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> | |
1da177e4 LT |
47 | #include <linux/workqueue.h> |
48 | #include <linux/bitops.h> | |
49 | ||
4ce79717 | 50 | #include "nfs4_fs.h" |
1da177e4 LT |
51 | #include "callback.h" |
52 | #include "delegation.h" | |
24c8dbbb | 53 | #include "internal.h" |
1da177e4 LT |
54 | |
55 | #define OPENOWNER_POOL_SIZE 8 | |
56 | ||
4ce79717 | 57 | const nfs4_stateid zero_stateid; |
1da177e4 LT |
58 | |
59 | static LIST_HEAD(nfs4_clientid_list); | |
60 | ||
adfa6f98 | 61 | static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred) |
1da177e4 | 62 | { |
286d7d6a TM |
63 | int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, |
64 | nfs_callback_tcpport, cred); | |
1da177e4 | 65 | if (status == 0) |
286d7d6a | 66 | status = nfs4_proc_setclientid_confirm(clp, cred); |
1da177e4 LT |
67 | if (status == 0) |
68 | nfs4_schedule_state_renewal(clp); | |
69 | return status; | |
70 | } | |
71 | ||
1da177e4 | 72 | u32 |
adfa6f98 | 73 | nfs4_alloc_lockowner_id(struct nfs_client *clp) |
1da177e4 LT |
74 | { |
75 | return clp->cl_lockowner_id ++; | |
76 | } | |
77 | ||
adfa6f98 | 78 | struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) |
b4454fe1 TM |
79 | { |
80 | struct nfs4_state_owner *sp; | |
81 | struct rpc_cred *cred = NULL; | |
82 | ||
83 | list_for_each_entry(sp, &clp->cl_state_owners, so_list) { | |
84 | if (list_empty(&sp->so_states)) | |
85 | continue; | |
86 | cred = get_rpccred(sp->so_cred); | |
87 | break; | |
88 | } | |
89 | return cred; | |
90 | } | |
91 | ||
10afec90 | 92 | static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp) |
286d7d6a TM |
93 | { |
94 | struct nfs4_state_owner *sp; | |
95 | ||
96 | if (!list_empty(&clp->cl_state_owners)) { | |
97 | sp = list_entry(clp->cl_state_owners.next, | |
98 | struct nfs4_state_owner, so_list); | |
99 | return get_rpccred(sp->so_cred); | |
100 | } | |
101 | return NULL; | |
102 | } | |
103 | ||
1da177e4 | 104 | static struct nfs4_state_owner * |
adfa6f98 | 105 | nfs4_find_state_owner(struct nfs_client *clp, struct rpc_cred *cred) |
1da177e4 LT |
106 | { |
107 | struct nfs4_state_owner *sp, *res = NULL; | |
108 | ||
109 | list_for_each_entry(sp, &clp->cl_state_owners, so_list) { | |
110 | if (sp->so_cred != cred) | |
111 | continue; | |
112 | atomic_inc(&sp->so_count); | |
113 | /* Move to the head of the list */ | |
114 | list_move(&sp->so_list, &clp->cl_state_owners); | |
115 | res = sp; | |
116 | break; | |
117 | } | |
118 | return res; | |
119 | } | |
120 | ||
121 | /* | |
122 | * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to | |
123 | * create a new state_owner. | |
124 | * | |
125 | */ | |
126 | static struct nfs4_state_owner * | |
127 | nfs4_alloc_state_owner(void) | |
128 | { | |
129 | struct nfs4_state_owner *sp; | |
130 | ||
cee54fc9 | 131 | sp = kzalloc(sizeof(*sp),GFP_KERNEL); |
1da177e4 LT |
132 | if (!sp) |
133 | return NULL; | |
ec073428 | 134 | spin_lock_init(&sp->so_lock); |
1da177e4 LT |
135 | INIT_LIST_HEAD(&sp->so_states); |
136 | INIT_LIST_HEAD(&sp->so_delegations); | |
cee54fc9 TM |
137 | rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); |
138 | sp->so_seqid.sequence = &sp->so_sequence; | |
139 | spin_lock_init(&sp->so_sequence.lock); | |
140 | INIT_LIST_HEAD(&sp->so_sequence.list); | |
1da177e4 LT |
141 | atomic_set(&sp->so_count, 1); |
142 | return sp; | |
143 | } | |
144 | ||
145 | void | |
146 | nfs4_drop_state_owner(struct nfs4_state_owner *sp) | |
147 | { | |
adfa6f98 | 148 | struct nfs_client *clp = sp->so_client; |
1da177e4 LT |
149 | spin_lock(&clp->cl_lock); |
150 | list_del_init(&sp->so_list); | |
151 | spin_unlock(&clp->cl_lock); | |
152 | } | |
153 | ||
154 | /* | |
155 | * Note: must be called with clp->cl_sem held in order to prevent races | |
156 | * with reboot recovery! | |
157 | */ | |
158 | struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred) | |
159 | { | |
7539bbab | 160 | struct nfs_client *clp = server->nfs_client; |
1da177e4 LT |
161 | struct nfs4_state_owner *sp, *new; |
162 | ||
1da177e4 LT |
163 | new = nfs4_alloc_state_owner(); |
164 | spin_lock(&clp->cl_lock); | |
165 | sp = nfs4_find_state_owner(clp, cred); | |
1da177e4 LT |
166 | if (sp == NULL && new != NULL) { |
167 | list_add(&new->so_list, &clp->cl_state_owners); | |
168 | new->so_client = clp; | |
169 | new->so_id = nfs4_alloc_lockowner_id(clp); | |
27b3f949 | 170 | new->so_cred = get_rpccred(cred); |
1da177e4 LT |
171 | sp = new; |
172 | new = NULL; | |
173 | } | |
174 | spin_unlock(&clp->cl_lock); | |
f99d49ad | 175 | kfree(new); |
1da177e4 LT |
176 | if (sp != NULL) |
177 | return sp; | |
1da177e4 LT |
178 | return NULL; |
179 | } | |
180 | ||
181 | /* | |
182 | * Must be called with clp->cl_sem held in order to avoid races | |
183 | * with state recovery... | |
184 | */ | |
185 | void nfs4_put_state_owner(struct nfs4_state_owner *sp) | |
186 | { | |
adfa6f98 | 187 | struct nfs_client *clp = sp->so_client; |
1da177e4 LT |
188 | struct rpc_cred *cred = sp->so_cred; |
189 | ||
190 | if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) | |
191 | return; | |
1da177e4 LT |
192 | list_del(&sp->so_list); |
193 | spin_unlock(&clp->cl_lock); | |
194 | put_rpccred(cred); | |
195 | kfree(sp); | |
196 | } | |
197 | ||
198 | static struct nfs4_state * | |
199 | nfs4_alloc_open_state(void) | |
200 | { | |
201 | struct nfs4_state *state; | |
202 | ||
e7616923 | 203 | state = kzalloc(sizeof(*state), GFP_KERNEL); |
1da177e4 LT |
204 | if (!state) |
205 | return NULL; | |
1da177e4 LT |
206 | atomic_set(&state->count, 1); |
207 | INIT_LIST_HEAD(&state->lock_states); | |
8d0a8a9d | 208 | spin_lock_init(&state->state_lock); |
1da177e4 LT |
209 | return state; |
210 | } | |
211 | ||
4cecb76f TM |
212 | void |
213 | nfs4_state_set_mode_locked(struct nfs4_state *state, mode_t mode) | |
214 | { | |
215 | if (state->state == mode) | |
216 | return; | |
217 | /* NB! List reordering - see the reclaim code for why. */ | |
218 | if ((mode & FMODE_WRITE) != (state->state & FMODE_WRITE)) { | |
219 | if (mode & FMODE_WRITE) | |
220 | list_move(&state->open_states, &state->owner->so_states); | |
221 | else | |
222 | list_move_tail(&state->open_states, &state->owner->so_states); | |
223 | } | |
224 | if (mode == 0) | |
225 | list_del_init(&state->inode_states); | |
226 | state->state = mode; | |
227 | } | |
228 | ||
1da177e4 LT |
229 | static struct nfs4_state * |
230 | __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner) | |
231 | { | |
232 | struct nfs_inode *nfsi = NFS_I(inode); | |
233 | struct nfs4_state *state; | |
234 | ||
235 | list_for_each_entry(state, &nfsi->open_states, inode_states) { | |
236 | /* Is this in the process of being freed? */ | |
4cecb76f | 237 | if (state->state == 0) |
1da177e4 LT |
238 | continue; |
239 | if (state->owner == owner) { | |
240 | atomic_inc(&state->count); | |
241 | return state; | |
242 | } | |
243 | } | |
244 | return NULL; | |
245 | } | |
246 | ||
1da177e4 LT |
247 | static void |
248 | nfs4_free_open_state(struct nfs4_state *state) | |
249 | { | |
250 | kfree(state); | |
251 | } | |
252 | ||
253 | struct nfs4_state * | |
254 | nfs4_get_open_state(struct inode *inode, struct nfs4_state_owner *owner) | |
255 | { | |
256 | struct nfs4_state *state, *new; | |
257 | struct nfs_inode *nfsi = NFS_I(inode); | |
258 | ||
259 | spin_lock(&inode->i_lock); | |
260 | state = __nfs4_find_state_byowner(inode, owner); | |
261 | spin_unlock(&inode->i_lock); | |
262 | if (state) | |
263 | goto out; | |
264 | new = nfs4_alloc_open_state(); | |
ec073428 | 265 | spin_lock(&owner->so_lock); |
1da177e4 LT |
266 | spin_lock(&inode->i_lock); |
267 | state = __nfs4_find_state_byowner(inode, owner); | |
268 | if (state == NULL && new != NULL) { | |
269 | state = new; | |
1da177e4 LT |
270 | state->owner = owner; |
271 | atomic_inc(&owner->so_count); | |
272 | list_add(&state->inode_states, &nfsi->open_states); | |
273 | state->inode = igrab(inode); | |
274 | spin_unlock(&inode->i_lock); | |
ec073428 TM |
275 | /* Note: The reclaim code dictates that we add stateless |
276 | * and read-only stateids to the end of the list */ | |
277 | list_add_tail(&state->open_states, &owner->so_states); | |
278 | spin_unlock(&owner->so_lock); | |
1da177e4 LT |
279 | } else { |
280 | spin_unlock(&inode->i_lock); | |
ec073428 | 281 | spin_unlock(&owner->so_lock); |
1da177e4 LT |
282 | if (new) |
283 | nfs4_free_open_state(new); | |
284 | } | |
285 | out: | |
286 | return state; | |
287 | } | |
288 | ||
289 | /* | |
290 | * Beware! Caller must be holding exactly one | |
e6dfa553 | 291 | * reference to clp->cl_sem! |
1da177e4 LT |
292 | */ |
293 | void nfs4_put_open_state(struct nfs4_state *state) | |
294 | { | |
295 | struct inode *inode = state->inode; | |
296 | struct nfs4_state_owner *owner = state->owner; | |
297 | ||
ec073428 | 298 | if (!atomic_dec_and_lock(&state->count, &owner->so_lock)) |
1da177e4 | 299 | return; |
ec073428 | 300 | spin_lock(&inode->i_lock); |
1da177e4 LT |
301 | if (!list_empty(&state->inode_states)) |
302 | list_del(&state->inode_states); | |
1da177e4 | 303 | list_del(&state->open_states); |
ec073428 TM |
304 | spin_unlock(&inode->i_lock); |
305 | spin_unlock(&owner->so_lock); | |
1da177e4 | 306 | iput(inode); |
1da177e4 LT |
307 | nfs4_free_open_state(state); |
308 | nfs4_put_state_owner(owner); | |
309 | } | |
310 | ||
311 | /* | |
83c9d41e | 312 | * Close the current file. |
1da177e4 | 313 | */ |
4a35bd41 | 314 | void nfs4_close_state(struct path *path, struct nfs4_state *state, mode_t mode) |
1da177e4 LT |
315 | { |
316 | struct inode *inode = state->inode; | |
317 | struct nfs4_state_owner *owner = state->owner; | |
4cecb76f | 318 | int oldstate, newstate = 0; |
1da177e4 LT |
319 | |
320 | atomic_inc(&owner->so_count); | |
1da177e4 | 321 | /* Protect against nfs4_find_state() */ |
ec073428 | 322 | spin_lock(&owner->so_lock); |
1da177e4 | 323 | spin_lock(&inode->i_lock); |
e7616923 TM |
324 | switch (mode & (FMODE_READ | FMODE_WRITE)) { |
325 | case FMODE_READ: | |
326 | state->n_rdonly--; | |
327 | break; | |
328 | case FMODE_WRITE: | |
329 | state->n_wronly--; | |
330 | break; | |
331 | case FMODE_READ|FMODE_WRITE: | |
332 | state->n_rdwr--; | |
333 | } | |
4cecb76f | 334 | oldstate = newstate = state->state; |
e7616923 TM |
335 | if (state->n_rdwr == 0) { |
336 | if (state->n_rdonly == 0) | |
337 | newstate &= ~FMODE_READ; | |
338 | if (state->n_wronly == 0) | |
339 | newstate &= ~FMODE_WRITE; | |
340 | } | |
4cecb76f TM |
341 | if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { |
342 | nfs4_state_set_mode_locked(state, newstate); | |
343 | oldstate = newstate; | |
1da177e4 LT |
344 | } |
345 | spin_unlock(&inode->i_lock); | |
ec073428 | 346 | spin_unlock(&owner->so_lock); |
4cecb76f | 347 | |
b39e625b TM |
348 | if (oldstate == newstate) { |
349 | nfs4_put_open_state(state); | |
350 | nfs4_put_state_owner(owner); | |
351 | } else | |
352 | nfs4_do_close(path, state); | |
1da177e4 LT |
353 | } |
354 | ||
355 | /* | |
356 | * Search the state->lock_states for an existing lock_owner | |
357 | * that is compatible with current->files | |
358 | */ | |
359 | static struct nfs4_lock_state * | |
360 | __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) | |
361 | { | |
362 | struct nfs4_lock_state *pos; | |
363 | list_for_each_entry(pos, &state->lock_states, ls_locks) { | |
364 | if (pos->ls_owner != fl_owner) | |
365 | continue; | |
366 | atomic_inc(&pos->ls_count); | |
367 | return pos; | |
368 | } | |
369 | return NULL; | |
370 | } | |
371 | ||
1da177e4 LT |
372 | /* |
373 | * Return a compatible lock_state. If no initialized lock_state structure | |
374 | * exists, return an uninitialized one. | |
375 | * | |
1da177e4 LT |
376 | */ |
377 | static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) | |
378 | { | |
379 | struct nfs4_lock_state *lsp; | |
adfa6f98 | 380 | struct nfs_client *clp = state->owner->so_client; |
1da177e4 | 381 | |
cee54fc9 | 382 | lsp = kzalloc(sizeof(*lsp), GFP_KERNEL); |
1da177e4 LT |
383 | if (lsp == NULL) |
384 | return NULL; | |
cee54fc9 | 385 | lsp->ls_seqid.sequence = &state->owner->so_sequence; |
1da177e4 LT |
386 | atomic_set(&lsp->ls_count, 1); |
387 | lsp->ls_owner = fl_owner; | |
1da177e4 LT |
388 | spin_lock(&clp->cl_lock); |
389 | lsp->ls_id = nfs4_alloc_lockowner_id(clp); | |
390 | spin_unlock(&clp->cl_lock); | |
8d0a8a9d | 391 | INIT_LIST_HEAD(&lsp->ls_locks); |
1da177e4 LT |
392 | return lsp; |
393 | } | |
394 | ||
395 | /* | |
396 | * Return a compatible lock_state. If no initialized lock_state structure | |
397 | * exists, return an uninitialized one. | |
398 | * | |
e6dfa553 | 399 | * The caller must be holding clp->cl_sem |
1da177e4 | 400 | */ |
8d0a8a9d | 401 | static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) |
1da177e4 | 402 | { |
8d0a8a9d | 403 | struct nfs4_lock_state *lsp, *new = NULL; |
1da177e4 | 404 | |
8d0a8a9d TM |
405 | for(;;) { |
406 | spin_lock(&state->state_lock); | |
407 | lsp = __nfs4_find_lock_state(state, owner); | |
408 | if (lsp != NULL) | |
409 | break; | |
410 | if (new != NULL) { | |
411 | new->ls_state = state; | |
412 | list_add(&new->ls_locks, &state->lock_states); | |
413 | set_bit(LK_STATE_IN_USE, &state->flags); | |
414 | lsp = new; | |
415 | new = NULL; | |
416 | break; | |
417 | } | |
418 | spin_unlock(&state->state_lock); | |
419 | new = nfs4_alloc_lock_state(state, owner); | |
420 | if (new == NULL) | |
421 | return NULL; | |
422 | } | |
423 | spin_unlock(&state->state_lock); | |
424 | kfree(new); | |
1da177e4 LT |
425 | return lsp; |
426 | } | |
427 | ||
428 | /* | |
8d0a8a9d TM |
429 | * Release reference to lock_state, and free it if we see that |
430 | * it is no longer in use | |
1da177e4 | 431 | */ |
faf5f49c | 432 | void nfs4_put_lock_state(struct nfs4_lock_state *lsp) |
1da177e4 | 433 | { |
8d0a8a9d | 434 | struct nfs4_state *state; |
1da177e4 | 435 | |
8d0a8a9d TM |
436 | if (lsp == NULL) |
437 | return; | |
438 | state = lsp->ls_state; | |
439 | if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock)) | |
440 | return; | |
441 | list_del(&lsp->ls_locks); | |
442 | if (list_empty(&state->lock_states)) | |
443 | clear_bit(LK_STATE_IN_USE, &state->flags); | |
444 | spin_unlock(&state->state_lock); | |
445 | kfree(lsp); | |
1da177e4 LT |
446 | } |
447 | ||
8d0a8a9d | 448 | static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src) |
1da177e4 | 449 | { |
8d0a8a9d | 450 | struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner; |
1da177e4 | 451 | |
8d0a8a9d TM |
452 | dst->fl_u.nfs4_fl.owner = lsp; |
453 | atomic_inc(&lsp->ls_count); | |
454 | } | |
1da177e4 | 455 | |
8d0a8a9d | 456 | static void nfs4_fl_release_lock(struct file_lock *fl) |
1da177e4 | 457 | { |
8d0a8a9d | 458 | nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner); |
1da177e4 LT |
459 | } |
460 | ||
8d0a8a9d TM |
461 | static struct file_lock_operations nfs4_fl_lock_ops = { |
462 | .fl_copy_lock = nfs4_fl_copy_lock, | |
463 | .fl_release_private = nfs4_fl_release_lock, | |
464 | }; | |
465 | ||
466 | int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl) | |
1da177e4 | 467 | { |
8d0a8a9d TM |
468 | struct nfs4_lock_state *lsp; |
469 | ||
470 | if (fl->fl_ops != NULL) | |
471 | return 0; | |
472 | lsp = nfs4_get_lock_state(state, fl->fl_owner); | |
473 | if (lsp == NULL) | |
474 | return -ENOMEM; | |
475 | fl->fl_u.nfs4_fl.owner = lsp; | |
476 | fl->fl_ops = &nfs4_fl_lock_ops; | |
477 | return 0; | |
1da177e4 LT |
478 | } |
479 | ||
8d0a8a9d TM |
480 | /* |
481 | * Byte-range lock aware utility to initialize the stateid of read/write | |
482 | * requests. | |
1da177e4 | 483 | */ |
8d0a8a9d | 484 | void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner) |
1da177e4 | 485 | { |
8d0a8a9d | 486 | struct nfs4_lock_state *lsp; |
1da177e4 | 487 | |
8d0a8a9d TM |
488 | memcpy(dst, &state->stateid, sizeof(*dst)); |
489 | if (test_bit(LK_STATE_IN_USE, &state->flags) == 0) | |
490 | return; | |
1da177e4 | 491 | |
8d0a8a9d TM |
492 | spin_lock(&state->state_lock); |
493 | lsp = __nfs4_find_lock_state(state, fl_owner); | |
494 | if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) | |
495 | memcpy(dst, &lsp->ls_stateid, sizeof(*dst)); | |
496 | spin_unlock(&state->state_lock); | |
1da177e4 LT |
497 | nfs4_put_lock_state(lsp); |
498 | } | |
499 | ||
cee54fc9 TM |
500 | struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) |
501 | { | |
36f20c6d | 502 | struct rpc_sequence *sequence = counter->sequence; |
cee54fc9 TM |
503 | struct nfs_seqid *new; |
504 | ||
505 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
506 | if (new != NULL) { | |
507 | new->sequence = counter; | |
36f20c6d TM |
508 | spin_lock(&sequence->lock); |
509 | list_add_tail(&new->list, &sequence->list); | |
510 | spin_unlock(&sequence->lock); | |
cee54fc9 TM |
511 | } |
512 | return new; | |
513 | } | |
514 | ||
515 | void nfs_free_seqid(struct nfs_seqid *seqid) | |
1da177e4 | 516 | { |
cee54fc9 | 517 | struct rpc_sequence *sequence = seqid->sequence->sequence; |
cee54fc9 | 518 | |
36f20c6d TM |
519 | spin_lock(&sequence->lock); |
520 | list_del(&seqid->list); | |
521 | spin_unlock(&sequence->lock); | |
522 | rpc_wake_up(&sequence->wait); | |
cee54fc9 | 523 | kfree(seqid); |
1da177e4 LT |
524 | } |
525 | ||
526 | /* | |
cee54fc9 TM |
527 | * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or |
528 | * failed with a seqid incrementing error - | |
529 | * see comments nfs_fs.h:seqid_mutating_error() | |
530 | */ | |
88d90939 | 531 | static void nfs_increment_seqid(int status, struct nfs_seqid *seqid) |
cee54fc9 TM |
532 | { |
533 | switch (status) { | |
534 | case 0: | |
535 | break; | |
536 | case -NFS4ERR_BAD_SEQID: | |
537 | case -NFS4ERR_STALE_CLIENTID: | |
538 | case -NFS4ERR_STALE_STATEID: | |
539 | case -NFS4ERR_BAD_STATEID: | |
540 | case -NFS4ERR_BADXDR: | |
541 | case -NFS4ERR_RESOURCE: | |
542 | case -NFS4ERR_NOFILEHANDLE: | |
543 | /* Non-seqid mutating errors */ | |
544 | return; | |
545 | }; | |
546 | /* | |
547 | * Note: no locking needed as we are guaranteed to be first | |
548 | * on the sequence list | |
549 | */ | |
550 | seqid->sequence->counter++; | |
551 | } | |
552 | ||
553 | void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) | |
554 | { | |
555 | if (status == -NFS4ERR_BAD_SEQID) { | |
556 | struct nfs4_state_owner *sp = container_of(seqid->sequence, | |
557 | struct nfs4_state_owner, so_seqid); | |
1da177e4 | 558 | nfs4_drop_state_owner(sp); |
cee54fc9 | 559 | } |
88d90939 | 560 | nfs_increment_seqid(status, seqid); |
cee54fc9 TM |
561 | } |
562 | ||
563 | /* | |
cee54fc9 TM |
564 | * Increment the seqid if the LOCK/LOCKU succeeded, or |
565 | * failed with a seqid incrementing error - | |
566 | * see comments nfs_fs.h:seqid_mutating_error() | |
567 | */ | |
568 | void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid) | |
569 | { | |
88d90939 | 570 | nfs_increment_seqid(status, seqid); |
cee54fc9 TM |
571 | } |
572 | ||
573 | int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) | |
574 | { | |
575 | struct rpc_sequence *sequence = seqid->sequence->sequence; | |
576 | int status = 0; | |
577 | ||
4e51336a TM |
578 | if (sequence->list.next == &seqid->list) |
579 | goto out; | |
cee54fc9 | 580 | spin_lock(&sequence->lock); |
36f20c6d | 581 | if (sequence->list.next != &seqid->list) { |
cee54fc9 TM |
582 | rpc_sleep_on(&sequence->wait, task, NULL, NULL); |
583 | status = -EAGAIN; | |
36f20c6d | 584 | } |
cee54fc9 | 585 | spin_unlock(&sequence->lock); |
4e51336a | 586 | out: |
cee54fc9 | 587 | return status; |
1da177e4 LT |
588 | } |
589 | ||
590 | static int reclaimer(void *); | |
1da177e4 | 591 | |
adfa6f98 | 592 | static inline void nfs4_clear_recover_bit(struct nfs_client *clp) |
433fbe4c TM |
593 | { |
594 | smp_mb__before_clear_bit(); | |
595 | clear_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state); | |
596 | smp_mb__after_clear_bit(); | |
597 | wake_up_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER); | |
598 | rpc_wake_up(&clp->cl_rpcwaitq); | |
599 | } | |
600 | ||
1da177e4 LT |
601 | /* |
602 | * State recovery routine | |
603 | */ | |
adfa6f98 | 604 | static void nfs4_recover_state(struct nfs_client *clp) |
1da177e4 | 605 | { |
5043e900 | 606 | struct task_struct *task; |
1da177e4 | 607 | |
5043e900 TM |
608 | __module_get(THIS_MODULE); |
609 | atomic_inc(&clp->cl_count); | |
610 | task = kthread_run(reclaimer, clp, "%u.%u.%u.%u-reclaim", | |
24c8dbbb | 611 | NIPQUAD(clp->cl_addr.sin_addr)); |
5043e900 TM |
612 | if (!IS_ERR(task)) |
613 | return; | |
433fbe4c | 614 | nfs4_clear_recover_bit(clp); |
24c8dbbb | 615 | nfs_put_client(clp); |
5043e900 | 616 | module_put(THIS_MODULE); |
1da177e4 LT |
617 | } |
618 | ||
619 | /* | |
620 | * Schedule a state recovery attempt | |
621 | */ | |
adfa6f98 | 622 | void nfs4_schedule_state_recovery(struct nfs_client *clp) |
1da177e4 LT |
623 | { |
624 | if (!clp) | |
625 | return; | |
433fbe4c | 626 | if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0) |
5043e900 | 627 | nfs4_recover_state(clp); |
1da177e4 LT |
628 | } |
629 | ||
630 | static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_state *state) | |
631 | { | |
632 | struct inode *inode = state->inode; | |
633 | struct file_lock *fl; | |
634 | int status = 0; | |
635 | ||
636 | for (fl = inode->i_flock; fl != 0; fl = fl->fl_next) { | |
43b2a33a | 637 | if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK))) |
1da177e4 LT |
638 | continue; |
639 | if (((struct nfs_open_context *)fl->fl_file->private_data)->state != state) | |
640 | continue; | |
641 | status = ops->recover_lock(state, fl); | |
642 | if (status >= 0) | |
643 | continue; | |
644 | switch (status) { | |
645 | default: | |
646 | printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n", | |
647 | __FUNCTION__, status); | |
648 | case -NFS4ERR_EXPIRED: | |
649 | case -NFS4ERR_NO_GRACE: | |
650 | case -NFS4ERR_RECLAIM_BAD: | |
651 | case -NFS4ERR_RECLAIM_CONFLICT: | |
43b2a33a | 652 | /* kill_proc(fl->fl_pid, SIGLOST, 1); */ |
1da177e4 LT |
653 | break; |
654 | case -NFS4ERR_STALE_CLIENTID: | |
655 | goto out_err; | |
656 | } | |
657 | } | |
658 | return 0; | |
659 | out_err: | |
660 | return status; | |
661 | } | |
662 | ||
663 | static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct nfs4_state_owner *sp) | |
664 | { | |
665 | struct nfs4_state *state; | |
666 | struct nfs4_lock_state *lock; | |
667 | int status = 0; | |
668 | ||
669 | /* Note: we rely on the sp->so_states list being ordered | |
670 | * so that we always reclaim open(O_RDWR) and/or open(O_WRITE) | |
671 | * states first. | |
672 | * This is needed to ensure that the server won't give us any | |
673 | * read delegations that we have to return if, say, we are | |
674 | * recovering after a network partition or a reboot from a | |
675 | * server that doesn't support a grace period. | |
676 | */ | |
677 | list_for_each_entry(state, &sp->so_states, open_states) { | |
678 | if (state->state == 0) | |
679 | continue; | |
680 | status = ops->recover_open(sp, state); | |
1da177e4 LT |
681 | if (status >= 0) { |
682 | status = nfs4_reclaim_locks(ops, state); | |
683 | if (status < 0) | |
684 | goto out_err; | |
685 | list_for_each_entry(lock, &state->lock_states, ls_locks) { | |
686 | if (!(lock->ls_flags & NFS_LOCK_INITIALIZED)) | |
687 | printk("%s: Lock reclaim failed!\n", | |
688 | __FUNCTION__); | |
689 | } | |
690 | continue; | |
691 | } | |
692 | switch (status) { | |
693 | default: | |
694 | printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n", | |
695 | __FUNCTION__, status); | |
696 | case -ENOENT: | |
697 | case -NFS4ERR_RECLAIM_BAD: | |
698 | case -NFS4ERR_RECLAIM_CONFLICT: | |
699 | /* | |
700 | * Open state on this file cannot be recovered | |
701 | * All we can do is revert to using the zero stateid. | |
702 | */ | |
703 | memset(state->stateid.data, 0, | |
704 | sizeof(state->stateid.data)); | |
705 | /* Mark the file as being 'closed' */ | |
706 | state->state = 0; | |
707 | break; | |
708 | case -NFS4ERR_EXPIRED: | |
709 | case -NFS4ERR_NO_GRACE: | |
710 | case -NFS4ERR_STALE_CLIENTID: | |
711 | goto out_err; | |
712 | } | |
713 | } | |
714 | return 0; | |
715 | out_err: | |
716 | return status; | |
717 | } | |
718 | ||
adfa6f98 | 719 | static void nfs4_state_mark_reclaim(struct nfs_client *clp) |
cee54fc9 TM |
720 | { |
721 | struct nfs4_state_owner *sp; | |
722 | struct nfs4_state *state; | |
723 | struct nfs4_lock_state *lock; | |
724 | ||
725 | /* Reset all sequence ids to zero */ | |
726 | list_for_each_entry(sp, &clp->cl_state_owners, so_list) { | |
727 | sp->so_seqid.counter = 0; | |
728 | sp->so_seqid.flags = 0; | |
ec073428 | 729 | spin_lock(&sp->so_lock); |
cee54fc9 TM |
730 | list_for_each_entry(state, &sp->so_states, open_states) { |
731 | list_for_each_entry(lock, &state->lock_states, ls_locks) { | |
732 | lock->ls_seqid.counter = 0; | |
733 | lock->ls_seqid.flags = 0; | |
734 | lock->ls_flags &= ~NFS_LOCK_INITIALIZED; | |
735 | } | |
736 | } | |
ec073428 | 737 | spin_unlock(&sp->so_lock); |
cee54fc9 TM |
738 | } |
739 | } | |
740 | ||
1da177e4 LT |
741 | static int reclaimer(void *ptr) |
742 | { | |
adfa6f98 | 743 | struct nfs_client *clp = ptr; |
1da177e4 LT |
744 | struct nfs4_state_owner *sp; |
745 | struct nfs4_state_recovery_ops *ops; | |
286d7d6a | 746 | struct rpc_cred *cred; |
1da177e4 LT |
747 | int status = 0; |
748 | ||
1da177e4 LT |
749 | allow_signal(SIGKILL); |
750 | ||
1da177e4 LT |
751 | /* Ensure exclusive access to NFSv4 state */ |
752 | lock_kernel(); | |
753 | down_write(&clp->cl_sem); | |
754 | /* Are there any NFS mounts out there? */ | |
755 | if (list_empty(&clp->cl_superblocks)) | |
756 | goto out; | |
757 | restart_loop: | |
286d7d6a TM |
758 | ops = &nfs4_network_partition_recovery_ops; |
759 | /* Are there any open files on this volume? */ | |
760 | cred = nfs4_get_renew_cred(clp); | |
761 | if (cred != NULL) { | |
762 | /* Yes there are: try to renew the old lease */ | |
763 | status = nfs4_proc_renew(clp, cred); | |
764 | switch (status) { | |
765 | case 0: | |
766 | case -NFS4ERR_CB_PATH_DOWN: | |
767 | put_rpccred(cred); | |
768 | goto out; | |
769 | case -NFS4ERR_STALE_CLIENTID: | |
770 | case -NFS4ERR_LEASE_MOVED: | |
771 | ops = &nfs4_reboot_recovery_ops; | |
772 | } | |
773 | } else { | |
774 | /* "reboot" to ensure we clear all state on the server */ | |
775 | clp->cl_boot_time = CURRENT_TIME; | |
776 | cred = nfs4_get_setclientid_cred(clp); | |
777 | } | |
778 | /* We're going to have to re-establish a clientid */ | |
cee54fc9 | 779 | nfs4_state_mark_reclaim(clp); |
286d7d6a TM |
780 | status = -ENOENT; |
781 | if (cred != NULL) { | |
782 | status = nfs4_init_client(clp, cred); | |
783 | put_rpccred(cred); | |
784 | } | |
1da177e4 LT |
785 | if (status) |
786 | goto out_error; | |
787 | /* Mark all delegations for reclaim */ | |
788 | nfs_delegation_mark_reclaim(clp); | |
789 | /* Note: list is protected by exclusive lock on cl->cl_sem */ | |
790 | list_for_each_entry(sp, &clp->cl_state_owners, so_list) { | |
791 | status = nfs4_reclaim_open_state(ops, sp); | |
792 | if (status < 0) { | |
793 | if (status == -NFS4ERR_NO_GRACE) { | |
794 | ops = &nfs4_network_partition_recovery_ops; | |
795 | status = nfs4_reclaim_open_state(ops, sp); | |
796 | } | |
797 | if (status == -NFS4ERR_STALE_CLIENTID) | |
798 | goto restart_loop; | |
799 | if (status == -NFS4ERR_EXPIRED) | |
800 | goto restart_loop; | |
801 | } | |
802 | } | |
803 | nfs_delegation_reap_unclaimed(clp); | |
804 | out: | |
1da177e4 LT |
805 | up_write(&clp->cl_sem); |
806 | unlock_kernel(); | |
1da177e4 LT |
807 | if (status == -NFS4ERR_CB_PATH_DOWN) |
808 | nfs_handle_cb_pathdown(clp); | |
433fbe4c | 809 | nfs4_clear_recover_bit(clp); |
24c8dbbb | 810 | nfs_put_client(clp); |
5043e900 | 811 | module_put_and_exit(0); |
1da177e4 LT |
812 | return 0; |
813 | out_error: | |
814 | printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n", | |
24c8dbbb | 815 | NIPQUAD(clp->cl_addr.sin_addr), -status); |
51581f3b | 816 | set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state); |
1da177e4 LT |
817 | goto out; |
818 | } | |
819 | ||
820 | /* | |
821 | * Local variables: | |
822 | * c-basic-offset: 8 | |
823 | * End: | |
824 | */ |