We do not negotiate the number of slots yet, so set the
1471 * maxslots to the session maxreqs which is used to encode
1472 * sr_highest_slotid and the sr_target_slot id to maxslots */
1473 seq->maxslots = session->se_fchannel.maxreqs;
1474
88e588d5 1475 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1476 if (status == nfserr_replay_cache) {
1477 cstate->slot = slot;
1478 cstate->session = session;
da3846a2 1479 /* Return the cached reply status and set cstate->status
557ce264 1480 * for nfsd4_proc_compound processing */
bf864a31 1481 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1482 cstate->status = nfserr_replay_cache;
aaf84eb9 1483 goto out;
b85d4c01
BH		 1484 }
1485 if (status)
1486 goto out;
1487
1488 /* Success! bump slot seqid */
1489 slot->sl_inuse = true;
1490 slot->sl_seqid = seq->seqid;
557ce264 1491 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1492
1493 cstate->slot = slot;
1494 cstate->session = session;
1495
b85d4c01 1496out:
26c0c75e 1497 /* Hold a session reference until done processing the compound. */
aaf84eb9 1498 if (cstate->session) {
36acb66b 1499 nfsd4_get_session(cstate->session);
d7682988 1500 atomic_inc(&session->se_client->cl_refcount);
aaf84eb9 1501 }
36acb66b 1502 spin_unlock(&client_lock);
b85d4c01
BH		 1503 dprintk("%s: return %d\n", __func__, ntohl(status));
1504 return status;
069b6ad4
AA		 1505}
1506
4dc6ec00
BF		 1507__be32
1508nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
1509{
1510 if (rc->rca_one_fs) {
1511 if (!cstate->current_fh.fh_dentry)
1512 return nfserr_nofilehandle;
1513 /*
1514 * We don't take advantage of the rca_one_fs case.
1515 * That's OK, it's optional, we can safely ignore it.
1516 */
1517 return nfs_ok;
1518 }
1519 nfs4_lock_state();
1520 if (is_client_expired(cstate->session->se_client)) {
1521 nfs4_unlock_state();
1522 /*
1523 * The following error isn't really legal.
1524 * But we only get here if the client just explicitly
1525 * destroyed the client. Surely it no longer cares what
1526 * error it gets back on an operation for the dead
1527 * client.
1528 */
1529 return nfserr_stale_clientid;
1530 }
1531 nfsd4_create_clid_dir(cstate->session->se_client);
1532 nfs4_unlock_state();
1533 return nfs_ok;
1534}
1535
b37ad28b 1536__be32
b591480b
BF		 1537nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1538 struct nfsd4_setclientid *setclid)
1da177e4 1539{
363168b4 1540 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1541 struct xdr_netobj clname = {
1542 .len = setclid->se_namelen,
1543 .data = setclid->se_name,
1544 };
1545 nfs4_verifier clverifier = setclid->se_verf;
1546 unsigned int strhashval;
28ce6054 1547 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1548 __be32 status;
a55370a3 1549 char dname[HEXDIR_LEN];
1da177e4 1550
1da177e4 1551 if (!check_name(clname))
73aea4ec 1552 return nfserr_inval;
1da177e4 1553
a55370a3
N		 1554 status = nfs4_make_rec_clidname(dname, &clname);
1555 if (status)
73aea4ec 1556 return status;
a55370a3 1557
1da177e4
LT		 1558 /*
1559 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1560 * We get here on a DRC miss.
1561 */
1562
a55370a3 1563 strhashval = clientstr_hashval(dname);
1da177e4 1564
1da177e4 1565 nfs4_lock_state();
a1bcecd2 1566 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1567 if (conf) {
a186e767 1568 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1569 status = nfserr_clid_inuse;
026722c2 1570 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1571 char addr_str[INET6_ADDRSTRLEN];
1572 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1573 sizeof(addr_str));
1574 dprintk("NFSD: setclientid: string in use by client "
1575 "at %s\n", addr_str);
1da177e4
LT		 1576 goto out;
1577 }
1da177e4 1578 }
a186e767
BF		 1579 /*
1580 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1581 * has a description of SETCLIENTID request processing consisting
1582 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1583 */
a1bcecd2 1584 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1585 status = nfserr_resource;
1586 if (!conf) {
a186e767
BF		 1587 /*
1588 * RFC 3530 14.2.33 CASE 4:
1589 * placed first, because it is the normal case
1da177e4
LT		 1590 */
1591 if (unconf)
1592 expire_client(unconf);
b09333c4 1593 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1594 if (new == NULL)
1da177e4 1595 goto out;
1da177e4 1596 gen_clid(new);
599e0a22 1597 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1598 /*
a186e767
BF		 1599 * RFC 3530 14.2.33 CASE 1:
1600 * probable callback update
1da177e4 1601 */
31f4a6c1
N		 1602 if (unconf) {
1603 /* Note this is removing unconfirmed {*x***},
1604 * which is stronger than RFC recommended {vxc**}.
1605 * This has the advantage that there is at most
1606 * one {*x***} in either list at any time.
1607 */
1608 expire_client(unconf);
1da177e4 1609 }
b09333c4 1610 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1611 if (new == NULL)
1da177e4 1612 goto out;
1da177e4 1613 copy_clid(new, conf);
1da177e4
LT		 1614 } else if (!unconf) {
1615 /*
a186e767
BF		 1616 * RFC 3530 14.2.33 CASE 2:
1617 * probable client reboot; state will be removed if
1618 * confirmed.
1da177e4 1619 */
b09333c4 1620 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1621 if (new == NULL)
1da177e4 1622 goto out;
1da177e4 1623 gen_clid(new);
49ba8781 1624 } else {
a186e767
BF		 1625 /*
1626 * RFC 3530 14.2.33 CASE 3:
1627 * probable client reboot; state will be removed if
1628 * confirmed.
1da177e4
LT		 1629 */
1630 expire_client(unconf);
b09333c4 1631 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1632 if (new == NULL)
1da177e4 1633 goto out;
1da177e4 1634 gen_clid(new);
1da177e4 1635 }
fbf4665f 1636 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1637 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1638 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1639 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1640 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1641 status = nfs_ok;
1642out:
1643 nfs4_unlock_state();
1644 return status;
1645}
1646
1647
1648/*
a186e767
BF		 1649 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1650 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1651 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1652 */
b37ad28b 1653__be32
b591480b
BF		 1654nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1655 struct nfsd4_compound_state *cstate,
1656 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1657{
363168b4 1658 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1659 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1660 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1661 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1662 __be32 status;
1da177e4
LT		 1663
1664 if (STALE_CLIENTID(clid))
1665 return nfserr_stale_clientid;
1666 /*
1667 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1668 * We get here on a DRC miss.
1669 */
1670
1671 nfs4_lock_state();
21ab45a4
N		 1672
1673 conf = find_confirmed_client(clid);
1674 unconf = find_unconfirmed_client(clid);
1675
1676 status = nfserr_clid_inuse;
363168b4 1677 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1678 goto out;
363168b4 1679 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1680 goto out;
1681
a186e767
BF		 1682 /*
1683 * section 14.2.34 of RFC 3530 has a description of
1684 * SETCLIENTID_CONFIRM request processing consisting
1685 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1686 */
366e0c1d 1687 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1688 /*
1689 * RFC 3530 14.2.34 CASE 1:
1690 * callback update
1691 */
599e0a22 1692 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1693 status = nfserr_clid_inuse;
1694 else {
2bf23875 1695 atomic_set(&conf->cl_cb_set, 0);
4b21d0de 1696 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1a69c179 1697 expire_client(unconf);
1da177e4 1698 status = nfs_ok;
1a69c179 1699
1da177e4 1700 }
f3aba4e5 1701 } else if (conf && !unconf) {
a186e767
BF		 1702 /*
1703 * RFC 3530 14.2.34 CASE 2:
1704 * probable retransmitted request; play it safe and
1705 * do nothing.
7c79f737 1706 */
599e0a22 1707 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1708 status = nfserr_clid_inuse;
21ab45a4 1709 else
1da177e4 1710 status = nfs_ok;
7c79f737 1711 } else if (!conf && unconf
599e0a22 1712 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1713 /*
1714 * RFC 3530 14.2.34 CASE 3:
1715 * Normal case; new or rebooted client:
7c79f737 1716 */
599e0a22 1717 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1718 status = nfserr_clid_inuse;
1719 } else {
1a69c179
N		 1720 unsigned int hash =
1721 clientstr_hashval(unconf->cl_recdir);
1722 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1723 hash, false);
1a69c179 1724 if (conf) {
c7b9a459 1725 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1726 expire_client(conf);
1727 }
1da177e4 1728 move_to_confirmed(unconf);
21ab45a4 1729 conf = unconf;
4b21d0de 1730 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1a69c179 1731 status = nfs_ok;
1da177e4 1732 }
599e0a22
BF		 1733 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1734 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1735 &confirm)))) {
a186e767
BF		 1736 /*
1737 * RFC 3530 14.2.34 CASE 4:
1738 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1739 */
1da177e4 1740 status = nfserr_stale_clientid;
7c79f737 1741 } else {
08e8987c
N		 1742 /* check that we have hit one of the cases...*/
1743 status = nfserr_clid_inuse;
1744 }
1da177e4 1745out:
1da177e4
LT		 1746 nfs4_unlock_state();
1747 return status;
1748}
1749
1da177e4
LT		 1750/* OPEN Share state helper functions */
1751static inline struct nfs4_file *
1752alloc_init_file(struct inode *ino)
1753{
1754 struct nfs4_file *fp;
1755 unsigned int hashval = file_hashval(ino);
1756
e60d4398
N		 1757 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1758 if (fp) {
8b671b80 1759 atomic_set(&fp->fi_ref, 1);
1da177e4 1760 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1761 INIT_LIST_HEAD(&fp->fi_stateids);
1762 INIT_LIST_HEAD(&fp->fi_delegations);
1da177e4
LT		 1763 fp->fi_inode = igrab(ino);
1764 fp->fi_id = current_fileid++;
47f9940c 1765 fp->fi_had_conflict = false;
47cee541
PE		 1766 spin_lock(&recall_lock);
1767 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
1768 spin_unlock(&recall_lock);
1da177e4
LT		 1769 return fp;
1770 }
1771 return NULL;
1772}
1773
e60d4398 1774static void
e18b890b 1775nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1776{
e60d4398
N		 1777 if (*slab == NULL)
1778 return;
1a1d92c1 1779 kmem_cache_destroy(*slab);
e60d4398 1780 *slab = NULL;
1da177e4
LT		 1781}
1782
e8ff2a84 1783void
1da177e4
LT		 1784nfsd4_free_slabs(void)
1785{
e60d4398
N		 1786 nfsd4_free_slab(&stateowner_slab);
1787 nfsd4_free_slab(&file_slab);
5ac049ac 1788 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1789 nfsd4_free_slab(&deleg_slab);
e60d4398 1790}
1da177e4 1791
e60d4398
N		 1792static int
1793nfsd4_init_slabs(void)
1794{
1795 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1796 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1797 if (stateowner_slab == NULL)
1798 goto out_nomem;
1799 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1800 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1801 if (file_slab == NULL)
1802 goto out_nomem;
5ac049ac 1803 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1804 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1805 if (stateid_slab == NULL)
1806 goto out_nomem;
5b2d21c1 1807 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1808 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1809 if (deleg_slab == NULL)
1810 goto out_nomem;
e60d4398
N		 1811 return 0;
1812out_nomem:
1813 nfsd4_free_slabs();
1814 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1815 return -ENOMEM;
1da177e4
LT		 1816}
1817
1818void
1819nfs4_free_stateowner(struct kref *kref)
1820{
1821 struct nfs4_stateowner *sop =
1822 container_of(kref, struct nfs4_stateowner, so_ref);
1823 kfree(sop->so_owner.data);
1824 kmem_cache_free(stateowner_slab, sop);
1825}
1826
1827static inline struct nfs4_stateowner *
1828alloc_stateowner(struct xdr_netobj *owner)
1829{
1830 struct nfs4_stateowner *sop;
1831
1832 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1833 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1834 memcpy(sop->so_owner.data, owner->data, owner->len);
1835 sop->so_owner.len = owner->len;
1836 kref_init(&sop->so_ref);
1837 return sop;
1838 }
1839 kmem_cache_free(stateowner_slab, sop);
1840 }
1841 return NULL;
1842}
1843
1844static struct nfs4_stateowner *
1845alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1846 struct nfs4_stateowner *sop;
1847 struct nfs4_replay *rp;
1848 unsigned int idhashval;
1849
1850 if (!(sop = alloc_stateowner(&open->op_owner)))
1851 return NULL;
1852 idhashval = ownerid_hashval(current_ownerid);
1853 INIT_LIST_HEAD(&sop->so_idhash);
1854 INIT_LIST_HEAD(&sop->so_strhash);
1855 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1856 INIT_LIST_HEAD(&sop->so_stateids);
1857 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1858 INIT_LIST_HEAD(&sop->so_close_lru);
1859 sop->so_time = 0;
1860 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1861 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1862 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1863 sop->so_is_open_owner = 1;
1864 sop->so_id = current_ownerid++;
1865 sop->so_client = clp;
1866 sop->so_seqid = open->op_seqid;
1867 sop->so_confirmed = 0;
1868 rp = &sop->so_replay;
de1ae286 1869 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1870 rp->rp_buflen = 0;
1871 rp->rp_buf = rp->rp_ibuf;
1872 return sop;
1873}
1874
1da177e4
LT		 1875static inline void
1876init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1877 struct nfs4_stateowner *sop = open->op_stateowner;
1878 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1879
1880 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1881 INIT_LIST_HEAD(&stp->st_perstateowner);
1882 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1883 INIT_LIST_HEAD(&stp->st_perfile);
1884 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1885 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1886 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1887 stp->st_stateowner = sop;
13cd2184 1888 get_nfs4_file(fp);
1da177e4 1889 stp->st_file = fp;
e4e83ea4 1890 stp->st_stateid.si_boot = boot_time;
1da177e4
LT		 1891 stp->st_stateid.si_stateownerid = sop->so_id;
1892 stp->st_stateid.si_fileid = fp->fi_id;
1893 stp->st_stateid.si_generation = 0;
1894 stp->st_access_bmap = 0;
1895 stp->st_deny_bmap = 0;
84459a11
AA		 1896 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1897 &stp->st_access_bmap);
1da177e4 1898 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1899 stp->st_openstp = NULL;
1da177e4
LT		 1900}
1901
fd39ca9a 1902static void
1da177e4
LT		 1903move_to_close_lru(struct nfs4_stateowner *sop)
1904{
1905 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1906
358dd55a 1907 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1908 sop->so_time = get_seconds();
1909}
1910
1da177e4 1911static int
599e0a22
BF		 1912same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1913 clientid_t *clid)
1914{
1915 return (sop->so_owner.len == owner->len) &&
1916 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1917 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1918}
1919
1920static struct nfs4_stateowner *
1921find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1922{
1923 struct nfs4_stateowner *so = NULL;
1924
1925 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1926 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1927 return so;
1928 }
1929 return NULL;
1930}
1931
1932/* search file_hashtbl[] for file */
1933static struct nfs4_file *
1934find_file(struct inode *ino)
1935{
1936 unsigned int hashval = file_hashval(ino);
1937 struct nfs4_file *fp;
1938
8b671b80 1939 spin_lock(&recall_lock);
1da177e4 1940 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1941 if (fp->fi_inode == ino) {
1942 get_nfs4_file(fp);
8b671b80 1943 spin_unlock(&recall_lock);
1da177e4 1944 return fp;
13cd2184 1945 }
1da177e4 1946 }
8b671b80 1947 spin_unlock(&recall_lock);
1da177e4
LT		 1948 return NULL;
1949}
1950
d87a8ade 1951static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1952{
d87a8ade 1953 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1954 return 0;
d87a8ade
AA		 1955 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1956 return 0;
1957 x &= ~NFS4_SHARE_ACCESS_MASK;
1958 if (minorversion && x) {
1959 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1960 return 0;
1961 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1962 return 0;
1963 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1964 }
1965 if (x)
8838dc43
BF		 1966 return 0;
1967 return 1;
ba5a6a19
BF		 1968}
1969
8838dc43 1970static inline int deny_valid(u32 x)
ba5a6a19 1971{
8838dc43
BF		 1972 /* Note: unlike access bits, deny bits may be zero. */
1973 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1974}
1da177e4 1975
4f83aa30
BF		 1976/*
1977 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1978 * st_{access,deny}_bmap field of the stateid, in order to track not
1979 * only what share bits are currently in force, but also what
1980 * combinations of share bits previous opens have used. This allows us
1981 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1982 * return an error if the client attempt to downgrade to a combination
1983 * of share bits not explicable by closing some of its previous opens.
1984 *
1985 * XXX: This enforcement is actually incomplete, since we don't keep
1986 * track of access/deny bit combinations; so, e.g., we allow:
1987 *
1988 * OPEN allow read, deny write
1989 * OPEN allow both, deny none
1990 * DOWNGRADE allow read, deny none
1991 *
1992 * which we should reject.
1993 */
fd39ca9a 1994static void
1da177e4
LT		 1995set_access(unsigned int *access, unsigned long bmap) {
1996 int i;
1997
1998 *access = 0;
1999 for (i = 1; i < 4; i++) {
2000 if (test_bit(i, &bmap))
2001 *access |= i;
2002 }
2003}
2004
fd39ca9a 2005static void
1da177e4
LT		 2006set_deny(unsigned int *deny, unsigned long bmap) {
2007 int i;
2008
2009 *deny = 0;
2010 for (i = 0; i < 4; i++) {
2011 if (test_bit(i, &bmap))
2012 *deny |= i ;
2013 }
2014}
2015
2016static int
2017test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2018 unsigned int access, deny;
2019
2020 set_access(&access, stp->st_access_bmap);
2021 set_deny(&deny, stp->st_deny_bmap);
2022 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2023 return 0;
2024 return 1;
2025}
2026
2027/*
2028 * Called to check deny when READ with all zero stateid or
2029 * WRITE with all zero or all one stateid
2030 */
b37ad28b 2031static __be32
1da177e4
LT		 2032nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2033{
2034 struct inode *ino = current_fh->fh_dentry->d_inode;
2035 struct nfs4_file *fp;
2036 struct nfs4_stateid *stp;
b37ad28b 2037 __be32 ret;
1da177e4
LT		 2038
2039 dprintk("NFSD: nfs4_share_conflict\n");
2040
2041 fp = find_file(ino);
13cd2184
N		 2042 if (!fp)
2043 return nfs_ok;
b700949b 2044 ret = nfserr_locked;
1da177e4 2045 /* Search for conflicting share reservations */
13cd2184
N		 2046 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2047 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2048 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2049 goto out;
1da177e4 2050 }
13cd2184
N		 2051 ret = nfs_ok;
2052out:
2053 put_nfs4_file(fp);
2054 return ret;
1da177e4
LT		 2055}
2056
2057static inline void
2058nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2059{
2060 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2061 drop_file_write_access(filp);
42e49608 2062 spin_lock(&filp->f_lock);
1da177e4 2063 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
42e49608 2064 spin_unlock(&filp->f_lock);
1da177e4
LT		 2065 }
2066}
2067
1da177e4
LT		 2068/*
2069 * Spawn a thread to perform a recall on the delegation represented
2070 * by the lease (file_lock)
2071 *
2072 * Called from break_lease() with lock_kernel() held.
2073 * Note: we assume break_lease will only call this *once* for any given
2074 * lease.
2075 */
2076static
2077void nfsd_break_deleg_cb(struct file_lock *fl)
2078{
63e4863f 2079 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2080
2081 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2082 if (!dp)
2083 return;
2084
2085 /* We're assuming the state code never drops its reference
2086 * without first removing the lease. Since we're in this lease
2087 * callback (and since the lease code is serialized by the kernel
2088 * lock) we know the server hasn't removed the lease yet, we know
2089 * it's safe to take a reference: */
2090 atomic_inc(&dp->dl_count);
2091
2092 spin_lock(&recall_lock);
2093 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2094 spin_unlock(&recall_lock);
2095
2096 /* only place dl_time is set. protected by lock_kernel*/
2097 dp->dl_time = get_seconds();
2098
0272e1fd
BF		 2099 /*
2100 * We don't want the locks code to timeout the lease for us;
2101 * we'll remove it ourself if the delegation isn't returned
2102 * in time.
2103 */
2104 fl->fl_break_time = 0;
1da177e4 2105
63e4863f
BF		 2106 dp->dl_file->fi_had_conflict = true;
2107 nfsd4_cb_recall(dp);
1da177e4
LT		 2108}
2109
2110/*
2111 * The file_lock is being reapd.
2112 *
2113 * Called by locks_free_lock() with lock_kernel() held.
2114 */
2115static
2116void nfsd_release_deleg_cb(struct file_lock *fl)
2117{
2118 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2119
2120 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2121
2122 if (!(fl->fl_flags & FL_LEASE) || !dp)
2123 return;
2124 dp->dl_flock = NULL;
2125}
2126
2127/*
2128 * Set the delegation file_lock back pointer.
2129 *
a9933cea 2130 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2131 */
2132static
2133void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2134{
2135 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2136
2137 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2138 if (!dp)
2139 return;
2140 dp->dl_flock = new;
2141}
2142
2143/*
a9933cea 2144 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2145 */
2146static
2147int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2148{
2149 struct nfs4_delegation *onlistd =
2150 (struct nfs4_delegation *)onlist->fl_owner;
2151 struct nfs4_delegation *tryd =
2152 (struct nfs4_delegation *)try->fl_owner;
2153
2154 if (onlist->fl_lmops != try->fl_lmops)
2155 return 0;
2156
2157 return onlistd->dl_client == tryd->dl_client;
2158}
2159
2160
2161static
2162int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2163{
2164 if (arg & F_UNLCK)
2165 return lease_modify(onlist, arg);
2166 else
2167 return -EAGAIN;
2168}
2169
7b021967 2170static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2171 .fl_break = nfsd_break_deleg_cb,
2172 .fl_release_private = nfsd_release_deleg_cb,
2173 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2174 .fl_mylease = nfsd_same_client_deleg_cb,
2175 .fl_change = nfsd_change_deleg_cb,
2176};
2177
2178
b37ad28b 2179__be32
6668958f
AA		 2180nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2181 struct nfsd4_open *open)
1da177e4 2182{
1da177e4
LT		 2183 clientid_t *clientid = &open->op_clientid;
2184 struct nfs4_client *clp = NULL;
2185 unsigned int strhashval;
2186 struct nfs4_stateowner *sop = NULL;
2187
1da177e4 2188 if (!check_name(open->op_owner))
0f442aa2 2189 return nfserr_inval;
1da177e4
LT		 2190
2191 if (STALE_CLIENTID(&open->op_clientid))
2192 return nfserr_stale_clientid;
2193
2194 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2195 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2196 open->op_stateowner = sop;
2197 if (!sop) {
2198 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2199 clp = find_confirmed_client(clientid);
2200 if (clp == NULL)
0f442aa2
BF		 2201 return nfserr_expired;
2202 goto renew;
1da177e4 2203 }
6668958f
AA		 2204 /* When sessions are used, skip open sequenceid processing */
2205 if (nfsd4_has_session(cstate))
2206 goto renew;
0f442aa2
BF		 2207 if (!sop->so_confirmed) {
2208 /* Replace unconfirmed owners without checking for replay. */
2209 clp = sop->so_client;
f044ff83 2210 release_openowner(sop);
0f442aa2
BF		 2211 open->op_stateowner = NULL;
2212 goto renew;
2213 }
2214 if (open->op_seqid == sop->so_seqid - 1) {
2215 if (sop->so_replay.rp_buflen)
a90b061c 2216 return nfserr_replay_me;
0f442aa2
BF		 2217 /* The original OPEN failed so spectacularly
2218 * that we don't even have replay data saved!
2219 * Therefore, we have no choice but to continue
2220 * processing this OPEN; presumably, we'll
2221 * fail again for the same reason.
2222 */
2223 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2224 goto renew;
2225 }
2226 if (open->op_seqid != sop->so_seqid)
2227 return nfserr_bad_seqid;
1da177e4 2228renew:
0f442aa2
BF		 2229 if (open->op_stateowner == NULL) {
2230 sop = alloc_init_open_stateowner(strhashval, clp, open);
2231 if (sop == NULL)
2232 return nfserr_resource;
2233 open->op_stateowner = sop;
2234 }
2235 list_del_init(&sop->so_close_lru);
1da177e4 2236 renew_client(sop->so_client);
0f442aa2 2237 return nfs_ok;
1da177e4
LT		 2238}
2239
b37ad28b 2240static inline __be32
4a6e43e6
N		 2241nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2242{
2243 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2244 return nfserr_openmode;
2245 else
2246 return nfs_ok;
2247}
2248
52f4fb43
N		 2249static struct nfs4_delegation *
2250find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2251{
2252 struct nfs4_delegation *dp;
2253
ea1da636 2254 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2255 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2256 return dp;
2257 }
2258 return NULL;
2259}
2260
24a0111e
BF		 2261int share_access_to_flags(u32 share_access)
2262{
2263 share_access &= ~NFS4_SHARE_WANT_MASK;
2264
2265 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2266}
2267
b37ad28b 2268static __be32
567d9829
N		 2269nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2270 struct nfs4_delegation **dp)
2271{
2272 int flags;
b37ad28b 2273 __be32 status = nfserr_bad_stateid;
567d9829
N		 2274
2275 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2276 if (*dp == NULL)
c44c5eeb 2277 goto out;
24a0111e 2278 flags = share_access_to_flags(open->op_share_access);
567d9829
N		 2279 status = nfs4_check_delegmode(*dp, flags);
2280 if (status)
2281 *dp = NULL;
c44c5eeb
N		 2282out:
2283 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2284 return nfs_ok;
2285 if (status)
2286 return status;
2287 open->op_stateowner->so_confirmed = 1;
2288 return nfs_ok;
567d9829
N		 2289}
2290
b37ad28b 2291static __be32
1da177e4
LT		 2292nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2293{
2294 struct nfs4_stateid *local;
b37ad28b 2295 __be32 status = nfserr_share_denied;
1da177e4
LT		 2296 struct nfs4_stateowner *sop = open->op_stateowner;
2297
8beefa24 2298 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2299 /* ignore lock owners */
2300 if (local->st_stateowner->so_is_open_owner == 0)
2301 continue;
2302 /* remember if we have seen this open owner */
2303 if (local->st_stateowner == sop)
2304 *stpp = local;
2305 /* check for conflicting share reservations */
2306 if (!test_share(local, open))
2307 goto out;
2308 }
2309 status = 0;
2310out:
2311 return status;
2312}
2313
5ac049ac
N		 2314static inline struct nfs4_stateid *
2315nfs4_alloc_stateid(void)
2316{
2317 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2318}
2319
21fb4016
BF		 2320static inline int nfs4_access_to_access(u32 nfs4_access)
2321{
2322 int flags = 0;
2323
2324 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2325 flags |= NFSD_MAY_READ;
2326 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2327 flags |= NFSD_MAY_WRITE;
2328 return flags;
2329}
2330
b37ad28b 2331static __be32
1da177e4 2332nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2333 struct nfs4_delegation *dp,
21fb4016 2334 struct svc_fh *cur_fh, struct nfsd4_open *open)
1da177e4
LT		 2335{
2336 struct nfs4_stateid *stp;
1da177e4 2337
5ac049ac 2338 stp = nfs4_alloc_stateid();
1da177e4
LT		 2339 if (stp == NULL)
2340 return nfserr_resource;
2341
567d9829
N		 2342 if (dp) {
2343 get_file(dp->dl_vfs_file);
2344 stp->st_vfs_file = dp->dl_vfs_file;
2345 } else {
b37ad28b 2346 __be32 status;
21fb4016
BF		 2347 int access = nfs4_access_to_access(open->op_share_access);
2348
2349 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2350 &stp->st_vfs_file);
567d9829
N		 2351 if (status) {
2352 if (status == nfserr_dropit)
2353 status = nfserr_jukebox;
5ac049ac 2354 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2355 return status;
2356 }
1da177e4 2357 }
1da177e4
LT		 2358 *stpp = stp;
2359 return 0;
2360}
2361
b37ad28b 2362static inline __be32
1da177e4
LT		 2363nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2364 struct nfsd4_open *open)
2365{
2366 struct iattr iattr = {
2367 .ia_valid = ATTR_SIZE,
2368 .ia_size = 0,
2369 };
2370 if (!open->op_truncate)
2371 return 0;
2372 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2373 return nfserr_inval;
1da177e4
LT		 2374 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2375}
2376
b37ad28b 2377static __be32
1da177e4
LT		 2378nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2379{
2380 struct file *filp = stp->st_vfs_file;
7eaa36e2 2381 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2382 unsigned int share_access, new_writer;
24a0111e 2383 u32 op_share_access;
b37ad28b 2384 __be32 status;
1da177e4
LT		 2385
2386 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2387 new_writer = (~share_access) & open->op_share_access
2388 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2389
6c26d08f 2390 if (new_writer) {
b37ad28b
AV		 2391 int err = get_write_access(inode);
2392 if (err)
2393 return nfserrno(err);
e518f056
BH		 2394 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2395 if (err)
2396 return nfserrno(err);
2397 file_take_write(filp);
6c26d08f 2398 }
1da177e4
LT		 2399 status = nfsd4_truncate(rqstp, cur_fh, open);
2400 if (status) {
6c26d08f
BF		 2401 if (new_writer)
2402 put_write_access(inode);
1da177e4
LT		 2403 return status;
2404 }
2405 /* remember the open */
24a0111e
BF		 2406 op_share_access = open->op_share_access & ~NFS4_SHARE_WANT_MASK;
2407 filp->f_mode |= op_share_access;
2408 __set_bit(op_share_access, &stp->st_access_bmap);
b55e0ba1 2409 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2410
2411 return nfs_ok;
2412}
2413
2414
1da177e4 2415static void
37515177 2416nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2417{
37515177
N		 2418 open->op_stateowner->so_confirmed = 1;
2419 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2420}
2421
2422/*
2423 * Attempt to hand out a delegation.
2424 */
2425static void
2426nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2427{
2428 struct nfs4_delegation *dp;
2429 struct nfs4_stateowner *sop = stp->st_stateowner;
2bf23875 2430 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
1da177e4
LT		 2431 struct file_lock fl, *flp = &fl;
2432 int status, flag = 0;
2433
2434 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2435 open->op_recall = 0;
2436 switch (open->op_claim_type) {
2437 case NFS4_OPEN_CLAIM_PREVIOUS:
2bf23875 2438 if (!cb_up)
7b190fec
N		 2439 open->op_recall = 1;
2440 flag = open->op_delegate_type;
2441 if (flag == NFS4_OPEN_DELEGATE_NONE)
2442 goto out;
2443 break;
2444 case NFS4_OPEN_CLAIM_NULL:
2445 /* Let's not give out any delegations till everyone's
2446 * had the chance to reclaim theirs.... */
af558e33 2447 if (locks_in_grace())
7b190fec 2448 goto out;
2bf23875 2449 if (!cb_up || !sop->so_confirmed)
7b190fec
N		 2450 goto out;
2451 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2452 flag = NFS4_OPEN_DELEGATE_WRITE;
2453 else
2454 flag = NFS4_OPEN_DELEGATE_READ;
2455 break;
2456 default:
2457 goto out;
2458 }
1da177e4
LT		 2459
2460 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2461 if (dp == NULL) {
2462 flag = NFS4_OPEN_DELEGATE_NONE;
2463 goto out;
2464 }
2465 locks_init_lock(&fl);
2466 fl.fl_lmops = &nfsd_lease_mng_ops;
2467 fl.fl_flags = FL_LEASE;
9167f501 2468 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2469 fl.fl_end = OFFSET_MAX;
2470 fl.fl_owner = (fl_owner_t)dp;
2471 fl.fl_file = stp->st_vfs_file;
2472 fl.fl_pid = current->tgid;
2473
a9933cea 2474 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2475 * the lock_manager callbacks fl_mylease and fl_change are used
2476 */
9167f501 2477 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2478 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2479 unhash_delegation(dp);
1da177e4
LT		 2480 flag = NFS4_OPEN_DELEGATE_NONE;
2481 goto out;
2482 }
2483
2484 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2485
8c10cbdb
BH		 2486 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2487 STATEID_VAL(&dp->dl_stateid));
1da177e4 2488out:
7b190fec
N		 2489 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2490 && flag == NFS4_OPEN_DELEGATE_NONE
2491 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2492 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2493 open->op_delegate_type = flag;
2494}
2495
2496/*
2497 * called with nfs4_lock_state() held.
2498 */
b37ad28b 2499__be32
1da177e4
LT		 2500nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2501{
6668958f 2502 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2503 struct nfs4_file *fp = NULL;
2504 struct inode *ino = current_fh->fh_dentry->d_inode;
2505 struct nfs4_stateid *stp = NULL;
567d9829 2506 struct nfs4_delegation *dp = NULL;
b37ad28b 2507 __be32 status;
1da177e4
LT		 2508
2509 status = nfserr_inval;
d87a8ade 2510 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2511 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2512 goto out;
2513 /*
2514 * Lookup file; if found, lookup stateid and check open request,
2515 * and check for delegations in the process of being recalled.
2516 * If not found, create the nfs4_file struct
2517 */
2518 fp = find_file(ino);
2519 if (fp) {
2520 if ((status = nfs4_check_open(fp, open, &stp)))
2521 goto out;
c44c5eeb
N		 2522 status = nfs4_check_deleg(fp, open, &dp);
2523 if (status)
2524 goto out;
1da177e4 2525 } else {
c44c5eeb
N		 2526 status = nfserr_bad_stateid;
2527 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2528 goto out;
1da177e4
LT		 2529 status = nfserr_resource;
2530 fp = alloc_init_file(ino);
2531 if (fp == NULL)
2532 goto out;
2533 }
2534
2535 /*
2536 * OPEN the file, or upgrade an existing OPEN.
2537 * If truncate fails, the OPEN fails.
2538 */
2539 if (stp) {
2540 /* Stateid was found, this is an OPEN upgrade */
2541 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2542 if (status)
2543 goto out;
444c2c07 2544 update_stateid(&stp->st_stateid);
1da177e4
LT		 2545 } else {
2546 /* Stateid was not found, this is a new OPEN */
21fb4016 2547 status = nfs4_new_open(rqstp, &stp, dp, current_fh, open);
567d9829 2548 if (status)
1da177e4
LT		 2549 goto out;
2550 init_stateid(stp, fp, open);
2551 status = nfsd4_truncate(rqstp, current_fh, open);
2552 if (status) {
2283963f 2553 release_open_stateid(stp);
1da177e4
LT		 2554 goto out;
2555 }
6668958f
AA		 2556 if (nfsd4_has_session(&resp->cstate))
2557 update_stateid(&stp->st_stateid);
1da177e4
LT		 2558 }
2559 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2560
4dc6ec00 2561 if (nfsd4_has_session(&resp->cstate))
6668958f
AA		 2562 open->op_stateowner->so_confirmed = 1;
2563
1da177e4
LT		 2564 /*
2565 * Attempt to hand out a delegation. No error return, because the
2566 * OPEN succeeds even if we fail.
2567 */
2568 nfs4_open_delegation(current_fh, open, stp);
2569
2570 status = nfs_ok;
2571
8c10cbdb
BH		 2572 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2573 STATEID_VAL(&stp->st_stateid));
1da177e4 2574out:
13cd2184
N		 2575 if (fp)
2576 put_nfs4_file(fp);
37515177
N		 2577 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2578 nfs4_set_claim_prev(open);
1da177e4
LT		 2579 /*
2580 * To finish the open response, we just need to set the rflags.
2581 */
2582 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2583 if (!open->op_stateowner->so_confirmed &&
2584 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2585 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2586
2587 return status;
2588}
2589
b37ad28b 2590__be32
b591480b
BF		 2591nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2592 clientid_t *clid)
1da177e4
LT		 2593{
2594 struct nfs4_client *clp;
b37ad28b 2595 __be32 status;
1da177e4
LT		 2596
2597 nfs4_lock_state();
2598 dprintk("process_renew(%08x/%08x): starting\n",
2599 clid->cl_boot, clid->cl_id);
2600 status = nfserr_stale_clientid;
2601 if (STALE_CLIENTID(clid))
2602 goto out;
2603 clp = find_confirmed_client(clid);
2604 status = nfserr_expired;
2605 if (clp == NULL) {
2606 /* We assume the client took too long to RENEW. */
2607 dprintk("nfsd4_renew: clientid not found!\n");
2608 goto out;
2609 }
2610 renew_client(clp);
2611 status = nfserr_cb_path_down;
ea1da636 2612 if (!list_empty(&clp->cl_delegations)
2bf23875 2613 && !atomic_read(&clp->cl_cb_set))
1da177e4
LT		 2614 goto out;
2615 status = nfs_ok;
2616out:
2617 nfs4_unlock_state();
2618 return status;
2619}
2620
af558e33
BF		 2621struct lock_manager nfsd4_manager = {
2622};
2623
a76b4319 2624static void
af558e33 2625nfsd4_end_grace(void)
a76b4319
N		 2626{
2627 dprintk("NFSD: end of grace period\n");
c7b9a459 2628 nfsd4_recdir_purge_old();
af558e33 2629 locks_end_grace(&nfsd4_manager);
e46b498c
BF		 2630 /*
2631 * Now that every NFSv4 client has had the chance to recover and
2632 * to see the (possibly new, possibly shorter) lease time, we
2633 * can safely set the next grace time to the current lease time:
2634 */
2635 nfsd4_grace = nfsd4_lease;
a76b4319
N		 2636}
2637
fd39ca9a 2638static time_t
1da177e4
LT		 2639nfs4_laundromat(void)
2640{
2641 struct nfs4_client *clp;
2642 struct nfs4_stateowner *sop;
2643 struct nfs4_delegation *dp;
2644 struct list_head *pos, *next, reaplist;
cf07d2ea
BF		 2645 time_t cutoff = get_seconds() - nfsd4_lease;
2646 time_t t, clientid_val = nfsd4_lease;
2647 time_t u, test_val = nfsd4_lease;
1da177e4
LT		 2648
2649 nfs4_lock_state();
2650
2651 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2652 if (locks_in_grace())
2653 nfsd4_end_grace();
36acb66b
BH		 2654 INIT_LIST_HEAD(&reaplist);
2655 spin_lock(&client_lock);
1da177e4
LT		 2656 list_for_each_safe(pos, next, &client_lru) {
2657 clp = list_entry(pos, struct nfs4_client, cl_lru);
2658 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2659 t = clp->cl_time - cutoff;
2660 if (clientid_val > t)
2661 clientid_val = t;
2662 break;
2663 }
d7682988
BH		 2664 if (atomic_read(&clp->cl_refcount)) {
2665 dprintk("NFSD: client in use (clientid %08x)\n",
2666 clp->cl_clientid.cl_id);
2667 continue;
2668 }
2669 unhash_client_locked(clp);
2670 list_add(&clp->cl_lru, &reaplist);
36acb66b
BH		 2671 }
2672 spin_unlock(&client_lock);
2673 list_for_each_safe(pos, next, &reaplist) {
2674 clp = list_entry(pos, struct nfs4_client, cl_lru);
1da177e4
LT		 2675 dprintk("NFSD: purging unused client (clientid %08x)\n",
2676 clp->cl_clientid.cl_id);
c7b9a459 2677 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2678 expire_client(clp);
2679 }
1da177e4
LT		 2680 spin_lock(&recall_lock);
2681 list_for_each_safe(pos, next, &del_recall_lru) {
2682 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2683 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2684 u = dp->dl_time - cutoff;
2685 if (test_val > u)
2686 test_val = u;
2687 break;
2688 }
2689 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2690 dp, dp->dl_flock);
2691 list_move(&dp->dl_recall_lru, &reaplist);
2692 }
2693 spin_unlock(&recall_lock);
2694 list_for_each_safe(pos, next, &reaplist) {
2695 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2696 list_del_init(&dp->dl_recall_lru);
2697 unhash_delegation(dp);
2698 }
cf07d2ea 2699 test_val = nfsd4_lease;
1da177e4
LT		 2700 list_for_each_safe(pos, next, &close_lru) {
2701 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2702 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2703 u = sop->so_time - cutoff;
2704 if (test_val > u)
2705 test_val = u;
2706 break;
2707 }
2708 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2709 sop->so_id);
f044ff83 2710 release_openowner(sop);
1da177e4
LT		 2711 }
2712 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2713 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2714 nfs4_unlock_state();
2715 return clientid_val;
2716}
2717
a254b246
HH		 2718static struct workqueue_struct *laundry_wq;
2719static void laundromat_main(struct work_struct *);
2720static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2721
2722static void
c4028958 2723laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2724{
2725 time_t t;
2726
2727 t = nfs4_laundromat();
2728 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2729 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2730}
2731
fd39ca9a 2732static struct nfs4_stateowner *
f8816512
N		 2733search_close_lru(u32 st_id, int flags)
2734{
1da177e4
LT		 2735 struct nfs4_stateowner *local = NULL;
2736
1da177e4
LT		 2737 if (flags & CLOSE_STATE) {
2738 list_for_each_entry(local, &close_lru, so_close_lru) {
2739 if (local->so_id == st_id)
2740 return local;
2741 }
2742 }
2743 return NULL;
2744}
2745
2746static inline int
2747nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2748{
7eaa36e2 2749 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2750}
2751
2752static int
2753STALE_STATEID(stateid_t *stateid)
2754{
e4e83ea4
BF		 2755 if (stateid->si_boot == boot_time)
2756 return 0;
2757 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
8c10cbdb 2758 STATEID_VAL(stateid));
e4e83ea4 2759 return 1;
1da177e4
LT		 2760}
2761
2762static inline int
2763access_permit_read(unsigned long access_bmap)
2764{
2765 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2766 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2767 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2768}
2769
2770static inline int
2771access_permit_write(unsigned long access_bmap)
2772{
2773 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2774 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2775}
2776
2777static
b37ad28b 2778__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2779{
b37ad28b 2780 __be32 status = nfserr_openmode;
1da177e4
LT		 2781
2782 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2783 goto out;
2784 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2785 goto out;
2786 status = nfs_ok;
2787out:
2788 return status;
2789}
2790
b37ad28b 2791static inline __be32
1da177e4
LT		 2792check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2793{
203a8c8e 2794 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2795 return nfs_ok;
af558e33 2796 else if (locks_in_grace()) {
1da177e4
LT		 2797 /* Answer in remaining cases depends on existance of
2798 * conflicting state; so we must wait out the grace period. */
2799 return nfserr_grace;
2800 } else if (flags & WR_STATE)
2801 return nfs4_share_conflict(current_fh,
2802 NFS4_SHARE_DENY_WRITE);
2803 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2804 return nfs4_share_conflict(current_fh,
2805 NFS4_SHARE_DENY_READ);
2806}
2807
2808/*
2809 * Allow READ/WRITE during grace period on recovered state only for files
2810 * that are not able to provide mandatory locking.
2811 */
2812static inline int
18f82731 2813grace_disallows_io(struct inode *inode)
1da177e4 2814{
203a8c8e 2815 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2816}
2817
6668958f 2818static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2819{
6668958f
AA		 2820 /*
2821 * When sessions are used the stateid generation number is ignored
2822 * when it is zero.
2823 */
2824 if ((flags & HAS_SESSION) && in->si_generation == 0)
2825 goto out;
2826
0836f587
BF		 2827 /* If the client sends us a stateid from the future, it's buggy: */
2828 if (in->si_generation > ref->si_generation)
2829 return nfserr_bad_stateid;
2830 /*
2831 * The following, however, can happen. For example, if the
2832 * client sends an open and some IO at the same time, the open
2833 * may bump si_generation while the IO is still in flight.
2834 * Thanks to hard links and renames, the client never knows what
2835 * file an open will affect. So it could avoid that situation
2836 * only by serializing all opens and IO from the same open
2837 * owner. To recover from the old_stateid error, the client
2838 * will just have to retry the IO:
2839 */
2840 if (in->si_generation < ref->si_generation)
2841 return nfserr_old_stateid;
6668958f 2842out:
0836f587
BF		 2843 return nfs_ok;
2844}
2845
3e633079
BF		 2846static int is_delegation_stateid(stateid_t *stateid)
2847{
2848 return stateid->si_fileid == 0;
2849}
2850
1da177e4
LT		 2851/*
2852* Checks for stateid operations
2853*/
b37ad28b 2854__be32
dd453dfd
BH		 2855nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2856 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2857{
2858 struct nfs4_stateid *stp = NULL;
2859 struct nfs4_delegation *dp = NULL;
dd453dfd 2860 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2861 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2862 __be32 status;
1da177e4 2863
1da177e4
LT		 2864 if (filpp)
2865 *filpp = NULL;
2866
18f82731 2867 if (grace_disallows_io(ino))
1da177e4
LT		 2868 return nfserr_grace;
2869
6668958f
AA		 2870 if (nfsd4_has_session(cstate))
2871 flags |= HAS_SESSION;
2872
1da177e4
LT		 2873 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2874 return check_special_stateids(current_fh, stateid, flags);
2875
1da177e4
LT		 2876 status = nfserr_stale_stateid;
2877 if (STALE_STATEID(stateid))
2878 goto out;
2879
1da177e4 2880 status = nfserr_bad_stateid;
3e633079 2881 if (is_delegation_stateid(stateid)) {
a4455be0 2882 dp = find_delegation_stateid(ino, stateid);
e4e83ea4 2883 if (!dp)
1da177e4 2884 goto out;
6668958f
AA		 2885 status = check_stateid_generation(stateid, &dp->dl_stateid,
2886 flags);
0c2a498f
BF		 2887 if (status)
2888 goto out;
dc9bf700
BF		 2889 status = nfs4_check_delegmode(dp, flags);
2890 if (status)
2891 goto out;
2892 renew_client(dp->dl_client);
dc9bf700
BF		 2893 if (filpp)
2894 *filpp = dp->dl_vfs_file;
1da177e4 2895 } else { /* open or lock stateid */
a4455be0 2896 stp = find_stateid(stateid, flags);
e4e83ea4 2897 if (!stp)
1da177e4 2898 goto out;
6150ef0d 2899 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2900 goto out;
2901 if (!stp->st_stateowner->so_confirmed)
2902 goto out;
6668958f
AA		 2903 status = check_stateid_generation(stateid, &stp->st_stateid,
2904 flags);
0c2a498f
BF		 2905 if (status)
2906 goto out;
a4455be0
BF		 2907 status = nfs4_check_openmode(stp, flags);
2908 if (status)
1da177e4
LT		 2909 goto out;
2910 renew_client(stp->st_stateowner->so_client);
2911 if (filpp)
2912 *filpp = stp->st_vfs_file;
1da177e4
LT		 2913 }
2914 status = nfs_ok;
2915out:
2916 return status;
2917}
2918
4c4cd222
N		 2919static inline int
2920setlkflg (int type)
2921{
2922 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2923 RD_STATE : WR_STATE;
2924}
1da177e4
LT		 2925
2926/*
2927 * Checks for sequence id mutating operations.
2928 */
b37ad28b 2929static __be32
dd453dfd
BH		 2930nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2931 stateid_t *stateid, int flags,
2932 struct nfs4_stateowner **sopp,
2933 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2934{
1da177e4
LT		 2935 struct nfs4_stateid *stp;
2936 struct nfs4_stateowner *sop;
dd453dfd 2937 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2938 __be32 status;
1da177e4 2939
8c10cbdb
BH		 2940 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2941 seqid, STATEID_VAL(stateid));
3a4f98bb 2942
1da177e4
LT		 2943 *stpp = NULL;
2944 *sopp = NULL;
2945
1da177e4 2946 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2947 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2948 return nfserr_bad_stateid;
1da177e4
LT		 2949 }
2950
1da177e4 2951 if (STALE_STATEID(stateid))
3a4f98bb 2952 return nfserr_stale_stateid;
6668958f
AA		 2953
2954 if (nfsd4_has_session(cstate))
2955 flags |= HAS_SESSION;
2956
1da177e4
LT		 2957 /*
2958 * We return BAD_STATEID if filehandle doesn't match stateid,
2959 * the confirmed flag is incorrecly set, or the generation
2960 * number is incorrect.
1da177e4 2961 */
f8816512
N		 2962 stp = find_stateid(stateid, flags);
2963 if (stp == NULL) {
2964 /*
2965 * Also, we should make sure this isn't just the result of
2966 * a replayed close:
2967 */
2968 sop = search_close_lru(stateid->si_stateownerid, flags);
2969 if (sop == NULL)
e4e83ea4 2970 return nfserr_bad_stateid;
f8816512
N		 2971 *sopp = sop;
2972 goto check_replay;
2973 }
1da177e4 2974
39325bd0
BF		 2975 *stpp = stp;
2976 *sopp = sop = stp->st_stateowner;
2977
4c4cd222 2978 if (lock) {
4c4cd222 2979 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2980 struct nfs4_client *clp = sop->so_client;
4c4cd222 2981 int lkflg = 0;
b37ad28b 2982 __be32 status;
4c4cd222
N		 2983
2984 lkflg = setlkflg(lock->lk_type);
2985
2986 if (lock->lk_is_new) {
599e0a22
BF		 2987 if (!sop->so_is_open_owner)
2988 return nfserr_bad_stateid;
60adfc50
AA		 2989 if (!(flags & HAS_SESSION) &&
2990 !same_clid(&clp->cl_clientid, lockclid))
2991 return nfserr_bad_stateid;
599e0a22
BF		 2992 /* stp is the open stateid */
2993 status = nfs4_check_openmode(stp, lkflg);
2994 if (status)
2995 return status;
2996 } else {
2997 /* stp is the lock stateid */
2998 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2999 if (status)
3000 return status;
4c4cd222 3001 }
1da177e4
LT		 3002 }
3003
f3362737 3004 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3005 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3006 return nfserr_bad_stateid;
1da177e4
LT		 3007 }
3008
1da177e4
LT		 3009 /*
3010 * We now validate the seqid and stateid generation numbers.
3011 * For the moment, we ignore the possibility of
3012 * generation number wraparound.
3013 */
6668958f 3014 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3015 goto check_replay;
3016
3a4f98bb 3017 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3018 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3019 " unconfirmed stateowner!\n");
3020 return nfserr_bad_stateid;
1da177e4 3021 }
3a4f98bb 3022 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3023 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3024 " confirmed yet!\n");
3025 return nfserr_bad_stateid;
1da177e4 3026 }
6668958f 3027 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3028 if (status)
3029 return status;
52fd004e 3030 renew_client(sop->so_client);
3a4f98bb 3031 return nfs_ok;
1da177e4 3032
1da177e4 3033check_replay:
bd9aac52 3034 if (seqid == sop->so_seqid - 1) {
849823c5 3035 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3036 /* indicate replay to calling function */
a90b061c 3037 return nfserr_replay_me;
1da177e4 3038 }
2fdada03 3039 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3040 sop->so_seqid, seqid);
3041 *sopp = NULL;
3042 return nfserr_bad_seqid;
1da177e4
LT		 3043}
3044
b37ad28b 3045__be32
ca364317 3046nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3047 struct nfsd4_open_confirm *oc)
1da177e4 3048{
b37ad28b 3049 __be32 status;
1da177e4
LT		 3050 struct nfs4_stateowner *sop;
3051 struct nfs4_stateid *stp;
3052
3053 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3054 (int)cstate->current_fh.fh_dentry->d_name.len,
3055 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3056
ca364317 3057 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3058 if (status)
3059 return status;
1da177e4
LT		 3060
3061 nfs4_lock_state();
3062
dd453dfd 3063 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3064 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3065 CONFIRM | OPEN_STATE,
1da177e4
LT		 3066 &oc->oc_stateowner, &stp, NULL)))
3067 goto out;
3068
3069 sop = oc->oc_stateowner;
3070 sop->so_confirmed = 1;
3071 update_stateid(&stp->st_stateid);
3072 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH		 3073 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3074 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N		 3075
3076 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3077out:
f2327d9a 3078 if (oc->oc_stateowner) {
1da177e4 3079 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3080 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3081 }
1da177e4
LT		 3082 nfs4_unlock_state();
3083 return status;
3084}
3085
3086
3087/*
3088 * unset all bits in union bitmap (bmap) that
3089 * do not exist in share (from successful OPEN_DOWNGRADE)
3090 */
3091static void
3092reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3093{
3094 int i;
3095 for (i = 1; i < 4; i++) {
3096 if ((i & access) != i)
3097 __clear_bit(i, bmap);
3098 }
3099}
3100
3101static void
3102reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3103{
3104 int i;
3105 for (i = 0; i < 4; i++) {
3106 if ((i & deny) != i)
3107 __clear_bit(i, bmap);
3108 }
3109}
3110
b37ad28b 3111__be32
ca364317
BF		 3112nfsd4_open_downgrade(struct svc_rqst *rqstp,
3113 struct nfsd4_compound_state *cstate,
a4f1706a 3114 struct nfsd4_open_downgrade *od)
1da177e4 3115{
b37ad28b 3116 __be32 status;
1da177e4
LT		 3117 struct nfs4_stateid *stp;
3118 unsigned int share_access;
3119
3120 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3121 (int)cstate->current_fh.fh_dentry->d_name.len,
3122 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3123
d87a8ade 3124 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3125 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3126 return nfserr_inval;
3127
3128 nfs4_lock_state();
dd453dfd 3129 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3130 od->od_seqid,
1da177e4 3131 &od->od_stateid,
f3362737 3132 OPEN_STATE,
1da177e4
LT		 3133 &od->od_stateowner, &stp, NULL)))
3134 goto out;
3135
3136 status = nfserr_inval;
3137 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3138 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3139 stp->st_access_bmap, od->od_share_access);
3140 goto out;
3141 }
3142 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3143 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3144 stp->st_deny_bmap, od->od_share_deny);
3145 goto out;
3146 }
3147 set_access(&share_access, stp->st_access_bmap);
3148 nfs4_file_downgrade(stp->st_vfs_file,
3149 share_access & ~od->od_share_access);
3150
3151 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3152 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3153
3154 update_stateid(&stp->st_stateid);
3155 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3156 status = nfs_ok;
3157out:
f2327d9a 3158 if (od->od_stateowner) {
1da177e4 3159 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3160 cstate->replay_owner = od->od_stateowner;
f2327d9a 3161 }
1da177e4
LT		 3162 nfs4_unlock_state();
3163 return status;
3164}
3165
3166/*
3167 * nfs4_unlock_state() called after encode
3168 */
b37ad28b 3169__be32
ca364317 3170nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3171 struct nfsd4_close *close)
1da177e4 3172{
b37ad28b 3173 __be32 status;
1da177e4
LT		 3174 struct nfs4_stateid *stp;
3175
3176 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3177 (int)cstate->current_fh.fh_dentry->d_name.len,
3178 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3179
3180 nfs4_lock_state();
3181 /* check close_lru for replay */
dd453dfd 3182 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3183 close->cl_seqid,
1da177e4 3184 &close->cl_stateid,
f3362737 3185 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3186 &close->cl_stateowner, &stp, NULL)))
3187 goto out;
1da177e4
LT		 3188 status = nfs_ok;
3189 update_stateid(&stp->st_stateid);
3190 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3191
04ef5954 3192 /* release_stateid() calls nfsd_close() if needed */
2283963f 3193 release_open_stateid(stp);
04ef5954
BF		 3194
3195 /* place unused nfs4_stateowners on so_close_lru list to be
3196 * released by the laundromat service after the lease period
3197 * to enable us to handle CLOSE replay
3198 */
3199 if (list_empty(&close->cl_stateowner->so_stateids))
3200 move_to_close_lru(close->cl_stateowner);
1da177e4 3201out:
f2327d9a 3202 if (close->cl_stateowner) {
1da177e4 3203 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3204 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3205 }
1da177e4
LT		 3206 nfs4_unlock_state();
3207 return status;
3208}
3209
b37ad28b 3210__be32
ca364317
BF		 3211nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3212 struct nfsd4_delegreturn *dr)
1da177e4 3213{
203a8c8e
BF		 3214 struct nfs4_delegation *dp;
3215 stateid_t *stateid = &dr->dr_stateid;
3216 struct inode *inode;
b37ad28b 3217 __be32 status;
6668958f 3218 int flags = 0;
1da177e4 3219
ca364317 3220 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3221 return status;
3222 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3223
6668958f
AA		 3224 if (nfsd4_has_session(cstate))
3225 flags |= HAS_SESSION;
1da177e4 3226 nfs4_lock_state();
203a8c8e
BF		 3227 status = nfserr_bad_stateid;
3228 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3229 goto out;
3230 status = nfserr_stale_stateid;
3231 if (STALE_STATEID(stateid))
3232 goto out;
7e0f7cf5 3233 status = nfserr_bad_stateid;
203a8c8e
BF		 3234 if (!is_delegation_stateid(stateid))
3235 goto out;
203a8c8e 3236 dp = find_delegation_stateid(inode, stateid);
e4e83ea4 3237 if (!dp)
203a8c8e 3238 goto out;
6668958f 3239 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3240 if (status)
3241 goto out;
3242 renew_client(dp->dl_client);
3243
3244 unhash_delegation(dp);
1da177e4 3245out:
203a8c8e
BF		 3246 nfs4_unlock_state();
3247
1da177e4
LT		 3248 return status;
3249}
3250
3251
3252/*
3253 * Lock owner state (byte-range locks)
3254 */
3255#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3256#define LOCK_HASH_BITS 8
3257#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3258#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3259
87df4de8
BH		 3260static inline u64
3261end_offset(u64 start, u64 len)
3262{
3263 u64 end;
3264
3265 end = start + len;
3266 return end >= start ? end: NFS4_MAX_UINT64;
3267}
3268
3269/* last octet in a range */
3270static inline u64
3271last_byte_offset(u64 start, u64 len)
3272{
3273 u64 end;
3274
3275 BUG_ON(!len);
3276 end = start + len;
3277 return end > start ? end - 1: NFS4_MAX_UINT64;
3278}
3279
1da177e4
LT		 3280#define lockownerid_hashval(id) \
3281 ((id) & LOCK_HASH_MASK)
3282
3283static inline unsigned int
3284lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3285 struct xdr_netobj *ownername)
3286{
3287 return (file_hashval(inode) + cl_id
3288 + opaque_hashval(ownername->data, ownername->len))
3289 & LOCK_HASH_MASK;
3290}
3291
3292static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3293static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3294static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3295
fd39ca9a 3296static struct nfs4_stateid *
1da177e4
LT		 3297find_stateid(stateid_t *stid, int flags)
3298{
9346eff0 3299 struct nfs4_stateid *local;
1da177e4
LT		 3300 u32 st_id = stid->si_stateownerid;
3301 u32 f_id = stid->si_fileid;
3302 unsigned int hashval;
3303
3304 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3305 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3306 hashval = stateid_hashval(st_id, f_id);
3307 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3308 if ((local->st_stateid.si_stateownerid == st_id) &&
3309 (local->st_stateid.si_fileid == f_id))
3310 return local;
3311 }
3312 }
9346eff0
KK		 3313
3314 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3315 hashval = stateid_hashval(st_id, f_id);
3316 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3317 if ((local->st_stateid.si_stateownerid == st_id) &&
3318 (local->st_stateid.si_fileid == f_id))
3319 return local;
3320 }
849823c5 3321 }
1da177e4
LT		 3322 return NULL;
3323}
3324
3325static struct nfs4_delegation *
3326find_delegation_stateid(struct inode *ino, stateid_t *stid)
3327{
13cd2184
N		 3328 struct nfs4_file *fp;
3329 struct nfs4_delegation *dl;
1da177e4 3330
8c10cbdb
BH		 3331 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3332 STATEID_VAL(stid));
1da177e4 3333
1da177e4 3334 fp = find_file(ino);
13cd2184
N		 3335 if (!fp)
3336 return NULL;
3337 dl = find_delegation_file(fp, stid);
3338 put_nfs4_file(fp);
3339 return dl;
1da177e4
LT		 3340}
3341
3342/*
3343 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3344 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3345 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3346 * locking, this prevents us from being completely protocol-compliant. The
3347 * real solution to this problem is to start using unsigned file offsets in
3348 * the VFS, but this is a very deep change!
3349 */
3350static inline void
3351nfs4_transform_lock_offset(struct file_lock *lock)
3352{
3353 if (lock->fl_start < 0)
3354 lock->fl_start = OFFSET_MAX;
3355 if (lock->fl_end < 0)
3356 lock->fl_end = OFFSET_MAX;
3357}
3358
d5b9026a
N		 3359/* Hack!: For now, we're defining this just so we can use a pointer to it
3360 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3361static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3362};
1da177e4
LT		 3363
3364static inline void
3365nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3366{
d5b9026a
N		 3367 struct nfs4_stateowner *sop;
3368 unsigned int hval;
1da177e4 3369
d5b9026a
N		 3370 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3371 sop = (struct nfs4_stateowner *) fl->fl_owner;
3372 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3373 kref_get(&sop->so_ref);
3374 deny->ld_sop = sop;
3375 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3376 } else {
3377 deny->ld_sop = NULL;
3378 deny->ld_clientid.cl_boot = 0;
3379 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3380 }
3381 deny->ld_start = fl->fl_start;
87df4de8
BH		 3382 deny->ld_length = NFS4_MAX_UINT64;
3383 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3384 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3385 deny->ld_type = NFS4_READ_LT;
3386 if (fl->fl_type != F_RDLCK)
3387 deny->ld_type = NFS4_WRITE_LT;
3388}
3389
1da177e4
LT		 3390static struct nfs4_stateowner *
3391find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3392 struct xdr_netobj *owner)
3393{
3394 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3395 struct nfs4_stateowner *op;
3396
3397 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3398 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3399 return op;
3400 }
3401 return NULL;
3402}
3403
3404/*
3405 * Alloc a lock owner structure.
3406 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3407 * occured.
3408 *
3409 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3410 */
3411
3412static struct nfs4_stateowner *
3413alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3414 struct nfs4_stateowner *sop;
3415 struct nfs4_replay *rp;
3416 unsigned int idhashval;
3417
3418 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3419 return NULL;
3420 idhashval = lockownerid_hashval(current_ownerid);
3421 INIT_LIST_HEAD(&sop->so_idhash);
3422 INIT_LIST_HEAD(&sop->so_strhash);
3423 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3424 INIT_LIST_HEAD(&sop->so_stateids);
3425 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3426 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3427 sop->so_time = 0;
3428 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3429 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3430 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3431 sop->so_is_open_owner = 0;
3432 sop->so_id = current_ownerid++;
3433 sop->so_client = clp;
b59e3c0e
NB		 3434 /* It is the openowner seqid that will be incremented in encode in the
3435 * case of new lockowners; so increment the lock seqid manually: */
3436 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3437 sop->so_confirmed = 1;
3438 rp = &sop->so_replay;
de1ae286 3439 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3440 rp->rp_buflen = 0;
3441 rp->rp_buf = rp->rp_ibuf;
3442 return sop;
3443}
3444
fd39ca9a 3445static struct nfs4_stateid *
1da177e4
LT		 3446alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3447{
3448 struct nfs4_stateid *stp;
3449 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3450
5ac049ac
N		 3451 stp = nfs4_alloc_stateid();
3452 if (stp == NULL)
1da177e4
LT		 3453 goto out;
3454 INIT_LIST_HEAD(&stp->st_hash);
3455 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3456 INIT_LIST_HEAD(&stp->st_perstateowner);
3457 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3458 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3459 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3460 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3461 stp->st_stateowner = sop;
13cd2184 3462 get_nfs4_file(fp);
1da177e4 3463 stp->st_file = fp;
e4e83ea4 3464 stp->st_stateid.si_boot = boot_time;
1da177e4
LT		 3465 stp->st_stateid.si_stateownerid = sop->so_id;
3466 stp->st_stateid.si_fileid = fp->fi_id;
3467 stp->st_stateid.si_generation = 0;
3468 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3469 stp->st_access_bmap = open_stp->st_access_bmap;
3470 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3471 stp->st_openstp = open_stp;
1da177e4
LT		 3472
3473out:
3474 return stp;
3475}
3476
fd39ca9a 3477static int
1da177e4
LT		 3478check_lock_length(u64 offset, u64 length)
3479{
87df4de8 3480 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3481 LOFF_OVERFLOW(offset, length)));
3482}
3483
3484/*
3485 * LOCK operation
3486 */
b37ad28b 3487__be32
ca364317 3488nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3489 struct nfsd4_lock *lock)
1da177e4 3490{
3e9e3dbe 3491 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3492 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3493 struct nfs4_stateid *lock_stp;
3494 struct file *filp;
3495 struct file_lock file_lock;
8dc7c311 3496 struct file_lock conflock;
b37ad28b 3497 __be32 status = 0;
1da177e4 3498 unsigned int strhashval;
150b3934 3499 unsigned int cmd;
b8dd7b9a 3500 int err;
1da177e4
LT		 3501
3502 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3503 (long long) lock->lk_offset,
3504 (long long) lock->lk_length);
3505
1da177e4
LT		 3506 if (check_lock_length(lock->lk_offset, lock->lk_length))
3507 return nfserr_inval;
3508
ca364317 3509 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3510 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3511 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3512 return status;
3513 }
3514
1da177e4
LT		 3515 nfs4_lock_state();
3516
3517 if (lock->lk_is_new) {
893f8770
N		 3518 /*
3519 * Client indicates that this is a new lockowner.
3520 * Use open owner and open stateid to create lock owner and
3521 * lock stateid.
3522 */
1da177e4
LT		 3523 struct nfs4_stateid *open_stp = NULL;
3524 struct nfs4_file *fp;
3525
3526 status = nfserr_stale_clientid;
60adfc50
AA		 3527 if (!nfsd4_has_session(cstate) &&
3528 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3529 goto out;
1da177e4 3530
1da177e4 3531 /* validate and update open stateid and open seqid */
dd453dfd 3532 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3533 lock->lk_new_open_seqid,
3534 &lock->lk_new_open_stateid,
f3362737 3535 OPEN_STATE,
3a65588a 3536 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3537 lock);
37515177 3538 if (status)
1da177e4 3539 goto out;
3a65588a 3540 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3541 /* create lockowner and lock stateid */
3542 fp = open_stp->st_file;
3543 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3544 open_sop->so_client->cl_clientid.cl_id,
3545 &lock->v.new.owner);
3e9e3dbe
N		 3546 /* XXX: Do we need to check for duplicate stateowners on
3547 * the same file, or should they just be allowed (and
3548 * create new stateids)? */
1da177e4 3549 status = nfserr_resource;
b59e3c0e
NB		 3550 lock_sop = alloc_init_lock_stateowner(strhashval,
3551 open_sop->so_client, open_stp, lock);
3552 if (lock_sop == NULL)
1da177e4 3553 goto out;
b59e3c0e 3554 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3555 if (lock_stp == NULL)
1da177e4 3556 goto out;
1da177e4
LT		 3557 } else {
3558 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3559 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3560 lock->lk_old_lock_seqid,
3561 &lock->lk_old_lock_stateid,
f3362737 3562 LOCK_STATE,
3a65588a 3563 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3564 if (status)
3565 goto out;
3a65588a 3566 lock_sop = lock->lk_replay_owner;
1da177e4 3567 }
3a65588a 3568 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3569 filp = lock_stp->st_vfs_file;
3570
0dd395dc 3571 status = nfserr_grace;
af558e33 3572 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3573 goto out;
3574 status = nfserr_no_grace;
af558e33 3575 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3576 goto out;
3577
1da177e4
LT		 3578 locks_init_lock(&file_lock);
3579 switch (lock->lk_type) {
3580 case NFS4_READ_LT:
3581 case NFS4_READW_LT:
3582 file_lock.fl_type = F_RDLCK;
150b3934 3583 cmd = F_SETLK;
1da177e4
LT		 3584 break;
3585 case NFS4_WRITE_LT:
3586 case NFS4_WRITEW_LT:
3587 file_lock.fl_type = F_WRLCK;
150b3934 3588 cmd = F_SETLK;
1da177e4
LT		 3589 break;
3590 default:
3591 status = nfserr_inval;
3592 goto out;
3593 }
b59e3c0e 3594 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3595 file_lock.fl_pid = current->tgid;
3596 file_lock.fl_file = filp;
3597 file_lock.fl_flags = FL_POSIX;
d5b9026a 3598 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3599
3600 file_lock.fl_start = lock->lk_offset;
87df4de8 3601 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3602 nfs4_transform_lock_offset(&file_lock);
3603
3604 /*
3605 * Try to lock the file in the VFS.
3606 * Note: locks.c uses the BKL to protect the inode's lock list.
3607 */
3608
fd85b817 3609 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3610 switch (-err) {
1da177e4
LT		 3611 case 0: /* success! */
3612 update_stateid(&lock_stp->st_stateid);
3613 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3614 sizeof(stateid_t));
b8dd7b9a 3615 status = 0;
eb76b3fd
AA		 3616 break;
3617 case (EAGAIN): /* conflock holds conflicting lock */
3618 status = nfserr_denied;
3619 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3620 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3621 break;
1da177e4
LT		 3622 case (EDEADLK):
3623 status = nfserr_deadlock;
eb76b3fd 3624 break;
1da177e4 3625 default:
fd85b817 3626 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3627 status = nfserr_resource;
3628 break;
1da177e4 3629 }
1da177e4 3630out:
8a280510 3631 if (status && lock->lk_is_new && lock_sop)
f044ff83 3632 release_lockowner(lock_sop);
3a65588a
BF		 3633 if (lock->lk_replay_owner) {
3634 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3635 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3636 }
1da177e4
LT		 3637 nfs4_unlock_state();
3638 return status;
3639}
3640
55ef1274
BF		 3641/*
3642 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3643 * so we do a temporary open here just to get an open file to pass to
3644 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3645 * inode operation.)
3646 */
3647static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3648{
3649 struct file *file;
3650 int err;
3651
3652 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3653 if (err)
3654 return err;
3655 err = vfs_test_lock(file, lock);
3656 nfsd_close(file);
3657 return err;
3658}
3659
1da177e4
LT		 3660/*
3661 * LOCKT operation
3662 */
b37ad28b 3663__be32
ca364317
BF		 3664nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3665 struct nfsd4_lockt *lockt)
1da177e4
LT		 3666{
3667 struct inode *inode;
1da177e4 3668 struct file_lock file_lock;
fd85b817 3669 int error;
b37ad28b 3670 __be32 status;
1da177e4 3671
af558e33 3672 if (locks_in_grace())
1da177e4
LT		 3673 return nfserr_grace;
3674
3675 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3676 return nfserr_inval;
3677
3678 lockt->lt_stateowner = NULL;
3679 nfs4_lock_state();
3680
3681 status = nfserr_stale_clientid;
60adfc50 3682 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3683 goto out;
1da177e4 3684
ca364317 3685 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3686 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3687 if (status == nfserr_symlink)
3688 status = nfserr_inval;
3689 goto out;
3690 }
3691
ca364317 3692 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3693 locks_init_lock(&file_lock);
3694 switch (lockt->lt_type) {
3695 case NFS4_READ_LT:
3696 case NFS4_READW_LT:
3697 file_lock.fl_type = F_RDLCK;
3698 break;
3699 case NFS4_WRITE_LT:
3700 case NFS4_WRITEW_LT:
3701 file_lock.fl_type = F_WRLCK;
3702 break;
3703 default:
2fdada03 3704 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3705 status = nfserr_inval;
3706 goto out;
3707 }
3708
3709 lockt->lt_stateowner = find_lockstateowner_str(inode,
3710 &lockt->lt_clientid, &lockt->lt_owner);
3711 if (lockt->lt_stateowner)
3712 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3713 file_lock.fl_pid = current->tgid;
3714 file_lock.fl_flags = FL_POSIX;
3715
3716 file_lock.fl_start = lockt->lt_offset;
87df4de8 3717 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3718
3719 nfs4_transform_lock_offset(&file_lock);
3720
1da177e4 3721 status = nfs_ok;
55ef1274 3722 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3723 if (error) {
3724 status = nfserrno(error);
3725 goto out;
3726 }
9d6a8c5c 3727 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3728 status = nfserr_denied;
9d6a8c5c 3729 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3730 }
3731out:
3732 nfs4_unlock_state();
3733 return status;
3734}
3735
b37ad28b 3736__be32
ca364317 3737nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3738 struct nfsd4_locku *locku)
1da177e4
LT		 3739{
3740 struct nfs4_stateid *stp;
3741 struct file *filp = NULL;
3742 struct file_lock file_lock;
b37ad28b 3743 __be32 status;
b8dd7b9a 3744 int err;
1da177e4
LT		 3745
3746 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3747 (long long) locku->lu_offset,
3748 (long long) locku->lu_length);
3749
3750 if (check_lock_length(locku->lu_offset, locku->lu_length))
3751 return nfserr_inval;
3752
3753 nfs4_lock_state();
3754
dd453dfd 3755 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3756 locku->lu_seqid,
3757 &locku->lu_stateid,
f3362737 3758 LOCK_STATE,
1da177e4
LT		 3759 &locku->lu_stateowner, &stp, NULL)))
3760 goto out;
3761
3762 filp = stp->st_vfs_file;
3763 BUG_ON(!filp);
3764 locks_init_lock(&file_lock);
3765 file_lock.fl_type = F_UNLCK;
3766 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3767 file_lock.fl_pid = current->tgid;
3768 file_lock.fl_file = filp;
3769 file_lock.fl_flags = FL_POSIX;
d5b9026a 3770 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3771 file_lock.fl_start = locku->lu_offset;
3772
87df4de8 3773 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3774 nfs4_transform_lock_offset(&file_lock);
3775
3776 /*
3777 * Try to unlock the file in the VFS.
3778 */
fd85b817 3779 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3780 if (err) {
fd85b817 3781 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3782 goto out_nfserr;
3783 }
3784 /*
3785 * OK, unlock succeeded; the only thing left to do is update the stateid.
3786 */
3787 update_stateid(&stp->st_stateid);
3788 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3789
3790out:
f2327d9a 3791 if (locku->lu_stateowner) {
1da177e4 3792 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3793 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3794 }
1da177e4
LT		 3795 nfs4_unlock_state();
3796 return status;
3797
3798out_nfserr:
b8dd7b9a 3799 status = nfserrno(err);
1da177e4
LT		 3800 goto out;
3801}
3802
3803/*
3804 * returns
3805 * 1: locks held by lockowner
3806 * 0: no locks held by lockowner
3807 */
3808static int
3809check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3810{
3811 struct file_lock **flpp;
7eaa36e2 3812 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3813 int status = 0;
3814
3815 lock_kernel();
3816 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3817 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3818 status = 1;
3819 goto out;
796dadfd 3820 }
1da177e4
LT		 3821 }
3822out:
3823 unlock_kernel();
3824 return status;
3825}
3826
b37ad28b 3827__be32
b591480b
BF		 3828nfsd4_release_lockowner(struct svc_rqst *rqstp,
3829 struct nfsd4_compound_state *cstate,
3830 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3831{
3832 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3833 struct nfs4_stateowner *sop;
3834 struct nfs4_stateid *stp;
1da177e4 3835 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3836 struct list_head matches;
3837 int i;
b37ad28b 3838 __be32 status;
1da177e4
LT		 3839
3840 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3841 clid->cl_boot, clid->cl_id);
3842
3843 /* XXX check for lease expiration */
3844
3845 status = nfserr_stale_clientid;
849823c5 3846 if (STALE_CLIENTID(clid))
1da177e4 3847 return status;
1da177e4
LT		 3848
3849 nfs4_lock_state();
3850
3e9e3dbe
N		 3851 status = nfserr_locks_held;
3852 /* XXX: we're doing a linear search through all the lockowners.
3853 * Yipes! For now we'll just hope clients aren't really using
3854 * release_lockowner much, but eventually we have to fix these
3855 * data structures. */
3856 INIT_LIST_HEAD(&matches);
3857 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3858 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3859 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3860 continue;
3861 list_for_each_entry(stp, &sop->so_stateids,
3862 st_perstateowner) {
3863 if (check_for_locks(stp->st_vfs_file, sop))
3864 goto out;
3865 /* Note: so_perclient unused for lockowners,
3866 * so it's OK to fool with here. */
3867 list_add(&sop->so_perclient, &matches);
3868 }
1da177e4 3869 }
3e9e3dbe
N		 3870 }
3871 /* Clients probably won't expect us to return with some (but not all)
3872 * of the lockowner state released; so don't release any until all
3873 * have been checked. */
3874 status = nfs_ok;
0fa822e4
N		 3875 while (!list_empty(&matches)) {
3876 sop = list_entry(matches.next, struct nfs4_stateowner,
3877 so_perclient);
3878 /* unhash_stateowner deletes so_perclient only
3879 * for openowners. */
3880 list_del(&sop->so_perclient);
f044ff83 3881 release_lockowner(sop);
1da177e4
LT		 3882 }
3883out:
3884 nfs4_unlock_state();
3885 return status;
3886}
3887
3888static inline struct nfs4_client_reclaim *
a55370a3 3889alloc_reclaim(void)
1da177e4 3890{
a55370a3 3891 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3892}
3893
c7b9a459 3894int
a1bcecd2 3895nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3896{
3897 unsigned int strhashval = clientstr_hashval(name);
3898 struct nfs4_client *clp;
3899
a1bcecd2 3900 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3901 return clp ? 1 : 0;
3902}
3903
1da177e4
LT		 3904/*
3905 * failure => all reset bets are off, nfserr_no_grace...
3906 */
190e4fbf
N		 3907int
3908nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3909{
3910 unsigned int strhashval;
3911 struct nfs4_client_reclaim *crp = NULL;
3912
a55370a3
N		 3913 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3914 crp = alloc_reclaim();
1da177e4
LT		 3915 if (!crp)
3916 return 0;
a55370a3 3917 strhashval = clientstr_hashval(name);
1da177e4
LT		 3918 INIT_LIST_HEAD(&crp->cr_strhash);
3919 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3920 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3921 reclaim_str_hashtbl_size++;
3922 return 1;
3923}
3924
3925static void
3926nfs4_release_reclaim(void)
3927{
3928 struct nfs4_client_reclaim *crp = NULL;
3929 int i;
3930
1da177e4
LT		 3931 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3932 while (!list_empty(&reclaim_str_hashtbl[i])) {
3933 crp = list_entry(reclaim_str_hashtbl[i].next,
3934 struct nfs4_client_reclaim, cr_strhash);
3935 list_del(&crp->cr_strhash);
1da177e4
LT		 3936 kfree(crp);
3937 reclaim_str_hashtbl_size--;
3938 }
3939 }
3940 BUG_ON(reclaim_str_hashtbl_size);
3941}
3942
3943/*
3944 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3945static struct nfs4_client_reclaim *
1da177e4
LT		 3946nfs4_find_reclaim_client(clientid_t *clid)
3947{
3948 unsigned int strhashval;
3949 struct nfs4_client *clp;
3950 struct nfs4_client_reclaim *crp = NULL;
3951
3952
3953 /* find clientid in conf_id_hashtbl */
3954 clp = find_confirmed_client(clid);
3955 if (clp == NULL)
3956 return NULL;
3957
a55370a3
N		 3958 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3959 clp->cl_name.len, clp->cl_name.data,
3960 clp->cl_recdir);
1da177e4
LT		 3961
3962 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3963 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3964 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3965 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3966 return crp;
3967 }
3968 }
3969 return NULL;
3970}
3971
3972/*
3973* Called from OPEN. Look for clientid in reclaim list.
3974*/
b37ad28b 3975__be32
1da177e4
LT		 3976nfs4_check_open_reclaim(clientid_t *clid)
3977{
dfc83565 3978 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3979}
3980
ac4d8ff2 3981/* initialization to perform at module load time: */
1da177e4 3982
e8ff2a84 3983int
ac4d8ff2 3984nfs4_state_init(void)
1da177e4 3985{
e8ff2a84 3986 int i, status;
1da177e4 3987
e8ff2a84
BF		 3988 status = nfsd4_init_slabs();
3989 if (status)
3990 return status;
1da177e4
LT		 3991 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3992 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3993 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3994 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3995 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3996 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3997 }
5282fd72
ME		 3998 for (i = 0; i < SESSION_HASH_SIZE; i++)
3999 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4000 for (i = 0; i < FILE_HASH_SIZE; i++) {
4001 INIT_LIST_HEAD(&file_hashtbl[i]);
4002 }
4003 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4004 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4005 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4006 }
4007 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4008 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4009 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4010 }
4011 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4012 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4013 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4014 }
1da177e4 4015 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4016 INIT_LIST_HEAD(&close_lru);
4017 INIT_LIST_HEAD(&client_lru);
4018 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4019 reclaim_str_hashtbl_size = 0;
e8ff2a84 4020 return 0;
ac4d8ff2
N		 4021}
4022
190e4fbf
N		 4023static void
4024nfsd4_load_reboot_recovery_data(void)
4025{
4026 int status;
4027
0964a3d3
N		 4028 nfs4_lock_state();
4029 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4030 status = nfsd4_recdir_load();
0964a3d3 4031 nfs4_unlock_state();
190e4fbf
N		 4032 if (status)
4033 printk("NFSD: Failure reading reboot recovery data\n");
4034}
4035
c2f1a551
MS		 4036/*
4037 * Since the lifetime of a delegation isn't limited to that of an open, a
4038 * client may quite reasonably hang on to a delegation as long as it has
4039 * the inode cached. This becomes an obvious problem the first time a
4040 * client's inode cache approaches the size of the server's total memory.
4041 *
4042 * For now we avoid this problem by imposing a hard limit on the number
4043 * of delegations, which varies according to the server's memory size.
4044 */
4045static void
4046set_max_delegations(void)
4047{
4048 /*
4049 * Allow at most 4 delegations per megabyte of RAM. Quick
4050 * estimates suggest that in the worst case (where every delegation
4051 * is for a different inode), a delegation could take about 1.5K,
4052 * giving a worst case usage of about 6% of memory.
4053 */
4054 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4055}
4056
ac4d8ff2
N		 4057/* initialization to perform when the nfsd service is started: */
4058
29ab23cc 4059static int
ac4d8ff2
N		 4060__nfs4_state_start(void)
4061{
b5a1a81e
BF		 4062 int ret;
4063
1da177e4 4064 boot_time = get_seconds();
af558e33 4065 locks_start_grace(&nfsd4_manager);
9a8db97e 4066 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
e46b498c 4067 nfsd4_grace);
b5a1a81e
BF		 4068 ret = set_callback_cred();
4069 if (ret)
4070 return -ENOMEM;
58da282b 4071 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 4072 if (laundry_wq == NULL)
4073 return -ENOMEM;
b5a1a81e
BF		 4074 ret = nfsd4_create_callback_queue();
4075 if (ret)
4076 goto out_free_laundry;
e46b498c 4077 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
c2f1a551 4078 set_max_delegations();
b5a1a81e
BF		 4079 return 0;
4080out_free_laundry:
4081 destroy_workqueue(laundry_wq);
4082 return ret;
1da177e4
LT		 4083}
4084
29ab23cc 4085int
76a3550e 4086nfs4_state_start(void)
1da177e4 4087{
190e4fbf 4088 nfsd4_load_reboot_recovery_data();
4ad9a344 4089 return __nfs4_state_start();
1da177e4
LT		 4090}
4091
1da177e4
LT		 4092static void
4093__nfs4_state_shutdown(void)
4094{
4095 int i;
4096 struct nfs4_client *clp = NULL;
4097 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4098 struct list_head *pos, *next, reaplist;
4099
1da177e4
LT		 4100 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4101 while (!list_empty(&conf_id_hashtbl[i])) {
4102 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4103 expire_client(clp);
4104 }
4105 while (!list_empty(&unconf_str_hashtbl[i])) {
4106 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4107 expire_client(clp);
4108 }
4109 }
4110 INIT_LIST_HEAD(&reaplist);
4111 spin_lock(&recall_lock);
4112 list_for_each_safe(pos, next, &del_recall_lru) {
4113 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4114 list_move(&dp->dl_recall_lru, &reaplist);
4115 }
4116 spin_unlock(&recall_lock);
4117 list_for_each_safe(pos, next, &reaplist) {
4118 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4119 list_del_init(&dp->dl_recall_lru);
4120 unhash_delegation(dp);
4121 }
4122
190e4fbf 4123 nfsd4_shutdown_recdir();
1da177e4
LT		 4124}
4125
4126void
4127nfs4_state_shutdown(void)
4128{
5e8d5c29
N		 4129 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4130 destroy_workqueue(laundry_wq);
2c5e7615 4131 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4132 nfs4_lock_state();
4133 nfs4_release_reclaim();
4134 __nfs4_state_shutdown();
b5a1a81e 4135 nfsd4_destroy_callback_queue();
1da177e4
LT		 4136 nfs4_unlock_state();
4137}
4138
3dd98a3b
JL		 4139/*
4140 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4141 * accessed when nfsd is starting.
4142 */
0964a3d3
N		 4143static void
4144nfs4_set_recdir(char *recdir)
4145{
0964a3d3 4146 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4147}
4148
4149/*
4150 * Change the NFSv4 recovery directory to recdir.
4151 */
4152int
4153nfs4_reset_recoverydir(char *recdir)
4154{
4155 int status;
a63bb996 4156 struct path path;
0964a3d3 4157
a63bb996 4158 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4159 if (status)
4160 return status;
4161 status = -ENOTDIR;
a63bb996 4162 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4163 nfs4_set_recdir(recdir);
4164 status = 0;
4165 }
a63bb996 4166 path_put(&path);
0964a3d3
N		 4167 return status;
4168}
4169
3dd98a3b
JL		 4170char *
4171nfs4_recoverydir(void)
4172{
4173 return user_recovery_dirname;
4174}
Linux kernel - Deliverables
This page took 0.759433 seconds and 5 git commands to generate.