Commit | Line | Data |
---|---|---|
2522fe45 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
e7fd4179 DT |
2 | /****************************************************************************** |
3 | ******************************************************************************* | |
4 | ** | |
5 | ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
60f98d18 | 6 | ** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved. |
e7fd4179 | 7 | ** |
e7fd4179 DT |
8 | ** |
9 | ******************************************************************************* | |
10 | ******************************************************************************/ | |
11 | ||
7963b8a5 PG |
12 | #include <linux/module.h> |
13 | ||
e7fd4179 DT |
14 | #include "dlm_internal.h" |
15 | #include "lockspace.h" | |
16 | #include "member.h" | |
17 | #include "recoverd.h" | |
e7fd4179 DT |
18 | #include "dir.h" |
19 | #include "lowcomms.h" | |
20 | #include "config.h" | |
21 | #include "memory.h" | |
22 | #include "lock.h" | |
c56b39cd | 23 | #include "recover.h" |
2896ee37 | 24 | #include "requestqueue.h" |
0f8e0d9a | 25 | #include "user.h" |
23e8e1aa | 26 | #include "ast.h" |
e7fd4179 | 27 | |
e7fd4179 | 28 | static int ls_count; |
90135925 | 29 | static struct mutex ls_lock; |
e7fd4179 DT |
30 | static struct list_head lslist; |
31 | static spinlock_t lslist_lock; | |
32 | static struct task_struct * scand_task; | |
33 | ||
34 | ||
35 | static ssize_t dlm_control_store(struct dlm_ls *ls, const char *buf, size_t len) | |
36 | { | |
37 | ssize_t ret = len; | |
6edb5687 FF |
38 | int n; |
39 | int rc = kstrtoint(buf, 0, &n); | |
e7fd4179 | 40 | |
6edb5687 FF |
41 | if (rc) |
42 | return rc; | |
e2de7f56 PC |
43 | ls = dlm_find_lockspace_local(ls->ls_local_handle); |
44 | if (!ls) | |
45 | return -EINVAL; | |
46 | ||
e7fd4179 DT |
47 | switch (n) { |
48 | case 0: | |
49 | dlm_ls_stop(ls); | |
50 | break; | |
51 | case 1: | |
52 | dlm_ls_start(ls); | |
53 | break; | |
54 | default: | |
55 | ret = -EINVAL; | |
56 | } | |
e2de7f56 | 57 | dlm_put_lockspace(ls); |
e7fd4179 DT |
58 | return ret; |
59 | } | |
60 | ||
61 | static ssize_t dlm_event_store(struct dlm_ls *ls, const char *buf, size_t len) | |
62 | { | |
6edb5687 FF |
63 | int rc = kstrtoint(buf, 0, &ls->ls_uevent_result); |
64 | ||
65 | if (rc) | |
66 | return rc; | |
e7fd4179 DT |
67 | set_bit(LSFL_UEVENT_WAIT, &ls->ls_flags); |
68 | wake_up(&ls->ls_uevent_wait); | |
69 | return len; | |
70 | } | |
71 | ||
72 | static ssize_t dlm_id_show(struct dlm_ls *ls, char *buf) | |
73 | { | |
a1d144c7 | 74 | return snprintf(buf, PAGE_SIZE, "%u\n", ls->ls_global_id); |
e7fd4179 DT |
75 | } |
76 | ||
77 | static ssize_t dlm_id_store(struct dlm_ls *ls, const char *buf, size_t len) | |
78 | { | |
6edb5687 FF |
79 | int rc = kstrtouint(buf, 0, &ls->ls_global_id); |
80 | ||
81 | if (rc) | |
82 | return rc; | |
e7fd4179 DT |
83 | return len; |
84 | } | |
85 | ||
4875647a DT |
86 | static ssize_t dlm_nodir_show(struct dlm_ls *ls, char *buf) |
87 | { | |
88 | return snprintf(buf, PAGE_SIZE, "%u\n", dlm_no_directory(ls)); | |
89 | } | |
90 | ||
91 | static ssize_t dlm_nodir_store(struct dlm_ls *ls, const char *buf, size_t len) | |
92 | { | |
6edb5687 FF |
93 | int val; |
94 | int rc = kstrtoint(buf, 0, &val); | |
95 | ||
96 | if (rc) | |
97 | return rc; | |
4875647a DT |
98 | if (val == 1) |
99 | set_bit(LSFL_NODIR, &ls->ls_flags); | |
100 | return len; | |
101 | } | |
102 | ||
c56b39cd DT |
103 | static ssize_t dlm_recover_status_show(struct dlm_ls *ls, char *buf) |
104 | { | |
105 | uint32_t status = dlm_recover_status(ls); | |
a1d144c7 | 106 | return snprintf(buf, PAGE_SIZE, "%x\n", status); |
c56b39cd DT |
107 | } |
108 | ||
faa0f267 DT |
109 | static ssize_t dlm_recover_nodeid_show(struct dlm_ls *ls, char *buf) |
110 | { | |
a1d144c7 | 111 | return snprintf(buf, PAGE_SIZE, "%d\n", ls->ls_recover_nodeid); |
faa0f267 DT |
112 | } |
113 | ||
e7fd4179 DT |
114 | struct dlm_attr { |
115 | struct attribute attr; | |
116 | ssize_t (*show)(struct dlm_ls *, char *); | |
117 | ssize_t (*store)(struct dlm_ls *, const char *, size_t); | |
118 | }; | |
119 | ||
120 | static struct dlm_attr dlm_attr_control = { | |
121 | .attr = {.name = "control", .mode = S_IWUSR}, | |
122 | .store = dlm_control_store | |
123 | }; | |
124 | ||
125 | static struct dlm_attr dlm_attr_event = { | |
126 | .attr = {.name = "event_done", .mode = S_IWUSR}, | |
127 | .store = dlm_event_store | |
128 | }; | |
129 | ||
130 | static struct dlm_attr dlm_attr_id = { | |
131 | .attr = {.name = "id", .mode = S_IRUGO | S_IWUSR}, | |
132 | .show = dlm_id_show, | |
133 | .store = dlm_id_store | |
134 | }; | |
135 | ||
4875647a DT |
136 | static struct dlm_attr dlm_attr_nodir = { |
137 | .attr = {.name = "nodir", .mode = S_IRUGO | S_IWUSR}, | |
138 | .show = dlm_nodir_show, | |
139 | .store = dlm_nodir_store | |
140 | }; | |
141 | ||
c56b39cd DT |
142 | static struct dlm_attr dlm_attr_recover_status = { |
143 | .attr = {.name = "recover_status", .mode = S_IRUGO}, | |
144 | .show = dlm_recover_status_show | |
145 | }; | |
146 | ||
faa0f267 DT |
147 | static struct dlm_attr dlm_attr_recover_nodeid = { |
148 | .attr = {.name = "recover_nodeid", .mode = S_IRUGO}, | |
149 | .show = dlm_recover_nodeid_show | |
150 | }; | |
151 | ||
e7fd4179 DT |
152 | static struct attribute *dlm_attrs[] = { |
153 | &dlm_attr_control.attr, | |
154 | &dlm_attr_event.attr, | |
155 | &dlm_attr_id.attr, | |
4875647a | 156 | &dlm_attr_nodir.attr, |
c56b39cd | 157 | &dlm_attr_recover_status.attr, |
faa0f267 | 158 | &dlm_attr_recover_nodeid.attr, |
e7fd4179 DT |
159 | NULL, |
160 | }; | |
c9c5b5e1 | 161 | ATTRIBUTE_GROUPS(dlm); |
e7fd4179 DT |
162 | |
163 | static ssize_t dlm_attr_show(struct kobject *kobj, struct attribute *attr, | |
164 | char *buf) | |
165 | { | |
166 | struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj); | |
167 | struct dlm_attr *a = container_of(attr, struct dlm_attr, attr); | |
168 | return a->show ? a->show(ls, buf) : 0; | |
169 | } | |
170 | ||
171 | static ssize_t dlm_attr_store(struct kobject *kobj, struct attribute *attr, | |
172 | const char *buf, size_t len) | |
173 | { | |
174 | struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj); | |
175 | struct dlm_attr *a = container_of(attr, struct dlm_attr, attr); | |
176 | return a->store ? a->store(ls, buf, len) : len; | |
177 | } | |
178 | ||
ba542e3b PC |
179 | static void lockspace_kobj_release(struct kobject *k) |
180 | { | |
181 | struct dlm_ls *ls = container_of(k, struct dlm_ls, ls_kobj); | |
182 | kfree(ls); | |
183 | } | |
184 | ||
52cf25d0 | 185 | static const struct sysfs_ops dlm_attr_ops = { |
e7fd4179 DT |
186 | .show = dlm_attr_show, |
187 | .store = dlm_attr_store, | |
188 | }; | |
189 | ||
190 | static struct kobj_type dlm_ktype = { | |
c9c5b5e1 | 191 | .default_groups = dlm_groups, |
e7fd4179 | 192 | .sysfs_ops = &dlm_attr_ops, |
ba542e3b | 193 | .release = lockspace_kobj_release, |
e7fd4179 DT |
194 | }; |
195 | ||
d405936b | 196 | static struct kset *dlm_kset; |
e7fd4179 | 197 | |
e7fd4179 DT |
198 | static int do_uevent(struct dlm_ls *ls, int in) |
199 | { | |
e7fd4179 DT |
200 | if (in) |
201 | kobject_uevent(&ls->ls_kobj, KOBJ_ONLINE); | |
202 | else | |
203 | kobject_uevent(&ls->ls_kobj, KOBJ_OFFLINE); | |
204 | ||
075f0177 | 205 | log_rinfo(ls, "%s the lockspace group...", in ? "joining" : "leaving"); |
8b0e7b2c DT |
206 | |
207 | /* dlm_controld will see the uevent, do the necessary group management | |
208 | and then write to sysfs to wake us */ | |
209 | ||
f084a4f4 RL |
210 | wait_event(ls->ls_uevent_wait, |
211 | test_and_clear_bit(LSFL_UEVENT_WAIT, &ls->ls_flags)); | |
8b0e7b2c | 212 | |
f084a4f4 | 213 | log_rinfo(ls, "group event done %d", ls->ls_uevent_result); |
e7fd4179 | 214 | |
f084a4f4 | 215 | return ls->ls_uevent_result; |
e7fd4179 DT |
216 | } |
217 | ||
b4a5d4bc SW |
218 | static int dlm_uevent(struct kset *kset, struct kobject *kobj, |
219 | struct kobj_uevent_env *env) | |
220 | { | |
221 | struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj); | |
222 | ||
223 | add_uevent_var(env, "LOCKSPACE=%s", ls->ls_name); | |
224 | return 0; | |
225 | } | |
226 | ||
417f7c59 | 227 | static const struct kset_uevent_ops dlm_uevent_ops = { |
b4a5d4bc SW |
228 | .uevent = dlm_uevent, |
229 | }; | |
e7fd4179 | 230 | |
30727174 | 231 | int __init dlm_lockspace_init(void) |
e7fd4179 | 232 | { |
e7fd4179 | 233 | ls_count = 0; |
90135925 | 234 | mutex_init(&ls_lock); |
e7fd4179 DT |
235 | INIT_LIST_HEAD(&lslist); |
236 | spin_lock_init(&lslist_lock); | |
237 | ||
b4a5d4bc | 238 | dlm_kset = kset_create_and_add("dlm", &dlm_uevent_ops, kernel_kobj); |
d405936b | 239 | if (!dlm_kset) { |
8e24eea7 | 240 | printk(KERN_WARNING "%s: can not create kset\n", __func__); |
d405936b GKH |
241 | return -ENOMEM; |
242 | } | |
243 | return 0; | |
e7fd4179 DT |
244 | } |
245 | ||
246 | void dlm_lockspace_exit(void) | |
247 | { | |
d405936b | 248 | kset_unregister(dlm_kset); |
e7fd4179 DT |
249 | } |
250 | ||
c1dcf65f DT |
251 | static struct dlm_ls *find_ls_to_scan(void) |
252 | { | |
253 | struct dlm_ls *ls; | |
254 | ||
255 | spin_lock(&lslist_lock); | |
256 | list_for_each_entry(ls, &lslist, ls_list) { | |
257 | if (time_after_eq(jiffies, ls->ls_scan_time + | |
258 | dlm_config.ci_scan_secs * HZ)) { | |
259 | spin_unlock(&lslist_lock); | |
260 | return ls; | |
261 | } | |
262 | } | |
263 | spin_unlock(&lslist_lock); | |
264 | return NULL; | |
265 | } | |
266 | ||
e7fd4179 DT |
267 | static int dlm_scand(void *data) |
268 | { | |
269 | struct dlm_ls *ls; | |
270 | ||
271 | while (!kthread_should_stop()) { | |
c1dcf65f DT |
272 | ls = find_ls_to_scan(); |
273 | if (ls) { | |
85e86edf | 274 | if (dlm_lock_recovery_try(ls)) { |
c1dcf65f | 275 | ls->ls_scan_time = jiffies; |
85e86edf | 276 | dlm_scan_rsbs(ls); |
3ae1acf9 | 277 | dlm_scan_timeout(ls); |
c6ff669b | 278 | dlm_scan_waiters(ls); |
85e86edf | 279 | dlm_unlock_recovery(ls); |
c1dcf65f DT |
280 | } else { |
281 | ls->ls_scan_time += HZ; | |
85e86edf | 282 | } |
c6ff669b | 283 | continue; |
85e86edf | 284 | } |
c6ff669b | 285 | schedule_timeout_interruptible(dlm_config.ci_scan_secs * HZ); |
e7fd4179 DT |
286 | } |
287 | return 0; | |
288 | } | |
289 | ||
290 | static int dlm_scand_start(void) | |
291 | { | |
292 | struct task_struct *p; | |
293 | int error = 0; | |
294 | ||
295 | p = kthread_run(dlm_scand, NULL, "dlm_scand"); | |
296 | if (IS_ERR(p)) | |
297 | error = PTR_ERR(p); | |
298 | else | |
299 | scand_task = p; | |
300 | return error; | |
301 | } | |
302 | ||
303 | static void dlm_scand_stop(void) | |
304 | { | |
305 | kthread_stop(scand_task); | |
306 | } | |
307 | ||
e7fd4179 DT |
308 | struct dlm_ls *dlm_find_lockspace_global(uint32_t id) |
309 | { | |
310 | struct dlm_ls *ls; | |
311 | ||
312 | spin_lock(&lslist_lock); | |
313 | ||
314 | list_for_each_entry(ls, &lslist, ls_list) { | |
315 | if (ls->ls_global_id == id) { | |
316 | ls->ls_count++; | |
317 | goto out; | |
318 | } | |
319 | } | |
320 | ls = NULL; | |
321 | out: | |
322 | spin_unlock(&lslist_lock); | |
323 | return ls; | |
324 | } | |
325 | ||
597d0cae | 326 | struct dlm_ls *dlm_find_lockspace_local(dlm_lockspace_t *lockspace) |
e7fd4179 | 327 | { |
597d0cae | 328 | struct dlm_ls *ls; |
e7fd4179 DT |
329 | |
330 | spin_lock(&lslist_lock); | |
597d0cae DT |
331 | list_for_each_entry(ls, &lslist, ls_list) { |
332 | if (ls->ls_local_handle == lockspace) { | |
333 | ls->ls_count++; | |
334 | goto out; | |
335 | } | |
336 | } | |
337 | ls = NULL; | |
338 | out: | |
339 | spin_unlock(&lslist_lock); | |
340 | return ls; | |
341 | } | |
342 | ||
343 | struct dlm_ls *dlm_find_lockspace_device(int minor) | |
344 | { | |
345 | struct dlm_ls *ls; | |
346 | ||
347 | spin_lock(&lslist_lock); | |
348 | list_for_each_entry(ls, &lslist, ls_list) { | |
349 | if (ls->ls_device.minor == minor) { | |
350 | ls->ls_count++; | |
351 | goto out; | |
352 | } | |
353 | } | |
354 | ls = NULL; | |
355 | out: | |
e7fd4179 DT |
356 | spin_unlock(&lslist_lock); |
357 | return ls; | |
358 | } | |
359 | ||
360 | void dlm_put_lockspace(struct dlm_ls *ls) | |
361 | { | |
362 | spin_lock(&lslist_lock); | |
363 | ls->ls_count--; | |
364 | spin_unlock(&lslist_lock); | |
365 | } | |
366 | ||
367 | static void remove_lockspace(struct dlm_ls *ls) | |
368 | { | |
369 | for (;;) { | |
370 | spin_lock(&lslist_lock); | |
371 | if (ls->ls_count == 0) { | |
0f8e0d9a | 372 | WARN_ON(ls->ls_create_count != 0); |
e7fd4179 DT |
373 | list_del(&ls->ls_list); |
374 | spin_unlock(&lslist_lock); | |
375 | return; | |
376 | } | |
377 | spin_unlock(&lslist_lock); | |
378 | ssleep(1); | |
379 | } | |
380 | } | |
381 | ||
382 | static int threads_start(void) | |
383 | { | |
384 | int error; | |
385 | ||
e7fd4179 DT |
386 | error = dlm_scand_start(); |
387 | if (error) { | |
388 | log_print("cannot start dlm_scand thread %d", error); | |
23e8e1aa | 389 | goto fail; |
e7fd4179 DT |
390 | } |
391 | ||
392 | /* Thread for sending/receiving messages for all lockspace's */ | |
393 | error = dlm_lowcomms_start(); | |
394 | if (error) { | |
395 | log_print("cannot start dlm lowcomms %d", error); | |
396 | goto scand_fail; | |
397 | } | |
398 | ||
399 | return 0; | |
400 | ||
401 | scand_fail: | |
402 | dlm_scand_stop(); | |
e7fd4179 DT |
403 | fail: |
404 | return error; | |
405 | } | |
406 | ||
60f98d18 DT |
407 | static int new_lockspace(const char *name, const char *cluster, |
408 | uint32_t flags, int lvblen, | |
409 | const struct dlm_lockspace_ops *ops, void *ops_arg, | |
410 | int *ops_result, dlm_lockspace_t **lockspace) | |
e7fd4179 DT |
411 | { |
412 | struct dlm_ls *ls; | |
0f8e0d9a | 413 | int i, size, error; |
79d72b54 | 414 | int do_unreg = 0; |
60f98d18 | 415 | int namelen = strlen(name); |
e7fd4179 | 416 | |
3f0806d2 | 417 | if (namelen > DLM_LOCKSPACE_LEN || namelen == 0) |
e7fd4179 DT |
418 | return -EINVAL; |
419 | ||
420 | if (!lvblen || (lvblen % 8)) | |
421 | return -EINVAL; | |
422 | ||
423 | if (!try_module_get(THIS_MODULE)) | |
424 | return -EINVAL; | |
425 | ||
dc68c7ed | 426 | if (!dlm_user_daemon_available()) { |
60f98d18 DT |
427 | log_print("dlm user daemon not available"); |
428 | error = -EUNATCH; | |
429 | goto out; | |
430 | } | |
431 | ||
432 | if (ops && ops_result) { | |
433 | if (!dlm_config.ci_recover_callbacks) | |
434 | *ops_result = -EOPNOTSUPP; | |
435 | else | |
436 | *ops_result = 0; | |
437 | } | |
438 | ||
3b0e761b ZL |
439 | if (!cluster) |
440 | log_print("dlm cluster name '%s' is being used without an application provided cluster name", | |
441 | dlm_config.ci_cluster_name); | |
442 | ||
60f98d18 DT |
443 | if (dlm_config.ci_recover_callbacks && cluster && |
444 | strncmp(cluster, dlm_config.ci_cluster_name, DLM_LOCKSPACE_LEN)) { | |
8e174374 GH |
445 | log_print("dlm cluster name '%s' does not match " |
446 | "the application cluster name '%s'", | |
60f98d18 DT |
447 | dlm_config.ci_cluster_name, cluster); |
448 | error = -EBADR; | |
449 | goto out; | |
dc68c7ed DT |
450 | } |
451 | ||
0f8e0d9a DT |
452 | error = 0; |
453 | ||
454 | spin_lock(&lslist_lock); | |
455 | list_for_each_entry(ls, &lslist, ls_list) { | |
456 | WARN_ON(ls->ls_create_count <= 0); | |
457 | if (ls->ls_namelen != namelen) | |
458 | continue; | |
459 | if (memcmp(ls->ls_name, name, namelen)) | |
460 | continue; | |
461 | if (flags & DLM_LSFL_NEWEXCL) { | |
462 | error = -EEXIST; | |
463 | break; | |
464 | } | |
465 | ls->ls_create_count++; | |
8511a272 DT |
466 | *lockspace = ls; |
467 | error = 1; | |
0f8e0d9a | 468 | break; |
e7fd4179 | 469 | } |
0f8e0d9a DT |
470 | spin_unlock(&lslist_lock); |
471 | ||
0f8e0d9a | 472 | if (error) |
8511a272 | 473 | goto out; |
0f8e0d9a DT |
474 | |
475 | error = -ENOMEM; | |
e7fd4179 | 476 | |
573c24c4 | 477 | ls = kzalloc(sizeof(struct dlm_ls) + namelen, GFP_NOFS); |
e7fd4179 DT |
478 | if (!ls) |
479 | goto out; | |
e7fd4179 DT |
480 | memcpy(ls->ls_name, name, namelen); |
481 | ls->ls_namelen = namelen; | |
e7fd4179 DT |
482 | ls->ls_lvblen = lvblen; |
483 | ls->ls_count = 0; | |
484 | ls->ls_flags = 0; | |
c1dcf65f | 485 | ls->ls_scan_time = jiffies; |
e7fd4179 | 486 | |
60f98d18 DT |
487 | if (ops && dlm_config.ci_recover_callbacks) { |
488 | ls->ls_ops = ops; | |
489 | ls->ls_ops_arg = ops_arg; | |
490 | } | |
491 | ||
3ae1acf9 DT |
492 | if (flags & DLM_LSFL_TIMEWARN) |
493 | set_bit(LSFL_TIMEWARN, &ls->ls_flags); | |
3ae1acf9 | 494 | |
fad59c13 | 495 | /* ls_exflags are forced to match among nodes, and we don't |
0f8e0d9a DT |
496 | need to require all nodes to have some flags set */ |
497 | ls->ls_exflags = (flags & ~(DLM_LSFL_TIMEWARN | DLM_LSFL_FS | | |
498 | DLM_LSFL_NEWEXCL)); | |
fad59c13 | 499 | |
68c817a1 | 500 | size = dlm_config.ci_rsbtbl_size; |
e7fd4179 DT |
501 | ls->ls_rsbtbl_size = size; |
502 | ||
42bc47b3 | 503 | ls->ls_rsbtbl = vmalloc(array_size(size, sizeof(struct dlm_rsbtable))); |
e7fd4179 DT |
504 | if (!ls->ls_rsbtbl) |
505 | goto out_lsfree; | |
506 | for (i = 0; i < size; i++) { | |
9beb3bf5 BP |
507 | ls->ls_rsbtbl[i].keep.rb_node = NULL; |
508 | ls->ls_rsbtbl[i].toss.rb_node = NULL; | |
c7be761a | 509 | spin_lock_init(&ls->ls_rsbtbl[i].lock); |
e7fd4179 DT |
510 | } |
511 | ||
05c32f47 DT |
512 | spin_lock_init(&ls->ls_remove_spin); |
513 | ||
514 | for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) { | |
515 | ls->ls_remove_names[i] = kzalloc(DLM_RESNAME_MAXLEN+1, | |
516 | GFP_KERNEL); | |
517 | if (!ls->ls_remove_names[i]) | |
518 | goto out_rsbtbl; | |
519 | } | |
520 | ||
3d6aa675 DT |
521 | idr_init(&ls->ls_lkbidr); |
522 | spin_lock_init(&ls->ls_lkbidr_spin); | |
e7fd4179 | 523 | |
e7fd4179 | 524 | INIT_LIST_HEAD(&ls->ls_waiters); |
90135925 | 525 | mutex_init(&ls->ls_waiters_mutex); |
ef0c2bb0 DT |
526 | INIT_LIST_HEAD(&ls->ls_orphans); |
527 | mutex_init(&ls->ls_orphans_mutex); | |
3ae1acf9 DT |
528 | INIT_LIST_HEAD(&ls->ls_timeout); |
529 | mutex_init(&ls->ls_timeout_mutex); | |
e7fd4179 | 530 | |
3881ac04 DT |
531 | INIT_LIST_HEAD(&ls->ls_new_rsb); |
532 | spin_lock_init(&ls->ls_new_rsb_spin); | |
533 | ||
e7fd4179 DT |
534 | INIT_LIST_HEAD(&ls->ls_nodes); |
535 | INIT_LIST_HEAD(&ls->ls_nodes_gone); | |
536 | ls->ls_num_nodes = 0; | |
537 | ls->ls_low_nodeid = 0; | |
538 | ls->ls_total_weight = 0; | |
539 | ls->ls_node_array = NULL; | |
540 | ||
541 | memset(&ls->ls_stub_rsb, 0, sizeof(struct dlm_rsb)); | |
542 | ls->ls_stub_rsb.res_ls = ls; | |
543 | ||
5de6319b DT |
544 | ls->ls_debug_rsb_dentry = NULL; |
545 | ls->ls_debug_waiters_dentry = NULL; | |
e7fd4179 DT |
546 | |
547 | init_waitqueue_head(&ls->ls_uevent_wait); | |
548 | ls->ls_uevent_result = 0; | |
8b0e7b2c DT |
549 | init_completion(&ls->ls_members_done); |
550 | ls->ls_members_result = -1; | |
e7fd4179 | 551 | |
23e8e1aa DT |
552 | mutex_init(&ls->ls_cb_mutex); |
553 | INIT_LIST_HEAD(&ls->ls_cb_delay); | |
554 | ||
e7fd4179 | 555 | ls->ls_recoverd_task = NULL; |
90135925 | 556 | mutex_init(&ls->ls_recoverd_active); |
e7fd4179 | 557 | spin_lock_init(&ls->ls_recover_lock); |
98f176fb DT |
558 | spin_lock_init(&ls->ls_rcom_spin); |
559 | get_random_bytes(&ls->ls_rcom_seq, sizeof(uint64_t)); | |
e7fd4179 DT |
560 | ls->ls_recover_status = 0; |
561 | ls->ls_recover_seq = 0; | |
562 | ls->ls_recover_args = NULL; | |
563 | init_rwsem(&ls->ls_in_recovery); | |
c36258b5 | 564 | init_rwsem(&ls->ls_recv_active); |
e7fd4179 | 565 | INIT_LIST_HEAD(&ls->ls_requestqueue); |
90135925 | 566 | mutex_init(&ls->ls_requestqueue_mutex); |
597d0cae | 567 | mutex_init(&ls->ls_clear_proc_locks); |
e7fd4179 | 568 | |
9f8f9c77 | 569 | ls->ls_recover_buf = kmalloc(LOWCOMMS_MAX_TX_BUFFER_LEN, GFP_NOFS); |
e7fd4179 | 570 | if (!ls->ls_recover_buf) |
05c32f47 | 571 | goto out_lkbidr; |
e7fd4179 | 572 | |
757a4271 DT |
573 | ls->ls_slot = 0; |
574 | ls->ls_num_slots = 0; | |
575 | ls->ls_slots_size = 0; | |
576 | ls->ls_slots = NULL; | |
577 | ||
e7fd4179 DT |
578 | INIT_LIST_HEAD(&ls->ls_recover_list); |
579 | spin_lock_init(&ls->ls_recover_list_lock); | |
1d7c484e DT |
580 | idr_init(&ls->ls_recover_idr); |
581 | spin_lock_init(&ls->ls_recover_idr_lock); | |
e7fd4179 | 582 | ls->ls_recover_list_count = 0; |
597d0cae | 583 | ls->ls_local_handle = ls; |
e7fd4179 DT |
584 | init_waitqueue_head(&ls->ls_wait_general); |
585 | INIT_LIST_HEAD(&ls->ls_root_list); | |
586 | init_rwsem(&ls->ls_root_sem); | |
587 | ||
5f88f1ea | 588 | spin_lock(&lslist_lock); |
0f8e0d9a | 589 | ls->ls_create_count = 1; |
5f88f1ea DT |
590 | list_add(&ls->ls_list, &lslist); |
591 | spin_unlock(&lslist_lock); | |
592 | ||
23e8e1aa DT |
593 | if (flags & DLM_LSFL_FS) { |
594 | error = dlm_callback_start(ls); | |
595 | if (error) { | |
596 | log_error(ls, "can't start dlm_callback %d", error); | |
597 | goto out_delist; | |
598 | } | |
599 | } | |
600 | ||
475f230c DT |
601 | init_waitqueue_head(&ls->ls_recover_lock_wait); |
602 | ||
603 | /* | |
604 | * Once started, dlm_recoverd first looks for ls in lslist, then | |
605 | * initializes ls_in_recovery as locked in "down" mode. We need | |
606 | * to wait for the wakeup from dlm_recoverd because in_recovery | |
607 | * has to start out in down mode. | |
608 | */ | |
609 | ||
e7fd4179 DT |
610 | error = dlm_recoverd_start(ls); |
611 | if (error) { | |
612 | log_error(ls, "can't start dlm_recoverd %d", error); | |
23e8e1aa | 613 | goto out_callback; |
e7fd4179 DT |
614 | } |
615 | ||
475f230c DT |
616 | wait_event(ls->ls_recover_lock_wait, |
617 | test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags)); | |
618 | ||
0ffddafc WH |
619 | /* let kobject handle freeing of ls if there's an error */ |
620 | do_unreg = 1; | |
621 | ||
901195ed GKH |
622 | ls->ls_kobj.kset = dlm_kset; |
623 | error = kobject_init_and_add(&ls->ls_kobj, &dlm_ktype, NULL, | |
624 | "%s", ls->ls_name); | |
e7fd4179 | 625 | if (error) |
23e8e1aa | 626 | goto out_recoverd; |
901195ed | 627 | kobject_uevent(&ls->ls_kobj, KOBJ_ADD); |
79d72b54 | 628 | |
8b0e7b2c DT |
629 | /* This uevent triggers dlm_controld in userspace to add us to the |
630 | group of nodes that are members of this lockspace (managed by the | |
631 | cluster infrastructure.) Once it's done that, it tells us who the | |
632 | current lockspace members are (via configfs) and then tells the | |
633 | lockspace to start running (via sysfs) in dlm_ls_start(). */ | |
634 | ||
e7fd4179 DT |
635 | error = do_uevent(ls, 1); |
636 | if (error) | |
23e8e1aa | 637 | goto out_recoverd; |
79d72b54 | 638 | |
8b0e7b2c DT |
639 | wait_for_completion(&ls->ls_members_done); |
640 | error = ls->ls_members_result; | |
641 | if (error) | |
642 | goto out_members; | |
643 | ||
79d72b54 DT |
644 | dlm_create_debug_file(ls); |
645 | ||
075f0177 | 646 | log_rinfo(ls, "join complete"); |
e7fd4179 DT |
647 | *lockspace = ls; |
648 | return 0; | |
649 | ||
8b0e7b2c DT |
650 | out_members: |
651 | do_uevent(ls, 0); | |
652 | dlm_clear_members(ls); | |
653 | kfree(ls->ls_node_array); | |
23e8e1aa | 654 | out_recoverd: |
5f88f1ea | 655 | dlm_recoverd_stop(ls); |
23e8e1aa DT |
656 | out_callback: |
657 | dlm_callback_stop(ls); | |
79d72b54 | 658 | out_delist: |
e7fd4179 DT |
659 | spin_lock(&lslist_lock); |
660 | list_del(&ls->ls_list); | |
661 | spin_unlock(&lslist_lock); | |
1d7c484e | 662 | idr_destroy(&ls->ls_recover_idr); |
e7fd4179 | 663 | kfree(ls->ls_recover_buf); |
05c32f47 | 664 | out_lkbidr: |
3d6aa675 | 665 | idr_destroy(&ls->ls_lkbidr); |
b982896c | 666 | out_rsbtbl: |
3456880f TM |
667 | for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) |
668 | kfree(ls->ls_remove_names[i]); | |
c282af49 | 669 | vfree(ls->ls_rsbtbl); |
e7fd4179 | 670 | out_lsfree: |
79d72b54 | 671 | if (do_unreg) |
197b12d6 | 672 | kobject_put(&ls->ls_kobj); |
79d72b54 DT |
673 | else |
674 | kfree(ls); | |
e7fd4179 DT |
675 | out: |
676 | module_put(THIS_MODULE); | |
677 | return error; | |
678 | } | |
679 | ||
60f98d18 DT |
680 | int dlm_new_lockspace(const char *name, const char *cluster, |
681 | uint32_t flags, int lvblen, | |
682 | const struct dlm_lockspace_ops *ops, void *ops_arg, | |
683 | int *ops_result, dlm_lockspace_t **lockspace) | |
e7fd4179 DT |
684 | { |
685 | int error = 0; | |
686 | ||
90135925 | 687 | mutex_lock(&ls_lock); |
e7fd4179 DT |
688 | if (!ls_count) |
689 | error = threads_start(); | |
690 | if (error) | |
691 | goto out; | |
692 | ||
60f98d18 DT |
693 | error = new_lockspace(name, cluster, flags, lvblen, ops, ops_arg, |
694 | ops_result, lockspace); | |
e7fd4179 DT |
695 | if (!error) |
696 | ls_count++; | |
8511a272 DT |
697 | if (error > 0) |
698 | error = 0; | |
9d232469 AA |
699 | if (!ls_count) { |
700 | dlm_scand_stop(); | |
701 | dlm_lowcomms_shutdown(); | |
702 | dlm_lowcomms_stop(); | |
703 | } | |
e7fd4179 | 704 | out: |
90135925 | 705 | mutex_unlock(&ls_lock); |
e7fd4179 DT |
706 | return error; |
707 | } | |
708 | ||
3d6aa675 | 709 | static int lkb_idr_is_local(int id, void *p, void *data) |
e7fd4179 | 710 | { |
3d6aa675 DT |
711 | struct dlm_lkb *lkb = p; |
712 | ||
a97f4a66 | 713 | return lkb->lkb_nodeid == 0 && lkb->lkb_grmode != DLM_LOCK_IV; |
3d6aa675 DT |
714 | } |
715 | ||
716 | static int lkb_idr_is_any(int id, void *p, void *data) | |
717 | { | |
718 | return 1; | |
719 | } | |
720 | ||
721 | static int lkb_idr_free(int id, void *p, void *data) | |
722 | { | |
723 | struct dlm_lkb *lkb = p; | |
724 | ||
3d6aa675 DT |
725 | if (lkb->lkb_lvbptr && lkb->lkb_flags & DLM_IFL_MSTCPY) |
726 | dlm_free_lvb(lkb->lkb_lvbptr); | |
727 | ||
728 | dlm_free_lkb(lkb); | |
729 | return 0; | |
730 | } | |
731 | ||
732 | /* NOTE: We check the lkbidr here rather than the resource table. | |
733 | This is because there may be LKBs queued as ASTs that have been unlinked | |
734 | from their RSBs and are pending deletion once the AST has been delivered */ | |
735 | ||
736 | static int lockspace_busy(struct dlm_ls *ls, int force) | |
737 | { | |
738 | int rv; | |
739 | ||
740 | spin_lock(&ls->ls_lkbidr_spin); | |
741 | if (force == 0) { | |
742 | rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_any, ls); | |
743 | } else if (force == 1) { | |
744 | rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_local, ls); | |
745 | } else { | |
746 | rv = 0; | |
e7fd4179 | 747 | } |
3d6aa675 DT |
748 | spin_unlock(&ls->ls_lkbidr_spin); |
749 | return rv; | |
e7fd4179 DT |
750 | } |
751 | ||
752 | static int release_lockspace(struct dlm_ls *ls, int force) | |
753 | { | |
e7fd4179 | 754 | struct dlm_rsb *rsb; |
9beb3bf5 | 755 | struct rb_node *n; |
0f8e0d9a DT |
756 | int i, busy, rv; |
757 | ||
3d6aa675 | 758 | busy = lockspace_busy(ls, force); |
0f8e0d9a DT |
759 | |
760 | spin_lock(&lslist_lock); | |
761 | if (ls->ls_create_count == 1) { | |
3d6aa675 | 762 | if (busy) { |
0f8e0d9a | 763 | rv = -EBUSY; |
3d6aa675 | 764 | } else { |
0f8e0d9a DT |
765 | /* remove_lockspace takes ls off lslist */ |
766 | ls->ls_create_count = 0; | |
767 | rv = 0; | |
768 | } | |
769 | } else if (ls->ls_create_count > 1) { | |
770 | rv = --ls->ls_create_count; | |
771 | } else { | |
772 | rv = -EINVAL; | |
773 | } | |
774 | spin_unlock(&lslist_lock); | |
775 | ||
776 | if (rv) { | |
777 | log_debug(ls, "release_lockspace no remove %d", rv); | |
778 | return rv; | |
779 | } | |
e7fd4179 | 780 | |
0f8e0d9a | 781 | dlm_device_deregister(ls); |
e7fd4179 | 782 | |
dc68c7ed | 783 | if (force < 3 && dlm_user_daemon_available()) |
e7fd4179 DT |
784 | do_uevent(ls, 0); |
785 | ||
786 | dlm_recoverd_stop(ls); | |
787 | ||
9d232469 AA |
788 | if (ls_count == 1) { |
789 | dlm_scand_stop(); | |
790 | dlm_lowcomms_shutdown(); | |
791 | } | |
792 | ||
23e8e1aa DT |
793 | dlm_callback_stop(ls); |
794 | ||
e7fd4179 DT |
795 | remove_lockspace(ls); |
796 | ||
797 | dlm_delete_debug_file(ls); | |
798 | ||
8fc6ed9a | 799 | idr_destroy(&ls->ls_recover_idr); |
e7fd4179 DT |
800 | kfree(ls->ls_recover_buf); |
801 | ||
e7fd4179 | 802 | /* |
3d6aa675 | 803 | * Free all lkb's in idr |
e7fd4179 DT |
804 | */ |
805 | ||
3d6aa675 | 806 | idr_for_each(&ls->ls_lkbidr, lkb_idr_free, ls); |
3d6aa675 | 807 | idr_destroy(&ls->ls_lkbidr); |
e7fd4179 | 808 | |
e7fd4179 DT |
809 | /* |
810 | * Free all rsb's on rsbtbl[] lists | |
811 | */ | |
812 | ||
813 | for (i = 0; i < ls->ls_rsbtbl_size; i++) { | |
9beb3bf5 BP |
814 | while ((n = rb_first(&ls->ls_rsbtbl[i].keep))) { |
815 | rsb = rb_entry(n, struct dlm_rsb, res_hashnode); | |
816 | rb_erase(n, &ls->ls_rsbtbl[i].keep); | |
52bda2b5 | 817 | dlm_free_rsb(rsb); |
e7fd4179 DT |
818 | } |
819 | ||
9beb3bf5 BP |
820 | while ((n = rb_first(&ls->ls_rsbtbl[i].toss))) { |
821 | rsb = rb_entry(n, struct dlm_rsb, res_hashnode); | |
822 | rb_erase(n, &ls->ls_rsbtbl[i].toss); | |
52bda2b5 | 823 | dlm_free_rsb(rsb); |
e7fd4179 DT |
824 | } |
825 | } | |
826 | ||
c282af49 | 827 | vfree(ls->ls_rsbtbl); |
e7fd4179 | 828 | |
05c32f47 DT |
829 | for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) |
830 | kfree(ls->ls_remove_names[i]); | |
831 | ||
3881ac04 DT |
832 | while (!list_empty(&ls->ls_new_rsb)) { |
833 | rsb = list_first_entry(&ls->ls_new_rsb, struct dlm_rsb, | |
834 | res_hashchain); | |
835 | list_del(&rsb->res_hashchain); | |
836 | dlm_free_rsb(rsb); | |
837 | } | |
838 | ||
e7fd4179 DT |
839 | /* |
840 | * Free structures on any other lists | |
841 | */ | |
842 | ||
2896ee37 | 843 | dlm_purge_requestqueue(ls); |
e7fd4179 | 844 | kfree(ls->ls_recover_args); |
e7fd4179 DT |
845 | dlm_clear_members(ls); |
846 | dlm_clear_members_gone(ls); | |
847 | kfree(ls->ls_node_array); | |
075f0177 | 848 | log_rinfo(ls, "release_lockspace final free"); |
197b12d6 | 849 | kobject_put(&ls->ls_kobj); |
79d72b54 | 850 | /* The ls structure will be freed when the kobject is done with */ |
e7fd4179 | 851 | |
e7fd4179 DT |
852 | module_put(THIS_MODULE); |
853 | return 0; | |
854 | } | |
855 | ||
856 | /* | |
857 | * Called when a system has released all its locks and is not going to use the | |
858 | * lockspace any longer. We free everything we're managing for this lockspace. | |
859 | * Remaining nodes will go through the recovery process as if we'd died. The | |
860 | * lockspace must continue to function as usual, participating in recoveries, | |
861 | * until this returns. | |
862 | * | |
863 | * Force has 4 possible values: | |
864 | * 0 - don't destroy locksapce if it has any LKBs | |
865 | * 1 - destroy lockspace if it has remote LKBs but not if it has local LKBs | |
866 | * 2 - destroy lockspace regardless of LKBs | |
867 | * 3 - destroy lockspace as part of a forced shutdown | |
868 | */ | |
869 | ||
870 | int dlm_release_lockspace(void *lockspace, int force) | |
871 | { | |
872 | struct dlm_ls *ls; | |
0f8e0d9a | 873 | int error; |
e7fd4179 DT |
874 | |
875 | ls = dlm_find_lockspace_local(lockspace); | |
876 | if (!ls) | |
877 | return -EINVAL; | |
878 | dlm_put_lockspace(ls); | |
0f8e0d9a DT |
879 | |
880 | mutex_lock(&ls_lock); | |
881 | error = release_lockspace(ls, force); | |
882 | if (!error) | |
883 | ls_count--; | |
278afcbf | 884 | if (!ls_count) |
9d232469 | 885 | dlm_lowcomms_stop(); |
0f8e0d9a DT |
886 | mutex_unlock(&ls_lock); |
887 | ||
888 | return error; | |
e7fd4179 DT |
889 | } |
890 | ||
dc68c7ed DT |
891 | void dlm_stop_lockspaces(void) |
892 | { | |
893 | struct dlm_ls *ls; | |
696b3d84 | 894 | int count; |
dc68c7ed DT |
895 | |
896 | restart: | |
696b3d84 | 897 | count = 0; |
dc68c7ed DT |
898 | spin_lock(&lslist_lock); |
899 | list_for_each_entry(ls, &lslist, ls_list) { | |
696b3d84 DT |
900 | if (!test_bit(LSFL_RUNNING, &ls->ls_flags)) { |
901 | count++; | |
dc68c7ed | 902 | continue; |
696b3d84 | 903 | } |
dc68c7ed DT |
904 | spin_unlock(&lslist_lock); |
905 | log_error(ls, "no userland control daemon, stopping lockspace"); | |
906 | dlm_ls_stop(ls); | |
907 | goto restart; | |
908 | } | |
909 | spin_unlock(&lslist_lock); | |
696b3d84 DT |
910 | |
911 | if (count) | |
912 | log_print("dlm user daemon left %d lockspaces", count); | |
dc68c7ed DT |
913 | } |
914 |