]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/dlm/lockspace.c
Merge remote-tracking branches 'asoc/topic/fsl-ssi', 'asoc/topic/hi6220' and 'asoc...
[mirror_ubuntu-artful-kernel.git] / fs / dlm / lockspace.c
1 /******************************************************************************
2 *******************************************************************************
3 **
4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
5 ** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
6 **
7 ** This copyrighted material is made available to anyone wishing to use,
8 ** modify, copy, or redistribute it subject to the terms and conditions
9 ** of the GNU General Public License v.2.
10 **
11 *******************************************************************************
12 ******************************************************************************/
13
14 #include <linux/module.h>
15
16 #include "dlm_internal.h"
17 #include "lockspace.h"
18 #include "member.h"
19 #include "recoverd.h"
20 #include "dir.h"
21 #include "lowcomms.h"
22 #include "config.h"
23 #include "memory.h"
24 #include "lock.h"
25 #include "recover.h"
26 #include "requestqueue.h"
27 #include "user.h"
28 #include "ast.h"
29
30 static int ls_count;
31 static struct mutex ls_lock;
32 static struct list_head lslist;
33 static spinlock_t lslist_lock;
34 static struct task_struct * scand_task;
35
36
37 static ssize_t dlm_control_store(struct dlm_ls *ls, const char *buf, size_t len)
38 {
39 ssize_t ret = len;
40 int n;
41 int rc = kstrtoint(buf, 0, &n);
42
43 if (rc)
44 return rc;
45 ls = dlm_find_lockspace_local(ls->ls_local_handle);
46 if (!ls)
47 return -EINVAL;
48
49 switch (n) {
50 case 0:
51 dlm_ls_stop(ls);
52 break;
53 case 1:
54 dlm_ls_start(ls);
55 break;
56 default:
57 ret = -EINVAL;
58 }
59 dlm_put_lockspace(ls);
60 return ret;
61 }
62
63 static ssize_t dlm_event_store(struct dlm_ls *ls, const char *buf, size_t len)
64 {
65 int rc = kstrtoint(buf, 0, &ls->ls_uevent_result);
66
67 if (rc)
68 return rc;
69 set_bit(LSFL_UEVENT_WAIT, &ls->ls_flags);
70 wake_up(&ls->ls_uevent_wait);
71 return len;
72 }
73
74 static ssize_t dlm_id_show(struct dlm_ls *ls, char *buf)
75 {
76 return snprintf(buf, PAGE_SIZE, "%u\n", ls->ls_global_id);
77 }
78
79 static ssize_t dlm_id_store(struct dlm_ls *ls, const char *buf, size_t len)
80 {
81 int rc = kstrtouint(buf, 0, &ls->ls_global_id);
82
83 if (rc)
84 return rc;
85 return len;
86 }
87
88 static ssize_t dlm_nodir_show(struct dlm_ls *ls, char *buf)
89 {
90 return snprintf(buf, PAGE_SIZE, "%u\n", dlm_no_directory(ls));
91 }
92
93 static ssize_t dlm_nodir_store(struct dlm_ls *ls, const char *buf, size_t len)
94 {
95 int val;
96 int rc = kstrtoint(buf, 0, &val);
97
98 if (rc)
99 return rc;
100 if (val == 1)
101 set_bit(LSFL_NODIR, &ls->ls_flags);
102 return len;
103 }
104
105 static ssize_t dlm_recover_status_show(struct dlm_ls *ls, char *buf)
106 {
107 uint32_t status = dlm_recover_status(ls);
108 return snprintf(buf, PAGE_SIZE, "%x\n", status);
109 }
110
111 static ssize_t dlm_recover_nodeid_show(struct dlm_ls *ls, char *buf)
112 {
113 return snprintf(buf, PAGE_SIZE, "%d\n", ls->ls_recover_nodeid);
114 }
115
116 struct dlm_attr {
117 struct attribute attr;
118 ssize_t (*show)(struct dlm_ls *, char *);
119 ssize_t (*store)(struct dlm_ls *, const char *, size_t);
120 };
121
122 static struct dlm_attr dlm_attr_control = {
123 .attr = {.name = "control", .mode = S_IWUSR},
124 .store = dlm_control_store
125 };
126
127 static struct dlm_attr dlm_attr_event = {
128 .attr = {.name = "event_done", .mode = S_IWUSR},
129 .store = dlm_event_store
130 };
131
132 static struct dlm_attr dlm_attr_id = {
133 .attr = {.name = "id", .mode = S_IRUGO | S_IWUSR},
134 .show = dlm_id_show,
135 .store = dlm_id_store
136 };
137
138 static struct dlm_attr dlm_attr_nodir = {
139 .attr = {.name = "nodir", .mode = S_IRUGO | S_IWUSR},
140 .show = dlm_nodir_show,
141 .store = dlm_nodir_store
142 };
143
144 static struct dlm_attr dlm_attr_recover_status = {
145 .attr = {.name = "recover_status", .mode = S_IRUGO},
146 .show = dlm_recover_status_show
147 };
148
149 static struct dlm_attr dlm_attr_recover_nodeid = {
150 .attr = {.name = "recover_nodeid", .mode = S_IRUGO},
151 .show = dlm_recover_nodeid_show
152 };
153
154 static struct attribute *dlm_attrs[] = {
155 &dlm_attr_control.attr,
156 &dlm_attr_event.attr,
157 &dlm_attr_id.attr,
158 &dlm_attr_nodir.attr,
159 &dlm_attr_recover_status.attr,
160 &dlm_attr_recover_nodeid.attr,
161 NULL,
162 };
163
164 static ssize_t dlm_attr_show(struct kobject *kobj, struct attribute *attr,
165 char *buf)
166 {
167 struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj);
168 struct dlm_attr *a = container_of(attr, struct dlm_attr, attr);
169 return a->show ? a->show(ls, buf) : 0;
170 }
171
172 static ssize_t dlm_attr_store(struct kobject *kobj, struct attribute *attr,
173 const char *buf, size_t len)
174 {
175 struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj);
176 struct dlm_attr *a = container_of(attr, struct dlm_attr, attr);
177 return a->store ? a->store(ls, buf, len) : len;
178 }
179
180 static void lockspace_kobj_release(struct kobject *k)
181 {
182 struct dlm_ls *ls = container_of(k, struct dlm_ls, ls_kobj);
183 kfree(ls);
184 }
185
186 static const struct sysfs_ops dlm_attr_ops = {
187 .show = dlm_attr_show,
188 .store = dlm_attr_store,
189 };
190
191 static struct kobj_type dlm_ktype = {
192 .default_attrs = dlm_attrs,
193 .sysfs_ops = &dlm_attr_ops,
194 .release = lockspace_kobj_release,
195 };
196
197 static struct kset *dlm_kset;
198
199 static int do_uevent(struct dlm_ls *ls, int in)
200 {
201 int error;
202
203 if (in)
204 kobject_uevent(&ls->ls_kobj, KOBJ_ONLINE);
205 else
206 kobject_uevent(&ls->ls_kobj, KOBJ_OFFLINE);
207
208 log_rinfo(ls, "%s the lockspace group...", in ? "joining" : "leaving");
209
210 /* dlm_controld will see the uevent, do the necessary group management
211 and then write to sysfs to wake us */
212
213 error = wait_event_interruptible(ls->ls_uevent_wait,
214 test_and_clear_bit(LSFL_UEVENT_WAIT, &ls->ls_flags));
215
216 log_rinfo(ls, "group event done %d %d", error, ls->ls_uevent_result);
217
218 if (error)
219 goto out;
220
221 error = ls->ls_uevent_result;
222 out:
223 if (error)
224 log_error(ls, "group %s failed %d %d", in ? "join" : "leave",
225 error, ls->ls_uevent_result);
226 return error;
227 }
228
229 static int dlm_uevent(struct kset *kset, struct kobject *kobj,
230 struct kobj_uevent_env *env)
231 {
232 struct dlm_ls *ls = container_of(kobj, struct dlm_ls, ls_kobj);
233
234 add_uevent_var(env, "LOCKSPACE=%s", ls->ls_name);
235 return 0;
236 }
237
238 static struct kset_uevent_ops dlm_uevent_ops = {
239 .uevent = dlm_uevent,
240 };
241
242 int __init dlm_lockspace_init(void)
243 {
244 ls_count = 0;
245 mutex_init(&ls_lock);
246 INIT_LIST_HEAD(&lslist);
247 spin_lock_init(&lslist_lock);
248
249 dlm_kset = kset_create_and_add("dlm", &dlm_uevent_ops, kernel_kobj);
250 if (!dlm_kset) {
251 printk(KERN_WARNING "%s: can not create kset\n", __func__);
252 return -ENOMEM;
253 }
254 return 0;
255 }
256
257 void dlm_lockspace_exit(void)
258 {
259 kset_unregister(dlm_kset);
260 }
261
262 static struct dlm_ls *find_ls_to_scan(void)
263 {
264 struct dlm_ls *ls;
265
266 spin_lock(&lslist_lock);
267 list_for_each_entry(ls, &lslist, ls_list) {
268 if (time_after_eq(jiffies, ls->ls_scan_time +
269 dlm_config.ci_scan_secs * HZ)) {
270 spin_unlock(&lslist_lock);
271 return ls;
272 }
273 }
274 spin_unlock(&lslist_lock);
275 return NULL;
276 }
277
278 static int dlm_scand(void *data)
279 {
280 struct dlm_ls *ls;
281
282 while (!kthread_should_stop()) {
283 ls = find_ls_to_scan();
284 if (ls) {
285 if (dlm_lock_recovery_try(ls)) {
286 ls->ls_scan_time = jiffies;
287 dlm_scan_rsbs(ls);
288 dlm_scan_timeout(ls);
289 dlm_scan_waiters(ls);
290 dlm_unlock_recovery(ls);
291 } else {
292 ls->ls_scan_time += HZ;
293 }
294 continue;
295 }
296 schedule_timeout_interruptible(dlm_config.ci_scan_secs * HZ);
297 }
298 return 0;
299 }
300
301 static int dlm_scand_start(void)
302 {
303 struct task_struct *p;
304 int error = 0;
305
306 p = kthread_run(dlm_scand, NULL, "dlm_scand");
307 if (IS_ERR(p))
308 error = PTR_ERR(p);
309 else
310 scand_task = p;
311 return error;
312 }
313
314 static void dlm_scand_stop(void)
315 {
316 kthread_stop(scand_task);
317 }
318
319 struct dlm_ls *dlm_find_lockspace_global(uint32_t id)
320 {
321 struct dlm_ls *ls;
322
323 spin_lock(&lslist_lock);
324
325 list_for_each_entry(ls, &lslist, ls_list) {
326 if (ls->ls_global_id == id) {
327 ls->ls_count++;
328 goto out;
329 }
330 }
331 ls = NULL;
332 out:
333 spin_unlock(&lslist_lock);
334 return ls;
335 }
336
337 struct dlm_ls *dlm_find_lockspace_local(dlm_lockspace_t *lockspace)
338 {
339 struct dlm_ls *ls;
340
341 spin_lock(&lslist_lock);
342 list_for_each_entry(ls, &lslist, ls_list) {
343 if (ls->ls_local_handle == lockspace) {
344 ls->ls_count++;
345 goto out;
346 }
347 }
348 ls = NULL;
349 out:
350 spin_unlock(&lslist_lock);
351 return ls;
352 }
353
354 struct dlm_ls *dlm_find_lockspace_device(int minor)
355 {
356 struct dlm_ls *ls;
357
358 spin_lock(&lslist_lock);
359 list_for_each_entry(ls, &lslist, ls_list) {
360 if (ls->ls_device.minor == minor) {
361 ls->ls_count++;
362 goto out;
363 }
364 }
365 ls = NULL;
366 out:
367 spin_unlock(&lslist_lock);
368 return ls;
369 }
370
371 void dlm_put_lockspace(struct dlm_ls *ls)
372 {
373 spin_lock(&lslist_lock);
374 ls->ls_count--;
375 spin_unlock(&lslist_lock);
376 }
377
378 static void remove_lockspace(struct dlm_ls *ls)
379 {
380 for (;;) {
381 spin_lock(&lslist_lock);
382 if (ls->ls_count == 0) {
383 WARN_ON(ls->ls_create_count != 0);
384 list_del(&ls->ls_list);
385 spin_unlock(&lslist_lock);
386 return;
387 }
388 spin_unlock(&lslist_lock);
389 ssleep(1);
390 }
391 }
392
393 static int threads_start(void)
394 {
395 int error;
396
397 error = dlm_scand_start();
398 if (error) {
399 log_print("cannot start dlm_scand thread %d", error);
400 goto fail;
401 }
402
403 /* Thread for sending/receiving messages for all lockspace's */
404 error = dlm_lowcomms_start();
405 if (error) {
406 log_print("cannot start dlm lowcomms %d", error);
407 goto scand_fail;
408 }
409
410 return 0;
411
412 scand_fail:
413 dlm_scand_stop();
414 fail:
415 return error;
416 }
417
418 static void threads_stop(void)
419 {
420 dlm_scand_stop();
421 dlm_lowcomms_stop();
422 }
423
424 static int new_lockspace(const char *name, const char *cluster,
425 uint32_t flags, int lvblen,
426 const struct dlm_lockspace_ops *ops, void *ops_arg,
427 int *ops_result, dlm_lockspace_t **lockspace)
428 {
429 struct dlm_ls *ls;
430 int i, size, error;
431 int do_unreg = 0;
432 int namelen = strlen(name);
433
434 if (namelen > DLM_LOCKSPACE_LEN)
435 return -EINVAL;
436
437 if (!lvblen || (lvblen % 8))
438 return -EINVAL;
439
440 if (!try_module_get(THIS_MODULE))
441 return -EINVAL;
442
443 if (!dlm_user_daemon_available()) {
444 log_print("dlm user daemon not available");
445 error = -EUNATCH;
446 goto out;
447 }
448
449 if (ops && ops_result) {
450 if (!dlm_config.ci_recover_callbacks)
451 *ops_result = -EOPNOTSUPP;
452 else
453 *ops_result = 0;
454 }
455
456 if (dlm_config.ci_recover_callbacks && cluster &&
457 strncmp(cluster, dlm_config.ci_cluster_name, DLM_LOCKSPACE_LEN)) {
458 log_print("dlm cluster name %s mismatch %s",
459 dlm_config.ci_cluster_name, cluster);
460 error = -EBADR;
461 goto out;
462 }
463
464 error = 0;
465
466 spin_lock(&lslist_lock);
467 list_for_each_entry(ls, &lslist, ls_list) {
468 WARN_ON(ls->ls_create_count <= 0);
469 if (ls->ls_namelen != namelen)
470 continue;
471 if (memcmp(ls->ls_name, name, namelen))
472 continue;
473 if (flags & DLM_LSFL_NEWEXCL) {
474 error = -EEXIST;
475 break;
476 }
477 ls->ls_create_count++;
478 *lockspace = ls;
479 error = 1;
480 break;
481 }
482 spin_unlock(&lslist_lock);
483
484 if (error)
485 goto out;
486
487 error = -ENOMEM;
488
489 ls = kzalloc(sizeof(struct dlm_ls) + namelen, GFP_NOFS);
490 if (!ls)
491 goto out;
492 memcpy(ls->ls_name, name, namelen);
493 ls->ls_namelen = namelen;
494 ls->ls_lvblen = lvblen;
495 ls->ls_count = 0;
496 ls->ls_flags = 0;
497 ls->ls_scan_time = jiffies;
498
499 if (ops && dlm_config.ci_recover_callbacks) {
500 ls->ls_ops = ops;
501 ls->ls_ops_arg = ops_arg;
502 }
503
504 if (flags & DLM_LSFL_TIMEWARN)
505 set_bit(LSFL_TIMEWARN, &ls->ls_flags);
506
507 /* ls_exflags are forced to match among nodes, and we don't
508 need to require all nodes to have some flags set */
509 ls->ls_exflags = (flags & ~(DLM_LSFL_TIMEWARN | DLM_LSFL_FS |
510 DLM_LSFL_NEWEXCL));
511
512 size = dlm_config.ci_rsbtbl_size;
513 ls->ls_rsbtbl_size = size;
514
515 ls->ls_rsbtbl = vmalloc(sizeof(struct dlm_rsbtable) * size);
516 if (!ls->ls_rsbtbl)
517 goto out_lsfree;
518 for (i = 0; i < size; i++) {
519 ls->ls_rsbtbl[i].keep.rb_node = NULL;
520 ls->ls_rsbtbl[i].toss.rb_node = NULL;
521 spin_lock_init(&ls->ls_rsbtbl[i].lock);
522 }
523
524 spin_lock_init(&ls->ls_remove_spin);
525
526 for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) {
527 ls->ls_remove_names[i] = kzalloc(DLM_RESNAME_MAXLEN+1,
528 GFP_KERNEL);
529 if (!ls->ls_remove_names[i])
530 goto out_rsbtbl;
531 }
532
533 idr_init(&ls->ls_lkbidr);
534 spin_lock_init(&ls->ls_lkbidr_spin);
535
536 INIT_LIST_HEAD(&ls->ls_waiters);
537 mutex_init(&ls->ls_waiters_mutex);
538 INIT_LIST_HEAD(&ls->ls_orphans);
539 mutex_init(&ls->ls_orphans_mutex);
540 INIT_LIST_HEAD(&ls->ls_timeout);
541 mutex_init(&ls->ls_timeout_mutex);
542
543 INIT_LIST_HEAD(&ls->ls_new_rsb);
544 spin_lock_init(&ls->ls_new_rsb_spin);
545
546 INIT_LIST_HEAD(&ls->ls_nodes);
547 INIT_LIST_HEAD(&ls->ls_nodes_gone);
548 ls->ls_num_nodes = 0;
549 ls->ls_low_nodeid = 0;
550 ls->ls_total_weight = 0;
551 ls->ls_node_array = NULL;
552
553 memset(&ls->ls_stub_rsb, 0, sizeof(struct dlm_rsb));
554 ls->ls_stub_rsb.res_ls = ls;
555
556 ls->ls_debug_rsb_dentry = NULL;
557 ls->ls_debug_waiters_dentry = NULL;
558
559 init_waitqueue_head(&ls->ls_uevent_wait);
560 ls->ls_uevent_result = 0;
561 init_completion(&ls->ls_members_done);
562 ls->ls_members_result = -1;
563
564 mutex_init(&ls->ls_cb_mutex);
565 INIT_LIST_HEAD(&ls->ls_cb_delay);
566
567 ls->ls_recoverd_task = NULL;
568 mutex_init(&ls->ls_recoverd_active);
569 spin_lock_init(&ls->ls_recover_lock);
570 spin_lock_init(&ls->ls_rcom_spin);
571 get_random_bytes(&ls->ls_rcom_seq, sizeof(uint64_t));
572 ls->ls_recover_status = 0;
573 ls->ls_recover_seq = 0;
574 ls->ls_recover_args = NULL;
575 init_rwsem(&ls->ls_in_recovery);
576 init_rwsem(&ls->ls_recv_active);
577 INIT_LIST_HEAD(&ls->ls_requestqueue);
578 mutex_init(&ls->ls_requestqueue_mutex);
579 mutex_init(&ls->ls_clear_proc_locks);
580
581 ls->ls_recover_buf = kmalloc(dlm_config.ci_buffer_size, GFP_NOFS);
582 if (!ls->ls_recover_buf)
583 goto out_lkbidr;
584
585 ls->ls_slot = 0;
586 ls->ls_num_slots = 0;
587 ls->ls_slots_size = 0;
588 ls->ls_slots = NULL;
589
590 INIT_LIST_HEAD(&ls->ls_recover_list);
591 spin_lock_init(&ls->ls_recover_list_lock);
592 idr_init(&ls->ls_recover_idr);
593 spin_lock_init(&ls->ls_recover_idr_lock);
594 ls->ls_recover_list_count = 0;
595 ls->ls_local_handle = ls;
596 init_waitqueue_head(&ls->ls_wait_general);
597 INIT_LIST_HEAD(&ls->ls_root_list);
598 init_rwsem(&ls->ls_root_sem);
599
600 spin_lock(&lslist_lock);
601 ls->ls_create_count = 1;
602 list_add(&ls->ls_list, &lslist);
603 spin_unlock(&lslist_lock);
604
605 if (flags & DLM_LSFL_FS) {
606 error = dlm_callback_start(ls);
607 if (error) {
608 log_error(ls, "can't start dlm_callback %d", error);
609 goto out_delist;
610 }
611 }
612
613 init_waitqueue_head(&ls->ls_recover_lock_wait);
614
615 /*
616 * Once started, dlm_recoverd first looks for ls in lslist, then
617 * initializes ls_in_recovery as locked in "down" mode. We need
618 * to wait for the wakeup from dlm_recoverd because in_recovery
619 * has to start out in down mode.
620 */
621
622 error = dlm_recoverd_start(ls);
623 if (error) {
624 log_error(ls, "can't start dlm_recoverd %d", error);
625 goto out_callback;
626 }
627
628 wait_event(ls->ls_recover_lock_wait,
629 test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags));
630
631 ls->ls_kobj.kset = dlm_kset;
632 error = kobject_init_and_add(&ls->ls_kobj, &dlm_ktype, NULL,
633 "%s", ls->ls_name);
634 if (error)
635 goto out_recoverd;
636 kobject_uevent(&ls->ls_kobj, KOBJ_ADD);
637
638 /* let kobject handle freeing of ls if there's an error */
639 do_unreg = 1;
640
641 /* This uevent triggers dlm_controld in userspace to add us to the
642 group of nodes that are members of this lockspace (managed by the
643 cluster infrastructure.) Once it's done that, it tells us who the
644 current lockspace members are (via configfs) and then tells the
645 lockspace to start running (via sysfs) in dlm_ls_start(). */
646
647 error = do_uevent(ls, 1);
648 if (error)
649 goto out_recoverd;
650
651 wait_for_completion(&ls->ls_members_done);
652 error = ls->ls_members_result;
653 if (error)
654 goto out_members;
655
656 dlm_create_debug_file(ls);
657
658 log_rinfo(ls, "join complete");
659 *lockspace = ls;
660 return 0;
661
662 out_members:
663 do_uevent(ls, 0);
664 dlm_clear_members(ls);
665 kfree(ls->ls_node_array);
666 out_recoverd:
667 dlm_recoverd_stop(ls);
668 out_callback:
669 dlm_callback_stop(ls);
670 out_delist:
671 spin_lock(&lslist_lock);
672 list_del(&ls->ls_list);
673 spin_unlock(&lslist_lock);
674 idr_destroy(&ls->ls_recover_idr);
675 kfree(ls->ls_recover_buf);
676 out_lkbidr:
677 idr_destroy(&ls->ls_lkbidr);
678 for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) {
679 if (ls->ls_remove_names[i])
680 kfree(ls->ls_remove_names[i]);
681 }
682 out_rsbtbl:
683 vfree(ls->ls_rsbtbl);
684 out_lsfree:
685 if (do_unreg)
686 kobject_put(&ls->ls_kobj);
687 else
688 kfree(ls);
689 out:
690 module_put(THIS_MODULE);
691 return error;
692 }
693
694 int dlm_new_lockspace(const char *name, const char *cluster,
695 uint32_t flags, int lvblen,
696 const struct dlm_lockspace_ops *ops, void *ops_arg,
697 int *ops_result, dlm_lockspace_t **lockspace)
698 {
699 int error = 0;
700
701 mutex_lock(&ls_lock);
702 if (!ls_count)
703 error = threads_start();
704 if (error)
705 goto out;
706
707 error = new_lockspace(name, cluster, flags, lvblen, ops, ops_arg,
708 ops_result, lockspace);
709 if (!error)
710 ls_count++;
711 if (error > 0)
712 error = 0;
713 if (!ls_count)
714 threads_stop();
715 out:
716 mutex_unlock(&ls_lock);
717 return error;
718 }
719
720 static int lkb_idr_is_local(int id, void *p, void *data)
721 {
722 struct dlm_lkb *lkb = p;
723
724 return lkb->lkb_nodeid == 0 && lkb->lkb_grmode != DLM_LOCK_IV;
725 }
726
727 static int lkb_idr_is_any(int id, void *p, void *data)
728 {
729 return 1;
730 }
731
732 static int lkb_idr_free(int id, void *p, void *data)
733 {
734 struct dlm_lkb *lkb = p;
735
736 if (lkb->lkb_lvbptr && lkb->lkb_flags & DLM_IFL_MSTCPY)
737 dlm_free_lvb(lkb->lkb_lvbptr);
738
739 dlm_free_lkb(lkb);
740 return 0;
741 }
742
743 /* NOTE: We check the lkbidr here rather than the resource table.
744 This is because there may be LKBs queued as ASTs that have been unlinked
745 from their RSBs and are pending deletion once the AST has been delivered */
746
747 static int lockspace_busy(struct dlm_ls *ls, int force)
748 {
749 int rv;
750
751 spin_lock(&ls->ls_lkbidr_spin);
752 if (force == 0) {
753 rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_any, ls);
754 } else if (force == 1) {
755 rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_local, ls);
756 } else {
757 rv = 0;
758 }
759 spin_unlock(&ls->ls_lkbidr_spin);
760 return rv;
761 }
762
763 static int release_lockspace(struct dlm_ls *ls, int force)
764 {
765 struct dlm_rsb *rsb;
766 struct rb_node *n;
767 int i, busy, rv;
768
769 busy = lockspace_busy(ls, force);
770
771 spin_lock(&lslist_lock);
772 if (ls->ls_create_count == 1) {
773 if (busy) {
774 rv = -EBUSY;
775 } else {
776 /* remove_lockspace takes ls off lslist */
777 ls->ls_create_count = 0;
778 rv = 0;
779 }
780 } else if (ls->ls_create_count > 1) {
781 rv = --ls->ls_create_count;
782 } else {
783 rv = -EINVAL;
784 }
785 spin_unlock(&lslist_lock);
786
787 if (rv) {
788 log_debug(ls, "release_lockspace no remove %d", rv);
789 return rv;
790 }
791
792 dlm_device_deregister(ls);
793
794 if (force < 3 && dlm_user_daemon_available())
795 do_uevent(ls, 0);
796
797 dlm_recoverd_stop(ls);
798
799 dlm_callback_stop(ls);
800
801 remove_lockspace(ls);
802
803 dlm_delete_debug_file(ls);
804
805 kfree(ls->ls_recover_buf);
806
807 /*
808 * Free all lkb's in idr
809 */
810
811 idr_for_each(&ls->ls_lkbidr, lkb_idr_free, ls);
812 idr_destroy(&ls->ls_lkbidr);
813
814 /*
815 * Free all rsb's on rsbtbl[] lists
816 */
817
818 for (i = 0; i < ls->ls_rsbtbl_size; i++) {
819 while ((n = rb_first(&ls->ls_rsbtbl[i].keep))) {
820 rsb = rb_entry(n, struct dlm_rsb, res_hashnode);
821 rb_erase(n, &ls->ls_rsbtbl[i].keep);
822 dlm_free_rsb(rsb);
823 }
824
825 while ((n = rb_first(&ls->ls_rsbtbl[i].toss))) {
826 rsb = rb_entry(n, struct dlm_rsb, res_hashnode);
827 rb_erase(n, &ls->ls_rsbtbl[i].toss);
828 dlm_free_rsb(rsb);
829 }
830 }
831
832 vfree(ls->ls_rsbtbl);
833
834 for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++)
835 kfree(ls->ls_remove_names[i]);
836
837 while (!list_empty(&ls->ls_new_rsb)) {
838 rsb = list_first_entry(&ls->ls_new_rsb, struct dlm_rsb,
839 res_hashchain);
840 list_del(&rsb->res_hashchain);
841 dlm_free_rsb(rsb);
842 }
843
844 /*
845 * Free structures on any other lists
846 */
847
848 dlm_purge_requestqueue(ls);
849 kfree(ls->ls_recover_args);
850 dlm_clear_members(ls);
851 dlm_clear_members_gone(ls);
852 kfree(ls->ls_node_array);
853 log_rinfo(ls, "release_lockspace final free");
854 kobject_put(&ls->ls_kobj);
855 /* The ls structure will be freed when the kobject is done with */
856
857 module_put(THIS_MODULE);
858 return 0;
859 }
860
861 /*
862 * Called when a system has released all its locks and is not going to use the
863 * lockspace any longer. We free everything we're managing for this lockspace.
864 * Remaining nodes will go through the recovery process as if we'd died. The
865 * lockspace must continue to function as usual, participating in recoveries,
866 * until this returns.
867 *
868 * Force has 4 possible values:
869 * 0 - don't destroy locksapce if it has any LKBs
870 * 1 - destroy lockspace if it has remote LKBs but not if it has local LKBs
871 * 2 - destroy lockspace regardless of LKBs
872 * 3 - destroy lockspace as part of a forced shutdown
873 */
874
875 int dlm_release_lockspace(void *lockspace, int force)
876 {
877 struct dlm_ls *ls;
878 int error;
879
880 ls = dlm_find_lockspace_local(lockspace);
881 if (!ls)
882 return -EINVAL;
883 dlm_put_lockspace(ls);
884
885 mutex_lock(&ls_lock);
886 error = release_lockspace(ls, force);
887 if (!error)
888 ls_count--;
889 if (!ls_count)
890 threads_stop();
891 mutex_unlock(&ls_lock);
892
893 return error;
894 }
895
896 void dlm_stop_lockspaces(void)
897 {
898 struct dlm_ls *ls;
899 int count;
900
901 restart:
902 count = 0;
903 spin_lock(&lslist_lock);
904 list_for_each_entry(ls, &lslist, ls_list) {
905 if (!test_bit(LSFL_RUNNING, &ls->ls_flags)) {
906 count++;
907 continue;
908 }
909 spin_unlock(&lslist_lock);
910 log_error(ls, "no userland control daemon, stopping lockspace");
911 dlm_ls_stop(ls);
912 goto restart;
913 }
914 spin_unlock(&lslist_lock);
915
916 if (count)
917 log_print("dlm user daemon left %d lockspaces", count);
918 }
919