]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * fs/nfs/nfs4state.c | |
3 | * | |
4 | * Client-side XDR for NFSv4. | |
5 | * | |
6 | * Copyright (c) 2002 The Regents of the University of Michigan. | |
7 | * All rights reserved. | |
8 | * | |
9 | * Kendrick Smith <kmsmith@umich.edu> | |
10 | * | |
11 | * Redistribution and use in source and binary forms, with or without | |
12 | * modification, are permitted provided that the following conditions | |
13 | * are met: | |
14 | * | |
15 | * 1. Redistributions of source code must retain the above copyright | |
16 | * notice, this list of conditions and the following disclaimer. | |
17 | * 2. Redistributions in binary form must reproduce the above copyright | |
18 | * notice, this list of conditions and the following disclaimer in the | |
19 | * documentation and/or other materials provided with the distribution. | |
20 | * 3. Neither the name of the University nor the names of its | |
21 | * contributors may be used to endorse or promote products derived | |
22 | * from this software without specific prior written permission. | |
23 | * | |
24 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED | |
25 | * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF | |
26 | * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE | |
27 | * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
28 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | |
29 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | |
30 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR | |
31 | * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF | |
32 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING | |
33 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS | |
34 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
35 | * | |
36 | * Implementation of the NFSv4 state model. For the time being, | |
37 | * this is minimal, but will be made much more complex in a | |
38 | * subsequent patch. | |
39 | */ | |
40 | ||
1da177e4 LT |
41 | #include <linux/slab.h> |
42 | #include <linux/smp_lock.h> | |
43 | #include <linux/nfs_fs.h> | |
44 | #include <linux/nfs_idmap.h> | |
5043e900 TM |
45 | #include <linux/kthread.h> |
46 | #include <linux/module.h> | |
9f958ab8 | 47 | #include <linux/random.h> |
1da177e4 LT |
48 | #include <linux/workqueue.h> |
49 | #include <linux/bitops.h> | |
50 | ||
4ce79717 | 51 | #include "nfs4_fs.h" |
1da177e4 LT |
52 | #include "callback.h" |
53 | #include "delegation.h" | |
24c8dbbb | 54 | #include "internal.h" |
1da177e4 LT |
55 | |
56 | #define OPENOWNER_POOL_SIZE 8 | |
57 | ||
4ce79717 | 58 | const nfs4_stateid zero_stateid; |
1da177e4 LT |
59 | |
60 | static LIST_HEAD(nfs4_clientid_list); | |
61 | ||
adfa6f98 | 62 | static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred) |
1da177e4 | 63 | { |
286d7d6a TM |
64 | int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, |
65 | nfs_callback_tcpport, cred); | |
1da177e4 | 66 | if (status == 0) |
286d7d6a | 67 | status = nfs4_proc_setclientid_confirm(clp, cred); |
1da177e4 LT |
68 | if (status == 0) |
69 | nfs4_schedule_state_renewal(clp); | |
70 | return status; | |
71 | } | |
72 | ||
adfa6f98 | 73 | struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) |
b4454fe1 TM |
74 | { |
75 | struct nfs4_state_owner *sp; | |
9f958ab8 | 76 | struct rb_node *pos; |
b4454fe1 TM |
77 | struct rpc_cred *cred = NULL; |
78 | ||
9f958ab8 TM |
79 | for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { |
80 | sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); | |
b4454fe1 TM |
81 | if (list_empty(&sp->so_states)) |
82 | continue; | |
83 | cred = get_rpccred(sp->so_cred); | |
84 | break; | |
85 | } | |
86 | return cred; | |
87 | } | |
88 | ||
10afec90 | 89 | static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp) |
286d7d6a TM |
90 | { |
91 | struct nfs4_state_owner *sp; | |
9f958ab8 | 92 | struct rb_node *pos; |
286d7d6a | 93 | |
9f958ab8 TM |
94 | pos = rb_first(&clp->cl_state_owners); |
95 | if (pos != NULL) { | |
96 | sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); | |
286d7d6a TM |
97 | return get_rpccred(sp->so_cred); |
98 | } | |
99 | return NULL; | |
100 | } | |
101 | ||
9f958ab8 TM |
102 | static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new, |
103 | __u64 minval, int maxbits) | |
104 | { | |
105 | struct rb_node **p, *parent; | |
106 | struct nfs_unique_id *pos; | |
107 | __u64 mask = ~0ULL; | |
108 | ||
109 | if (maxbits < 64) | |
110 | mask = (1ULL << maxbits) - 1ULL; | |
111 | ||
112 | /* Ensure distribution is more or less flat */ | |
113 | get_random_bytes(&new->id, sizeof(new->id)); | |
114 | new->id &= mask; | |
115 | if (new->id < minval) | |
116 | new->id += minval; | |
117 | retry: | |
118 | p = &root->rb_node; | |
119 | parent = NULL; | |
120 | ||
121 | while (*p != NULL) { | |
122 | parent = *p; | |
123 | pos = rb_entry(parent, struct nfs_unique_id, rb_node); | |
124 | ||
125 | if (new->id < pos->id) | |
126 | p = &(*p)->rb_left; | |
127 | else if (new->id > pos->id) | |
128 | p = &(*p)->rb_right; | |
129 | else | |
130 | goto id_exists; | |
131 | } | |
132 | rb_link_node(&new->rb_node, parent, p); | |
133 | rb_insert_color(&new->rb_node, root); | |
134 | return; | |
135 | id_exists: | |
136 | for (;;) { | |
137 | new->id++; | |
138 | if (new->id < minval || (new->id & mask) != new->id) { | |
139 | new->id = minval; | |
140 | break; | |
141 | } | |
142 | parent = rb_next(parent); | |
143 | if (parent == NULL) | |
144 | break; | |
145 | pos = rb_entry(parent, struct nfs_unique_id, rb_node); | |
146 | if (new->id < pos->id) | |
147 | break; | |
148 | } | |
149 | goto retry; | |
150 | } | |
151 | ||
152 | static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id) | |
153 | { | |
154 | rb_erase(&id->rb_node, root); | |
155 | } | |
156 | ||
1da177e4 | 157 | static struct nfs4_state_owner * |
adfa6f98 | 158 | nfs4_find_state_owner(struct nfs_client *clp, struct rpc_cred *cred) |
1da177e4 | 159 | { |
9f958ab8 TM |
160 | struct rb_node **p = &clp->cl_state_owners.rb_node, |
161 | *parent = NULL; | |
1da177e4 LT |
162 | struct nfs4_state_owner *sp, *res = NULL; |
163 | ||
9f958ab8 TM |
164 | while (*p != NULL) { |
165 | parent = *p; | |
166 | sp = rb_entry(parent, struct nfs4_state_owner, so_client_node); | |
167 | ||
168 | if (cred < sp->so_cred) | |
169 | p = &parent->rb_left; | |
170 | else if (cred > sp->so_cred) | |
171 | p = &parent->rb_right; | |
172 | else { | |
173 | atomic_inc(&sp->so_count); | |
174 | res = sp; | |
175 | break; | |
176 | } | |
1da177e4 LT |
177 | } |
178 | return res; | |
179 | } | |
180 | ||
9f958ab8 TM |
181 | static struct nfs4_state_owner * |
182 | nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new) | |
183 | { | |
184 | struct rb_node **p = &clp->cl_state_owners.rb_node, | |
185 | *parent = NULL; | |
186 | struct nfs4_state_owner *sp; | |
187 | ||
188 | while (*p != NULL) { | |
189 | parent = *p; | |
190 | sp = rb_entry(parent, struct nfs4_state_owner, so_client_node); | |
191 | ||
192 | if (new->so_cred < sp->so_cred) | |
193 | p = &parent->rb_left; | |
194 | else if (new->so_cred > sp->so_cred) | |
195 | p = &parent->rb_right; | |
196 | else { | |
197 | atomic_inc(&sp->so_count); | |
198 | return sp; | |
199 | } | |
200 | } | |
201 | nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64); | |
202 | rb_link_node(&new->so_client_node, parent, p); | |
203 | rb_insert_color(&new->so_client_node, &clp->cl_state_owners); | |
204 | return new; | |
205 | } | |
206 | ||
207 | static void | |
208 | nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp) | |
209 | { | |
210 | if (!RB_EMPTY_NODE(&sp->so_client_node)) | |
211 | rb_erase(&sp->so_client_node, &clp->cl_state_owners); | |
212 | nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id); | |
213 | } | |
214 | ||
1da177e4 LT |
215 | /* |
216 | * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to | |
217 | * create a new state_owner. | |
218 | * | |
219 | */ | |
220 | static struct nfs4_state_owner * | |
221 | nfs4_alloc_state_owner(void) | |
222 | { | |
223 | struct nfs4_state_owner *sp; | |
224 | ||
cee54fc9 | 225 | sp = kzalloc(sizeof(*sp),GFP_KERNEL); |
1da177e4 LT |
226 | if (!sp) |
227 | return NULL; | |
ec073428 | 228 | spin_lock_init(&sp->so_lock); |
1da177e4 LT |
229 | INIT_LIST_HEAD(&sp->so_states); |
230 | INIT_LIST_HEAD(&sp->so_delegations); | |
cee54fc9 TM |
231 | rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); |
232 | sp->so_seqid.sequence = &sp->so_sequence; | |
233 | spin_lock_init(&sp->so_sequence.lock); | |
234 | INIT_LIST_HEAD(&sp->so_sequence.list); | |
1da177e4 LT |
235 | atomic_set(&sp->so_count, 1); |
236 | return sp; | |
237 | } | |
238 | ||
239 | void | |
240 | nfs4_drop_state_owner(struct nfs4_state_owner *sp) | |
241 | { | |
9f958ab8 TM |
242 | if (!RB_EMPTY_NODE(&sp->so_client_node)) { |
243 | struct nfs_client *clp = sp->so_client; | |
244 | ||
245 | spin_lock(&clp->cl_lock); | |
246 | rb_erase(&sp->so_client_node, &clp->cl_state_owners); | |
247 | RB_CLEAR_NODE(&sp->so_client_node); | |
248 | spin_unlock(&clp->cl_lock); | |
249 | } | |
1da177e4 LT |
250 | } |
251 | ||
252 | /* | |
253 | * Note: must be called with clp->cl_sem held in order to prevent races | |
254 | * with reboot recovery! | |
255 | */ | |
256 | struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred) | |
257 | { | |
7539bbab | 258 | struct nfs_client *clp = server->nfs_client; |
1da177e4 LT |
259 | struct nfs4_state_owner *sp, *new; |
260 | ||
1da177e4 LT |
261 | spin_lock(&clp->cl_lock); |
262 | sp = nfs4_find_state_owner(clp, cred); | |
1da177e4 | 263 | spin_unlock(&clp->cl_lock); |
1da177e4 LT |
264 | if (sp != NULL) |
265 | return sp; | |
9f958ab8 TM |
266 | new = nfs4_alloc_state_owner(); |
267 | if (new == NULL) | |
268 | return NULL; | |
269 | new->so_client = clp; | |
270 | new->so_cred = cred; | |
271 | spin_lock(&clp->cl_lock); | |
272 | sp = nfs4_insert_state_owner(clp, new); | |
273 | spin_unlock(&clp->cl_lock); | |
274 | if (sp == new) | |
275 | get_rpccred(cred); | |
276 | else | |
277 | kfree(new); | |
278 | return sp; | |
1da177e4 LT |
279 | } |
280 | ||
281 | /* | |
282 | * Must be called with clp->cl_sem held in order to avoid races | |
283 | * with state recovery... | |
284 | */ | |
285 | void nfs4_put_state_owner(struct nfs4_state_owner *sp) | |
286 | { | |
adfa6f98 | 287 | struct nfs_client *clp = sp->so_client; |
1da177e4 LT |
288 | struct rpc_cred *cred = sp->so_cred; |
289 | ||
290 | if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) | |
291 | return; | |
9f958ab8 | 292 | nfs4_remove_state_owner(clp, sp); |
1da177e4 LT |
293 | spin_unlock(&clp->cl_lock); |
294 | put_rpccred(cred); | |
295 | kfree(sp); | |
296 | } | |
297 | ||
298 | static struct nfs4_state * | |
299 | nfs4_alloc_open_state(void) | |
300 | { | |
301 | struct nfs4_state *state; | |
302 | ||
e7616923 | 303 | state = kzalloc(sizeof(*state), GFP_KERNEL); |
1da177e4 LT |
304 | if (!state) |
305 | return NULL; | |
1da177e4 LT |
306 | atomic_set(&state->count, 1); |
307 | INIT_LIST_HEAD(&state->lock_states); | |
8d0a8a9d | 308 | spin_lock_init(&state->state_lock); |
1da177e4 LT |
309 | return state; |
310 | } | |
311 | ||
4cecb76f TM |
312 | void |
313 | nfs4_state_set_mode_locked(struct nfs4_state *state, mode_t mode) | |
314 | { | |
315 | if (state->state == mode) | |
316 | return; | |
317 | /* NB! List reordering - see the reclaim code for why. */ | |
318 | if ((mode & FMODE_WRITE) != (state->state & FMODE_WRITE)) { | |
319 | if (mode & FMODE_WRITE) | |
320 | list_move(&state->open_states, &state->owner->so_states); | |
321 | else | |
322 | list_move_tail(&state->open_states, &state->owner->so_states); | |
323 | } | |
324 | if (mode == 0) | |
325 | list_del_init(&state->inode_states); | |
326 | state->state = mode; | |
327 | } | |
328 | ||
1da177e4 LT |
329 | static struct nfs4_state * |
330 | __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner) | |
331 | { | |
332 | struct nfs_inode *nfsi = NFS_I(inode); | |
333 | struct nfs4_state *state; | |
334 | ||
335 | list_for_each_entry(state, &nfsi->open_states, inode_states) { | |
1c816efa | 336 | if (state->owner != owner) |
1da177e4 | 337 | continue; |
1c816efa | 338 | if (atomic_inc_not_zero(&state->count)) |
1da177e4 | 339 | return state; |
1da177e4 LT |
340 | } |
341 | return NULL; | |
342 | } | |
343 | ||
1da177e4 LT |
344 | static void |
345 | nfs4_free_open_state(struct nfs4_state *state) | |
346 | { | |
347 | kfree(state); | |
348 | } | |
349 | ||
350 | struct nfs4_state * | |
351 | nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner) | |
352 | { | |
353 | struct nfs4_state *state, *new; | |
354 | struct nfs_inode *nfsi = NFS_I(inode); | |
355 | ||
356 | spin_lock(&inode->i_lock); | |
357 | state = __nfs4_find_state_byowner(inode, owner); | |
358 | spin_unlock(&inode->i_lock); | |
359 | if (state) | |
360 | goto out; | |
361 | new = nfs4_alloc_open_state(); | |
ec073428 | 362 | spin_lock(&owner->so_lock); |
1da177e4 LT |
363 | spin_lock(&inode->i_lock); |
364 | state = __nfs4_find_state_byowner(inode, owner); | |
365 | if (state == NULL && new != NULL) { | |
366 | state = new; | |
1da177e4 LT |
367 | state->owner = owner; |
368 | atomic_inc(&owner->so_count); | |
369 | list_add(&state->inode_states, &nfsi->open_states); | |
370 | state->inode = igrab(inode); | |
371 | spin_unlock(&inode->i_lock); | |
ec073428 TM |
372 | /* Note: The reclaim code dictates that we add stateless |
373 | * and read-only stateids to the end of the list */ | |
374 | list_add_tail(&state->open_states, &owner->so_states); | |
375 | spin_unlock(&owner->so_lock); | |
1da177e4 LT |
376 | } else { |
377 | spin_unlock(&inode->i_lock); | |
ec073428 | 378 | spin_unlock(&owner->so_lock); |
1da177e4 LT |
379 | if (new) |
380 | nfs4_free_open_state(new); | |
381 | } | |
382 | out: | |
383 | return state; | |
384 | } | |
385 | ||
386 | /* | |
387 | * Beware! Caller must be holding exactly one | |
e6dfa553 | 388 | * reference to clp->cl_sem! |
1da177e4 LT |
389 | */ |
390 | void nfs4_put_open_state(struct nfs4_state *state) | |
391 | { | |
392 | struct inode *inode = state->inode; | |
393 | struct nfs4_state_owner *owner = state->owner; | |
394 | ||
ec073428 | 395 | if (!atomic_dec_and_lock(&state->count, &owner->so_lock)) |
1da177e4 | 396 | return; |
ec073428 | 397 | spin_lock(&inode->i_lock); |
1da177e4 LT |
398 | if (!list_empty(&state->inode_states)) |
399 | list_del(&state->inode_states); | |
1da177e4 | 400 | list_del(&state->open_states); |
ec073428 TM |
401 | spin_unlock(&inode->i_lock); |
402 | spin_unlock(&owner->so_lock); | |
1da177e4 | 403 | iput(inode); |
1da177e4 LT |
404 | nfs4_free_open_state(state); |
405 | nfs4_put_state_owner(owner); | |
406 | } | |
407 | ||
408 | /* | |
83c9d41e | 409 | * Close the current file. |
1da177e4 | 410 | */ |
4a35bd41 | 411 | void nfs4_close_state(struct path *path, struct nfs4_state *state, mode_t mode) |
1da177e4 LT |
412 | { |
413 | struct inode *inode = state->inode; | |
414 | struct nfs4_state_owner *owner = state->owner; | |
4cecb76f | 415 | int oldstate, newstate = 0; |
1da177e4 LT |
416 | |
417 | atomic_inc(&owner->so_count); | |
1da177e4 | 418 | /* Protect against nfs4_find_state() */ |
ec073428 | 419 | spin_lock(&owner->so_lock); |
1da177e4 | 420 | spin_lock(&inode->i_lock); |
e7616923 TM |
421 | switch (mode & (FMODE_READ | FMODE_WRITE)) { |
422 | case FMODE_READ: | |
423 | state->n_rdonly--; | |
424 | break; | |
425 | case FMODE_WRITE: | |
426 | state->n_wronly--; | |
427 | break; | |
428 | case FMODE_READ|FMODE_WRITE: | |
429 | state->n_rdwr--; | |
430 | } | |
4cecb76f | 431 | oldstate = newstate = state->state; |
e7616923 TM |
432 | if (state->n_rdwr == 0) { |
433 | if (state->n_rdonly == 0) | |
434 | newstate &= ~FMODE_READ; | |
435 | if (state->n_wronly == 0) | |
436 | newstate &= ~FMODE_WRITE; | |
437 | } | |
4cecb76f TM |
438 | if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { |
439 | nfs4_state_set_mode_locked(state, newstate); | |
440 | oldstate = newstate; | |
1da177e4 LT |
441 | } |
442 | spin_unlock(&inode->i_lock); | |
ec073428 | 443 | spin_unlock(&owner->so_lock); |
4cecb76f | 444 | |
b39e625b TM |
445 | if (oldstate == newstate) { |
446 | nfs4_put_open_state(state); | |
447 | nfs4_put_state_owner(owner); | |
448 | } else | |
449 | nfs4_do_close(path, state); | |
1da177e4 LT |
450 | } |
451 | ||
452 | /* | |
453 | * Search the state->lock_states for an existing lock_owner | |
454 | * that is compatible with current->files | |
455 | */ | |
456 | static struct nfs4_lock_state * | |
457 | __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) | |
458 | { | |
459 | struct nfs4_lock_state *pos; | |
460 | list_for_each_entry(pos, &state->lock_states, ls_locks) { | |
461 | if (pos->ls_owner != fl_owner) | |
462 | continue; | |
463 | atomic_inc(&pos->ls_count); | |
464 | return pos; | |
465 | } | |
466 | return NULL; | |
467 | } | |
468 | ||
1da177e4 LT |
469 | /* |
470 | * Return a compatible lock_state. If no initialized lock_state structure | |
471 | * exists, return an uninitialized one. | |
472 | * | |
1da177e4 LT |
473 | */ |
474 | static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) | |
475 | { | |
476 | struct nfs4_lock_state *lsp; | |
adfa6f98 | 477 | struct nfs_client *clp = state->owner->so_client; |
1da177e4 | 478 | |
cee54fc9 | 479 | lsp = kzalloc(sizeof(*lsp), GFP_KERNEL); |
1da177e4 LT |
480 | if (lsp == NULL) |
481 | return NULL; | |
cee54fc9 | 482 | lsp->ls_seqid.sequence = &state->owner->so_sequence; |
1da177e4 LT |
483 | atomic_set(&lsp->ls_count, 1); |
484 | lsp->ls_owner = fl_owner; | |
1da177e4 | 485 | spin_lock(&clp->cl_lock); |
9f958ab8 | 486 | nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64); |
1da177e4 | 487 | spin_unlock(&clp->cl_lock); |
8d0a8a9d | 488 | INIT_LIST_HEAD(&lsp->ls_locks); |
1da177e4 LT |
489 | return lsp; |
490 | } | |
491 | ||
9f958ab8 TM |
492 | static void nfs4_free_lock_state(struct nfs4_lock_state *lsp) |
493 | { | |
494 | struct nfs_client *clp = lsp->ls_state->owner->so_client; | |
495 | ||
496 | spin_lock(&clp->cl_lock); | |
497 | nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id); | |
498 | spin_unlock(&clp->cl_lock); | |
499 | kfree(lsp); | |
500 | } | |
501 | ||
1da177e4 LT |
502 | /* |
503 | * Return a compatible lock_state. If no initialized lock_state structure | |
504 | * exists, return an uninitialized one. | |
505 | * | |
e6dfa553 | 506 | * The caller must be holding clp->cl_sem |
1da177e4 | 507 | */ |
8d0a8a9d | 508 | static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) |
1da177e4 | 509 | { |
8d0a8a9d | 510 | struct nfs4_lock_state *lsp, *new = NULL; |
1da177e4 | 511 | |
8d0a8a9d TM |
512 | for(;;) { |
513 | spin_lock(&state->state_lock); | |
514 | lsp = __nfs4_find_lock_state(state, owner); | |
515 | if (lsp != NULL) | |
516 | break; | |
517 | if (new != NULL) { | |
518 | new->ls_state = state; | |
519 | list_add(&new->ls_locks, &state->lock_states); | |
520 | set_bit(LK_STATE_IN_USE, &state->flags); | |
521 | lsp = new; | |
522 | new = NULL; | |
523 | break; | |
524 | } | |
525 | spin_unlock(&state->state_lock); | |
526 | new = nfs4_alloc_lock_state(state, owner); | |
527 | if (new == NULL) | |
528 | return NULL; | |
529 | } | |
530 | spin_unlock(&state->state_lock); | |
9f958ab8 TM |
531 | if (new != NULL) |
532 | nfs4_free_lock_state(new); | |
1da177e4 LT |
533 | return lsp; |
534 | } | |
535 | ||
536 | /* | |
8d0a8a9d TM |
537 | * Release reference to lock_state, and free it if we see that |
538 | * it is no longer in use | |
1da177e4 | 539 | */ |
faf5f49c | 540 | void nfs4_put_lock_state(struct nfs4_lock_state *lsp) |
1da177e4 | 541 | { |
8d0a8a9d | 542 | struct nfs4_state *state; |
1da177e4 | 543 | |
8d0a8a9d TM |
544 | if (lsp == NULL) |
545 | return; | |
546 | state = lsp->ls_state; | |
547 | if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock)) | |
548 | return; | |
549 | list_del(&lsp->ls_locks); | |
550 | if (list_empty(&state->lock_states)) | |
551 | clear_bit(LK_STATE_IN_USE, &state->flags); | |
552 | spin_unlock(&state->state_lock); | |
9f958ab8 | 553 | nfs4_free_lock_state(lsp); |
1da177e4 LT |
554 | } |
555 | ||
8d0a8a9d | 556 | static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src) |
1da177e4 | 557 | { |
8d0a8a9d | 558 | struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner; |
1da177e4 | 559 | |
8d0a8a9d TM |
560 | dst->fl_u.nfs4_fl.owner = lsp; |
561 | atomic_inc(&lsp->ls_count); | |
562 | } | |
1da177e4 | 563 | |
8d0a8a9d | 564 | static void nfs4_fl_release_lock(struct file_lock *fl) |
1da177e4 | 565 | { |
8d0a8a9d | 566 | nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner); |
1da177e4 LT |
567 | } |
568 | ||
8d0a8a9d TM |
569 | static struct file_lock_operations nfs4_fl_lock_ops = { |
570 | .fl_copy_lock = nfs4_fl_copy_lock, | |
571 | .fl_release_private = nfs4_fl_release_lock, | |
572 | }; | |
573 | ||
574 | int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl) | |
1da177e4 | 575 | { |
8d0a8a9d TM |
576 | struct nfs4_lock_state *lsp; |
577 | ||
578 | if (fl->fl_ops != NULL) | |
579 | return 0; | |
580 | lsp = nfs4_get_lock_state(state, fl->fl_owner); | |
581 | if (lsp == NULL) | |
582 | return -ENOMEM; | |
583 | fl->fl_u.nfs4_fl.owner = lsp; | |
584 | fl->fl_ops = &nfs4_fl_lock_ops; | |
585 | return 0; | |
1da177e4 LT |
586 | } |
587 | ||
8d0a8a9d TM |
588 | /* |
589 | * Byte-range lock aware utility to initialize the stateid of read/write | |
590 | * requests. | |
1da177e4 | 591 | */ |
8d0a8a9d | 592 | void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner) |
1da177e4 | 593 | { |
8d0a8a9d | 594 | struct nfs4_lock_state *lsp; |
1da177e4 | 595 | |
8d0a8a9d TM |
596 | memcpy(dst, &state->stateid, sizeof(*dst)); |
597 | if (test_bit(LK_STATE_IN_USE, &state->flags) == 0) | |
598 | return; | |
1da177e4 | 599 | |
8d0a8a9d TM |
600 | spin_lock(&state->state_lock); |
601 | lsp = __nfs4_find_lock_state(state, fl_owner); | |
602 | if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) | |
603 | memcpy(dst, &lsp->ls_stateid, sizeof(*dst)); | |
604 | spin_unlock(&state->state_lock); | |
1da177e4 LT |
605 | nfs4_put_lock_state(lsp); |
606 | } | |
607 | ||
cee54fc9 TM |
608 | struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) |
609 | { | |
36f20c6d | 610 | struct rpc_sequence *sequence = counter->sequence; |
cee54fc9 TM |
611 | struct nfs_seqid *new; |
612 | ||
613 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
614 | if (new != NULL) { | |
615 | new->sequence = counter; | |
36f20c6d TM |
616 | spin_lock(&sequence->lock); |
617 | list_add_tail(&new->list, &sequence->list); | |
618 | spin_unlock(&sequence->lock); | |
cee54fc9 TM |
619 | } |
620 | return new; | |
621 | } | |
622 | ||
623 | void nfs_free_seqid(struct nfs_seqid *seqid) | |
1da177e4 | 624 | { |
cee54fc9 | 625 | struct rpc_sequence *sequence = seqid->sequence->sequence; |
cee54fc9 | 626 | |
36f20c6d TM |
627 | spin_lock(&sequence->lock); |
628 | list_del(&seqid->list); | |
629 | spin_unlock(&sequence->lock); | |
630 | rpc_wake_up(&sequence->wait); | |
cee54fc9 | 631 | kfree(seqid); |
1da177e4 LT |
632 | } |
633 | ||
634 | /* | |
cee54fc9 TM |
635 | * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or |
636 | * failed with a seqid incrementing error - | |
637 | * see comments nfs_fs.h:seqid_mutating_error() | |
638 | */ | |
88d90939 | 639 | static void nfs_increment_seqid(int status, struct nfs_seqid *seqid) |
cee54fc9 TM |
640 | { |
641 | switch (status) { | |
642 | case 0: | |
643 | break; | |
644 | case -NFS4ERR_BAD_SEQID: | |
645 | case -NFS4ERR_STALE_CLIENTID: | |
646 | case -NFS4ERR_STALE_STATEID: | |
647 | case -NFS4ERR_BAD_STATEID: | |
648 | case -NFS4ERR_BADXDR: | |
649 | case -NFS4ERR_RESOURCE: | |
650 | case -NFS4ERR_NOFILEHANDLE: | |
651 | /* Non-seqid mutating errors */ | |
652 | return; | |
653 | }; | |
654 | /* | |
655 | * Note: no locking needed as we are guaranteed to be first | |
656 | * on the sequence list | |
657 | */ | |
658 | seqid->sequence->counter++; | |
659 | } | |
660 | ||
661 | void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) | |
662 | { | |
663 | if (status == -NFS4ERR_BAD_SEQID) { | |
664 | struct nfs4_state_owner *sp = container_of(seqid->sequence, | |
665 | struct nfs4_state_owner, so_seqid); | |
1da177e4 | 666 | nfs4_drop_state_owner(sp); |
cee54fc9 | 667 | } |
88d90939 | 668 | nfs_increment_seqid(status, seqid); |
cee54fc9 TM |
669 | } |
670 | ||
671 | /* | |
cee54fc9 TM |
672 | * Increment the seqid if the LOCK/LOCKU succeeded, or |
673 | * failed with a seqid incrementing error - | |
674 | * see comments nfs_fs.h:seqid_mutating_error() | |
675 | */ | |
676 | void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid) | |
677 | { | |
88d90939 | 678 | nfs_increment_seqid(status, seqid); |
cee54fc9 TM |
679 | } |
680 | ||
681 | int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) | |
682 | { | |
683 | struct rpc_sequence *sequence = seqid->sequence->sequence; | |
684 | int status = 0; | |
685 | ||
4e51336a TM |
686 | if (sequence->list.next == &seqid->list) |
687 | goto out; | |
cee54fc9 | 688 | spin_lock(&sequence->lock); |
36f20c6d | 689 | if (sequence->list.next != &seqid->list) { |
cee54fc9 TM |
690 | rpc_sleep_on(&sequence->wait, task, NULL, NULL); |
691 | status = -EAGAIN; | |
36f20c6d | 692 | } |
cee54fc9 | 693 | spin_unlock(&sequence->lock); |
4e51336a | 694 | out: |
cee54fc9 | 695 | return status; |
1da177e4 LT |
696 | } |
697 | ||
698 | static int reclaimer(void *); | |
1da177e4 | 699 | |
adfa6f98 | 700 | static inline void nfs4_clear_recover_bit(struct nfs_client *clp) |
433fbe4c TM |
701 | { |
702 | smp_mb__before_clear_bit(); | |
703 | clear_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state); | |
704 | smp_mb__after_clear_bit(); | |
705 | wake_up_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER); | |
706 | rpc_wake_up(&clp->cl_rpcwaitq); | |
707 | } | |
708 | ||
1da177e4 LT |
709 | /* |
710 | * State recovery routine | |
711 | */ | |
adfa6f98 | 712 | static void nfs4_recover_state(struct nfs_client *clp) |
1da177e4 | 713 | { |
5043e900 | 714 | struct task_struct *task; |
1da177e4 | 715 | |
5043e900 TM |
716 | __module_get(THIS_MODULE); |
717 | atomic_inc(&clp->cl_count); | |
718 | task = kthread_run(reclaimer, clp, "%u.%u.%u.%u-reclaim", | |
24c8dbbb | 719 | NIPQUAD(clp->cl_addr.sin_addr)); |
5043e900 TM |
720 | if (!IS_ERR(task)) |
721 | return; | |
433fbe4c | 722 | nfs4_clear_recover_bit(clp); |
24c8dbbb | 723 | nfs_put_client(clp); |
5043e900 | 724 | module_put(THIS_MODULE); |
1da177e4 LT |
725 | } |
726 | ||
727 | /* | |
728 | * Schedule a state recovery attempt | |
729 | */ | |
adfa6f98 | 730 | void nfs4_schedule_state_recovery(struct nfs_client *clp) |
1da177e4 LT |
731 | { |
732 | if (!clp) | |
733 | return; | |
433fbe4c | 734 | if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0) |
5043e900 | 735 | nfs4_recover_state(clp); |
1da177e4 LT |
736 | } |
737 | ||
738 | static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_state *state) | |
739 | { | |
740 | struct inode *inode = state->inode; | |
741 | struct file_lock *fl; | |
742 | int status = 0; | |
743 | ||
744 | for (fl = inode->i_flock; fl != 0; fl = fl->fl_next) { | |
43b2a33a | 745 | if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK))) |
1da177e4 LT |
746 | continue; |
747 | if (((struct nfs_open_context *)fl->fl_file->private_data)->state != state) | |
748 | continue; | |
749 | status = ops->recover_lock(state, fl); | |
750 | if (status >= 0) | |
751 | continue; | |
752 | switch (status) { | |
753 | default: | |
754 | printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n", | |
755 | __FUNCTION__, status); | |
756 | case -NFS4ERR_EXPIRED: | |
757 | case -NFS4ERR_NO_GRACE: | |
758 | case -NFS4ERR_RECLAIM_BAD: | |
759 | case -NFS4ERR_RECLAIM_CONFLICT: | |
43b2a33a | 760 | /* kill_proc(fl->fl_pid, SIGLOST, 1); */ |
1da177e4 LT |
761 | break; |
762 | case -NFS4ERR_STALE_CLIENTID: | |
763 | goto out_err; | |
764 | } | |
765 | } | |
766 | return 0; | |
767 | out_err: | |
768 | return status; | |
769 | } | |
770 | ||
771 | static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct nfs4_state_owner *sp) | |
772 | { | |
773 | struct nfs4_state *state; | |
774 | struct nfs4_lock_state *lock; | |
775 | int status = 0; | |
776 | ||
777 | /* Note: we rely on the sp->so_states list being ordered | |
778 | * so that we always reclaim open(O_RDWR) and/or open(O_WRITE) | |
779 | * states first. | |
780 | * This is needed to ensure that the server won't give us any | |
781 | * read delegations that we have to return if, say, we are | |
782 | * recovering after a network partition or a reboot from a | |
783 | * server that doesn't support a grace period. | |
784 | */ | |
785 | list_for_each_entry(state, &sp->so_states, open_states) { | |
786 | if (state->state == 0) | |
787 | continue; | |
788 | status = ops->recover_open(sp, state); | |
1da177e4 LT |
789 | if (status >= 0) { |
790 | status = nfs4_reclaim_locks(ops, state); | |
791 | if (status < 0) | |
792 | goto out_err; | |
793 | list_for_each_entry(lock, &state->lock_states, ls_locks) { | |
794 | if (!(lock->ls_flags & NFS_LOCK_INITIALIZED)) | |
795 | printk("%s: Lock reclaim failed!\n", | |
796 | __FUNCTION__); | |
797 | } | |
798 | continue; | |
799 | } | |
800 | switch (status) { | |
801 | default: | |
802 | printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n", | |
803 | __FUNCTION__, status); | |
804 | case -ENOENT: | |
805 | case -NFS4ERR_RECLAIM_BAD: | |
806 | case -NFS4ERR_RECLAIM_CONFLICT: | |
807 | /* | |
808 | * Open state on this file cannot be recovered | |
809 | * All we can do is revert to using the zero stateid. | |
810 | */ | |
811 | memset(state->stateid.data, 0, | |
812 | sizeof(state->stateid.data)); | |
813 | /* Mark the file as being 'closed' */ | |
814 | state->state = 0; | |
815 | break; | |
816 | case -NFS4ERR_EXPIRED: | |
817 | case -NFS4ERR_NO_GRACE: | |
818 | case -NFS4ERR_STALE_CLIENTID: | |
819 | goto out_err; | |
820 | } | |
821 | } | |
822 | return 0; | |
823 | out_err: | |
824 | return status; | |
825 | } | |
826 | ||
adfa6f98 | 827 | static void nfs4_state_mark_reclaim(struct nfs_client *clp) |
cee54fc9 TM |
828 | { |
829 | struct nfs4_state_owner *sp; | |
9f958ab8 | 830 | struct rb_node *pos; |
cee54fc9 TM |
831 | struct nfs4_state *state; |
832 | struct nfs4_lock_state *lock; | |
833 | ||
834 | /* Reset all sequence ids to zero */ | |
9f958ab8 TM |
835 | for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { |
836 | sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); | |
cee54fc9 TM |
837 | sp->so_seqid.counter = 0; |
838 | sp->so_seqid.flags = 0; | |
ec073428 | 839 | spin_lock(&sp->so_lock); |
cee54fc9 TM |
840 | list_for_each_entry(state, &sp->so_states, open_states) { |
841 | list_for_each_entry(lock, &state->lock_states, ls_locks) { | |
842 | lock->ls_seqid.counter = 0; | |
843 | lock->ls_seqid.flags = 0; | |
844 | lock->ls_flags &= ~NFS_LOCK_INITIALIZED; | |
845 | } | |
846 | } | |
ec073428 | 847 | spin_unlock(&sp->so_lock); |
cee54fc9 TM |
848 | } |
849 | } | |
850 | ||
1da177e4 LT |
851 | static int reclaimer(void *ptr) |
852 | { | |
adfa6f98 | 853 | struct nfs_client *clp = ptr; |
1da177e4 | 854 | struct nfs4_state_owner *sp; |
9f958ab8 | 855 | struct rb_node *pos; |
1da177e4 | 856 | struct nfs4_state_recovery_ops *ops; |
286d7d6a | 857 | struct rpc_cred *cred; |
1da177e4 LT |
858 | int status = 0; |
859 | ||
1da177e4 LT |
860 | allow_signal(SIGKILL); |
861 | ||
1da177e4 LT |
862 | /* Ensure exclusive access to NFSv4 state */ |
863 | lock_kernel(); | |
864 | down_write(&clp->cl_sem); | |
865 | /* Are there any NFS mounts out there? */ | |
866 | if (list_empty(&clp->cl_superblocks)) | |
867 | goto out; | |
868 | restart_loop: | |
286d7d6a TM |
869 | ops = &nfs4_network_partition_recovery_ops; |
870 | /* Are there any open files on this volume? */ | |
871 | cred = nfs4_get_renew_cred(clp); | |
872 | if (cred != NULL) { | |
873 | /* Yes there are: try to renew the old lease */ | |
874 | status = nfs4_proc_renew(clp, cred); | |
875 | switch (status) { | |
876 | case 0: | |
877 | case -NFS4ERR_CB_PATH_DOWN: | |
878 | put_rpccred(cred); | |
879 | goto out; | |
880 | case -NFS4ERR_STALE_CLIENTID: | |
881 | case -NFS4ERR_LEASE_MOVED: | |
882 | ops = &nfs4_reboot_recovery_ops; | |
883 | } | |
884 | } else { | |
885 | /* "reboot" to ensure we clear all state on the server */ | |
886 | clp->cl_boot_time = CURRENT_TIME; | |
887 | cred = nfs4_get_setclientid_cred(clp); | |
888 | } | |
889 | /* We're going to have to re-establish a clientid */ | |
cee54fc9 | 890 | nfs4_state_mark_reclaim(clp); |
286d7d6a TM |
891 | status = -ENOENT; |
892 | if (cred != NULL) { | |
893 | status = nfs4_init_client(clp, cred); | |
894 | put_rpccred(cred); | |
895 | } | |
1da177e4 LT |
896 | if (status) |
897 | goto out_error; | |
898 | /* Mark all delegations for reclaim */ | |
899 | nfs_delegation_mark_reclaim(clp); | |
900 | /* Note: list is protected by exclusive lock on cl->cl_sem */ | |
9f958ab8 TM |
901 | for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { |
902 | sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); | |
1da177e4 LT |
903 | status = nfs4_reclaim_open_state(ops, sp); |
904 | if (status < 0) { | |
905 | if (status == -NFS4ERR_NO_GRACE) { | |
906 | ops = &nfs4_network_partition_recovery_ops; | |
907 | status = nfs4_reclaim_open_state(ops, sp); | |
908 | } | |
909 | if (status == -NFS4ERR_STALE_CLIENTID) | |
910 | goto restart_loop; | |
911 | if (status == -NFS4ERR_EXPIRED) | |
912 | goto restart_loop; | |
913 | } | |
914 | } | |
915 | nfs_delegation_reap_unclaimed(clp); | |
916 | out: | |
1da177e4 LT |
917 | up_write(&clp->cl_sem); |
918 | unlock_kernel(); | |
1da177e4 LT |
919 | if (status == -NFS4ERR_CB_PATH_DOWN) |
920 | nfs_handle_cb_pathdown(clp); | |
433fbe4c | 921 | nfs4_clear_recover_bit(clp); |
24c8dbbb | 922 | nfs_put_client(clp); |
5043e900 | 923 | module_put_and_exit(0); |
1da177e4 LT |
924 | return 0; |
925 | out_error: | |
926 | printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n", | |
24c8dbbb | 927 | NIPQUAD(clp->cl_addr.sin_addr), -status); |
51581f3b | 928 | set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state); |
1da177e4 LT |
929 | goto out; |
930 | } | |
931 | ||
932 | /* | |
933 | * Local variables: | |
934 | * c-basic-offset: 8 | |
935 | * End: | |
936 | */ |