nfsd4: nfsd4_destroy_session must set callback client under the state lock
[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;
54static u32 nfs4_init;
fd39ca9a
N
55static stateid_t zerostateid; /* bits all 0 */
56static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 57static u64 current_sessionid = 1;
fd39ca9a
N
58
59#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 61
1da177e4 62/* forward declarations */
fd39ca9a 63static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 64static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N
65static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
66static void nfs4_set_recdir(char *recdir);
1da177e4 67
8b671b80
BF
68/* Locking: */
69
70/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 71static DEFINE_MUTEX(client_mutex);
1da177e4 72
8b671b80
BF
73/*
74 * Currently used for the del_recall_lru and file hash table. In an
75 * effort to decrease the scope of the client_mutex, this spinlock may
76 * eventually cover more:
77 */
78static DEFINE_SPINLOCK(recall_lock);
79
e18b890b
CL
80static struct kmem_cache *stateowner_slab = NULL;
81static struct kmem_cache *file_slab = NULL;
82static struct kmem_cache *stateid_slab = NULL;
83static struct kmem_cache *deleg_slab = NULL;
e60d4398 84
1da177e4
LT
85void
86nfs4_lock_state(void)
87{
353ab6e9 88 mutex_lock(&client_mutex);
1da177e4
LT
89}
90
91void
92nfs4_unlock_state(void)
93{
353ab6e9 94 mutex_unlock(&client_mutex);
1da177e4
LT
95}
96
97static inline u32
98opaque_hashval(const void *ptr, int nbytes)
99{
100 unsigned char *cptr = (unsigned char *) ptr;
101
102 u32 x = 0;
103 while (nbytes--) {
104 x *= 37;
105 x += *cptr++;
106 }
107 return x;
108}
109
1da177e4
LT
110static struct list_head del_recall_lru;
111
13cd2184
N
112static inline void
113put_nfs4_file(struct nfs4_file *fi)
114{
8b671b80
BF
115 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
116 list_del(&fi->fi_hash);
117 spin_unlock(&recall_lock);
118 iput(fi->fi_inode);
119 kmem_cache_free(file_slab, fi);
120 }
13cd2184
N
121}
122
123static inline void
124get_nfs4_file(struct nfs4_file *fi)
125{
8b671b80 126 atomic_inc(&fi->fi_ref);
13cd2184
N
127}
128
ef0f3390 129static int num_delegations;
c2f1a551 130unsigned int max_delegations;
ef0f3390
N
131
132/*
133 * Open owner state (share locks)
134 */
135
136/* hash tables for nfs4_stateowner */
137#define OWNER_HASH_BITS 8
138#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
139#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140
141#define ownerid_hashval(id) \
142 ((id) & OWNER_HASH_MASK)
143#define ownerstr_hashval(clientid, ownername) \
144 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145
146static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
147static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
148
149/* hash table for nfs4_file */
150#define FILE_HASH_BITS 8
151#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
153/* hash table for (open)nfs4_stateid */
154#define STATEID_HASH_BITS 10
155#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
156#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157
158#define file_hashval(x) \
159 hash_ptr(x, FILE_HASH_BITS)
160#define stateid_hashval(owner_id, file_id) \
161 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162
163static struct list_head file_hashtbl[FILE_HASH_SIZE];
164static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
165
1da177e4
LT
166static struct nfs4_delegation *
167alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
168{
169 struct nfs4_delegation *dp;
170 struct nfs4_file *fp = stp->st_file;
c237dc03 171 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT
172
173 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS
174 if (fp->fi_had_conflict)
175 return NULL;
c2f1a551 176 if (num_delegations > max_delegations)
ef0f3390 177 return NULL;
5b2d21c1
N
178 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
179 if (dp == NULL)
1da177e4 180 return dp;
ef0f3390 181 num_delegations++;
ea1da636
N
182 INIT_LIST_HEAD(&dp->dl_perfile);
183 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT
184 INIT_LIST_HEAD(&dp->dl_recall_lru);
185 dp->dl_client = clp;
13cd2184 186 get_nfs4_file(fp);
1da177e4
LT
187 dp->dl_file = fp;
188 dp->dl_flock = NULL;
189 get_file(stp->st_vfs_file);
190 dp->dl_vfs_file = stp->st_vfs_file;
191 dp->dl_type = type;
b53d40c5 192 dp->dl_ident = cb->cb_ident;
78155ed7 193 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT
194 dp->dl_stateid.si_stateownerid = current_delegid++;
195 dp->dl_stateid.si_fileid = 0;
196 dp->dl_stateid.si_generation = 0;
6c02eaa1 197 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT
198 dp->dl_time = 0;
199 atomic_set(&dp->dl_count, 1);
ea1da636
N
200 list_add(&dp->dl_perfile, &fp->fi_delegations);
201 list_add(&dp->dl_perclnt, &clp->cl_delegations);
b5a1a81e 202 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
1da177e4
LT
203 return dp;
204}
205
206void
207nfs4_put_delegation(struct nfs4_delegation *dp)
208{
209 if (atomic_dec_and_test(&dp->dl_count)) {
210 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 211 put_nfs4_file(dp->dl_file);
5b2d21c1 212 kmem_cache_free(deleg_slab, dp);
ef0f3390 213 num_delegations--;
1da177e4
LT
214 }
215}
216
217/* Remove the associated file_lock first, then remove the delegation.
218 * lease_modify() is called to remove the FS_LEASE file_lock from
219 * the i_flock list, eventually calling nfsd's lock_manager
220 * fl_release_callback.
221 */
222static void
223nfs4_close_delegation(struct nfs4_delegation *dp)
224{
225 struct file *filp = dp->dl_vfs_file;
226
227 dprintk("NFSD: close_delegation dp %p\n",dp);
228 dp->dl_vfs_file = NULL;
229 /* The following nfsd_close may not actually close the file,
230 * but we want to remove the lease in any case. */
c907132d 231 if (dp->dl_flock)
a9933cea 232 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 233 nfsd_close(filp);
1da177e4
LT
234}
235
236/* Called under the state lock. */
237static void
238unhash_delegation(struct nfs4_delegation *dp)
239{
ea1da636
N
240 list_del_init(&dp->dl_perfile);
241 list_del_init(&dp->dl_perclnt);
1da177e4
LT
242 spin_lock(&recall_lock);
243 list_del_init(&dp->dl_recall_lru);
244 spin_unlock(&recall_lock);
245 nfs4_close_delegation(dp);
246 nfs4_put_delegation(dp);
247}
248
249/*
250 * SETCLIENTID state
251 */
252
36acb66b 253/* client_lock protects the client lru list and session hash table */
9089f1b4
BH
254static DEFINE_SPINLOCK(client_lock);
255
1da177e4
LT
256/* Hash tables for nfs4_clientid state */
257#define CLIENT_HASH_BITS 4
258#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
260
261#define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
a55370a3
N
263#define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT
265/*
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
268 *
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
271 *
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
273 * setclientid info.
274 *
275 * client_lru holds client queue ordered by nfs4_client.cl_time
276 * for lease renewal.
277 *
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
280 */
281static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
282static int reclaim_str_hashtbl_size = 0;
283static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
284static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
285static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
286static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
287static struct list_head client_lru;
288static struct list_head close_lru;
289
2283963f
BF
290static void unhash_generic_stateid(struct nfs4_stateid *stp)
291{
292 list_del(&stp->st_hash);
293 list_del(&stp->st_perfile);
294 list_del(&stp->st_perstateowner);
295}
296
297static void free_generic_stateid(struct nfs4_stateid *stp)
298{
299 put_nfs4_file(stp->st_file);
300 kmem_cache_free(stateid_slab, stp);
301}
302
303static void release_lock_stateid(struct nfs4_stateid *stp)
304{
305 unhash_generic_stateid(stp);
306 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
307 free_generic_stateid(stp);
308}
309
f044ff83
BF
310static void unhash_lockowner(struct nfs4_stateowner *sop)
311{
312 struct nfs4_stateid *stp;
313
314 list_del(&sop->so_idhash);
315 list_del(&sop->so_strhash);
316 list_del(&sop->so_perstateid);
317 while (!list_empty(&sop->so_stateids)) {
318 stp = list_first_entry(&sop->so_stateids,
319 struct nfs4_stateid, st_perstateowner);
320 release_lock_stateid(stp);
321 }
322}
323
324static void release_lockowner(struct nfs4_stateowner *sop)
325{
326 unhash_lockowner(sop);
327 nfs4_put_stateowner(sop);
328}
329
f1d110ca
BF
330static void
331release_stateid_lockowners(struct nfs4_stateid *open_stp)
332{
333 struct nfs4_stateowner *lock_sop;
334
335 while (!list_empty(&open_stp->st_lockowners)) {
336 lock_sop = list_entry(open_stp->st_lockowners.next,
337 struct nfs4_stateowner, so_perstateid);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 340 release_lockowner(lock_sop);
f1d110ca
BF
341 }
342}
343
2283963f
BF
344static void release_open_stateid(struct nfs4_stateid *stp)
345{
346 unhash_generic_stateid(stp);
347 release_stateid_lockowners(stp);
348 nfsd_close(stp->st_vfs_file);
349 free_generic_stateid(stp);
350}
351
f044ff83 352static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF
353{
354 struct nfs4_stateid *stp;
355
356 list_del(&sop->so_idhash);
357 list_del(&sop->so_strhash);
f044ff83
BF
358 list_del(&sop->so_perclient);
359 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 360 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF
361 stp = list_first_entry(&sop->so_stateids,
362 struct nfs4_stateid, st_perstateowner);
363 release_open_stateid(stp);
f1d110ca
BF
364 }
365}
366
f044ff83 367static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 368{
f044ff83 369 unhash_openowner(sop);
f1d110ca
BF
370 list_del(&sop->so_close_lru);
371 nfs4_put_stateowner(sop);
372}
373
5282fd72
ME
374#define SESSION_HASH_SIZE 512
375static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
376
377static inline int
378hash_sessionid(struct nfs4_sessionid *sessionid)
379{
380 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
381
382 return sid->sequence % SESSION_HASH_SIZE;
383}
384
385static inline void
386dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
387{
388 u32 *ptr = (u32 *)(&sessionid->data[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
390}
391
ec6b5d7b
AA
392static void
393gen_sessionid(struct nfsd4_session *ses)
394{
395 struct nfs4_client *clp = ses->se_client;
396 struct nfsd4_sessionid *sid;
397
398 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
399 sid->clientid = clp->cl_clientid;
400 sid->sequence = current_sessionid++;
401 sid->reserved = 0;
402}
403
404/*
a649637c
AA
405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
415 */
416#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
417
418/*
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 423 *
a649637c
AA
424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 427 */
a649637c 428static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 429{
a649637c 430 int mem, size = fchan->maxresp_cached;
ec6b5d7b 431
5d77ddfb
AA
432 if (fchan->maxreqs < 1)
433 return nfserr_inval;
5d77ddfb 434
a649637c
AA
435 if (size < NFSD_MIN_HDR_SEQ_SZ)
436 size = NFSD_MIN_HDR_SEQ_SZ;
437 size -= NFSD_MIN_HDR_SEQ_SZ;
438 if (size > NFSD_SLOT_CACHE_SIZE)
439 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 440
a649637c
AA
441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem = fchan->maxreqs * size;
443 if (mem > NFSD_MAX_MEM_PER_SESSION) {
444 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
445 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
446 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
447 mem = fchan->maxreqs * size;
448 }
ec6b5d7b 449
4bd9b0f4 450 spin_lock(&nfsd_drc_lock);
a649637c
AA
451 /* bound the total session drc memory ussage */
452 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
453 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
454 mem = fchan->maxreqs * size;
455 }
0c193054 456 nfsd_drc_mem_used += mem;
4bd9b0f4 457 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 458
b101ebbc 459 if (fchan->maxreqs == 0)
bdac86e2 460 return nfserr_serverfault;
a649637c
AA
461
462 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 463 return 0;
ec6b5d7b
AA
464}
465
466/*
467 * fchan holds the client values on input, and the server values on output
ddc04fd4 468 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA
469 */
470static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB
471 struct nfsd4_channel_attrs *session_fchan,
472 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA
473{
474 int status = 0;
ddc04fd4 475 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA
476
477 /* headerpadsz set to zero in encode routine */
478
479 /* Use the client's max request and max response size if possible */
480 if (fchan->maxreq_sz > maxcount)
481 fchan->maxreq_sz = maxcount;
6c18ba9f 482 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA
483
484 if (fchan->maxresp_sz > maxcount)
485 fchan->maxresp_sz = maxcount;
6c18ba9f 486 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 487
ec6b5d7b
AA
488 /* Use the client's maxops if possible */
489 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
490 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 491 session_fchan->maxops = fchan->maxops;
ec6b5d7b 492
ec6b5d7b
AA
493 /* FIXME: Error means no more DRC pages so the server should
494 * recover pages from existing sessions. For now fail session
495 * creation.
496 */
a649637c 497 status = set_forechannel_drc_size(fchan);
ec6b5d7b 498
a649637c 499 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 500 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA
501
502 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA
503 return status;
504}
505
557ce264
AA
506static void
507free_session_slots(struct nfsd4_session *ses)
508{
509 int i;
510
511 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
512 kfree(ses->se_slots[i]);
513}
514
efe0cb6d
BF
515/*
516 * We don't actually need to cache the rpc and session headers, so we
517 * can allocate a little less for each slot:
518 */
519static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
520{
521 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
522}
523
ec6b5d7b
AA
524static int
525alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
526 struct nfsd4_create_session *cses)
527{
528 struct nfsd4_session *new, tmp;
557ce264
AA
529 struct nfsd4_slot *sp;
530 int idx, slotsize, cachesize, i;
531 int status;
ec6b5d7b
AA
532
533 memset(&tmp, 0, sizeof(tmp));
534
535 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB
536 tmp.se_bchannel = cses->back_channel;
537 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
538 &cses->fore_channel);
ec6b5d7b
AA
539 if (status)
540 goto out;
541
557ce264
AA
542 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
543 + sizeof(struct nfsd4_session) > PAGE_SIZE);
544
545 status = nfserr_serverfault;
546 /* allocate struct nfsd4_session and slot table pointers in one piece */
547 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA
548 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
549 if (!new)
550 goto out;
551
552 memcpy(new, &tmp, sizeof(*new));
553
557ce264 554 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 555 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA
556 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
557 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
558 if (!sp)
559 goto out_free;
560 new->se_slots[i] = sp;
561 }
562
ec6b5d7b
AA
563 new->se_client = clp;
564 gen_sessionid(new);
565 idx = hash_sessionid(&new->se_sessionid);
566 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
567 NFS4_MAX_SESSIONID_LEN);
568
569 new->se_flags = cses->flags;
570 kref_init(&new->se_ref);
9089f1b4 571 spin_lock(&client_lock);
ec6b5d7b
AA
572 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
573 list_add(&new->se_perclnt, &clp->cl_sessions);
9089f1b4 574 spin_unlock(&client_lock);
ec6b5d7b
AA
575
576 status = nfs_ok;
577out:
578 return status;
557ce264
AA
579out_free:
580 free_session_slots(new);
581 kfree(new);
582 goto out;
ec6b5d7b
AA
583}
584
9089f1b4 585/* caller must hold client_lock */
5282fd72
ME
586static struct nfsd4_session *
587find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
588{
589 struct nfsd4_session *elem;
590 int idx;
591
592 dump_sessionid(__func__, sessionid);
593 idx = hash_sessionid(sessionid);
594 dprintk("%s: idx is %d\n", __func__, idx);
595 /* Search in the appropriate list */
596 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
597 dump_sessionid("list traversal", &elem->se_sessionid);
598 if (!memcmp(elem->se_sessionid.data, sessionid->data,
599 NFS4_MAX_SESSIONID_LEN)) {
600 return elem;
601 }
602 }
603
604 dprintk("%s: session not found\n", __func__);
605 return NULL;
606}
607
9089f1b4 608/* caller must hold client_lock */
7116ed6b 609static void
5282fd72 610unhash_session(struct nfsd4_session *ses)
7116ed6b
AA
611{
612 list_del(&ses->se_hash);
613 list_del(&ses->se_perclnt);
5282fd72
ME
614}
615
7116ed6b
AA
616void
617free_session(struct kref *kref)
618{
619 struct nfsd4_session *ses;
dd829c45 620 int mem;
7116ed6b
AA
621
622 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 623 spin_lock(&nfsd_drc_lock);
efe0cb6d 624 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 625 nfsd_drc_mem_used -= mem;
be98d1bb 626 spin_unlock(&nfsd_drc_lock);
557ce264 627 free_session_slots(ses);
7116ed6b
AA
628 kfree(ses);
629}
630
36acb66b 631/* must be called under the client_lock */
1da177e4 632static inline void
36acb66b 633renew_client_locked(struct nfs4_client *clp)
1da177e4 634{
07cd4909
BH
635 if (is_client_expired(clp)) {
636 dprintk("%s: client (clientid %08x/%08x) already expired\n",
637 __func__,
638 clp->cl_clientid.cl_boot,
639 clp->cl_clientid.cl_id);
640 return;
641 }
642
1da177e4
LT
643 /*
644 * Move client to the end to the LRU list.
645 */
646 dprintk("renewing client (clientid %08x/%08x)\n",
647 clp->cl_clientid.cl_boot,
648 clp->cl_clientid.cl_id);
649 list_move_tail(&clp->cl_lru, &client_lru);
650 clp->cl_time = get_seconds();
651}
652
36acb66b
BH
653static inline void
654renew_client(struct nfs4_client *clp)
655{
656 spin_lock(&client_lock);
657 renew_client_locked(clp);
658 spin_unlock(&client_lock);
659}
660
1da177e4
LT
661/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
662static int
663STALE_CLIENTID(clientid_t *clid)
664{
665 if (clid->cl_boot == boot_time)
666 return 0;
60adfc50
AA
667 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
668 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT
669 return 1;
670}
671
672/*
673 * XXX Should we use a slab cache ?
674 * This type of memory management is somewhat inefficient, but we use it
675 * anyway since SETCLIENTID is not a common operation.
676 */
35bba9a3 677static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT
678{
679 struct nfs4_client *clp;
680
35bba9a3
BF
681 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
682 if (clp == NULL)
683 return NULL;
684 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
685 if (clp->cl_name.data == NULL) {
686 kfree(clp);
687 return NULL;
1da177e4 688 }
35bba9a3
BF
689 memcpy(clp->cl_name.data, name.data, name.len);
690 clp->cl_name.len = name.len;
1da177e4
LT
691 return clp;
692}
693
694static inline void
695free_client(struct nfs4_client *clp)
696{
697 if (clp->cl_cred.cr_group_info)
698 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 699 kfree(clp->cl_principal);
1da177e4
LT
700 kfree(clp->cl_name.data);
701 kfree(clp);
702}
703
d7682988
BH
704void
705release_session_client(struct nfsd4_session *session)
706{
707 struct nfs4_client *clp = session->se_client;
708
709 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
710 return;
711 if (is_client_expired(clp)) {
712 free_client(clp);
713 session->se_client = NULL;
714 } else
715 renew_client_locked(clp);
716 spin_unlock(&client_lock);
717 nfsd4_put_session(session);
718}
719
84d38ac9
BH
720/* must be called under the client_lock */
721static inline void
722unhash_client_locked(struct nfs4_client *clp)
723{
07cd4909 724 mark_client_expired(clp);
84d38ac9
BH
725 list_del(&clp->cl_lru);
726 while (!list_empty(&clp->cl_sessions)) {
727 struct nfsd4_session *ses;
728 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
729 se_perclnt);
730 unhash_session(ses);
731 nfsd4_put_session(ses);
732 }
733}
734
1da177e4
LT
735static void
736expire_client(struct nfs4_client *clp)
737{
738 struct nfs4_stateowner *sop;
739 struct nfs4_delegation *dp;
1da177e4
LT
740 struct list_head reaplist;
741
1da177e4
LT
742 INIT_LIST_HEAD(&reaplist);
743 spin_lock(&recall_lock);
ea1da636
N
744 while (!list_empty(&clp->cl_delegations)) {
745 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT
746 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
747 dp->dl_flock);
ea1da636 748 list_del_init(&dp->dl_perclnt);
1da177e4
LT
749 list_move(&dp->dl_recall_lru, &reaplist);
750 }
751 spin_unlock(&recall_lock);
752 while (!list_empty(&reaplist)) {
753 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
754 list_del_init(&dp->dl_recall_lru);
755 unhash_delegation(dp);
756 }
ea1da636
N
757 while (!list_empty(&clp->cl_openowners)) {
758 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 759 release_openowner(sop);
1da177e4 760 }
84d38ac9
BH
761 nfsd4_set_callback_client(clp, NULL);
762 if (clp->cl_cb_conn.cb_xprt)
763 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
36acb66b
BH
764 list_del(&clp->cl_idhash);
765 list_del(&clp->cl_strhash);
be1fdf6c 766 spin_lock(&client_lock);
84d38ac9 767 unhash_client_locked(clp);
46583e25
BH
768 if (atomic_read(&clp->cl_refcount) == 0)
769 free_client(clp);
be1fdf6c 770 spin_unlock(&client_lock);
1da177e4
LT
771}
772
35bba9a3
BF
773static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
774{
775 memcpy(target->cl_verifier.data, source->data,
776 sizeof(target->cl_verifier.data));
1da177e4
LT
777}
778
35bba9a3
BF
779static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
780{
1da177e4
LT
781 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
782 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
783}
784
35bba9a3
BF
785static void copy_cred(struct svc_cred *target, struct svc_cred *source)
786{
1da177e4
LT
787 target->cr_uid = source->cr_uid;
788 target->cr_gid = source->cr_gid;
789 target->cr_group_info = source->cr_group_info;
790 get_group_info(target->cr_group_info);
791}
792
35bba9a3 793static int same_name(const char *n1, const char *n2)
599e0a22 794{
a55370a3 795 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT
796}
797
798static int
599e0a22
BF
799same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
800{
801 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT
802}
803
804static int
599e0a22
BF
805same_clid(clientid_t *cl1, clientid_t *cl2)
806{
807 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT
808}
809
810/* XXX what about NGROUP */
811static int
599e0a22
BF
812same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
813{
814 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT
815}
816
5ec7b46c
BF
817static void gen_clid(struct nfs4_client *clp)
818{
819 static u32 current_clientid = 1;
820
1da177e4
LT
821 clp->cl_clientid.cl_boot = boot_time;
822 clp->cl_clientid.cl_id = current_clientid++;
823}
824
deda2faa
BF
825static void gen_confirm(struct nfs4_client *clp)
826{
827 static u32 i;
828 u32 *p;
1da177e4 829
1da177e4 830 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF
831 *p++ = get_seconds();
832 *p++ = i++;
1da177e4
LT
833}
834
b09333c4
RL
835static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
836 struct svc_rqst *rqstp, nfs4_verifier *verf)
837{
838 struct nfs4_client *clp;
839 struct sockaddr *sa = svc_addr(rqstp);
840 char *princ;
841
842 clp = alloc_client(name);
843 if (clp == NULL)
844 return NULL;
845
846 princ = svc_gss_principal(rqstp);
847 if (princ) {
848 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
849 if (clp->cl_principal == NULL) {
850 free_client(clp);
851 return NULL;
852 }
853 }
854
855 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
46583e25 856 atomic_set(&clp->cl_refcount, 0);
2bf23875 857 atomic_set(&clp->cl_cb_set, 0);
b09333c4
RL
858 INIT_LIST_HEAD(&clp->cl_idhash);
859 INIT_LIST_HEAD(&clp->cl_strhash);
860 INIT_LIST_HEAD(&clp->cl_openowners);
861 INIT_LIST_HEAD(&clp->cl_delegations);
862 INIT_LIST_HEAD(&clp->cl_sessions);
863 INIT_LIST_HEAD(&clp->cl_lru);
07cd4909 864 clp->cl_time = get_seconds();
b09333c4
RL
865 clear_bit(0, &clp->cl_cb_slot_busy);
866 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
867 copy_verf(clp, verf);
868 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
869 clp->cl_flavor = rqstp->rq_flavor;
870 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
871 gen_confirm(clp);
872
873 return clp;
874}
875
35bba9a3
BF
876static int check_name(struct xdr_netobj name)
877{
1da177e4
LT
878 if (name.len == 0)
879 return 0;
880 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 881 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT
882 return 0;
883 }
884 return 1;
885}
886
fd39ca9a 887static void
1da177e4
LT
888add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
889{
890 unsigned int idhashval;
891
892 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
893 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
894 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
36acb66b 895 renew_client(clp);
1da177e4
LT
896}
897
fd39ca9a 898static void
1da177e4
LT
899move_to_confirmed(struct nfs4_client *clp)
900{
901 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
902 unsigned int strhashval;
903
904 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
f116629d 905 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 906 strhashval = clientstr_hashval(clp->cl_recdir);
328efbab 907 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1da177e4
LT
908 renew_client(clp);
909}
910
911static struct nfs4_client *
912find_confirmed_client(clientid_t *clid)
913{
914 struct nfs4_client *clp;
915 unsigned int idhashval = clientid_hashval(clid->cl_id);
916
917 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 918 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
919 return clp;
920 }
921 return NULL;
922}
923
924static struct nfs4_client *
925find_unconfirmed_client(clientid_t *clid)
926{
927 struct nfs4_client *clp;
928 unsigned int idhashval = clientid_hashval(clid->cl_id);
929
930 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 931 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
932 return clp;
933 }
934 return NULL;
935}
936
a1bcecd2
AA
937/*
938 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
939 * parameter. Matching is based on the fact the at least one of the
940 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
941 *
942 * FIXME: we need to unify the clientid namespaces for nfsv4.x
943 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
944 * and SET_CLIENTID{,_CONFIRM}
945 */
946static inline int
947match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
948{
949 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
950 return use_exchange_id == has_exchange_flags;
951}
952
28ce6054 953static struct nfs4_client *
a1bcecd2
AA
954find_confirmed_client_by_str(const char *dname, unsigned int hashval,
955 bool use_exchange_id)
28ce6054
N
956{
957 struct nfs4_client *clp;
958
959 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
960 if (same_name(clp->cl_recdir, dname) &&
961 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
962 return clp;
963 }
964 return NULL;
965}
966
967static struct nfs4_client *
a1bcecd2
AA
968find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
969 bool use_exchange_id)
28ce6054
N
970{
971 struct nfs4_client *clp;
972
973 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA
974 if (same_name(clp->cl_recdir, dname) &&
975 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N
976 return clp;
977 }
978 return NULL;
979}
980
fd39ca9a 981static void
fbf4665f 982gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 983{
c237dc03 984 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL
985 unsigned short expected_family;
986
987 /* Currently, we only support tcp and tcp6 for the callback channel */
988 if (se->se_callback_netid_len == 3 &&
989 !memcmp(se->se_callback_netid_val, "tcp", 3))
990 expected_family = AF_INET;
991 else if (se->se_callback_netid_len == 4 &&
992 !memcmp(se->se_callback_netid_val, "tcp6", 4))
993 expected_family = AF_INET6;
994 else
1da177e4
LT
995 goto out_err;
996
aa9a4ec7
JL
997 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
998 se->se_callback_addr_len,
999 (struct sockaddr *) &cb->cb_addr,
1000 sizeof(cb->cb_addr));
1001
7077ecba 1002 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 1003 goto out_err;
aa9a4ec7 1004
fbf4665f
JL
1005 if (cb->cb_addr.ss_family == AF_INET6)
1006 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
1007
ab52ae6d 1008 cb->cb_minorversion = 0;
1da177e4
LT
1009 cb->cb_prog = se->se_callback_prog;
1010 cb->cb_ident = se->se_callback_ident;
1da177e4
LT
1011 return;
1012out_err:
aa9a4ec7
JL
1013 cb->cb_addr.ss_family = AF_UNSPEC;
1014 cb->cb_addrlen = 0;
849823c5 1015 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT
1016 "will not receive delegations\n",
1017 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1018
1da177e4
LT
1019 return;
1020}
1021
074fe897 1022/*
557ce264 1023 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 1024 */
074fe897
AA
1025void
1026nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 1027{
557ce264
AA
1028 struct nfsd4_slot *slot = resp->cstate.slot;
1029 unsigned int base;
074fe897 1030
557ce264 1031 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1032
557ce264
AA
1033 slot->sl_opcnt = resp->opcnt;
1034 slot->sl_status = resp->cstate.status;
074fe897 1035
bf864a31 1036 if (nfsd4_not_cached(resp)) {
557ce264 1037 slot->sl_datalen = 0;
bf864a31 1038 return;
074fe897 1039 }
557ce264
AA
1040 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1041 base = (char *)resp->cstate.datap -
1042 (char *)resp->xbuf->head[0].iov_base;
1043 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1044 slot->sl_datalen))
1045 WARN("%s: sessions DRC could not cache compound\n", __func__);
1046 return;
074fe897
AA
1047}
1048
1049/*
abfabf8c
AA
1050 * Encode the replay sequence operation from the slot values.
1051 * If cachethis is FALSE encode the uncached rep error on the next
1052 * operation which sets resp->p and increments resp->opcnt for
1053 * nfs4svc_encode_compoundres.
074fe897 1054 *
074fe897 1055 */
abfabf8c
AA
1056static __be32
1057nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1058 struct nfsd4_compoundres *resp)
074fe897 1059{
abfabf8c
AA
1060 struct nfsd4_op *op;
1061 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1062
abfabf8c 1063 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1064 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1065
abfabf8c
AA
1066 /* Encode the replayed sequence operation */
1067 op = &args->ops[resp->opcnt - 1];
1068 nfsd4_encode_operation(resp, op);
bf864a31 1069
abfabf8c 1070 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1071 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA
1072 op = &args->ops[resp->opcnt++];
1073 op->status = nfserr_retry_uncached_rep;
1074 nfsd4_encode_operation(resp, op);
074fe897 1075 }
abfabf8c 1076 return op->status;
074fe897
AA
1077}
1078
1079/*
557ce264
AA
1080 * The sequence operation is not cached because we can use the slot and
1081 * session values.
074fe897
AA
1082 */
1083__be32
bf864a31
AA
1084nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1085 struct nfsd4_sequence *seq)
074fe897 1086{
557ce264 1087 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA
1088 __be32 status;
1089
557ce264 1090 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1091
abfabf8c
AA
1092 /* Either returns 0 or nfserr_retry_uncached */
1093 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1094 if (status == nfserr_retry_uncached_rep)
1095 return status;
074fe897 1096
557ce264
AA
1097 /* The sequence operation has been encoded, cstate->datap set. */
1098 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1099
557ce264
AA
1100 resp->opcnt = slot->sl_opcnt;
1101 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1102 status = slot->sl_status;
074fe897
AA
1103
1104 return status;
1105}
1106
0733d213
AA
1107/*
1108 * Set the exchange_id flags returned by the server.
1109 */
1110static void
1111nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1112{
1113 /* pNFS is not supported */
1114 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1115
1116 /* Referrals are supported, Migration is not. */
1117 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1118
1119 /* set the wire flags to return to client. */
1120 clid->flags = new->cl_exchange_flags;
1121}
1122
069b6ad4
AA
1123__be32
1124nfsd4_exchange_id(struct svc_rqst *rqstp,
1125 struct nfsd4_compound_state *cstate,
1126 struct nfsd4_exchange_id *exid)
1127{
0733d213
AA
1128 struct nfs4_client *unconf, *conf, *new;
1129 int status;
1130 unsigned int strhashval;
1131 char dname[HEXDIR_LEN];
363168b4 1132 char addr_str[INET6_ADDRSTRLEN];
0733d213 1133 nfs4_verifier verf = exid->verifier;
363168b4 1134 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1135
363168b4 1136 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1137 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1138 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1139 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1140 addr_str, exid->flags, exid->spa_how);
0733d213
AA
1141
1142 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1143 return nfserr_inval;
1144
1145 /* Currently only support SP4_NONE */
1146 switch (exid->spa_how) {
1147 case SP4_NONE:
1148 break;
1149 case SP4_SSV:
1150 return nfserr_encr_alg_unsupp;
1151 default:
1152 BUG(); /* checked by xdr code */
1153 case SP4_MACH_CRED:
1154 return nfserr_serverfault; /* no excuse :-/ */
1155 }
1156
1157 status = nfs4_make_rec_clidname(dname, &exid->clname);
1158
1159 if (status)
1160 goto error;
1161
1162 strhashval = clientstr_hashval(dname);
1163
1164 nfs4_lock_state();
1165 status = nfs_ok;
1166
a1bcecd2 1167 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1168 if (conf) {
1169 if (!same_verf(&verf, &conf->cl_verifier)) {
1170 /* 18.35.4 case 8 */
1171 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1172 status = nfserr_not_same;
1173 goto out;
1174 }
1175 /* Client reboot: destroy old state */
1176 expire_client(conf);
1177 goto out_new;
1178 }
1179 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1180 /* 18.35.4 case 9 */
1181 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1182 status = nfserr_perm;
1183 goto out;
1184 }
1185 expire_client(conf);
1186 goto out_new;
1187 }
0733d213
AA
1188 /*
1189 * Set bit when the owner id and verifier map to an already
1190 * confirmed client id (18.35.3).
1191 */
1192 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1193
1194 /*
1195 * Falling into 18.35.4 case 2, possible router replay.
1196 * Leave confirmed record intact and return same result.
1197 */
1198 copy_verf(conf, &verf);
1199 new = conf;
1200 goto out_copy;
6ddbbbfe
MS
1201 }
1202
1203 /* 18.35.4 case 7 */
1204 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1205 status = nfserr_noent;
1206 goto out;
0733d213
AA
1207 }
1208
a1bcecd2 1209 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA
1210 if (unconf) {
1211 /*
1212 * Possible retry or client restart. Per 18.35.4 case 4,
1213 * a new unconfirmed record should be generated regardless
1214 * of whether any properties have changed.
1215 */
1216 expire_client(unconf);
1217 }
1218
1219out_new:
1220 /* Normal case */
b09333c4 1221 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1222 if (new == NULL) {
bdac86e2 1223 status = nfserr_serverfault;
0733d213
AA
1224 goto out;
1225 }
1226
0733d213 1227 gen_clid(new);
0733d213
AA
1228 add_to_unconfirmed(new, strhashval);
1229out_copy:
1230 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1231 exid->clientid.cl_id = new->cl_clientid.cl_id;
1232
38eb76a5 1233 exid->seqid = 1;
0733d213
AA
1234 nfsd4_set_ex_flags(new, exid);
1235
1236 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1237 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA
1238 status = nfs_ok;
1239
1240out:
1241 nfs4_unlock_state();
1242error:
1243 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1244 return status;
069b6ad4
AA
1245}
1246
b85d4c01 1247static int
88e588d5 1248check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1249{
88e588d5
AA
1250 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1251 slot_seqid);
b85d4c01
BH
1252
1253 /* The slot is in use, and no response has been sent. */
88e588d5
AA
1254 if (slot_inuse) {
1255 if (seqid == slot_seqid)
b85d4c01
BH
1256 return nfserr_jukebox;
1257 else
1258 return nfserr_seq_misordered;
1259 }
1260 /* Normal */
88e588d5 1261 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH
1262 return nfs_ok;
1263 /* Replay */
88e588d5 1264 if (seqid == slot_seqid)
b85d4c01
BH
1265 return nfserr_replay_cache;
1266 /* Wraparound */
88e588d5 1267 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH
1268 return nfs_ok;
1269 /* Misordered replay or misordered new request */
1270 return nfserr_seq_misordered;
1271}
1272
49557cc7
AA
1273/*
1274 * Cache the create session result into the create session single DRC
1275 * slot cache by saving the xdr structure. sl_seqid has been set.
1276 * Do this for solo or embedded create session operations.
1277 */
1278static void
1279nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1280 struct nfsd4_clid_slot *slot, int nfserr)
1281{
1282 slot->sl_status = nfserr;
1283 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1284}
1285
1286static __be32
1287nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1288 struct nfsd4_clid_slot *slot)
1289{
1290 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1291 return slot->sl_status;
1292}
1293
069b6ad4
AA
1294__be32
1295nfsd4_create_session(struct svc_rqst *rqstp,
1296 struct nfsd4_compound_state *cstate,
1297 struct nfsd4_create_session *cr_ses)
1298{
363168b4 1299 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1300 struct nfs4_client *conf, *unconf;
49557cc7 1301 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA
1302 int status = 0;
1303
1304 nfs4_lock_state();
1305 unconf = find_unconfirmed_client(&cr_ses->clientid);
1306 conf = find_confirmed_client(&cr_ses->clientid);
1307
1308 if (conf) {
49557cc7
AA
1309 cs_slot = &conf->cl_cs_slot;
1310 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1311 if (status == nfserr_replay_cache) {
ec6b5d7b 1312 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1313 cs_slot->sl_seqid);
38eb76a5 1314 /* Return the cached reply status */
49557cc7 1315 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1316 goto out;
49557cc7 1317 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA
1318 status = nfserr_seq_misordered;
1319 dprintk("Sequence misordered!\n");
1320 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1321 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA
1322 goto out;
1323 }
49557cc7 1324 cs_slot->sl_seqid++;
ec6b5d7b
AA
1325 } else if (unconf) {
1326 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1327 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA
1328 status = nfserr_clid_inuse;
1329 goto out;
1330 }
1331
49557cc7
AA
1332 cs_slot = &unconf->cl_cs_slot;
1333 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA
1334 if (status) {
1335 /* an unconfirmed replay returns misordered */
ec6b5d7b 1336 status = nfserr_seq_misordered;
49557cc7 1337 goto out_cache;
ec6b5d7b
AA
1338 }
1339
49557cc7 1340 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA
1341 move_to_confirmed(unconf);
1342
38524ab3 1343 if (cr_ses->flags & SESSION4_BACK_CHAN) {
2bf23875
BF
1344 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1345 svc_xprt_get(rqstp->rq_xprt);
38524ab3
AA
1346 rpc_copy_addr(
1347 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1348 sa);
1349 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1350 unconf->cl_cb_conn.cb_minorversion =
1351 cstate->minorversion;
1352 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1353 unconf->cl_cb_seq_nr = 1;
4b21d0de 1354 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
38524ab3 1355 }
ec6b5d7b
AA
1356 conf = unconf;
1357 } else {
1358 status = nfserr_stale_clientid;
1359 goto out;
1360 }
1361
408b79bc
BF
1362 /*
1363 * We do not support RDMA or persistent sessions
1364 */
1365 cr_ses->flags &= ~SESSION4_PERSIST;
1366 cr_ses->flags &= ~SESSION4_RDMA;
1367
ec6b5d7b
AA
1368 status = alloc_init_session(rqstp, conf, cr_ses);
1369 if (status)
1370 goto out;
1371
ec6b5d7b
AA
1372 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1373 NFS4_MAX_SESSIONID_LEN);
49557cc7 1374 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1375
49557cc7
AA
1376out_cache:
1377 /* cache solo and embedded create sessions under the state lock */
1378 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA
1379out:
1380 nfs4_unlock_state();
1381 dprintk("%s returns %d\n", __func__, ntohl(status));
1382 return status;
069b6ad4
AA
1383}
1384
57716355
BF
1385static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1386{
1387 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1388 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1389
1390 return argp->opcnt == resp->opcnt;
1391}
1392
5d4cec2f
BF
1393static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1394{
1395 if (!session)
1396 return 0;
1397 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1398}
1399
069b6ad4
AA
1400__be32
1401nfsd4_destroy_session(struct svc_rqst *r,
1402 struct nfsd4_compound_state *cstate,
1403 struct nfsd4_destroy_session *sessionid)
1404{
e10e0cfc
BH
1405 struct nfsd4_session *ses;
1406 u32 status = nfserr_badsession;
1407
1408 /* Notes:
1409 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1410 * - Should we return nfserr_back_chan_busy if waiting for
1411 * callbacks on to-be-destroyed session?
1412 * - Do we need to clear any callback info from previous session?
1413 */
1414
5d4cec2f 1415 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
57716355
BF
1416 if (!nfsd4_last_compound_op(r))
1417 return nfserr_not_only_op;
1418 }
e10e0cfc 1419 dump_sessionid(__func__, &sessionid->sessionid);
9089f1b4 1420 spin_lock(&client_lock);
e10e0cfc
BH
1421 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1422 if (!ses) {
9089f1b4 1423 spin_unlock(&client_lock);
e10e0cfc
BH
1424 goto out;
1425 }
1426
1427 unhash_session(ses);
9089f1b4 1428 spin_unlock(&client_lock);
e10e0cfc 1429
ab707e15 1430 nfs4_lock_state();
e10e0cfc 1431 /* wait for callbacks */
b5a1a81e 1432 nfsd4_set_callback_client(ses->se_client, NULL);
ab707e15 1433 nfs4_unlock_state();
e10e0cfc
BH
1434 nfsd4_put_session(ses);
1435 status = nfs_ok;
1436out:
1437 dprintk("%s returns %d\n", __func__, ntohl(status));
1438 return status;
069b6ad4
AA
1439}
1440
1441__be32
b85d4c01 1442nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA
1443 struct nfsd4_compound_state *cstate,
1444 struct nfsd4_sequence *seq)
1445{
f9bb94c4 1446 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH
1447 struct nfsd4_session *session;
1448 struct nfsd4_slot *slot;
1449 int status;
1450
f9bb94c4
AA
1451 if (resp->opcnt != 1)
1452 return nfserr_sequence_pos;
1453
9089f1b4 1454 spin_lock(&client_lock);
b85d4c01
BH
1455 status = nfserr_badsession;
1456 session = find_in_sessionid_hashtbl(&seq->sessionid);
1457 if (!session)
1458 goto out;
1459
1460 status = nfserr_badslot;
6c18ba9f 1461 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH
1462 goto out;
1463
557ce264 1464 slot = session->se_slots[seq->slotid];
b85d4c01
BH
1465 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1466
a8dfdaeb
AA
1467 /* We do not negotiate the number of slots yet, so set the
1468 * maxslots to the session maxreqs which is used to encode
1469 * sr_highest_slotid and the sr_target_slot id to maxslots */
1470 seq->maxslots = session->se_fchannel.maxreqs;
1471
88e588d5 1472 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH
1473 if (status == nfserr_replay_cache) {
1474 cstate->slot = slot;
1475 cstate->session = session;
da3846a2 1476 /* Return the cached reply status and set cstate->status
557ce264 1477 * for nfsd4_proc_compound processing */
bf864a31 1478 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1479 cstate->status = nfserr_replay_cache;
aaf84eb9 1480 goto out;
b85d4c01
BH
1481 }
1482 if (status)
1483 goto out;
1484
1485 /* Success! bump slot seqid */
1486 slot->sl_inuse = true;
1487 slot->sl_seqid = seq->seqid;
557ce264 1488 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH
1489
1490 cstate->slot = slot;
1491 cstate->session = session;
1492
b85d4c01 1493out:
26c0c75e 1494 /* Hold a session reference until done processing the compound. */
aaf84eb9 1495 if (cstate->session) {
36acb66b 1496 nfsd4_get_session(cstate->session);
d7682988 1497 atomic_inc(&session->se_client->cl_refcount);
aaf84eb9 1498 }
36acb66b 1499 spin_unlock(&client_lock);
b85d4c01
BH
1500 dprintk("%s: return %d\n", __func__, ntohl(status));
1501 return status;
069b6ad4
AA
1502}
1503
b37ad28b 1504__be32
b591480b
BF
1505nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1506 struct nfsd4_setclientid *setclid)
1da177e4 1507{
363168b4 1508 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT
1509 struct xdr_netobj clname = {
1510 .len = setclid->se_namelen,
1511 .data = setclid->se_name,
1512 };
1513 nfs4_verifier clverifier = setclid->se_verf;
1514 unsigned int strhashval;
28ce6054 1515 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1516 __be32 status;
a55370a3 1517 char dname[HEXDIR_LEN];
1da177e4 1518
1da177e4 1519 if (!check_name(clname))
73aea4ec 1520 return nfserr_inval;
1da177e4 1521
a55370a3
N
1522 status = nfs4_make_rec_clidname(dname, &clname);
1523 if (status)
73aea4ec 1524 return status;
a55370a3 1525
1da177e4
LT
1526 /*
1527 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1528 * We get here on a DRC miss.
1529 */
1530
a55370a3 1531 strhashval = clientstr_hashval(dname);
1da177e4 1532
1da177e4 1533 nfs4_lock_state();
a1bcecd2 1534 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1535 if (conf) {
a186e767 1536 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1537 status = nfserr_clid_inuse;
026722c2 1538 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL
1539 char addr_str[INET6_ADDRSTRLEN];
1540 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1541 sizeof(addr_str));
1542 dprintk("NFSD: setclientid: string in use by client "
1543 "at %s\n", addr_str);
1da177e4
LT
1544 goto out;
1545 }
1da177e4 1546 }
a186e767
BF
1547 /*
1548 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1549 * has a description of SETCLIENTID request processing consisting
1550 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1551 */
a1bcecd2 1552 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT
1553 status = nfserr_resource;
1554 if (!conf) {
a186e767
BF
1555 /*
1556 * RFC 3530 14.2.33 CASE 4:
1557 * placed first, because it is the normal case
1da177e4
LT
1558 */
1559 if (unconf)
1560 expire_client(unconf);
b09333c4 1561 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1562 if (new == NULL)
1da177e4 1563 goto out;
1da177e4 1564 gen_clid(new);
599e0a22 1565 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1566 /*
a186e767
BF
1567 * RFC 3530 14.2.33 CASE 1:
1568 * probable callback update
1da177e4 1569 */
31f4a6c1
N
1570 if (unconf) {
1571 /* Note this is removing unconfirmed {*x***},
1572 * which is stronger than RFC recommended {vxc**}.
1573 * This has the advantage that there is at most
1574 * one {*x***} in either list at any time.
1575 */
1576 expire_client(unconf);
1da177e4 1577 }
b09333c4 1578 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1579 if (new == NULL)
1da177e4 1580 goto out;
1da177e4 1581 copy_clid(new, conf);
1da177e4
LT
1582 } else if (!unconf) {
1583 /*
a186e767
BF
1584 * RFC 3530 14.2.33 CASE 2:
1585 * probable client reboot; state will be removed if
1586 * confirmed.
1da177e4 1587 */
b09333c4 1588 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1589 if (new == NULL)
1da177e4 1590 goto out;
1da177e4 1591 gen_clid(new);
49ba8781 1592 } else {
a186e767
BF
1593 /*
1594 * RFC 3530 14.2.33 CASE 3:
1595 * probable client reboot; state will be removed if
1596 * confirmed.
1da177e4
LT
1597 */
1598 expire_client(unconf);
b09333c4 1599 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1600 if (new == NULL)
1da177e4 1601 goto out;
1da177e4 1602 gen_clid(new);
1da177e4 1603 }
fbf4665f 1604 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1605 add_to_unconfirmed(new, strhashval);
1da177e4
LT
1606 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1607 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1608 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1609 status = nfs_ok;
1610out:
1611 nfs4_unlock_state();
1612 return status;
1613}
1614
1615
1616/*
a186e767
BF
1617 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1618 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1619 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1620 */
b37ad28b 1621__be32
b591480b
BF
1622nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1623 struct nfsd4_compound_state *cstate,
1624 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1625{
363168b4 1626 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1627 struct nfs4_client *conf, *unconf;
1da177e4
LT
1628 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1629 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1630 __be32 status;
1da177e4
LT
1631
1632 if (STALE_CLIENTID(clid))
1633 return nfserr_stale_clientid;
1634 /*
1635 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1636 * We get here on a DRC miss.
1637 */
1638
1639 nfs4_lock_state();
21ab45a4
N
1640
1641 conf = find_confirmed_client(clid);
1642 unconf = find_unconfirmed_client(clid);
1643
1644 status = nfserr_clid_inuse;
363168b4 1645 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1646 goto out;
363168b4 1647 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N
1648 goto out;
1649
a186e767
BF
1650 /*
1651 * section 14.2.34 of RFC 3530 has a description of
1652 * SETCLIENTID_CONFIRM request processing consisting
1653 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1654 */
366e0c1d 1655 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF
1656 /*
1657 * RFC 3530 14.2.34 CASE 1:
1658 * callback update
1659 */
599e0a22 1660 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT
1661 status = nfserr_clid_inuse;
1662 else {
2bf23875 1663 atomic_set(&conf->cl_cb_set, 0);
4b21d0de 1664 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1a69c179 1665 expire_client(unconf);
1da177e4 1666 status = nfs_ok;
1a69c179 1667
1da177e4 1668 }
f3aba4e5 1669 } else if (conf && !unconf) {
a186e767
BF
1670 /*
1671 * RFC 3530 14.2.34 CASE 2:
1672 * probable retransmitted request; play it safe and
1673 * do nothing.
7c79f737 1674 */
599e0a22 1675 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1676 status = nfserr_clid_inuse;
21ab45a4 1677 else
1da177e4 1678 status = nfs_ok;
7c79f737 1679 } else if (!conf && unconf
599e0a22 1680 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF
1681 /*
1682 * RFC 3530 14.2.34 CASE 3:
1683 * Normal case; new or rebooted client:
7c79f737 1684 */
599e0a22 1685 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT
1686 status = nfserr_clid_inuse;
1687 } else {
1a69c179
N
1688 unsigned int hash =
1689 clientstr_hashval(unconf->cl_recdir);
1690 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1691 hash, false);
1a69c179 1692 if (conf) {
c7b9a459 1693 nfsd4_remove_clid_dir(conf);
1a69c179
N
1694 expire_client(conf);
1695 }
1da177e4 1696 move_to_confirmed(unconf);
21ab45a4 1697 conf = unconf;
4b21d0de 1698 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1a69c179 1699 status = nfs_ok;
1da177e4 1700 }
599e0a22
BF
1701 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1702 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1703 &confirm)))) {
a186e767
BF
1704 /*
1705 * RFC 3530 14.2.34 CASE 4:
1706 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1707 */
1da177e4 1708 status = nfserr_stale_clientid;
7c79f737 1709 } else {
08e8987c
N
1710 /* check that we have hit one of the cases...*/
1711 status = nfserr_clid_inuse;
1712 }
1da177e4 1713out:
1da177e4
LT
1714 nfs4_unlock_state();
1715 return status;
1716}
1717
1da177e4
LT
1718/* OPEN Share state helper functions */
1719static inline struct nfs4_file *
1720alloc_init_file(struct inode *ino)
1721{
1722 struct nfs4_file *fp;
1723 unsigned int hashval = file_hashval(ino);
1724
e60d4398
N
1725 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1726 if (fp) {
8b671b80 1727 atomic_set(&fp->fi_ref, 1);
1da177e4 1728 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N
1729 INIT_LIST_HEAD(&fp->fi_stateids);
1730 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1731 spin_lock(&recall_lock);
1da177e4 1732 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1733 spin_unlock(&recall_lock);
1da177e4
LT
1734 fp->fi_inode = igrab(ino);
1735 fp->fi_id = current_fileid++;
47f9940c 1736 fp->fi_had_conflict = false;
1da177e4
LT
1737 return fp;
1738 }
1739 return NULL;
1740}
1741
e60d4398 1742static void
e18b890b 1743nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1744{
e60d4398
N
1745 if (*slab == NULL)
1746 return;
1a1d92c1 1747 kmem_cache_destroy(*slab);
e60d4398 1748 *slab = NULL;
1da177e4
LT
1749}
1750
e8ff2a84 1751void
1da177e4
LT
1752nfsd4_free_slabs(void)
1753{
e60d4398
N
1754 nfsd4_free_slab(&stateowner_slab);
1755 nfsd4_free_slab(&file_slab);
5ac049ac 1756 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1757 nfsd4_free_slab(&deleg_slab);
e60d4398 1758}
1da177e4 1759
e60d4398
N
1760static int
1761nfsd4_init_slabs(void)
1762{
1763 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1764 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N
1765 if (stateowner_slab == NULL)
1766 goto out_nomem;
1767 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1768 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N
1769 if (file_slab == NULL)
1770 goto out_nomem;
5ac049ac 1771 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1772 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N
1773 if (stateid_slab == NULL)
1774 goto out_nomem;
5b2d21c1 1775 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1776 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N
1777 if (deleg_slab == NULL)
1778 goto out_nomem;
e60d4398
N
1779 return 0;
1780out_nomem:
1781 nfsd4_free_slabs();
1782 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1783 return -ENOMEM;
1da177e4
LT
1784}
1785
1786void
1787nfs4_free_stateowner(struct kref *kref)
1788{
1789 struct nfs4_stateowner *sop =
1790 container_of(kref, struct nfs4_stateowner, so_ref);
1791 kfree(sop->so_owner.data);
1792 kmem_cache_free(stateowner_slab, sop);
1793}
1794
1795static inline struct nfs4_stateowner *
1796alloc_stateowner(struct xdr_netobj *owner)
1797{
1798 struct nfs4_stateowner *sop;
1799
1800 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1801 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1802 memcpy(sop->so_owner.data, owner->data, owner->len);
1803 sop->so_owner.len = owner->len;
1804 kref_init(&sop->so_ref);
1805 return sop;
1806 }
1807 kmem_cache_free(stateowner_slab, sop);
1808 }
1809 return NULL;
1810}
1811
1812static struct nfs4_stateowner *
1813alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1814 struct nfs4_stateowner *sop;
1815 struct nfs4_replay *rp;
1816 unsigned int idhashval;
1817
1818 if (!(sop = alloc_stateowner(&open->op_owner)))
1819 return NULL;
1820 idhashval = ownerid_hashval(current_ownerid);
1821 INIT_LIST_HEAD(&sop->so_idhash);
1822 INIT_LIST_HEAD(&sop->so_strhash);
1823 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
1824 INIT_LIST_HEAD(&sop->so_stateids);
1825 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT
1826 INIT_LIST_HEAD(&sop->so_close_lru);
1827 sop->so_time = 0;
1828 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1829 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1830 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT
1831 sop->so_is_open_owner = 1;
1832 sop->so_id = current_ownerid++;
1833 sop->so_client = clp;
1834 sop->so_seqid = open->op_seqid;
1835 sop->so_confirmed = 0;
1836 rp = &sop->so_replay;
de1ae286 1837 rp->rp_status = nfserr_serverfault;
1da177e4
LT
1838 rp->rp_buflen = 0;
1839 rp->rp_buf = rp->rp_ibuf;
1840 return sop;
1841}
1842
1da177e4
LT
1843static inline void
1844init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1845 struct nfs4_stateowner *sop = open->op_stateowner;
1846 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1847
1848 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N
1849 INIT_LIST_HEAD(&stp->st_perstateowner);
1850 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT
1851 INIT_LIST_HEAD(&stp->st_perfile);
1852 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1853 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1854 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1855 stp->st_stateowner = sop;
13cd2184 1856 get_nfs4_file(fp);
1da177e4 1857 stp->st_file = fp;
78155ed7 1858 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
1859 stp->st_stateid.si_stateownerid = sop->so_id;
1860 stp->st_stateid.si_fileid = fp->fi_id;
1861 stp->st_stateid.si_generation = 0;
1862 stp->st_access_bmap = 0;
1863 stp->st_deny_bmap = 0;
84459a11
AA
1864 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1865 &stp->st_access_bmap);
1da177e4 1866 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1867 stp->st_openstp = NULL;
1da177e4
LT
1868}
1869
fd39ca9a 1870static void
1da177e4
LT
1871move_to_close_lru(struct nfs4_stateowner *sop)
1872{
1873 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1874
358dd55a 1875 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT
1876 sop->so_time = get_seconds();
1877}
1878
1da177e4 1879static int
599e0a22
BF
1880same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1881 clientid_t *clid)
1882{
1883 return (sop->so_owner.len == owner->len) &&
1884 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1885 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT
1886}
1887
1888static struct nfs4_stateowner *
1889find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1890{
1891 struct nfs4_stateowner *so = NULL;
1892
1893 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1894 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT
1895 return so;
1896 }
1897 return NULL;
1898}
1899
1900/* search file_hashtbl[] for file */
1901static struct nfs4_file *
1902find_file(struct inode *ino)
1903{
1904 unsigned int hashval = file_hashval(ino);
1905 struct nfs4_file *fp;
1906
8b671b80 1907 spin_lock(&recall_lock);
1da177e4 1908 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N
1909 if (fp->fi_inode == ino) {
1910 get_nfs4_file(fp);
8b671b80 1911 spin_unlock(&recall_lock);
1da177e4 1912 return fp;
13cd2184 1913 }
1da177e4 1914 }
8b671b80 1915 spin_unlock(&recall_lock);
1da177e4
LT
1916 return NULL;
1917}
1918
d87a8ade 1919static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1920{
d87a8ade 1921 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1922 return 0;
d87a8ade
AA
1923 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1924 return 0;
1925 x &= ~NFS4_SHARE_ACCESS_MASK;
1926 if (minorversion && x) {
1927 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1928 return 0;
1929 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1930 return 0;
1931 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1932 }
1933 if (x)
8838dc43
BF
1934 return 0;
1935 return 1;
ba5a6a19
BF
1936}
1937
8838dc43 1938static inline int deny_valid(u32 x)
ba5a6a19 1939{
8838dc43
BF
1940 /* Note: unlike access bits, deny bits may be zero. */
1941 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1942}
1da177e4 1943
4f83aa30
BF
1944/*
1945 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1946 * st_{access,deny}_bmap field of the stateid, in order to track not
1947 * only what share bits are currently in force, but also what
1948 * combinations of share bits previous opens have used. This allows us
1949 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1950 * return an error if the client attempt to downgrade to a combination
1951 * of share bits not explicable by closing some of its previous opens.
1952 *
1953 * XXX: This enforcement is actually incomplete, since we don't keep
1954 * track of access/deny bit combinations; so, e.g., we allow:
1955 *
1956 * OPEN allow read, deny write
1957 * OPEN allow both, deny none
1958 * DOWNGRADE allow read, deny none
1959 *
1960 * which we should reject.
1961 */
fd39ca9a 1962static void
1da177e4
LT
1963set_access(unsigned int *access, unsigned long bmap) {
1964 int i;
1965
1966 *access = 0;
1967 for (i = 1; i < 4; i++) {
1968 if (test_bit(i, &bmap))
1969 *access |= i;
1970 }
1971}
1972
fd39ca9a 1973static void
1da177e4
LT
1974set_deny(unsigned int *deny, unsigned long bmap) {
1975 int i;
1976
1977 *deny = 0;
1978 for (i = 0; i < 4; i++) {
1979 if (test_bit(i, &bmap))
1980 *deny |= i ;
1981 }
1982}
1983
1984static int
1985test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1986 unsigned int access, deny;
1987
1988 set_access(&access, stp->st_access_bmap);
1989 set_deny(&deny, stp->st_deny_bmap);
1990 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1991 return 0;
1992 return 1;
1993}
1994
1995/*
1996 * Called to check deny when READ with all zero stateid or
1997 * WRITE with all zero or all one stateid
1998 */
b37ad28b 1999static __be32
1da177e4
LT
2000nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2001{
2002 struct inode *ino = current_fh->fh_dentry->d_inode;
2003 struct nfs4_file *fp;
2004 struct nfs4_stateid *stp;
b37ad28b 2005 __be32 ret;
1da177e4
LT
2006
2007 dprintk("NFSD: nfs4_share_conflict\n");
2008
2009 fp = find_file(ino);
13cd2184
N
2010 if (!fp)
2011 return nfs_ok;
b700949b 2012 ret = nfserr_locked;
1da177e4 2013 /* Search for conflicting share reservations */
13cd2184
N
2014 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2015 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2016 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2017 goto out;
1da177e4 2018 }
13cd2184
N
2019 ret = nfs_ok;
2020out:
2021 put_nfs4_file(fp);
2022 return ret;
1da177e4
LT
2023}
2024
2025static inline void
2026nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2027{
2028 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2029 drop_file_write_access(filp);
42e49608 2030 spin_lock(&filp->f_lock);
1da177e4 2031 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
42e49608 2032 spin_unlock(&filp->f_lock);
1da177e4
LT
2033 }
2034}
2035
1da177e4
LT
2036/*
2037 * Spawn a thread to perform a recall on the delegation represented
2038 * by the lease (file_lock)
2039 *
2040 * Called from break_lease() with lock_kernel() held.
2041 * Note: we assume break_lease will only call this *once* for any given
2042 * lease.
2043 */
2044static
2045void nfsd_break_deleg_cb(struct file_lock *fl)
2046{
63e4863f 2047 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT
2048
2049 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2050 if (!dp)
2051 return;
2052
2053 /* We're assuming the state code never drops its reference
2054 * without first removing the lease. Since we're in this lease
2055 * callback (and since the lease code is serialized by the kernel
2056 * lock) we know the server hasn't removed the lease yet, we know
2057 * it's safe to take a reference: */
2058 atomic_inc(&dp->dl_count);
2059
2060 spin_lock(&recall_lock);
2061 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2062 spin_unlock(&recall_lock);
2063
2064 /* only place dl_time is set. protected by lock_kernel*/
2065 dp->dl_time = get_seconds();
2066
0272e1fd
BF
2067 /*
2068 * We don't want the locks code to timeout the lease for us;
2069 * we'll remove it ourself if the delegation isn't returned
2070 * in time.
2071 */
2072 fl->fl_break_time = 0;
1da177e4 2073
63e4863f
BF
2074 dp->dl_file->fi_had_conflict = true;
2075 nfsd4_cb_recall(dp);
1da177e4
LT
2076}
2077
2078/*
2079 * The file_lock is being reapd.
2080 *
2081 * Called by locks_free_lock() with lock_kernel() held.
2082 */
2083static
2084void nfsd_release_deleg_cb(struct file_lock *fl)
2085{
2086 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2087
2088 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2089
2090 if (!(fl->fl_flags & FL_LEASE) || !dp)
2091 return;
2092 dp->dl_flock = NULL;
2093}
2094
2095/*
2096 * Set the delegation file_lock back pointer.
2097 *
a9933cea 2098 * Called from setlease() with lock_kernel() held.
1da177e4
LT
2099 */
2100static
2101void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2102{
2103 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2104
2105 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2106 if (!dp)
2107 return;
2108 dp->dl_flock = new;
2109}
2110
2111/*
a9933cea 2112 * Called from setlease() with lock_kernel() held
1da177e4
LT
2113 */
2114static
2115int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2116{
2117 struct nfs4_delegation *onlistd =
2118 (struct nfs4_delegation *)onlist->fl_owner;
2119 struct nfs4_delegation *tryd =
2120 (struct nfs4_delegation *)try->fl_owner;
2121
2122 if (onlist->fl_lmops != try->fl_lmops)
2123 return 0;
2124
2125 return onlistd->dl_client == tryd->dl_client;
2126}
2127
2128
2129static
2130int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2131{
2132 if (arg & F_UNLCK)
2133 return lease_modify(onlist, arg);
2134 else
2135 return -EAGAIN;
2136}
2137
7b021967 2138static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT
2139 .fl_break = nfsd_break_deleg_cb,
2140 .fl_release_private = nfsd_release_deleg_cb,
2141 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2142 .fl_mylease = nfsd_same_client_deleg_cb,
2143 .fl_change = nfsd_change_deleg_cb,
2144};
2145
2146
b37ad28b 2147__be32
6668958f
AA
2148nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2149 struct nfsd4_open *open)
1da177e4 2150{
1da177e4
LT
2151 clientid_t *clientid = &open->op_clientid;
2152 struct nfs4_client *clp = NULL;
2153 unsigned int strhashval;
2154 struct nfs4_stateowner *sop = NULL;
2155
1da177e4 2156 if (!check_name(open->op_owner))
0f442aa2 2157 return nfserr_inval;
1da177e4
LT
2158
2159 if (STALE_CLIENTID(&open->op_clientid))
2160 return nfserr_stale_clientid;
2161
2162 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2163 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF
2164 open->op_stateowner = sop;
2165 if (!sop) {
2166 /* Make sure the client's lease hasn't expired. */
1da177e4
LT
2167 clp = find_confirmed_client(clientid);
2168 if (clp == NULL)
0f442aa2
BF
2169 return nfserr_expired;
2170 goto renew;
1da177e4 2171 }
6668958f
AA
2172 /* When sessions are used, skip open sequenceid processing */
2173 if (nfsd4_has_session(cstate))
2174 goto renew;
0f442aa2
BF
2175 if (!sop->so_confirmed) {
2176 /* Replace unconfirmed owners without checking for replay. */
2177 clp = sop->so_client;
f044ff83 2178 release_openowner(sop);
0f442aa2
BF
2179 open->op_stateowner = NULL;
2180 goto renew;
2181 }
2182 if (open->op_seqid == sop->so_seqid - 1) {
2183 if (sop->so_replay.rp_buflen)
a90b061c 2184 return nfserr_replay_me;
0f442aa2
BF
2185 /* The original OPEN failed so spectacularly
2186 * that we don't even have replay data saved!
2187 * Therefore, we have no choice but to continue
2188 * processing this OPEN; presumably, we'll
2189 * fail again for the same reason.
2190 */
2191 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2192 goto renew;
2193 }
2194 if (open->op_seqid != sop->so_seqid)
2195 return nfserr_bad_seqid;
1da177e4 2196renew:
0f442aa2
BF
2197 if (open->op_stateowner == NULL) {
2198 sop = alloc_init_open_stateowner(strhashval, clp, open);
2199 if (sop == NULL)
2200 return nfserr_resource;
2201 open->op_stateowner = sop;
2202 }
2203 list_del_init(&sop->so_close_lru);
1da177e4 2204 renew_client(sop->so_client);
0f442aa2 2205 return nfs_ok;
1da177e4
LT
2206}
2207
b37ad28b 2208static inline __be32
4a6e43e6
N
2209nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2210{
2211 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2212 return nfserr_openmode;
2213 else
2214 return nfs_ok;
2215}
2216
52f4fb43
N
2217static struct nfs4_delegation *
2218find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2219{
2220 struct nfs4_delegation *dp;
2221
ea1da636 2222 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N
2223 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2224 return dp;
2225 }
2226 return NULL;
2227}
2228
b37ad28b 2229static __be32
567d9829
N
2230nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2231 struct nfs4_delegation **dp)
2232{
2233 int flags;
b37ad28b 2234 __be32 status = nfserr_bad_stateid;
567d9829
N
2235
2236 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2237 if (*dp == NULL)
c44c5eeb 2238 goto out;
567d9829
N
2239 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2240 RD_STATE : WR_STATE;
2241 status = nfs4_check_delegmode(*dp, flags);
2242 if (status)
2243 *dp = NULL;
c44c5eeb
N
2244out:
2245 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2246 return nfs_ok;
2247 if (status)
2248 return status;
2249 open->op_stateowner->so_confirmed = 1;
2250 return nfs_ok;
567d9829
N
2251}
2252
b37ad28b 2253static __be32
1da177e4
LT
2254nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2255{
2256 struct nfs4_stateid *local;
b37ad28b 2257 __be32 status = nfserr_share_denied;
1da177e4
LT
2258 struct nfs4_stateowner *sop = open->op_stateowner;
2259
8beefa24 2260 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT
2261 /* ignore lock owners */
2262 if (local->st_stateowner->so_is_open_owner == 0)
2263 continue;
2264 /* remember if we have seen this open owner */
2265 if (local->st_stateowner == sop)
2266 *stpp = local;
2267 /* check for conflicting share reservations */
2268 if (!test_share(local, open))
2269 goto out;
2270 }
2271 status = 0;
2272out:
2273 return status;
2274}
2275
5ac049ac
N
2276static inline struct nfs4_stateid *
2277nfs4_alloc_stateid(void)
2278{
2279 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2280}
2281
b37ad28b 2282static __be32
1da177e4 2283nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2284 struct nfs4_delegation *dp,
1da177e4
LT
2285 struct svc_fh *cur_fh, int flags)
2286{
2287 struct nfs4_stateid *stp;
1da177e4 2288
5ac049ac 2289 stp = nfs4_alloc_stateid();
1da177e4
LT
2290 if (stp == NULL)
2291 return nfserr_resource;
2292
567d9829
N
2293 if (dp) {
2294 get_file(dp->dl_vfs_file);
2295 stp->st_vfs_file = dp->dl_vfs_file;
2296 } else {
b37ad28b 2297 __be32 status;
567d9829
N
2298 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2299 &stp->st_vfs_file);
2300 if (status) {
2301 if (status == nfserr_dropit)
2302 status = nfserr_jukebox;
5ac049ac 2303 kmem_cache_free(stateid_slab, stp);
567d9829
N
2304 return status;
2305 }
1da177e4 2306 }
1da177e4
LT
2307 *stpp = stp;
2308 return 0;
2309}
2310
b37ad28b 2311static inline __be32
1da177e4
LT
2312nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2313 struct nfsd4_open *open)
2314{
2315 struct iattr iattr = {
2316 .ia_valid = ATTR_SIZE,
2317 .ia_size = 0,
2318 };
2319 if (!open->op_truncate)
2320 return 0;
2321 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2322 return nfserr_inval;
1da177e4
LT
2323 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2324}
2325
b37ad28b 2326static __be32
1da177e4
LT
2327nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2328{
2329 struct file *filp = stp->st_vfs_file;
7eaa36e2 2330 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2331 unsigned int share_access, new_writer;
b37ad28b 2332 __be32 status;
1da177e4
LT
2333
2334 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF
2335 new_writer = (~share_access) & open->op_share_access
2336 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2337
6c26d08f 2338 if (new_writer) {
b37ad28b
AV
2339 int err = get_write_access(inode);
2340 if (err)
2341 return nfserrno(err);
e518f056
BH
2342 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2343 if (err)
2344 return nfserrno(err);
2345 file_take_write(filp);
6c26d08f 2346 }
1da177e4
LT
2347 status = nfsd4_truncate(rqstp, cur_fh, open);
2348 if (status) {
6c26d08f
BF
2349 if (new_writer)
2350 put_write_access(inode);
1da177e4
LT
2351 return status;
2352 }
2353 /* remember the open */
6c26d08f 2354 filp->f_mode |= open->op_share_access;
b55e0ba1
BF
2355 __set_bit(open->op_share_access, &stp->st_access_bmap);
2356 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT
2357
2358 return nfs_ok;
2359}
2360
2361
1da177e4 2362static void
37515177 2363nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2364{
37515177
N
2365 open->op_stateowner->so_confirmed = 1;
2366 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT
2367}
2368
2369/*
2370 * Attempt to hand out a delegation.
2371 */
2372static void
2373nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2374{
2375 struct nfs4_delegation *dp;
2376 struct nfs4_stateowner *sop = stp->st_stateowner;
2bf23875 2377 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
1da177e4
LT
2378 struct file_lock fl, *flp = &fl;
2379 int status, flag = 0;
2380
2381 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N
2382 open->op_recall = 0;
2383 switch (open->op_claim_type) {
2384 case NFS4_OPEN_CLAIM_PREVIOUS:
2bf23875 2385 if (!cb_up)
7b190fec
N
2386 open->op_recall = 1;
2387 flag = open->op_delegate_type;
2388 if (flag == NFS4_OPEN_DELEGATE_NONE)
2389 goto out;
2390 break;
2391 case NFS4_OPEN_CLAIM_NULL:
2392 /* Let's not give out any delegations till everyone's
2393 * had the chance to reclaim theirs.... */
af558e33 2394 if (locks_in_grace())
7b190fec 2395 goto out;
2bf23875 2396 if (!cb_up || !sop->so_confirmed)
7b190fec
N
2397 goto out;
2398 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2399 flag = NFS4_OPEN_DELEGATE_WRITE;
2400 else
2401 flag = NFS4_OPEN_DELEGATE_READ;
2402 break;
2403 default:
2404 goto out;
2405 }
1da177e4
LT
2406
2407 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2408 if (dp == NULL) {
2409 flag = NFS4_OPEN_DELEGATE_NONE;
2410 goto out;
2411 }
2412 locks_init_lock(&fl);
2413 fl.fl_lmops = &nfsd_lease_mng_ops;
2414 fl.fl_flags = FL_LEASE;
9167f501 2415 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT
2416 fl.fl_end = OFFSET_MAX;
2417 fl.fl_owner = (fl_owner_t)dp;
2418 fl.fl_file = stp->st_vfs_file;
2419 fl.fl_pid = current->tgid;
2420
a9933cea 2421 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT
2422 * the lock_manager callbacks fl_mylease and fl_change are used
2423 */
9167f501 2424 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2425 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2426 unhash_delegation(dp);
1da177e4
LT
2427 flag = NFS4_OPEN_DELEGATE_NONE;
2428 goto out;
2429 }
2430
2431 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2432
8c10cbdb
BH
2433 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2434 STATEID_VAL(&dp->dl_stateid));
1da177e4 2435out:
7b190fec
N
2436 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2437 && flag == NFS4_OPEN_DELEGATE_NONE
2438 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2439 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT
2440 open->op_delegate_type = flag;
2441}
2442
2443/*
2444 * called with nfs4_lock_state() held.
2445 */
b37ad28b 2446__be32
1da177e4
LT
2447nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2448{
6668958f 2449 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT
2450 struct nfs4_file *fp = NULL;
2451 struct inode *ino = current_fh->fh_dentry->d_inode;
2452 struct nfs4_stateid *stp = NULL;
567d9829 2453 struct nfs4_delegation *dp = NULL;
b37ad28b 2454 __be32 status;
1da177e4
LT
2455
2456 status = nfserr_inval;
d87a8ade 2457 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2458 || !deny_valid(open->op_share_deny))
1da177e4
LT
2459 goto out;
2460 /*
2461 * Lookup file; if found, lookup stateid and check open request,
2462 * and check for delegations in the process of being recalled.
2463 * If not found, create the nfs4_file struct
2464 */
2465 fp = find_file(ino);
2466 if (fp) {
2467 if ((status = nfs4_check_open(fp, open, &stp)))
2468 goto out;
c44c5eeb
N
2469 status = nfs4_check_deleg(fp, open, &dp);
2470 if (status)
2471 goto out;
1da177e4 2472 } else {
c44c5eeb
N
2473 status = nfserr_bad_stateid;
2474 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2475 goto out;
1da177e4
LT
2476 status = nfserr_resource;
2477 fp = alloc_init_file(ino);
2478 if (fp == NULL)
2479 goto out;
2480 }
2481
2482 /*
2483 * OPEN the file, or upgrade an existing OPEN.
2484 * If truncate fails, the OPEN fails.
2485 */
2486 if (stp) {
2487 /* Stateid was found, this is an OPEN upgrade */
2488 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2489 if (status)
2490 goto out;
444c2c07 2491 update_stateid(&stp->st_stateid);
1da177e4
LT
2492 } else {
2493 /* Stateid was not found, this is a new OPEN */
2494 int flags = 0;
9ecb6a08 2495 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2496 flags |= NFSD_MAY_READ;
1da177e4 2497 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2498 flags |= NFSD_MAY_WRITE;
567d9829
N
2499 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2500 if (status)
1da177e4
LT
2501 goto out;
2502 init_stateid(stp, fp, open);
2503 status = nfsd4_truncate(rqstp, current_fh, open);
2504 if (status) {
2283963f 2505 release_open_stateid(stp);
1da177e4
LT
2506 goto out;
2507 }
6668958f
AA
2508 if (nfsd4_has_session(&resp->cstate))
2509 update_stateid(&stp->st_stateid);
1da177e4
LT
2510 }
2511 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2512
8b8aae40 2513 if (nfsd4_has_session(&resp->cstate)) {
6668958f 2514 open->op_stateowner->so_confirmed = 1;
8b8aae40
RL
2515 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2516 }
6668958f 2517
1da177e4
LT
2518 /*
2519 * Attempt to hand out a delegation. No error return, because the
2520 * OPEN succeeds even if we fail.
2521 */
2522 nfs4_open_delegation(current_fh, open, stp);
2523
2524 status = nfs_ok;
2525
8c10cbdb
BH
2526 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2527 STATEID_VAL(&stp->st_stateid));
1da177e4 2528out:
13cd2184
N
2529 if (fp)
2530 put_nfs4_file(fp);
37515177
N
2531 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2532 nfs4_set_claim_prev(open);
1da177e4
LT
2533 /*
2534 * To finish the open response, we just need to set the rflags.
2535 */
2536 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA
2537 if (!open->op_stateowner->so_confirmed &&
2538 !nfsd4_has_session(&resp->cstate))
1da177e4
LT
2539 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2540
2541 return status;
2542}
2543
b37ad28b 2544__be32
b591480b
BF
2545nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2546 clientid_t *clid)
1da177e4
LT
2547{
2548 struct nfs4_client *clp;
b37ad28b 2549 __be32 status;
1da177e4
LT
2550
2551 nfs4_lock_state();
2552 dprintk("process_renew(%08x/%08x): starting\n",
2553 clid->cl_boot, clid->cl_id);
2554 status = nfserr_stale_clientid;
2555 if (STALE_CLIENTID(clid))
2556 goto out;
2557 clp = find_confirmed_client(clid);
2558 status = nfserr_expired;
2559 if (clp == NULL) {
2560 /* We assume the client took too long to RENEW. */
2561 dprintk("nfsd4_renew: clientid not found!\n");
2562 goto out;
2563 }
2564 renew_client(clp);
2565 status = nfserr_cb_path_down;
ea1da636 2566 if (!list_empty(&clp->cl_delegations)
2bf23875 2567 && !atomic_read(&clp->cl_cb_set))
1da177e4
LT
2568 goto out;
2569 status = nfs_ok;
2570out:
2571 nfs4_unlock_state();
2572 return status;
2573}
2574
af558e33
BF
2575struct lock_manager nfsd4_manager = {
2576};
2577
a76b4319 2578static void
af558e33 2579nfsd4_end_grace(void)
a76b4319
N
2580{
2581 dprintk("NFSD: end of grace period\n");
c7b9a459 2582 nfsd4_recdir_purge_old();
af558e33 2583 locks_end_grace(&nfsd4_manager);
e46b498c
BF
2584 /*
2585 * Now that every NFSv4 client has had the chance to recover and
2586 * to see the (possibly new, possibly shorter) lease time, we
2587 * can safely set the next grace time to the current lease time:
2588 */
2589 nfsd4_grace = nfsd4_lease;
a76b4319
N
2590}
2591
fd39ca9a 2592static time_t
1da177e4
LT
2593nfs4_laundromat(void)
2594{
2595 struct nfs4_client *clp;
2596 struct nfs4_stateowner *sop;
2597 struct nfs4_delegation *dp;
2598 struct list_head *pos, *next, reaplist;
cf07d2ea
BF
2599 time_t cutoff = get_seconds() - nfsd4_lease;
2600 time_t t, clientid_val = nfsd4_lease;
2601 time_t u, test_val = nfsd4_lease;
1da177e4
LT
2602
2603 nfs4_lock_state();
2604
2605 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF
2606 if (locks_in_grace())
2607 nfsd4_end_grace();
36acb66b
BH
2608 INIT_LIST_HEAD(&reaplist);
2609 spin_lock(&client_lock);
1da177e4
LT
2610 list_for_each_safe(pos, next, &client_lru) {
2611 clp = list_entry(pos, struct nfs4_client, cl_lru);
2612 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2613 t = clp->cl_time - cutoff;
2614 if (clientid_val > t)
2615 clientid_val = t;
2616 break;
2617 }
d7682988
BH
2618 if (atomic_read(&clp->cl_refcount)) {
2619 dprintk("NFSD: client in use (clientid %08x)\n",
2620 clp->cl_clientid.cl_id);
2621 continue;
2622 }
2623 unhash_client_locked(clp);
2624 list_add(&clp->cl_lru, &reaplist);
36acb66b
BH
2625 }
2626 spin_unlock(&client_lock);
2627 list_for_each_safe(pos, next, &reaplist) {
2628 clp = list_entry(pos, struct nfs4_client, cl_lru);
1da177e4
LT
2629 dprintk("NFSD: purging unused client (clientid %08x)\n",
2630 clp->cl_clientid.cl_id);
c7b9a459 2631 nfsd4_remove_clid_dir(clp);
1da177e4
LT
2632 expire_client(clp);
2633 }
1da177e4
LT
2634 spin_lock(&recall_lock);
2635 list_for_each_safe(pos, next, &del_recall_lru) {
2636 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2637 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2638 u = dp->dl_time - cutoff;
2639 if (test_val > u)
2640 test_val = u;
2641 break;
2642 }
2643 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2644 dp, dp->dl_flock);
2645 list_move(&dp->dl_recall_lru, &reaplist);
2646 }
2647 spin_unlock(&recall_lock);
2648 list_for_each_safe(pos, next, &reaplist) {
2649 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2650 list_del_init(&dp->dl_recall_lru);
2651 unhash_delegation(dp);
2652 }
cf07d2ea 2653 test_val = nfsd4_lease;
1da177e4
LT
2654 list_for_each_safe(pos, next, &close_lru) {
2655 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2656 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2657 u = sop->so_time - cutoff;
2658 if (test_val > u)
2659 test_val = u;
2660 break;
2661 }
2662 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2663 sop->so_id);
f044ff83 2664 release_openowner(sop);
1da177e4
LT
2665 }
2666 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2667 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2668 nfs4_unlock_state();
2669 return clientid_val;
2670}
2671
a254b246
HH
2672static struct workqueue_struct *laundry_wq;
2673static void laundromat_main(struct work_struct *);
2674static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2675
2676static void
c4028958 2677laundromat_main(struct work_struct *not_used)
1da177e4
LT
2678{
2679 time_t t;
2680
2681 t = nfs4_laundromat();
2682 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2683 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT
2684}
2685
fd39ca9a 2686static struct nfs4_stateowner *
f8816512
N
2687search_close_lru(u32 st_id, int flags)
2688{
1da177e4
LT
2689 struct nfs4_stateowner *local = NULL;
2690
1da177e4
LT
2691 if (flags & CLOSE_STATE) {
2692 list_for_each_entry(local, &close_lru, so_close_lru) {
2693 if (local->so_id == st_id)
2694 return local;
2695 }
2696 }
2697 return NULL;
2698}
2699
2700static inline int
2701nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2702{
7eaa36e2 2703 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT
2704}
2705
2706static int
2707STALE_STATEID(stateid_t *stateid)
2708{
78155ed7
BN
2709 if (time_after((unsigned long)boot_time,
2710 (unsigned long)stateid->si_boot)) {
8c10cbdb
BH
2711 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2712 STATEID_VAL(stateid));
78155ed7
BN
2713 return 1;
2714 }
2715 return 0;
2716}
2717
2718static int
2719EXPIRED_STATEID(stateid_t *stateid)
2720{
2721 if (time_before((unsigned long)boot_time,
2722 ((unsigned long)stateid->si_boot)) &&
cf07d2ea 2723 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
8c10cbdb
BH
2724 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2725 STATEID_VAL(stateid));
78155ed7
BN
2726 return 1;
2727 }
2728 return 0;
2729}
2730
2731static __be32
2732stateid_error_map(stateid_t *stateid)
2733{
2734 if (STALE_STATEID(stateid))
2735 return nfserr_stale_stateid;
2736 if (EXPIRED_STATEID(stateid))
2737 return nfserr_expired;
2738
8c10cbdb
BH
2739 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2740 STATEID_VAL(stateid));
78155ed7 2741 return nfserr_bad_stateid;
1da177e4
LT
2742}
2743
2744static inline int
2745access_permit_read(unsigned long access_bmap)
2746{
2747 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2748 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2749 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2750}
2751
2752static inline int
2753access_permit_write(unsigned long access_bmap)
2754{
2755 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2756 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2757}
2758
2759static
b37ad28b 2760__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2761{
b37ad28b 2762 __be32 status = nfserr_openmode;
1da177e4
LT
2763
2764 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2765 goto out;
2766 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2767 goto out;
2768 status = nfs_ok;
2769out:
2770 return status;
2771}
2772
b37ad28b 2773static inline __be32
1da177e4
LT
2774check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2775{
203a8c8e 2776 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2777 return nfs_ok;
af558e33 2778 else if (locks_in_grace()) {
1da177e4
LT
2779 /* Answer in remaining cases depends on existance of
2780 * conflicting state; so we must wait out the grace period. */
2781 return nfserr_grace;
2782 } else if (flags & WR_STATE)
2783 return nfs4_share_conflict(current_fh,
2784 NFS4_SHARE_DENY_WRITE);
2785 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2786 return nfs4_share_conflict(current_fh,
2787 NFS4_SHARE_DENY_READ);
2788}
2789
2790/*
2791 * Allow READ/WRITE during grace period on recovered state only for files
2792 * that are not able to provide mandatory locking.
2793 */
2794static inline int
18f82731 2795grace_disallows_io(struct inode *inode)
1da177e4 2796{
203a8c8e 2797 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT
2798}
2799
6668958f 2800static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2801{
6668958f
AA
2802 /*
2803 * When sessions are used the stateid generation number is ignored
2804 * when it is zero.
2805 */
2806 if ((flags & HAS_SESSION) && in->si_generation == 0)
2807 goto out;
2808
0836f587
BF
2809 /* If the client sends us a stateid from the future, it's buggy: */
2810 if (in->si_generation > ref->si_generation)
2811 return nfserr_bad_stateid;
2812 /*
2813 * The following, however, can happen. For example, if the
2814 * client sends an open and some IO at the same time, the open
2815 * may bump si_generation while the IO is still in flight.
2816 * Thanks to hard links and renames, the client never knows what
2817 * file an open will affect. So it could avoid that situation
2818 * only by serializing all opens and IO from the same open
2819 * owner. To recover from the old_stateid error, the client
2820 * will just have to retry the IO:
2821 */
2822 if (in->si_generation < ref->si_generation)
2823 return nfserr_old_stateid;
6668958f 2824out:
0836f587
BF
2825 return nfs_ok;
2826}
2827
3e633079
BF
2828static int is_delegation_stateid(stateid_t *stateid)
2829{
2830 return stateid->si_fileid == 0;
2831}
2832
1da177e4
LT
2833/*
2834* Checks for stateid operations
2835*/
b37ad28b 2836__be32
dd453dfd
BH
2837nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2838 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT
2839{
2840 struct nfs4_stateid *stp = NULL;
2841 struct nfs4_delegation *dp = NULL;
dd453dfd 2842 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2843 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2844 __be32 status;
1da177e4 2845
1da177e4
LT
2846 if (filpp)
2847 *filpp = NULL;
2848
18f82731 2849 if (grace_disallows_io(ino))
1da177e4
LT
2850 return nfserr_grace;
2851
6668958f
AA
2852 if (nfsd4_has_session(cstate))
2853 flags |= HAS_SESSION;
2854
1da177e4
LT
2855 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2856 return check_special_stateids(current_fh, stateid, flags);
2857
1da177e4
LT
2858 status = nfserr_stale_stateid;
2859 if (STALE_STATEID(stateid))
2860 goto out;
2861
1da177e4 2862 status = nfserr_bad_stateid;
3e633079 2863 if (is_delegation_stateid(stateid)) {
a4455be0 2864 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN
2865 if (!dp) {
2866 status = stateid_error_map(stateid);
1da177e4 2867 goto out;
78155ed7 2868 }
6668958f
AA
2869 status = check_stateid_generation(stateid, &dp->dl_stateid,
2870 flags);
0c2a498f
BF
2871 if (status)
2872 goto out;
dc9bf700
BF
2873 status = nfs4_check_delegmode(dp, flags);
2874 if (status)
2875 goto out;
2876 renew_client(dp->dl_client);
dc9bf700
BF
2877 if (filpp)
2878 *filpp = dp->dl_vfs_file;
1da177e4 2879 } else { /* open or lock stateid */
a4455be0 2880 stp = find_stateid(stateid, flags);
78155ed7
BN
2881 if (!stp) {
2882 status = stateid_error_map(stateid);
1da177e4 2883 goto out;
78155ed7 2884 }
6150ef0d 2885 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT
2886 goto out;
2887 if (!stp->st_stateowner->so_confirmed)
2888 goto out;
6668958f
AA
2889 status = check_stateid_generation(stateid, &stp->st_stateid,
2890 flags);
0c2a498f
BF
2891 if (status)
2892 goto out;
a4455be0
BF
2893 status = nfs4_check_openmode(stp, flags);
2894 if (status)
1da177e4
LT
2895 goto out;
2896 renew_client(stp->st_stateowner->so_client);
2897 if (filpp)
2898 *filpp = stp->st_vfs_file;
1da177e4
LT
2899 }
2900 status = nfs_ok;
2901out:
2902 return status;
2903}
2904
4c4cd222
N
2905static inline int
2906setlkflg (int type)
2907{
2908 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2909 RD_STATE : WR_STATE;
2910}
1da177e4
LT
2911
2912/*
2913 * Checks for sequence id mutating operations.
2914 */
b37ad28b 2915static __be32
dd453dfd
BH
2916nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2917 stateid_t *stateid, int flags,
2918 struct nfs4_stateowner **sopp,
2919 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2920{
1da177e4
LT
2921 struct nfs4_stateid *stp;
2922 struct nfs4_stateowner *sop;
dd453dfd 2923 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2924 __be32 status;
1da177e4 2925
8c10cbdb
BH
2926 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2927 seqid, STATEID_VAL(stateid));
3a4f98bb 2928
1da177e4
LT
2929 *stpp = NULL;
2930 *sopp = NULL;
2931
1da177e4 2932 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2933 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2934 return nfserr_bad_stateid;
1da177e4
LT
2935 }
2936
1da177e4 2937 if (STALE_STATEID(stateid))
3a4f98bb 2938 return nfserr_stale_stateid;
6668958f
AA
2939
2940 if (nfsd4_has_session(cstate))
2941 flags |= HAS_SESSION;
2942
1da177e4
LT
2943 /*
2944 * We return BAD_STATEID if filehandle doesn't match stateid,
2945 * the confirmed flag is incorrecly set, or the generation
2946 * number is incorrect.
1da177e4 2947 */
f8816512
N
2948 stp = find_stateid(stateid, flags);
2949 if (stp == NULL) {
2950 /*
2951 * Also, we should make sure this isn't just the result of
2952 * a replayed close:
2953 */
2954 sop = search_close_lru(stateid->si_stateownerid, flags);
2955 if (sop == NULL)
78155ed7 2956 return stateid_error_map(stateid);
f8816512
N
2957 *sopp = sop;
2958 goto check_replay;
2959 }
1da177e4 2960
39325bd0
BF
2961 *stpp = stp;
2962 *sopp = sop = stp->st_stateowner;
2963
4c4cd222 2964 if (lock) {
4c4cd222 2965 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2966 struct nfs4_client *clp = sop->so_client;
4c4cd222 2967 int lkflg = 0;
b37ad28b 2968 __be32 status;
4c4cd222
N
2969
2970 lkflg = setlkflg(lock->lk_type);
2971
2972 if (lock->lk_is_new) {
599e0a22
BF
2973 if (!sop->so_is_open_owner)
2974 return nfserr_bad_stateid;
60adfc50
AA
2975 if (!(flags & HAS_SESSION) &&
2976 !same_clid(&clp->cl_clientid, lockclid))
2977 return nfserr_bad_stateid;
599e0a22
BF
2978 /* stp is the open stateid */
2979 status = nfs4_check_openmode(stp, lkflg);
2980 if (status)
2981 return status;
2982 } else {
2983 /* stp is the lock stateid */
2984 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2985 if (status)
2986 return status;
4c4cd222 2987 }
1da177e4
LT
2988 }
2989
f3362737 2990 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2991 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2992 return nfserr_bad_stateid;
1da177e4
LT
2993 }
2994
1da177e4
LT
2995 /*
2996 * We now validate the seqid and stateid generation numbers.
2997 * For the moment, we ignore the possibility of
2998 * generation number wraparound.
2999 */
6668958f 3000 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT
3001 goto check_replay;
3002
3a4f98bb 3003 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3004 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N
3005 " unconfirmed stateowner!\n");
3006 return nfserr_bad_stateid;
1da177e4 3007 }
3a4f98bb 3008 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3009 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N
3010 " confirmed yet!\n");
3011 return nfserr_bad_stateid;
1da177e4 3012 }
6668958f 3013 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF
3014 if (status)
3015 return status;
52fd004e 3016 renew_client(sop->so_client);
3a4f98bb 3017 return nfs_ok;
1da177e4 3018
1da177e4 3019check_replay:
bd9aac52 3020 if (seqid == sop->so_seqid - 1) {
849823c5 3021 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3022 /* indicate replay to calling function */
a90b061c 3023 return nfserr_replay_me;
1da177e4 3024 }
2fdada03 3025 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N
3026 sop->so_seqid, seqid);
3027 *sopp = NULL;
3028 return nfserr_bad_seqid;
1da177e4
LT
3029}
3030
b37ad28b 3031__be32
ca364317 3032nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3033 struct nfsd4_open_confirm *oc)
1da177e4 3034{
b37ad28b 3035 __be32 status;
1da177e4
LT
3036 struct nfs4_stateowner *sop;
3037 struct nfs4_stateid *stp;
3038
3039 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF
3040 (int)cstate->current_fh.fh_dentry->d_name.len,
3041 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3042
ca364317 3043 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF
3044 if (status)
3045 return status;
1da177e4
LT
3046
3047 nfs4_lock_state();
3048
dd453dfd 3049 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3050 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3051 CONFIRM | OPEN_STATE,
1da177e4
LT
3052 &oc->oc_stateowner, &stp, NULL)))
3053 goto out;
3054
3055 sop = oc->oc_stateowner;
3056 sop->so_confirmed = 1;
3057 update_stateid(&stp->st_stateid);
3058 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH
3059 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3060 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N
3061
3062 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3063out:
f2327d9a 3064 if (oc->oc_stateowner) {
1da177e4 3065 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3066 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3067 }
1da177e4
LT
3068 nfs4_unlock_state();
3069 return status;
3070}
3071
3072
3073/*
3074 * unset all bits in union bitmap (bmap) that
3075 * do not exist in share (from successful OPEN_DOWNGRADE)
3076 */
3077static void
3078reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3079{
3080 int i;
3081 for (i = 1; i < 4; i++) {
3082 if ((i & access) != i)
3083 __clear_bit(i, bmap);
3084 }
3085}
3086
3087static void
3088reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3089{
3090 int i;
3091 for (i = 0; i < 4; i++) {
3092 if ((i & deny) != i)
3093 __clear_bit(i, bmap);
3094 }
3095}
3096
b37ad28b 3097__be32
ca364317
BF
3098nfsd4_open_downgrade(struct svc_rqst *rqstp,
3099 struct nfsd4_compound_state *cstate,
a4f1706a 3100 struct nfsd4_open_downgrade *od)
1da177e4 3101{
b37ad28b 3102 __be32 status;
1da177e4
LT
3103 struct nfs4_stateid *stp;
3104 unsigned int share_access;
3105
3106 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF
3107 (int)cstate->current_fh.fh_dentry->d_name.len,
3108 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3109
d87a8ade 3110 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3111 || !deny_valid(od->od_share_deny))
1da177e4
LT
3112 return nfserr_inval;
3113
3114 nfs4_lock_state();
dd453dfd 3115 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3116 od->od_seqid,
1da177e4 3117 &od->od_stateid,
f3362737 3118 OPEN_STATE,
1da177e4
LT
3119 &od->od_stateowner, &stp, NULL)))
3120 goto out;
3121
3122 status = nfserr_inval;
3123 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3124 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3125 stp->st_access_bmap, od->od_share_access);
3126 goto out;
3127 }
3128 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3129 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3130 stp->st_deny_bmap, od->od_share_deny);
3131 goto out;
3132 }
3133 set_access(&share_access, stp->st_access_bmap);
3134 nfs4_file_downgrade(stp->st_vfs_file,
3135 share_access & ~od->od_share_access);
3136
3137 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3138 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3139
3140 update_stateid(&stp->st_stateid);
3141 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3142 status = nfs_ok;
3143out:
f2327d9a 3144 if (od->od_stateowner) {
1da177e4 3145 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3146 cstate->replay_owner = od->od_stateowner;
f2327d9a 3147 }
1da177e4
LT
3148 nfs4_unlock_state();
3149 return status;
3150}
3151
3152/*
3153 * nfs4_unlock_state() called after encode
3154 */
b37ad28b 3155__be32
ca364317 3156nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3157 struct nfsd4_close *close)
1da177e4 3158{
b37ad28b 3159 __be32 status;
1da177e4
LT
3160 struct nfs4_stateid *stp;
3161
3162 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF
3163 (int)cstate->current_fh.fh_dentry->d_name.len,
3164 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT
3165
3166 nfs4_lock_state();
3167 /* check close_lru for replay */
dd453dfd 3168 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3169 close->cl_seqid,
1da177e4 3170 &close->cl_stateid,
f3362737 3171 OPEN_STATE | CLOSE_STATE,
1da177e4
LT
3172 &close->cl_stateowner, &stp, NULL)))
3173 goto out;
1da177e4
LT
3174 status = nfs_ok;
3175 update_stateid(&stp->st_stateid);
3176 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3177
04ef5954 3178 /* release_stateid() calls nfsd_close() if needed */
2283963f 3179 release_open_stateid(stp);
04ef5954
BF
3180
3181 /* place unused nfs4_stateowners on so_close_lru list to be
3182 * released by the laundromat service after the lease period
3183 * to enable us to handle CLOSE replay
3184 */
3185 if (list_empty(&close->cl_stateowner->so_stateids))
3186 move_to_close_lru(close->cl_stateowner);
1da177e4 3187out:
f2327d9a 3188 if (close->cl_stateowner) {
1da177e4 3189 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3190 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3191 }
1da177e4
LT
3192 nfs4_unlock_state();
3193 return status;
3194}
3195
b37ad28b 3196__be32
ca364317
BF
3197nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3198 struct nfsd4_delegreturn *dr)
1da177e4 3199{
203a8c8e
BF
3200 struct nfs4_delegation *dp;
3201 stateid_t *stateid = &dr->dr_stateid;
3202 struct inode *inode;
b37ad28b 3203 __be32 status;
6668958f 3204 int flags = 0;
1da177e4 3205
ca364317 3206 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF
3207 return status;
3208 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3209
6668958f
AA
3210 if (nfsd4_has_session(cstate))
3211 flags |= HAS_SESSION;
1da177e4 3212 nfs4_lock_state();
203a8c8e
BF
3213 status = nfserr_bad_stateid;
3214 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3215 goto out;
3216 status = nfserr_stale_stateid;
3217 if (STALE_STATEID(stateid))
3218 goto out;
7e0f7cf5 3219 status = nfserr_bad_stateid;
203a8c8e
BF
3220 if (!is_delegation_stateid(stateid))
3221 goto out;
203a8c8e 3222 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN
3223 if (!dp) {
3224 status = stateid_error_map(stateid);
203a8c8e 3225 goto out;
78155ed7 3226 }
6668958f 3227 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF
3228 if (status)
3229 goto out;
3230 renew_client(dp->dl_client);
3231
3232 unhash_delegation(dp);
1da177e4 3233out:
203a8c8e
BF
3234 nfs4_unlock_state();
3235
1da177e4
LT
3236 return status;
3237}
3238
3239
3240/*
3241 * Lock owner state (byte-range locks)
3242 */
3243#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3244#define LOCK_HASH_BITS 8
3245#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3246#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3247
87df4de8
BH
3248static inline u64
3249end_offset(u64 start, u64 len)
3250{
3251 u64 end;
3252
3253 end = start + len;
3254 return end >= start ? end: NFS4_MAX_UINT64;
3255}
3256
3257/* last octet in a range */
3258static inline u64
3259last_byte_offset(u64 start, u64 len)
3260{
3261 u64 end;
3262
3263 BUG_ON(!len);
3264 end = start + len;
3265 return end > start ? end - 1: NFS4_MAX_UINT64;
3266}
3267
1da177e4
LT
3268#define lockownerid_hashval(id) \
3269 ((id) & LOCK_HASH_MASK)
3270
3271static inline unsigned int
3272lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3273 struct xdr_netobj *ownername)
3274{
3275 return (file_hashval(inode) + cl_id
3276 + opaque_hashval(ownername->data, ownername->len))
3277 & LOCK_HASH_MASK;
3278}
3279
3280static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3281static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3282static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3283
fd39ca9a 3284static struct nfs4_stateid *
1da177e4
LT
3285find_stateid(stateid_t *stid, int flags)
3286{
9346eff0 3287 struct nfs4_stateid *local;
1da177e4
LT
3288 u32 st_id = stid->si_stateownerid;
3289 u32 f_id = stid->si_fileid;
3290 unsigned int hashval;
3291
3292 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3293 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3294 hashval = stateid_hashval(st_id, f_id);
3295 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3296 if ((local->st_stateid.si_stateownerid == st_id) &&
3297 (local->st_stateid.si_fileid == f_id))
3298 return local;
3299 }
3300 }
9346eff0
KK
3301
3302 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT
3303 hashval = stateid_hashval(st_id, f_id);
3304 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3305 if ((local->st_stateid.si_stateownerid == st_id) &&
3306 (local->st_stateid.si_fileid == f_id))
3307 return local;
3308 }
849823c5 3309 }
1da177e4
LT
3310 return NULL;
3311}
3312
3313static struct nfs4_delegation *
3314find_delegation_stateid(struct inode *ino, stateid_t *stid)
3315{
13cd2184
N
3316 struct nfs4_file *fp;
3317 struct nfs4_delegation *dl;
1da177e4 3318
8c10cbdb
BH
3319 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3320 STATEID_VAL(stid));
1da177e4 3321
1da177e4 3322 fp = find_file(ino);
13cd2184
N
3323 if (!fp)
3324 return NULL;
3325 dl = find_delegation_file(fp, stid);
3326 put_nfs4_file(fp);
3327 return dl;
1da177e4
LT
3328}
3329
3330/*
3331 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3332 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3333 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3334 * locking, this prevents us from being completely protocol-compliant. The
3335 * real solution to this problem is to start using unsigned file offsets in
3336 * the VFS, but this is a very deep change!
3337 */
3338static inline void
3339nfs4_transform_lock_offset(struct file_lock *lock)
3340{
3341 if (lock->fl_start < 0)
3342 lock->fl_start = OFFSET_MAX;
3343 if (lock->fl_end < 0)
3344 lock->fl_end = OFFSET_MAX;
3345}
3346
d5b9026a
N
3347/* Hack!: For now, we're defining this just so we can use a pointer to it
3348 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3349static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3350};
1da177e4
LT
3351
3352static inline void
3353nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3354{
d5b9026a
N
3355 struct nfs4_stateowner *sop;
3356 unsigned int hval;
1da177e4 3357
d5b9026a
N
3358 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3359 sop = (struct nfs4_stateowner *) fl->fl_owner;
3360 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT
3361 kref_get(&sop->so_ref);
3362 deny->ld_sop = sop;
3363 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N
3364 } else {
3365 deny->ld_sop = NULL;
3366 deny->ld_clientid.cl_boot = 0;
3367 deny->ld_clientid.cl_id = 0;
1da177e4
LT
3368 }
3369 deny->ld_start = fl->fl_start;
87df4de8
BH
3370 deny->ld_length = NFS4_MAX_UINT64;
3371 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT
3372 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3373 deny->ld_type = NFS4_READ_LT;
3374 if (fl->fl_type != F_RDLCK)
3375 deny->ld_type = NFS4_WRITE_LT;
3376}
3377
1da177e4
LT
3378static struct nfs4_stateowner *
3379find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3380 struct xdr_netobj *owner)
3381{
3382 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3383 struct nfs4_stateowner *op;
3384
3385 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3386 if (same_owner_str(op, owner, clid))
1da177e4
LT
3387 return op;
3388 }
3389 return NULL;
3390}
3391
3392/*
3393 * Alloc a lock owner structure.
3394 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3395 * occured.
3396 *
3397 * strhashval = lock_ownerstr_hashval
1da177e4
LT
3398 */
3399
3400static struct nfs4_stateowner *
3401alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3402 struct nfs4_stateowner *sop;
3403 struct nfs4_replay *rp;
3404 unsigned int idhashval;
3405
3406 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3407 return NULL;
3408 idhashval = lockownerid_hashval(current_ownerid);
3409 INIT_LIST_HEAD(&sop->so_idhash);
3410 INIT_LIST_HEAD(&sop->so_strhash);
3411 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
3412 INIT_LIST_HEAD(&sop->so_stateids);
3413 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT
3414 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3415 sop->so_time = 0;
3416 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3417 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3418 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT
3419 sop->so_is_open_owner = 0;
3420 sop->so_id = current_ownerid++;
3421 sop->so_client = clp;
b59e3c0e
NB
3422 /* It is the openowner seqid that will be incremented in encode in the
3423 * case of new lockowners; so increment the lock seqid manually: */
3424 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT
3425 sop->so_confirmed = 1;
3426 rp = &sop->so_replay;
de1ae286 3427 rp->rp_status = nfserr_serverfault;
1da177e4
LT
3428 rp->rp_buflen = 0;
3429 rp->rp_buf = rp->rp_ibuf;
3430 return sop;
3431}
3432
fd39ca9a 3433static struct nfs4_stateid *
1da177e4
LT
3434alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3435{
3436 struct nfs4_stateid *stp;
3437 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3438
5ac049ac
N
3439 stp = nfs4_alloc_stateid();
3440 if (stp == NULL)
1da177e4
LT
3441 goto out;
3442 INIT_LIST_HEAD(&stp->st_hash);
3443 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N
3444 INIT_LIST_HEAD(&stp->st_perstateowner);
3445 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3446 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3447 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3448 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3449 stp->st_stateowner = sop;
13cd2184 3450 get_nfs4_file(fp);
1da177e4 3451 stp->st_file = fp;
78155ed7 3452 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT
3453 stp->st_stateid.si_stateownerid = sop->so_id;
3454 stp->st_stateid.si_fileid = fp->fi_id;
3455 stp->st_stateid.si_generation = 0;
3456 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3457 stp->st_access_bmap = open_stp->st_access_bmap;
3458 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3459 stp->st_openstp = open_stp;
1da177e4
LT
3460
3461out:
3462 return stp;
3463}
3464
fd39ca9a 3465static int
1da177e4
LT
3466check_lock_length(u64 offset, u64 length)
3467{
87df4de8 3468 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT
3469 LOFF_OVERFLOW(offset, length)));
3470}
3471
3472/*
3473 * LOCK operation
3474 */
b37ad28b 3475__be32
ca364317 3476nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3477 struct nfsd4_lock *lock)
1da177e4 3478{
3e9e3dbe 3479 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3480 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT
3481 struct nfs4_stateid *lock_stp;
3482 struct file *filp;
3483 struct file_lock file_lock;
8dc7c311 3484 struct file_lock conflock;
b37ad28b 3485 __be32 status = 0;
1da177e4 3486 unsigned int strhashval;
150b3934 3487 unsigned int cmd;
b8dd7b9a 3488 int err;
1da177e4
LT
3489
3490 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3491 (long long) lock->lk_offset,
3492 (long long) lock->lk_length);
3493
1da177e4
LT
3494 if (check_lock_length(lock->lk_offset, lock->lk_length))
3495 return nfserr_inval;
3496
ca364317 3497 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3498 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA
3499 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3500 return status;
3501 }
3502
1da177e4
LT
3503 nfs4_lock_state();
3504
3505 if (lock->lk_is_new) {
893f8770
N
3506 /*
3507 * Client indicates that this is a new lockowner.
3508 * Use open owner and open stateid to create lock owner and
3509 * lock stateid.
3510 */
1da177e4
LT
3511 struct nfs4_stateid *open_stp = NULL;
3512 struct nfs4_file *fp;
3513
3514 status = nfserr_stale_clientid;
60adfc50
AA
3515 if (!nfsd4_has_session(cstate) &&
3516 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3517 goto out;
1da177e4 3518
1da177e4 3519 /* validate and update open stateid and open seqid */
dd453dfd 3520 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3521 lock->lk_new_open_seqid,
3522 &lock->lk_new_open_stateid,
f3362737 3523 OPEN_STATE,
3a65588a 3524 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3525 lock);
37515177 3526 if (status)
1da177e4 3527 goto out;
3a65588a 3528 open_sop = lock->lk_replay_owner;
1da177e4
LT
3529 /* create lockowner and lock stateid */
3530 fp = open_stp->st_file;
3531 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3532 open_sop->so_client->cl_clientid.cl_id,
3533 &lock->v.new.owner);
3e9e3dbe
N
3534 /* XXX: Do we need to check for duplicate stateowners on
3535 * the same file, or should they just be allowed (and
3536 * create new stateids)? */
1da177e4 3537 status = nfserr_resource;
b59e3c0e
NB
3538 lock_sop = alloc_init_lock_stateowner(strhashval,
3539 open_sop->so_client, open_stp, lock);
3540 if (lock_sop == NULL)
1da177e4 3541 goto out;
b59e3c0e 3542 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3543 if (lock_stp == NULL)
1da177e4 3544 goto out;
1da177e4
LT
3545 } else {
3546 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3547 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3548 lock->lk_old_lock_seqid,
3549 &lock->lk_old_lock_stateid,
f3362737 3550 LOCK_STATE,
3a65588a 3551 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT
3552 if (status)
3553 goto out;
3a65588a 3554 lock_sop = lock->lk_replay_owner;
1da177e4 3555 }
3a65588a 3556 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT
3557 filp = lock_stp->st_vfs_file;
3558
0dd395dc 3559 status = nfserr_grace;
af558e33 3560 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N
3561 goto out;
3562 status = nfserr_no_grace;
af558e33 3563 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N
3564 goto out;
3565
1da177e4
LT
3566 locks_init_lock(&file_lock);
3567 switch (lock->lk_type) {
3568 case NFS4_READ_LT:
3569 case NFS4_READW_LT:
3570 file_lock.fl_type = F_RDLCK;
150b3934 3571 cmd = F_SETLK;
1da177e4
LT
3572 break;
3573 case NFS4_WRITE_LT:
3574 case NFS4_WRITEW_LT:
3575 file_lock.fl_type = F_WRLCK;
150b3934 3576 cmd = F_SETLK;
1da177e4
LT
3577 break;
3578 default:
3579 status = nfserr_inval;
3580 goto out;
3581 }
b59e3c0e 3582 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT
3583 file_lock.fl_pid = current->tgid;
3584 file_lock.fl_file = filp;
3585 file_lock.fl_flags = FL_POSIX;
d5b9026a 3586 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3587
3588 file_lock.fl_start = lock->lk_offset;
87df4de8 3589 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT
3590 nfs4_transform_lock_offset(&file_lock);
3591
3592 /*
3593 * Try to lock the file in the VFS.
3594 * Note: locks.c uses the BKL to protect the inode's lock list.
3595 */
3596
fd85b817 3597 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3598 switch (-err) {
1da177e4
LT
3599 case 0: /* success! */
3600 update_stateid(&lock_stp->st_stateid);
3601 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3602 sizeof(stateid_t));
b8dd7b9a 3603 status = 0;
eb76b3fd
AA
3604 break;
3605 case (EAGAIN): /* conflock holds conflicting lock */
3606 status = nfserr_denied;
3607 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3608 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3609 break;
1da177e4
LT
3610 case (EDEADLK):
3611 status = nfserr_deadlock;
eb76b3fd 3612 break;
1da177e4 3613 default:
fd85b817 3614 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA
3615 status = nfserr_resource;
3616 break;
1da177e4 3617 }
1da177e4 3618out:
8a280510 3619 if (status && lock->lk_is_new && lock_sop)
f044ff83 3620 release_lockowner(lock_sop);
3a65588a
BF
3621 if (lock->lk_replay_owner) {
3622 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3623 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3624 }
1da177e4
LT
3625 nfs4_unlock_state();
3626 return status;
3627}
3628
55ef1274
BF
3629/*
3630 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3631 * so we do a temporary open here just to get an open file to pass to
3632 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3633 * inode operation.)
3634 */
3635static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3636{
3637 struct file *file;
3638 int err;
3639
3640 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3641 if (err)
3642 return err;
3643 err = vfs_test_lock(file, lock);
3644 nfsd_close(file);
3645 return err;
3646}
3647
1da177e4
LT
3648/*
3649 * LOCKT operation
3650 */
b37ad28b 3651__be32
ca364317
BF
3652nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3653 struct nfsd4_lockt *lockt)
1da177e4
LT
3654{
3655 struct inode *inode;
1da177e4 3656 struct file_lock file_lock;
fd85b817 3657 int error;
b37ad28b 3658 __be32 status;
1da177e4 3659
af558e33 3660 if (locks_in_grace())
1da177e4
LT
3661 return nfserr_grace;
3662
3663 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3664 return nfserr_inval;
3665
3666 lockt->lt_stateowner = NULL;
3667 nfs4_lock_state();
3668
3669 status = nfserr_stale_clientid;
60adfc50 3670 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3671 goto out;
1da177e4 3672
ca364317 3673 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3674 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT
3675 if (status == nfserr_symlink)
3676 status = nfserr_inval;
3677 goto out;
3678 }
3679
ca364317 3680 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
3681 locks_init_lock(&file_lock);
3682 switch (lockt->lt_type) {
3683 case NFS4_READ_LT:
3684 case NFS4_READW_LT:
3685 file_lock.fl_type = F_RDLCK;
3686 break;
3687 case NFS4_WRITE_LT:
3688 case NFS4_WRITEW_LT:
3689 file_lock.fl_type = F_WRLCK;
3690 break;
3691 default:
2fdada03 3692 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT
3693 status = nfserr_inval;
3694 goto out;
3695 }
3696
3697 lockt->lt_stateowner = find_lockstateowner_str(inode,
3698 &lockt->lt_clientid, &lockt->lt_owner);
3699 if (lockt->lt_stateowner)
3700 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3701 file_lock.fl_pid = current->tgid;
3702 file_lock.fl_flags = FL_POSIX;
3703
3704 file_lock.fl_start = lockt->lt_offset;
87df4de8 3705 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT
3706
3707 nfs4_transform_lock_offset(&file_lock);
3708
1da177e4 3709 status = nfs_ok;
55ef1274 3710 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME
3711 if (error) {
3712 status = nfserrno(error);
3713 goto out;
3714 }
9d6a8c5c 3715 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3716 status = nfserr_denied;
9d6a8c5c 3717 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT
3718 }
3719out:
3720 nfs4_unlock_state();
3721 return status;
3722}
3723
b37ad28b 3724__be32
ca364317 3725nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3726 struct nfsd4_locku *locku)
1da177e4
LT
3727{
3728 struct nfs4_stateid *stp;
3729 struct file *filp = NULL;
3730 struct file_lock file_lock;
b37ad28b 3731 __be32 status;
b8dd7b9a 3732 int err;
1da177e4
LT
3733
3734 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3735 (long long) locku->lu_offset,
3736 (long long) locku->lu_length);
3737
3738 if (check_lock_length(locku->lu_offset, locku->lu_length))
3739 return nfserr_inval;
3740
3741 nfs4_lock_state();
3742
dd453dfd 3743 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT
3744 locku->lu_seqid,
3745 &locku->lu_stateid,
f3362737 3746 LOCK_STATE,
1da177e4
LT
3747 &locku->lu_stateowner, &stp, NULL)))
3748 goto out;
3749
3750 filp = stp->st_vfs_file;
3751 BUG_ON(!filp);
3752 locks_init_lock(&file_lock);
3753 file_lock.fl_type = F_UNLCK;
3754 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3755 file_lock.fl_pid = current->tgid;
3756 file_lock.fl_file = filp;
3757 file_lock.fl_flags = FL_POSIX;
d5b9026a 3758 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
3759 file_lock.fl_start = locku->lu_offset;
3760
87df4de8 3761 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT
3762 nfs4_transform_lock_offset(&file_lock);
3763
3764 /*
3765 * Try to unlock the file in the VFS.
3766 */
fd85b817 3767 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3768 if (err) {
fd85b817 3769 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT
3770 goto out_nfserr;
3771 }
3772 /*
3773 * OK, unlock succeeded; the only thing left to do is update the stateid.
3774 */
3775 update_stateid(&stp->st_stateid);
3776 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3777
3778out:
f2327d9a 3779 if (locku->lu_stateowner) {
1da177e4 3780 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3781 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3782 }
1da177e4
LT
3783 nfs4_unlock_state();
3784 return status;
3785
3786out_nfserr:
b8dd7b9a 3787 status = nfserrno(err);
1da177e4
LT
3788 goto out;
3789}
3790
3791/*
3792 * returns
3793 * 1: locks held by lockowner
3794 * 0: no locks held by lockowner
3795 */
3796static int
3797check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3798{
3799 struct file_lock **flpp;
7eaa36e2 3800 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT
3801 int status = 0;
3802
3803 lock_kernel();
3804 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3805 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT
3806 status = 1;
3807 goto out;
796dadfd 3808 }
1da177e4
LT
3809 }
3810out:
3811 unlock_kernel();
3812 return status;
3813}
3814
b37ad28b 3815__be32
b591480b
BF
3816nfsd4_release_lockowner(struct svc_rqst *rqstp,
3817 struct nfsd4_compound_state *cstate,
3818 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT
3819{
3820 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N
3821 struct nfs4_stateowner *sop;
3822 struct nfs4_stateid *stp;
1da177e4 3823 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N
3824 struct list_head matches;
3825 int i;
b37ad28b 3826 __be32 status;
1da177e4
LT
3827
3828 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3829 clid->cl_boot, clid->cl_id);
3830
3831 /* XXX check for lease expiration */
3832
3833 status = nfserr_stale_clientid;
849823c5 3834 if (STALE_CLIENTID(clid))
1da177e4 3835 return status;
1da177e4
LT
3836
3837 nfs4_lock_state();
3838
3e9e3dbe
N
3839 status = nfserr_locks_held;
3840 /* XXX: we're doing a linear search through all the lockowners.
3841 * Yipes! For now we'll just hope clients aren't really using
3842 * release_lockowner much, but eventually we have to fix these
3843 * data structures. */
3844 INIT_LIST_HEAD(&matches);
3845 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3846 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3847 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N
3848 continue;
3849 list_for_each_entry(stp, &sop->so_stateids,
3850 st_perstateowner) {
3851 if (check_for_locks(stp->st_vfs_file, sop))
3852 goto out;
3853 /* Note: so_perclient unused for lockowners,
3854 * so it's OK to fool with here. */
3855 list_add(&sop->so_perclient, &matches);
3856 }
1da177e4 3857 }
3e9e3dbe
N
3858 }
3859 /* Clients probably won't expect us to return with some (but not all)
3860 * of the lockowner state released; so don't release any until all
3861 * have been checked. */
3862 status = nfs_ok;
0fa822e4
N
3863 while (!list_empty(&matches)) {
3864 sop = list_entry(matches.next, struct nfs4_stateowner,
3865 so_perclient);
3866 /* unhash_stateowner deletes so_perclient only
3867 * for openowners. */
3868 list_del(&sop->so_perclient);
f044ff83 3869 release_lockowner(sop);
1da177e4
LT
3870 }
3871out:
3872 nfs4_unlock_state();
3873 return status;
3874}
3875
3876static inline struct nfs4_client_reclaim *
a55370a3 3877alloc_reclaim(void)
1da177e4 3878{
a55370a3 3879 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT
3880}
3881
c7b9a459 3882int
a1bcecd2 3883nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N
3884{
3885 unsigned int strhashval = clientstr_hashval(name);
3886 struct nfs4_client *clp;
3887
a1bcecd2 3888 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N
3889 return clp ? 1 : 0;
3890}
3891
1da177e4
LT
3892/*
3893 * failure => all reset bets are off, nfserr_no_grace...
3894 */
190e4fbf
N
3895int
3896nfs4_client_to_reclaim(const char *name)
1da177e4
LT
3897{
3898 unsigned int strhashval;
3899 struct nfs4_client_reclaim *crp = NULL;
3900
a55370a3
N
3901 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3902 crp = alloc_reclaim();
1da177e4
LT
3903 if (!crp)
3904 return 0;
a55370a3 3905 strhashval = clientstr_hashval(name);
1da177e4
LT
3906 INIT_LIST_HEAD(&crp->cr_strhash);
3907 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3908 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT
3909 reclaim_str_hashtbl_size++;
3910 return 1;
3911}
3912
3913static void
3914nfs4_release_reclaim(void)
3915{
3916 struct nfs4_client_reclaim *crp = NULL;
3917 int i;
3918
1da177e4
LT
3919 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3920 while (!list_empty(&reclaim_str_hashtbl[i])) {
3921 crp = list_entry(reclaim_str_hashtbl[i].next,
3922 struct nfs4_client_reclaim, cr_strhash);
3923 list_del(&crp->cr_strhash);
1da177e4
LT
3924 kfree(crp);
3925 reclaim_str_hashtbl_size--;
3926 }
3927 }
3928 BUG_ON(reclaim_str_hashtbl_size);
3929}
3930
3931/*
3932 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3933static struct nfs4_client_reclaim *
1da177e4
LT
3934nfs4_find_reclaim_client(clientid_t *clid)
3935{
3936 unsigned int strhashval;
3937 struct nfs4_client *clp;
3938 struct nfs4_client_reclaim *crp = NULL;
3939
3940
3941 /* find clientid in conf_id_hashtbl */
3942 clp = find_confirmed_client(clid);
3943 if (clp == NULL)
3944 return NULL;
3945
a55370a3
N
3946 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3947 clp->cl_name.len, clp->cl_name.data,
3948 clp->cl_recdir);
1da177e4
LT
3949
3950 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3951 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3952 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3953 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT
3954 return crp;
3955 }
3956 }
3957 return NULL;
3958}
3959
3960/*
3961* Called from OPEN. Look for clientid in reclaim list.
3962*/
b37ad28b 3963__be32
1da177e4
LT
3964nfs4_check_open_reclaim(clientid_t *clid)
3965{
dfc83565 3966 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT
3967}
3968
ac4d8ff2 3969/* initialization to perform at module load time: */
1da177e4 3970
e8ff2a84 3971int
ac4d8ff2 3972nfs4_state_init(void)
1da177e4 3973{
e8ff2a84 3974 int i, status;
1da177e4 3975
e8ff2a84
BF
3976 status = nfsd4_init_slabs();
3977 if (status)
3978 return status;
1da177e4
LT
3979 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3980 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3981 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3982 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3983 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3984 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3985 }
5282fd72
ME
3986 for (i = 0; i < SESSION_HASH_SIZE; i++)
3987 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT
3988 for (i = 0; i < FILE_HASH_SIZE; i++) {
3989 INIT_LIST_HEAD(&file_hashtbl[i]);
3990 }
3991 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3992 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3993 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3994 }
3995 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3996 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3997 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3998 }
3999 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4000 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4001 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4002 }
1da177e4 4003 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT
4004 INIT_LIST_HEAD(&close_lru);
4005 INIT_LIST_HEAD(&client_lru);
4006 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4007 reclaim_str_hashtbl_size = 0;
e8ff2a84 4008 return 0;
ac4d8ff2
N
4009}
4010
190e4fbf
N
4011static void
4012nfsd4_load_reboot_recovery_data(void)
4013{
4014 int status;
4015
0964a3d3
N
4016 nfs4_lock_state();
4017 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4018 status = nfsd4_recdir_load();
0964a3d3 4019 nfs4_unlock_state();
190e4fbf
N
4020 if (status)
4021 printk("NFSD: Failure reading reboot recovery data\n");
4022}
4023
c2f1a551
MS
4024/*
4025 * Since the lifetime of a delegation isn't limited to that of an open, a
4026 * client may quite reasonably hang on to a delegation as long as it has
4027 * the inode cached. This becomes an obvious problem the first time a
4028 * client's inode cache approaches the size of the server's total memory.
4029 *
4030 * For now we avoid this problem by imposing a hard limit on the number
4031 * of delegations, which varies according to the server's memory size.
4032 */
4033static void
4034set_max_delegations(void)
4035{
4036 /*
4037 * Allow at most 4 delegations per megabyte of RAM. Quick
4038 * estimates suggest that in the worst case (where every delegation
4039 * is for a different inode), a delegation could take about 1.5K,
4040 * giving a worst case usage of about 6% of memory.
4041 */
4042 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4043}
4044
ac4d8ff2
N
4045/* initialization to perform when the nfsd service is started: */
4046
29ab23cc 4047static int
ac4d8ff2
N
4048__nfs4_state_start(void)
4049{
b5a1a81e
BF
4050 int ret;
4051
1da177e4 4052 boot_time = get_seconds();
af558e33 4053 locks_start_grace(&nfsd4_manager);
9a8db97e 4054 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
e46b498c 4055 nfsd4_grace);
b5a1a81e
BF
4056 ret = set_callback_cred();
4057 if (ret)
4058 return -ENOMEM;
58da282b 4059 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF
4060 if (laundry_wq == NULL)
4061 return -ENOMEM;
b5a1a81e
BF
4062 ret = nfsd4_create_callback_queue();
4063 if (ret)
4064 goto out_free_laundry;
e46b498c 4065 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
c2f1a551 4066 set_max_delegations();
b5a1a81e
BF
4067 return 0;
4068out_free_laundry:
4069 destroy_workqueue(laundry_wq);
4070 return ret;
1da177e4
LT
4071}
4072
29ab23cc 4073int
76a3550e 4074nfs4_state_start(void)
1da177e4 4075{
29ab23cc
BF
4076 int ret;
4077
1da177e4 4078 if (nfs4_init)
29ab23cc 4079 return 0;
190e4fbf 4080 nfsd4_load_reboot_recovery_data();
29ab23cc
BF
4081 ret = __nfs4_state_start();
4082 if (ret)
4083 return ret;
1da177e4 4084 nfs4_init = 1;
29ab23cc 4085 return 0;
1da177e4
LT
4086}
4087
1da177e4
LT
4088static void
4089__nfs4_state_shutdown(void)
4090{
4091 int i;
4092 struct nfs4_client *clp = NULL;
4093 struct nfs4_delegation *dp = NULL;
1da177e4
LT
4094 struct list_head *pos, *next, reaplist;
4095
1da177e4
LT
4096 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4097 while (!list_empty(&conf_id_hashtbl[i])) {
4098 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4099 expire_client(clp);
4100 }
4101 while (!list_empty(&unconf_str_hashtbl[i])) {
4102 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4103 expire_client(clp);
4104 }
4105 }
4106 INIT_LIST_HEAD(&reaplist);
4107 spin_lock(&recall_lock);
4108 list_for_each_safe(pos, next, &del_recall_lru) {
4109 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4110 list_move(&dp->dl_recall_lru, &reaplist);
4111 }
4112 spin_unlock(&recall_lock);
4113 list_for_each_safe(pos, next, &reaplist) {
4114 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4115 list_del_init(&dp->dl_recall_lru);
4116 unhash_delegation(dp);
4117 }
4118
190e4fbf 4119 nfsd4_shutdown_recdir();
1da177e4 4120 nfs4_init = 0;
1da177e4
LT
4121}
4122
4123void
4124nfs4_state_shutdown(void)
4125{
5e8d5c29
N
4126 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4127 destroy_workqueue(laundry_wq);
2c5e7615 4128 locks_end_grace(&nfsd4_manager);
1da177e4
LT
4129 nfs4_lock_state();
4130 nfs4_release_reclaim();
4131 __nfs4_state_shutdown();
b5a1a81e 4132 nfsd4_destroy_callback_queue();
1da177e4
LT
4133 nfs4_unlock_state();
4134}
4135
3dd98a3b
JL
4136/*
4137 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4138 * accessed when nfsd is starting.
4139 */
0964a3d3
N
4140static void
4141nfs4_set_recdir(char *recdir)
4142{
0964a3d3 4143 strcpy(user_recovery_dirname, recdir);
0964a3d3
N
4144}
4145
4146/*
4147 * Change the NFSv4 recovery directory to recdir.
4148 */
4149int
4150nfs4_reset_recoverydir(char *recdir)
4151{
4152 int status;
a63bb996 4153 struct path path;
0964a3d3 4154
a63bb996 4155 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N
4156 if (status)
4157 return status;
4158 status = -ENOTDIR;
a63bb996 4159 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N
4160 nfs4_set_recdir(recdir);
4161 status = 0;
4162 }
a63bb996 4163 path_put(&path);
0964a3d3
N
4164 return status;
4165}
4166
3dd98a3b
JL
4167char *
4168nfs4_recoverydir(void)
4169{
4170 return user_recovery_dirname;
4171}
This page took 0.994264 seconds and 5 git commands to generate.